Public and Private Productivity Commission

Hospitals Research Report

December 2009
© COMMONWEALTH OF AUSTRALIA 2009

ISBN 978-1-74037-297-8

This work is copyright. Apart from any use as permitted under the Copyright Act 1968, the
work may be reproduced in whole or in part for study or training purposes, subject to the
inclusion of an acknowledgment of the source. Reproduction for commercial use or sale
requires prior written permission from the Commonwealth. Requests and inquiries
concerning reproduction and rights should be addressed to the Commonwealth Copyright
Administration, Attorney-General's Department, 3-5 National Circuit, Canberra ACT 2600
or posted at www.ag.gov.au/cca.
This publication is available in hard copy or PDF format from the Productivity
Commission website at www.pc.gov.au. If you require part or all of this publication in a
different format, please contact Media and Publications (see below).

Publications Inquiries:
Media and Publications
Productivity Commission
Locked Bag 2 Collins Street East
Melbourne VIC 8003
Tel: (03) 9653 2244
Fax: (03) 9653 2303
Email: [email protected]

General Inquiries:
Tel: (03) 9653 2100 or (02) 6240 3200

An appropriate citation for this paper is:

Productivity Commission 2009, Public and Private Hospitals, Research Report, Canberra.

The Productivity Commission

The Productivity Commission is the Australian Government’s independent research
and advisory body on a range of economic, social and environmental issues affecting
the welfare of Australians. Its role, expressed most simply, is to help governments
make better policies, in the long term interest of the Australian community.
The Commission’s independence is underpinned by an Act of Parliament. Its
processes and outputs are open to public scrutiny and are driven by concern for the
wellbeing of the community as a whole.
Further information on the Productivity Commission can be obtained from the
Commission’s website (www.pc.gov.au) or by contacting Media and Publications on
(03) 9653 2244 or email: [email protected]
Foreword

Australia spends about 9 per cent of its gross domestic product on health care, and
hospital services account for around one-third of this. This report examines three
aspects of the health care system — the relative performance of public and private
hospitals; rates of informed financial consent for privately-insured patients; and the
indexation factor used for the Medicare Levy Surcharge income thresholds.

The first task — comparing the relative performance of hospitals — has been the
most challenging part of the study. Diversity within and between the public and
private hospital sectors makes like-for-like comparisons difficult, and existing data
collections are limited by inconsistent collection methods and missing information.

Foreshadowed developments under the National Healthcare Agreement are

expected to improve the quality of hospital-related data, but the Commission has
identified further potential improvements, such as consistent national reporting of
costs and infections for both public and private hospitals. This report also discusses
measures to improve data accessibility, which could ultimately facilitate
improvements in health care, especially as competitive markets only have a limited
role in the health sector.

A wide range of parties assisted the Commission in conducting the study, including
through participation in roundtables, providing technical advice, and making written
submissions. The Commission thanks all of those who contributed to the report.

The study was overseen by Commissioner David Kalisch. The staff research team
was headed by Greg Murtough and based in the Commission’s Melbourne office.

Gary Banks AO
Chairman

December 2009

FOREWORD III
Terms of reference

Productivity Commission research study into public and private

hospitals

The Productivity Commission is requested to undertake a study into public and

private hospitals, and report back within 6 months of receipt of this reference.

Context

The Government is committed to improving transparency, accountability and

performance reporting within the health system. This commitment is evident in the
new National Healthcare Agreement and in COAG’s agreement to introduce a
nationally consistent approach to activity-based funding in public hospitals. It is
also shown by the Government's commitment to move towards nationally consistent
performance reporting for public and private hospitals.

In furthering the Government’s commitment in this area, the Productivity

Commission is requested to examine and report on the relative performance of the
public and private hospital systems, and related data issues. As part of its study, the
Commission should consider:

a) comparative hospital and medical costs for clinically similar procedures

performed by public and private hospitals, using baseline data to be provided by
states and territories under the new National Healthcare Agreement, and
existing data provided to the Government by private hospitals. The analysis is
to take into account the costs of capital, FBT exemptions and other relevant
factors.

b) the rate of hospital-acquired infections, by type, reported by public and private

hospitals, using baseline data to be provided by states and territories under the
new National Healthcare Agreement, and existing data provided to the
Government by private hospitals.

c) rates of fully informed financial consent for privately insured patients treated as
private patients in both public and private hospitals, categorised by type of
provider (that is, public hospital, private hospital, medical practitioner [by
Speciality]), and by Statistical Local Area (SLA) or equivalent, including:

IV TERMS OF
REFERENCE
 c (i) the average cost of out of pocket expenses for patients who do not receive
enough financial information from the provider to give fully informed
financial consent, the range of these costs and the maximum out of pocket
cost incurred by in-hospital patients categorised by type of provider (as
detailed above).

c (ii) best practice examples where fully informed financial consent is provided
for every procedure, (with a specific emphasis on any best practice
examples occurring in specialties where lack of fully informed financial
consent is most common).

d) other relevant performance indicators, including the ability of such indicators to

inform comparisons of hospital performance and efficiency.

e) If any of the foregoing tasks prove not fully possible because of conceptual
problems and data limitations, the Commission should propose any
developments that would improve the feasibility of future comparisons.

The Commission will also provide advice to the Government on the most
appropriate indexation factor for the Medicare Levy Surcharge thresholds.

The Commission is to consult with relevant experts and others as necessary and
produce a final report within six months of receipt of this reference. The report will
be published.

CHRIS BOWEN
[Received 15 May 2009]

TERMS OF REFERENCE V
VI TERMS OF
REFERENCE
Contents

Foreword III
Terms of Reference IV
Abbreviations and explanations XVIII
Glossary XXII
Overview XXXI
Findings LV
1 Introduction 1
1.1 What the Commission has been asked to do 2
1.2 Report structure and study approach 2
1.3 Future data improvements 7
1.4 Conduct of the study 14
2 Australia’s public hospital sector 17
2.1 Role and structure of public hospitals 18
2.2 Characteristics of public hospitals 23
2.3 Services provided by public hospitals 31
2.4 Workforce characteristics 39
2.5 Recent developments in public hospitals 40
3 Australia’s private hospital sector 45
3.1 Structure of private hospitals 46
3.2 Characteristics of private hospitals 51
3.3 Services provided by private hospitals 56
3.4 Workforce characteristics 61
3.5 Private freestanding day hospitals 63
3.6 Recent developments in private hospitals 65
4 Public and private hospitals in the health system 69
4.1 Similarities and differences 70
4.2 Relationship between the two sectors 82
4.3 Possible directions for hospitals 85
CONTENTS VII
5 Hospital and medical costs 91
5.1 Cost indicators 93
5.2 Data sources and estimation methods 94
5.3 Cost per casemix-adjusted separation 102
5.4 Average cost of individual DRGs 112
5.5 Improving future cost comparisons 118
6 Hospital-acquired infections 123
6.1 Types of hospital-acquired infections 124
6.2 How should infection rates be measured and compared? 130
6.3 Available evidence on hospital-acquired infections 133
6.4 Developments to improve future comparisons 137
7 Other partial indicators 141
7.1 Productivity 144
7.2 Access to hospital services 153
7.3 Quality and patient safety 167
7.4 Developments to improve future comparisons 177
8 Multivariate analysis 181
8.1 About the Commission’s multivariate analysis 182
8.2 Profile of hospitals in the sample 184
8.3 Factors affecting hospital performance 191
8.4 Factors contributing to best-practice benchmarks 197
8.5 Hospital efficiencies 203
8.6 Further analysis 206
9 Informed financial consent 209
9.1 Potential impediments to the provision of informed financial
consent 211
9.2 Informed financial consent data sources and their suitability 212
9.3 Rates of informed financial consent 215
9.4 Out-of-pocket expenses 222
9.5 Future data improvements 229
9.6 Best-practice examples of IFC 230
10 Indexation of the Medicare Levy Surcharge income thresholds 235
10.1 Background to the Medicare Levy Surcharge 236
10.2 Why index the MLS thresholds? 238

VIII CONTENTS
 10.3 Possible indexation factors 238
10.4 Assessment of potential indexation factors 241
A Public consultation 247
B National Healthcare Agreement performance indicators 253
B.1 The National Healthcare Agreement 253
B.2 Monitoring and reporting 258
C Other health performance monitoring frameworks 263
C.1 National Health Performance Framework 263
C.2 Review of Government Service Provision 268
D Constructing estimates of hospital and medical costs 277
D.1 National Hospital Cost Data Collection 277
D.2 Hospital Casemix Protocol 285
D.3 Cost components 288
D.4 Cost indicators 293
D.5 Tax exemptions 295
D.6 Capital costs 300
D.7 Relative complexity 313
D.8 Prostheses costs 315
D.9 Costs for patients funded by the Department of Veterans’
Affairs 317
E Multivariate analysis in detail 325
E.1 Previous studies 325
E.2 Commission’s approach to modelling hospital performance 338
E.3 Data sources 347
E.4 Variables 350
E.5 The results 359
E.6 Proposed future analysis 377
F State-level data on hospital-acquired infections 379
F.1 Victoria 380
F.2 Queensland 385
F.3 South Australia 388
F.4 Western Australia 390
F.5 Tasmania 395

CONTENTS IX
G Referee reports on modelling 397
G.1 Report from Adjunct Professor Tim Coelli 397
G.2 Report from Professor Jim Butler 399
References 401
Boxes
1.1 Components of economic efficiency 6
1.2 Privacy legislation in each jurisdiction 10
2.1 Measures of hospital activity 24
2.2 Australian Refined Diagnosis-Related Group (AR-DRG) 32
5.1 Some of the deficiencies of existing cost data 96
6.1 NHSN/NNIS risk index for SSIs 132
7.1 Summary of partial indicators 143
7.2 Quality and patient safety indicators 168
9.1 Meditrust and IFC 232
10.1 Recent changes to the Medicare Levy Surcharge 237
10.2 Income distribution data used in MLS analysis 242
B.1 National Agreement Reporting 254
B.2 Policy directions and priority reform areas 259
C.1 Selection criteria used by the NHPC for health performance
indicators 266
C.2 Aims of the Review of Government Service Provision 269
D.1 NHCDC collection and reporting process 278
D.2 Allocation of ungroupable HCP medical and diagnostics costs across
DRGs 287
D.3 Classifying episodes of care — Diagnosis-Related Groups 294
D.4 Cost per casemix-adjusted separation 295
D.5 SCRGSP methodology for calculating public hospital capital costs 301
D.6 Estimating asset values 306
D.7 Enterprise value 309
D.8 Health care arrangements for veterans and their dependants 318
Figures
1.1 Governance arrangements for national health data 8
2.1 Funding sources for public hospital services, 2007-08 22
2.2 Share of public hospital separations by patient funding source,
2007-08 29

X CONTENTS
2.3 Share of public hospital separations by socioeconomic status of
patients, 2007-08 30
2.4 Share of public hospital separations by MDC, 2007-08 33
2.5 Share of public hospital separations by AR-DRG partition, 2007-08 33
2.6 Most frequent same-day public hospital separations by AR-DRG,
2007-08 34
2.7 Most frequent overnight public hospital separations by AR-DRG,
2007-08 35
2.8 Ten fastest increasing public hospital separations by AR-DRG,
2003-04 to 2007-08 42
3.1 Share of private hospital separations by patient funding source,
2007-08 54
3.2 Share of population with private health insurance, 2008 55
3.3 Share of private hospital separations by socioeconomic status of
patients, 2007-08 55
3.4 Share of private hospital separations by AR-DRG partition, 2007-08 56
3.5 Most frequent same-day private hospital separations by AR-DRG,
2007-08 57
3.6 Most frequent overnight private hospital separations by AR-DRG,
2007-08 58
3.7 Private hospital separations per 1000 residents, 2003-04 to 2007-08 66
3.8 Private hospital patient separations by funding source, 2003-04 to
2007-08 66
4.1 Share of separations by sector and patient age, 2007-08 81
4.2 Share of separations by sector and socioeconomic status of patients,
2007-08 81
4.3 Funding sources of public and private hospitals, 2007-08 86
5.1 Composition of general hospital costs by sector, 2007-08 104
5.2 Comparison of cost per separation for individual DRGs in public and
private hospitals, 2007-08 113
5.3 Cumulative distribution of DRGs ranked by public-private cost ratio,
2007-08 114
5.4 DRG partitions by public cost relative to private cost, 2007-08 115
5.5 Cumulative distribution of separations in each sector ranked by DRG
cost weight, 2007-08 117
7.1 Rates of caesarean section by hospital sector, 1991–2006 177
9.1 Informed financial consent rates by jurisdiction 219

CONTENTS XI
9.2 Sample cumulative distribution function of gaps with no IFC, 2007 224
9.3 Distribution of gaps with no IFC, 2007 224
10.1 MLS income thresholds if there had been indexation, singles 241
10.2 Proportion of single taxpayers subject to the MLS under alternative
indexation options 243
10.3 Proportion of taxpayers who were members of a family subject to
the MLS under alternative indexation options 244
C.1 Report on Government Services general framework 270
C.2 Performance indicator framework for health services 272
C.3 Performance indicators for public hospitals 274
C.4 Performance indicators for maternity services 275
D.1 Maximum effect of the capped fringe-benefits tax exemption on
post-tax remuneration, by occupation 297
D.2 Estimated benefit to public hospitals of payroll-tax exemptions,
2007-08 300
D.3 Public hospital asset values (excluding land), 2007-08 304
D.4 Sensitivity analysis for private hospital user cost of capital 312
D.5 Sensitivity analysis for public hospital user cost of capital 312
E.1 Illustration of SFA production model 341
F.1 Hospital-acquired MRSA infections in Victoria by sector,
2005–2008 383
F.2 Hospital-acquired VRE infections in Victoria by sector, 2005–2008 384
F.3 Hospital-acquired BSIs in South Australia by sector, 2003–2008 389
F.4 Hospital-acquired MRSA infections in South Australia by sector and
ICU status, 2003–2008 390
F.5 Hip SSIs in Western Australia by risk category and sector,
2006–2008 394
F.6 Knee SSIs in Western Australia by risk category and sector,
2006–2008 394
Tables
2.1 Number and activity of public hospitals, 2007-08 25
2.2 Number of public hospitals by size, 2007-08 26
2.3 Number of public hospitals by location, 2009 27
2.4 Number of public hospital beds per 1000 residents by location,
2007-08 28
2.5 Public hospital separations by patient profile, 2007-08 30

XII CONTENTS
2.6 Number of public hospital separations by type of care, 2007-08 31
2.7 Number of specialist service units in public hospitals, 2007-08 36
2.8 Services to non-admitted patients in public hospitals, 2007-08 37
2.9 Number of staff in public hospitals, by occupation, 2007-08 39
2.10 Average salaries of staff in public hospitals, 2007-08 40
2.11 Changes in acute public hospitals, 2003-04 to 2007-08 41
2.12 Public hospital emergency department activity, 2003-04 to 2007-08 43
3.1 Number of private acute and psychiatric hospitals, 2006-07 47
3.2 Income and expenditure of private hospitals, 2006-07 48
3.3 Number and activity of private hospitals, 2007-08 51
3.4 Number of private hospitals by size, 2006-07 52
3.5 Number of private hospitals and beds by location, 2006-07 53
3.6 Share of private hospital separations by patient profile, 2007-08 53
3.7 Number of private hospital separations by care type, 2007-08 59
3.8 Accident and emergency treatment in private hospitals, 2006-07 60
3.9 Number of staff in private hospitals, 2006-07 62
3.10 Private freestanding day hospital facilities, 2007-08 64
3.11 Number of private freestanding day hospital facilities by type of
centre, 2006-07 64
3.12 Number of staff in private freestanding day hospitals, 2006-07 65
3.13 Number of separations for the most common private hospital
AR-DRGs, 2003-04 to 2007-08 67
3.14 Private hospitals with teaching roles, 2005-06 to 2006-07 68
4.1 Share of public and private hospitals by size 71
4.2 Public and private hospitals by location, 2009 72
4.3 Share of hospital separations by sector, 2007-08 73
4.4 Thirty most frequent overnight separations in public and private
hospitals by AR-DRG, 2007-08 75
4.5 Thirty most frequent same-day separations in public and private
hospitals by AR-DRG, 2007-08 78
4.6 Share of patient separations by sector and funding source, 2007-08 82
5.1 Cost components 97
5.2 Cost per casemix-adjusted separation by jurisdiction and sector,
2007-08 103
5.3 Cost per casemix-adjusted separation by region and sector, 2007-08 109

CONTENTS XIII
5.4 Cost per casemix-adjusted separation by hospital size and sector,
2007-08 110
6.1 ACHS infection indicators that differed between public and private
hospitals, 2007 135
7.1 Labour productivity, 2002-03 and 2007-08 146
7.2 Bed productivity, 2002-03 and 2007-08 147
7.3 Drug, medical and surgical supplies productivity, 2002-03 and
2007-08 149
7.4 Relative stay index, public and private hospitals, by DRG procedure
partitions, 2002-03 and 2007-08 150
7.5 Labour intensity of public and private hospitals, 2002-03 and
2007-08 152
7.6 Public hospital emergency department waiting times, 2002-03 and
2007-08 155
7.7 Additions and removals from public hospital elective surgery
waiting lists, 2002-03 to 2007-08 158
7.8 Elective surgery waiting times, public hospitals, 2002-03 and
2007-08 159
7.9 Elective surgery waiting times by clinical category, public hospitals,
2002-03 and 2006-07 160
7.10 Occupancy rates, 2002-03 to 2007-08 162
7.11 Average cost and population coverage of private hospital insurance,
2002-03 to 2007-08 164
7.12 Elective surgery separation statistics, 2007-08 165
7.13 Hospital separations with an adverse event, 2002-03 to 2007-08 172
7.14 Rates of foetal, neonatal and perinatal deaths by hospital sector, rate
per 1000 births, 2006 176
8.1 Profile of sample hospitals by location and size, 2006-07 185
8.2 Profile of sample hospitals, output and partial productivity measures,
2006-07 186
8.3 Profile of sample hospitals, by service and patient characteristics,
2006-07 187
8.4 Coefficient results of stochastic frontier analysis, 2006-07 199
8.5 Summary of hospital technical efficiency scores, 2006-07 204
9.1 Comparison of Ipsos and PHIAC data, 2007 214
9.2 Informed financial consent rates for pre-planned and emergency
admissions, 2007 216

XIV CONTENTS
9.3 Informed financial consent rates, 2004–2007 217
9.4 Informed financial consent rates by hospital provider 218
9.5 Informed financial consent rates by jurisdiction, 2007 219
9.6 Informed financial consent rates by location and type of provider 220
9.7 Informed financial consent rates by medical practitioner or service
provider, 2007 221
9.8 Informed financial consent rates for patients with a lead time of at
least five days, 2007 222
9.9 Average gap 223
9.10 Average gap by hospital provider, 2004–2007 225
9.11 Average gap by jurisdiction, 2007 226
9.12 Average gap by region, 2004–2007 227
9.13 Average gap by medical practitioner or service provider, 2007 228
10.1 Possible indexation factors for the Medicare Levy Surcharge 239
A.1 Submissions received 247
A.2 Visits 249
A.3 Participants in initial roundtable 250
A.4 Participants in Discussion Draft roundtable 251
A.5 Teleconference participants 252
B.1 Objectives of the National Healthcare Agreement 255
B.2 National Healthcare Agreement Reporting Structure 256
B.3 Hospital and related care performance indicators 260
B.4 Other NHA indicators related to hospital performance 261
C.1 The National Health Performance Framework 264
C.2 Indicators reported in the National Report on Health Sector
Performance, 2003 267
D.1 NHCDC sample by jurisdiction and region, 2007-08 279
D.2 NHCDC sample by jurisdiction and hospital size, 2007-08 280
D.3 NHCDC sample by region and hospital size, 2007-08 281
D.4 Differences between Victoria and other jurisdictions for NHCDC
cost buckets 285
D.5 Hospital Casemix Protocol descriptive statistics, 2007-08 286
D.6 Ungroupable separations for private patients by sector, 2007-08 288
D.7 Components of total cost estimates 289
D.8 Distribution of NHCDC pharmacy and medical costs for selected
patient-costed public hospitals, by cost bucket, 2007-08 292
CONTENTS XV
D.9 Public hospital medical salaries and wages included in other
NHCDC cost buckets, 2007-08 293
D.10 Distribution of benefits from the capped FBT exemption by sector 298
D.11 Summary of sources and methods used to estimate capital costs 310
D.12 Estimated capital costs per casemix-adjusted separation, 2007-08 311
D.13 Renal dialysis and chemotherapy separations as a percentage of all
separations by sector, 2007-08 313
D.14 Impact of renal dialysis and chemotherapy separations on average
cost weights by sector, 2007-08 314
D.15 Cost per casemix-adjusted separation for adjacent DRGs, Australia,
2007-08 315
D.16 Cost per casemix-adjusted separation for adjacent DRGs, Australia,
2007-08 315
D.17 Prosthesis costs for selected DRGs, 2007-08 317
D.18 Comparison of average length of stay for DVA and NHCDC
patients, selected DRGs, 2006-07 320
D.19 Separations and episode costs for DVA patients, selected DRGs,
2006-07 322
E.1 Selected literature review 326
E.2 Hospital sample by size, region and sector, 2006-07 350
E.3 Description and summary statistics of variables, 2006-07 360
E.4 Results of Tobit regression of mortality rates, 2006-07 364
E.5 Predicted mortality rates and risk-adjusted mortality ratios, by
sector, 2006-07 365
E.6 Risk-adjusted mortality ratios, by sector and hospital size, 2006-07 366
E.7 Results of Cobb-Douglas stochastic frontier analysis, 2006-07 367
E.8 Results of translog stochastic frontier analysis, 2006-07 369
E.9 Technical efficiency scores, all hospitals, 2006-07 373
E.10 Technical efficiency scores, large and very large hospitals, 2006-07 374
E.11 Technical efficiency scores, small and very small, and medium
hospitals, 2006-07 375
E.12 Correlation coefficients between selected variables and technical
efficiency scores 376
F.1 SSI rates for Victorian public hospitals by procedure and risk
category, 2007 381
F.2 MRSA and BSI rates for Victorian public hospitals by hospital size,
2004–2007 381

XVI CONTENTS
F.3 Comparison of VAED and VICNISS data for public-hospital MRSA
infections 383
F.4 Hospital-acquired MRSA infections in Victoria by region and ICU
status, 2005-06 to 2007-08 384
F.5 Hospital-acquired VRE infections in Victoria by region and ICU
status, 2005-06 to 2007-08 385
F.6 Selected hospital-acquired infections in Queensland,
July–December 2008 386
F.7 SSI rates for Queensland public hospitals by surgical procedure,
2004–2008 388
F.8 Hospital-acquired MRSA in Western Australia by sector, 2006–2008 392
F.9 Hospital-acquired SAB BSIs in Western Australia by sector,
2007–2008 393
F.10 Rate of hospital-acquired infections in Tasmanian public hospitals
by organism, 2005–2008 395

CONTENTS XVII
Abbreviations and explanations

Abbreviations
ABF Activity-based funding
ABS Australian Bureau of Statistics
ACEM Australasian College for Emergency Medicine
ACERH Australian Centre for Economic Research on Health
ACHI Australian Classification of Health Interventions
ACHS Australian Council on Healthcare Standards
ACSQHC Australian Commission on Safety and Quality in Health Care
AHCA Australian Health Care Agreement
AHHA Australian Healthcare and Hospitals Association
AHIA Australian Health Insurance Association
AHMC Australian Health Ministers’ Conference
AHSA Australian Health Service Alliance
AIHW Australian Institute of Health and Welfare
AIMS Australian Incident Monitoring System
ALOS Average length of stay
AMA Australian Medical Association
ANF Australian Nursing Federation
APHA Australian Private Hospitals Association
AR-DRG Australian Refined Diagnosis-Related Group
ASA Australian Society of Anaesthetists
ASGC Australian Standard Geographical Classification
ASX Australian Securities Exchange
ATO Australian Taxation Office
AWE Average Weekly Earnings

XVIII ABBREVIATIONS AND

EXPLANATIONS
AWOTE Average Weekly Ordinary Time Earnings
AWTE Average Weekly Total Earnings
BSI Blood stream infection
CEPA Centre for Efficiency and Productivity Analysis
CHA Catholic Health Australia
CHE Centre for Health Economics (Monash University)
CHF Consumers’ Health Forum
CHI Centre for Healthcare Improvement
CHRISP Centre for Healthcare Related Infection Surveillance and
Prevention
CIP Clinical Indicator Program
COAG Council of Australian Governments
CPI Consumer Price Index
CRC COAG Reform Council
DEA Data Envelopment Analysis
DOHA Department of Health and Ageing (Australian Government)
DRG Diagnosis-related group
DVA Department of Veterans’ Affairs
FBT Fringe-benefits tax
FTE Full-time equivalent
HCAIU Health Care Associated Infection Unit (WA Government)
HCI Hospital cost index
HCP Hospital Casemix Protocol
HDU High-dependency unit
HISWA Healthcare Infection Surveillance Western Australia
HQCC Health Quality and Complaints Commission (Queensland
Government)
ICD-10-AM International Classification of Diseases, Version 10,
Australian Modification
ICS Infection Control Service (SA Government)
ICU Intensive-care unit

ABBREVIATIONS AND XIX

EXPLANATIONS
IFC Informed financial consent
IGA Intergovernmental Agreement on Federal Financial Relations
IPHA Independent Private Hospitals of Australia
MBS Medicare Benefits Schedule
MDC Major Diagnostic Category
MIAESR Melbourne Institute of Applied Economic and Social
Research
MLS Medicare Levy Surcharge
MRO Multiresistant organism
MRSA Methicillin-resistant Staphylococcus aureus
MSSA Methicillin-sensitive Staphylococcus aureus
NATSEM National Centre for Social and Economic Modelling
NHA National Healthcare Agreement
NHCDC National Hospital Cost Data Collection
NHHRC National Health and Hospitals Reform Commission
NHMD National Hospital Morbidity Database
NHPC National Health Performance Committee
NHPF National Health Performance Framework
NHS National Health Survey
NHSN National Healthcare Safety Network
NNIS National Nosocomial Infections Surveillance
NP National Partnership
NPHED National Public Hospital Establishments Database
OLS Ordinary least squares
PC Productivity Commission
PHEC Private hospital establishments collection
PHI Private health insurance
PHIAC Private Health Insurance Administration Council
PHIO Private Health Insurance Ombudsman
RACS Royal Australasian College of Surgeons
RAMR Risk-adjusted mortality ratio
XX ABBREVIATIONS AND
EXPLANATIONS
RCN Royal College of Nursing
ROGS Report on Government Services
RSI Relative stay index
SAB Staphylococcus aureus bacteraemia
SCRGSP Steering Committee for the Review of Government Service
Provision
SDF Stochastic distance function
SEIFA Socio-Economic Indexes for Areas (Relative
Socio-economic Advantage and Disadvantage)
SFA Stochastic frontier analysis
SLA Statistical local area
SIR Standardised Infection Ratio
SPP Specific Purpose Payment
SSI Surgical-site infection
TFP Total factor productivity
TIPCU Tasmanian Infection Prevention and Control Unit
UCC User cost of capital
VAED Victorian Admitted Episodes Dataset
VICNISS Victorian Nosocomial Infection Surveillance System
VRE Vancomycin-resistant enterococci
WPI Wage Price Index

Explanations
Billion The convention used for a billion is a thousand million (109).
Findings Findings in the body of the report are paragraphs high-
lighted using italics, as this is.

ABBREVIATIONS AND XXI

EXPLANATIONS
Glossary

Activity-based A method of funding hospitals in which hospitals are paid for

funding each episode of care they provide. The amount (price) paid
for each episode of care is calculated in advance so that the
risk of managing costs is borne by the hospital.

Acute care Clinical services provided to admitted or non-admitted

patients, including managing labour, curing illness or treating
injury, performing surgery, relieving symptoms and/or
reducing the severity of illness or injury, and performing
diagnostic and therapeutic procedures. Most episodes involve
a relatively short hospital stay.

Acute hospital Establishments which provide at least minimal medical,

surgical or obstetric services for admitted patient treatment
and/or care, and which provide round-the-clock
comprehensive qualified nursing service as well as other
necessary professional services. They must be licensed by a
state/territory health department, or controlled by government
departments. It also includes hospitals specialising in dental,
ophthalmic aids and other specialised medical or surgical
care.

Admitted patient A patient who has undergone a formal admission process in a

hospital to begin an episode of care. Admitted patients may
receive acute, sub-acute or non-acute care services.

Admitted patient The proportion of total (or operating) costs that are attributed
cost proportion to admitted patients. Also known as the inpatient fraction.

Adverse event The unintentional harm arising from an episode of healthcare

and not due to the disease process itself.

Allocative How well resources are allocated across different uses so as

efficiency to generate the greatest community wellbeing at a given point
in time.

XXII GLOSSARY
Apgar score A numerical score used to indicate the baby’s condition at
one minute and five minutes after birth. Between 0 and 2
points are given for each of five characteristics: heart rate,
breathing, colour, muscle tone and reflex irritability, and the
total score is between 0 and 10.

Average length of The average number of patient days per admitted patient
stay episode. Patients admitted and separated on the same day are
allocated a length of stay of one day.

Australian An Australian system of Diagnosis-related Groups (DRGs).

Refined Version 5.0/5.1 is based on the fifth edition of ICD-10-AM.
Diagnosis-related See Diagnosis-related groups.
Groups

AR-DRG Individual AR-DRGs are assigned to a ‘surgical’, ‘medical’

partitions or ‘other’ partition on the basis of the type of procedure
involved. A separation is classified as ‘surgical’ if it includes
an operating room procedure. A separation is classified as
‘other’ if it includes a procedure performed outside of an
operating room. A separation is classified as ‘medical’ if it
does not include any type of procedure.

Capital city A geographic classification of state and territory capital cities

statistical defined used by the Australian Bureau of Statistics for
divisions general statistical purposes. Each division contains its
anticipated urban development (and its associated urban
centres) for a period of at least twenty years.

Casemix The range and types of episodes of care (the mix of cases)
treated by a hospital.

Casemix-adjusted The adjustment of data to account for differences in the

number and type (complexity) of cases. Casemix adjustment
is an important step to achieving comparable measures of
efficiency across hospitals and jurisdictions.

Clostridium A species of gram-positive bacteria of the genus Clostridium,

difficile that normally reside in the body. Overpopulation of
Clostridium difficile is harmful because the release of toxins
can lead to bloating, constipation and diarrhoea.

GLOSSARY XXIII
 Cost weight The average costliness of an AR-DRG relative to all other
AR-DRGs, such that the average cost weight for all
separations is 1.

Data A linear programming technique used to identify the

Envelopment best-practice frontiers of entities, such as hospitals and firms.
Analysis

Depreciation Reduction in the value of an asset due to usage or

obsolescence.

Diagnosis-related A system used to classify hospital admissions into groups

group with similar clinical conditions (related diagnoses) and
similar resource usage (hospital services).

Dynamic How well resources are allocated to achieve the greatest

efficiency possible community wellbeing over time.

Effectiveness How successful a hospital is in achieving a particular

objective, such as avoiding hospital-acquired infections.

Efficiency Economic efficiency, in its broadest sense, refers to how well

resources are used to benefit the wellbeing of the community
as a whole. It comprises productive efficiency, allocative
efficiency, and dynamic efficiency.

Elective surgery Any surgery that a patient’s doctor or health professional

considers to be necessary but which can be delayed by at
least 24 hours. In Australia, elective surgical procedures are
defined in the Medicare Benefits Schedule.

Episode of care The period of admitted patient care between a formal or

statistical admission and a formal or statistical separation,
characterised by only one care type.

High-dependency An area or environment in a hospital that provides a higher

unit level of critical care and monitoring than is provided in a
general ward, but a lower level of care provided by an
intensive-care unit.

XXIV GLOSSARY
Hospital A healthcare facility established under Commonwealth, state
or territory legislation as a hospital or a freestanding day
procedure unit and authorised to provide treatment and/or
care to patients.

Hospital-acquired An infection that appears during the course of care at a

infection hospital or healthcare facility and is the result of that care.
Also referred to as nosocomial infection.

Hospital Casemix A data collection of the episodes of admitted patient care,

Protocol benefits and charges for privately insured patients. It includes
clinical, demographic and financial information for privately
insured admitted patient services.

Hospital cost An index of hospital costs published by the Australian

index Bureau of Statistics.

Informed The provision of cost information to patients, including

financial consent notification of likely out-of-pocket expenses (gaps), desirably
prior to agreeing to treatment.

Inpatient fraction See admitted patient cost proportion.

Intensive-care An area or environment in a hospital that provides the highest

unit level of critical care and monitoring.

International The World Health Organisation’s internationally-accepted

Classification of classification of diseases and related health conditions. The
Diseases current version ICD-10 forms the basis of Australia’s
ICD-10-AM.

Labour A measure of episodes of care per full-time equivalent

productivity employee in a hospital or ward.

Length of stay The period from admission to separation, less any days spent
away from the hospital.

Maintenance care Care in which the treatment goal is prevention of

deterioration in the functional and current health status of a
patient with a disability or severe level of functional
impairment. Following assessment or treatment the patient
does not require further complex assessment or stabilisation,
and requires care over an indefinite period.

GLOSSARY XXV
 Major Diagnostic A high-level grouping of patients used in the AR-DRG
Categories classification. They correspond generally to the major organ
systems of the body.

Medicare Levy A levy on Australian taxpayers who earn above a specified

Surcharge income threshold and do not have private health insurance.

Methicillin- A strain of the Staphylococcus aureus bacterium that is

resistant resistant to a number of antibiotics including penicillins (such
Staphylococcus as methicillin). Patients with open wounds and weakened
aureus immune systems are at risk to Methicillin-resistant
Staphylococcus aureus from hospital staff that do not follow
proper sanitary procedures.

Methicillin- A strain of the Staphylococcus aureus bacterium that is

sensitive sensitive (not resistant) to a number of antibiotics.
Staphylococcus
aureus

Multiresistant A class of gram-negative bacteria that are resistant to a

gram-negative multiple range of antibiotics (such as ciprofloxacin,
bacteria trimethoprim-sulfamethoxazole and amipicillin/sulbactam),
and which can lead to bloodstream infections.

Multiresistant Bacteria that are resistant to multiple antibiotics. Such

organism organisms include Methicillin-resistant Staphylococcus
aureus and vancomycin-resistant enterococci.

Multivariate A statistical analysis technique in which two or more

analysis variables are analysed simultaneously.

Non-admitted Occasion of examination, consultation, treatment or other

occasion of service provided to a non-admitted patient in a functional unit
service of a health service establishment. Services may include
emergency department visits, outpatient services (such as
pathology, radiology and imaging, and allied health services,
including speech therapy and family planning) and other
services to non-admitted patients.

Non-admitted A patient who has not undergone a formal admission process,

patient but who may receive care through an emergency department,
outpatient or other non-admitted service.

XXVI GLOSSARY
Partial An indicator of a particular aspect of a hospital’s
performance performance that does not take account of other aspects of
indicator performance.

Patient days The aggregate number of days of stay (calculated as

separation date minus admission date) for all overnight-stay
patients who were separated from hospital during the year.
Periods of approved leave are subtracted from these
calculations. Same-day patients are each counted as having a
stay of one day.

Peer group A group of hospitals that are broadly similar in terms of their
volume of admitted-patient activity and geographical
location.

Principal The diagnosis chiefly responsible for occasioning an episode

diagnosis of admitted-patient care.

Private hospital A privately owned and operated institution, catering for

patients who are treated by a doctor of their own choice.
Patients are charged fees for accommodation and other
services provided by the hospital and relevant medical and
paramedical practitioners.

Private patients Patients admitted to a hospital who decide to choose the

doctor(s) who will treat them and/or to have private ward
accommodation. They are charged for medical services, food
and accommodation.

Procedure A clinical intervention that is surgical in nature, carries a

procedural risk, carries an anaesthetic risk, requires
specialised training, and/or requires special facilities or
equipment available only in an acute-care setting.

Productive The degree to which outputs are produced at least possible

efficiency cost. It incorporates technical efficiency.

Public hospital A health care provider facility that has been established under
state or territory legislation as a hospital or as a freestanding
day procedure unit. Public hospitals are operated by, or on
behalf of, the government of the state or territory in which
they are established. Public hospitals provide hospital
services free of charge to all eligible patients.
GLOSSARY XXVII
 Public patient A patient admitted to a hospital who has agreed to be treated
by doctors of the hospital’s choice and to accept shared
accommodation. This means the patient is not charged.

Recurrent Expenditure on goods and services which are used up during

expenditure the year. Includes salaries and wages, payments to visiting
medical officers, expenditure on drug, medical and surgical
supplies, and repairs and maintenance. Does not include
investment expenditure.

Relative stay The actual number of patient days for acute-care separations
index in selected AR-DRGs divided by the expected number of
patient days adjusted for casemix.

Same-day Day centres, hospitals and freestanding day surgery centres

establishments that provide a course of acute treatment on a full-day or
part-day non-residential attendance basis at specified
intervals over a period of time. Freestanding day surgery
centres are approved by the Commonwealth for the purposes
of basic table health insurance benefits.

Sentinel events A limited number of serious adverse events, in which death

or serious harm to a patient has occurred.

Separation An episode of care for an admitted patient, which can be a

total hospital stay (from admission to discharge, transfer or
death), or a portion of a hospital stay beginning or ending in a
change of type of care (for example, from acute to
rehabilitation).

Separation also means the process by which an admitted

patient completes an episode of care either by being
discharged, dying, transferring to another hospital or
changing type of care.

Statistical local The smallest geographic area used by the Australian Bureau
area of Statistics (in non-Census years) in its Australian Standard
Geographic Classification system.

Stochastic A statistical regression technique used to determine the

frontier analysis frontier of best-practice entities such as firms and hospitals.

XXVIII GLOSSARY
Sub-acute and Clinical services provided to patients suffering from chronic
non-acute care illnesses or recovering from such illnesses. Services include
rehabilitation, planned geriatric care, palliative care, geriatric
care evaluation and management, and services for nursing
home type patients. Clinical services delivered by designated
psychogeriatric units, designated rehabilitation units and
mothercraft services are considered non-acute.

Surgical-site The number of surgical-site infections for a selected

infection rate procedure performed during the surveillance period divided
by the total number of the selected procedures performed
during the surveillance period.

Technical The extent to which the quantity of inputs can be reduced

efficiency without also reducing the quantity of outputs.

Total factor A measure of output relative to all inputs.

productivity

Univariate The separate exploration of each variable in a data set. It

analysis looks at the range of values, as well as the central tendency of
the values. It describes the pattern of response to the variable.

Unplanned An unexpected hospital admission for treatment of: the same

hospital condition for which the patient was previously hospitalised; a
readmission condition related to one for which the patient was previously
hospitalised; or a complication of the condition for which the
patient was previously hospitalised.

Unplanned The number of unplanned readmissions to the same hospital

hospital that occur within a given period after separation, divided by
readmission rate the total number of separations (excluding deaths), including
day stay patients.

User cost of The opportunity cost of the capital used to deliver hospital
capital services. That is, the return that could be generated if the
funds were employed in their next best use.

Vancomycin- A group of bacterial species of the genus Enterococcus,

resistant found in the digestive and urinary tract, that is resistant to the
enterococci antibiotic vancomycin.

GLOSSARY XXIX
OVERVIEW
 Key points
• Although there is significant diversity within and between the public and private
hospital sectors, there are sufficient similarities to warrant comparing them, ideally in
a way that takes account of differences in the services provided and patients treated.
• Existing datasets on hospital costs are limited by inconsistent collection methods and
missing information. The Commission has sought to address these limitations by
drawing on various data sources and incorporating adjustments to make the data
more comparable. Nevertheless, the resulting estimates should be considered
experimental.
• The Commission’s experimental cost estimates suggest that, at a national level,
public and private hospitals have similar average costs. However, significant
differences were found in the composition of costs. General hospital costs were
higher in public hospitals. Medical and diagnostics costs and prostheses costs were
higher in private hospitals. Capital costs were higher in public hospitals, but this
result is particularly reliant on a range of data sources and adjustments to make the
data comparable.
• Australia does not have a robust nationally-consistent data collection on
hospital-acquired infections. The limited available evidence suggests that private
hospitals have lower infection rates than public hospitals, but this result could be
misleading because private hospitals on average treat patients who have a lower risk
of infection.
• Other partial indicators show that:
– private hospitals have higher labour productivity and shorter lengths of stay than
public hospitals, but this is at least partly due to casemix and patient differences
between the public and private sectors
– elective surgery in public hospitals is more accessible for disadvantaged
socioeconomic groups, but tends to be less timely than in the private sector.
• A multivariate analysis of hospital-level data suggests that the efficiency of public
and private hospitals is, on average, similar. The output of individual hospitals in both
sectors is, on average, estimated to be around 20 per cent below best practice.
• Improvements could be made to data collections to improve the feasibility of future
comparisons. Foreshadowed changes under the National Healthcare Agreement will
help in this regard, but more improvements could be made, such as consistent
national reporting of costs and infections for both public and private hospitals.
• Only a small proportion of patients incur out-of-pocket expenses without receiving
sufficient prior information to give informed financial consent. The medical profession
has facilitated best practice by educating practitioners and using internet-based
packages to inform consumers.
• The most appropriate indexation factor for the Medicare Levy Surcharge income
thresholds is average weekly ordinary time earnings.

XXXII PUBLIC AND PRIVATE

HOSPITALS
Overview

This study examines three aspects of Australia’s health care system:

• the relative performance of public and private hospitals, with particular regard to
the cost of performing clinically-similar procedures and the rate of hospital-
acquired infections
• rates of informed financial consent and out-of-pocket expenses for privately-
insured patients in public and private hospitals
• the most appropriate factor for indexing the Medicare Levy Surcharge income
thresholds.

The first task — comparing the relative performance of hospitals — has been the
most challenging part of the study, particularly in the short time available. This is
because hospital complexity and diversity make like-for-like comparisons difficult,
and existing data collections are not well suited to the task.

Hospitals are complex organisations, with many essential services delivered by a

range of health professionals in a location supported by available technologies, with
management oversight and administrative support. This makes comparisons
particularly challenging, especially to distinguish genuine differences in
performance from variation caused by differences in what hospitals do and who
they treat.

Study participants emphasised the importance of taking account of:

• variations in the types of services that hospitals provide, recognising that some
hospitals provide more complex health services that are relatively costly to
provide and are inherently more risky for patients
• the resource and service implications of providing emergency services
(including the need to keep staff and facilities on standby) and clinical training
(which can affect the rate of patient throughput), with its potential impact on
efficiency and service quality
• the impact of patient characteristics on the performance of public and private
hospitals, recognising that patients with more complex conditions and those
from lower socioeconomic groups are likely to require more intensive and

OVERVIEW XXXIII
 expensive health treatment and be more susceptible to hospital-acquired
infections
• the additional tax burden that for-profit hospitals face compared to public and
not-for-profit hospitals.

Existing datasets have made the task of robust comparison more difficult, with data
often defined and collected in different ways between the public and private sectors.
To the greatest extent possible, the Commission has sought to adjust for these
differences in providing estimates of hospital performance that are as robust and
comparable as possible. However, the Commission also readily acknowledges that a
number of significant data shortcomings has limited its ability to construct fully
comparable estimates.

A common theme throughout this report is that improvements could be made to data
collections to improve the feasibility of future comparisons. Foreshadowed changes
— such as strengthened national reporting under the new National Healthcare
Agreement (NHA) between the Australian, State and Territory Governments — will
help in this regard. However, more improvements could be made, such as adopting
consistent national reporting of costs and infections across both public and private
hospitals.

The Commission encountered significant delays in accessing hospital-related data

for this study beyond what could reasonably be expected to address legitimate
privacy or confidentiality concerns. Making these data more accessible to a range of
users could drive improvements in health care, especially as competitive markets
only have a limited role in the health sector. It could also encourage future
improvements in data collections.

The community fully meets the cost of data collections in the public hospital sector
— including national statistical collections compiled by the Australian Bureau of
Statistics (ABS) — and, through public and private contributions to the cost of
private hospital care, also contributes to the cost of data collections in the private
sector. There is a strong case for maximising the benefits that the community
achieves from the data it has paid for. More extensive research and analysis of these
data collections could deliver significant improvements in the efficiency and
effectiveness of health care.

Australia’s public and private hospital systems

Australian hospitals are part of a comprehensive system of services that together
contribute to the nation’s health outcomes. Australia spends about 9 per cent of its

XXXIV PUBLIC AND PRIVATE

HOSPITALS
gross domestic product on health care, and hospital services account for around
one-third of this. There are currently 768 public hospitals and 556 private hospitals,
providing 4.7 and 3.1 million episodes of care respectively for admitted patients in
2007-08. The main service provided by hospitals is the treatment of acutely ill
people, but many also provide a range of services to the wider community,
including radiology and pathology and outpatient care (such as rehabilitation and
physiotherapy).

Governments have assumed responsibility for delivering public hospital services in

Australia, largely to ensure equity of access. While the provision of services is
undertaken by state and territory governments, funding is shared with the
Commonwealth Government. Some public hospitals are operated by religious or
charitable organisations. Public hospitals are required to provide free treatment to
public patients, and are also assigned specific functions under the NHA, including
the provision of emergency services, clinical teaching and research, and equity of
access. Given that they are funded primarily by governments, the operational
incentive for public hospitals is therefore to manage demand within the given
budget constraint. Non-price factors, such as elective surgery waiting lists, serve as
the ‘escape valve’ that absorbs excess demand. Additionally, public hospitals face
growing community expectations concerning access to hospital services.

Private hospitals do not have the same degree of service obligations as public
hospitals, and have more scope to raise revenue from fees. The incentive for private
(particularly for-profit) hospitals is to generate returns on their capital investment
and labour force, for the benefit of owners/shareholders. However, not-for-profit
private hospitals may be more strongly driven by other objectives. Around 60 per
cent of private acute and psychiatric hospitals operate on a for-profit basis, while
the remainder are run by not-for-profit bodies, such as religious and charitable
groups.

Diversity exists not just between the public and private sectors, but also within
them. For example, while many large metropolitan public hospitals provide a full
range of services and have a large teaching role, many small public hospitals in
remote areas offer fewer acute services and may be called upon to deliver other
health services, such as primary care and aged care, to regional and remote
communities. Many private hospitals specialise in a limited range of surgical
procedures, although some offer services akin to the large public hospitals,
including an increasing share of the clinical teaching load.

Private hospitals tend to be concentrated in major cities. In comparison, public

hospitals are more widely distributed across the country (table 1).

OVERVIEW XXXV
Table 1 Number of public and private hospitals by region, 2009a
Private hospitals
Regionb Public hospitals Day hospitals Otherc Total

Major city 164 233 203 436

Inner regional 205 31 64 95
Outer regional 234 7 18 25
Remote 79 – – –
Very remote 81 – – –
Unable to be classifiedd 5 .. .. ..
Australia 768 271 285 556
a September 2009. b Regional classifications are based on the Australian Standard Geographical
Classification. c ‘Other’ comprises private acute and psychiatric hospitals. d Unable to be classified due to
missing postcodes. – Nil or rounded to zero. .. Not applicable.

Around half of Australia’s hospitals have 50 beds or fewer. Although the

concentration of small-scale hospitals is higher in the public sector, there is
diversity in hospital sizes in both sectors (figure 1). Almost three-quarters of public
hospitals have 50 beds or fewer (predominantly in regional and remote areas), while
almost 20 per cent have more than 100 beds. In the private sector, over 40 per cent
of hospitals have 50 beds or fewer, while over one-quarter have more than
100 beds.

Figure 1 Number of hospitals by sector and bed numbersa

600
Number of hospitals

500

400
Public
300
Private
200

100

0
0-50 beds 51-100 beds 101-200 beds Over 200 beds
Number of beds

a Acute and psychiatric hospitals. Data for private hospitals are for 2006-07, and for public hospitals are for
2007-08.

Although most patients in public hospitals are treated as public patients, around
14 per cent are treated as private patients, the majority of whom have private health
insurance (table 2). Most patients in private hospitals have private insurance or are
self-funded. Several state governments have arrangements that allow a small

XXXVI PUBLIC AND PRIVATE

HOSPITALS
number of public patients to be treated in contracted private hospitals, particularly
for elective surgery procedures.

Table 2 Hospital separations by sector and patient election status,

2007-08a
Per cent of separations in sector
Patient election status
Public Private Self-funded Dept of Compensation Total
insurance Veterans’ or other b
Sector Affairs
Public hospitals 85.9 8.8 1.2 2.6 1.5 100.0
Private hospitals 2.4 79.8 8.5 6.4 2.9 100.0
All hospitals 52.7 37.0 4.1 4.1 2.1 100.0
a Includes same-day facilities. b Compensation or other includes workers compensation, other compensation,
motor vehicle third party personal claim, other public authorities and other funding sources.

As many study participants commented, public and private hospitals complement

each other to some extent by specialising in the provision of different services and
treating different patient populations. For example:
• Public hospitals handled around three-quarters of all medical separations
(episodes of care), while private hospitals performed around 60 per cent of all
surgeries and nearly 70 per cent of other procedures in 2007-08 (table 3).
• Around 95 per cent of outpatient occasions of service (including emergency
department presentations) were handled by public hospitals in 2007-08, while
nearly two-thirds of elective surgery separations were performed by private
hospitals.
• Patients treated in public hospitals are, on average, from lower socioeconomic
groups, and have more complex medical conditions.
• A greater proportion of patients in public hospitals are aged under 35 years,
while private hospitals treat proportionately more patients aged 35–64 years.

However, there is also overlap between the two sectors, which suggests that, to
some extent, public and private hospitals compete to offer substitutable services.
For example, the most frequent types of same-day separations in both sectors are
renal dialysis, chemotherapy, non-complex colonoscopy and lens procedures
(although the respective order of frequency in each sector varies slightly). In
addition, a number of private hospitals display features typical of larger public
hospitals. In 2006-07, 47 private hospitals treated accident and emergency cases, of
which 24 had formal emergency departments, and 47 provided teaching to medical
staff and undergraduates.

OVERVIEW XXXVII
Table 3 Types of treatments in public and private hospitals, 2007-08a
Public hospitals Private hospitals

Number of Per cent of Number of Per cent of

separations separations separations separations

Surgical 919 109 20 1 232 428 41

Medical 3 397 595 74 1 132 851 38
Other 291 297 6 634 350 21
Total 4 608 001 100 2 999 629 100
a Includes same-day facilities. Separations are assigned to the surgical, medical or other categories on the
basis of the type of procedure involved. A procedure is a clinical intervention that carries a procedural or
anaesthetic risk, and/or requires specialised training, facilities or equipment. A separation is classified as
surgical if it involves at least one operating-room procedure; medical if there is no procedure; and other if it
involves a procedure performed outside of an operating room, such as dental extractions and colonoscopies.

Many study participants commented that the boundaries between the two sectors are
not clear cut, and are complicated by the fact that the two sectors do not operate in
isolation from each other, but are inter-related. For example, some public and
private hospitals share resources in co-located establishments, a single provider may
deliver services in both sectors, and medical staff can work across both sectors.

Costs, productivity and access

The terms of reference ask the Commission to report comparative hospital and
medical costs using data to be provided by the states and territories under the NHA
and already reported by private hospitals. However, recent policy developments —
such as a foreshadowed move to nationally-consistent activity-based funding —
have yet to lead to the reporting of all costs on a consistent basis between
jurisdictions, or between public and private hospitals. It has therefore been a major
challenge to report comparable cost data. It should also be noted that costs are a
partial indicator of hospital performance, since they do not include information on
other aspects of performance, such as quality and patient safety.

Two commonly-used measures of hospital costs were estimated for this study:
• cost per casemix-adjusted separation — the average cost of treating a range of
different diagnoses, after controlling for differences in the complexity of
required treatments (casemix adjustment)
• cost per separation — the average cost of treating a group of diagnoses that are
clinically similar.

Clinically-similar diagnoses were grouped according to the widely-accepted system

of Diagnosis-Related Groups (DRGs), which provides a clinically-meaningful way

XXXVIII PUBLIC AND PRIVATE

HOSPITALS
of relating types of patients treated to required resources. The DRG system only
applies to admitted-patient services, and so costs for other services were excluded.
Admitted-patient services accounted for 71 per cent of the costs incurred by
overnight acute-care hospitals in 2007-08. Furthermore, around 12 per cent of
DRGs (about 3 per cent of total annual separations) were excluded from the analysis
because there were few separations in at least one sector, and/or activity was
confined to less than three hospitals.

Most of the cost data were sourced from the Australian Government Department of
Health and Ageing, which has gathered the best available information as part of its
regulatory and oversight functions. However, the data collections have significant
limitations for this study (box 1). The Commission sought to address these
limitations by drawing on various data sources and incorporating adjustments to
make the data more comparable, including in response to comments made by study
participants on the Discussion Draft. Nevertheless, the cost estimates presented in
this report should be treated as experimental.

Overall costs were estimated by summing the various items that contribute to an
episode of care. Cost data on these items have varying degrees of accuracy and
comparability, and so the Commission distinguished between them using six
categories:
• general hospital — ward nursing, ward supplies and other overheads, allied
health, critical care, operating rooms, specialist procedure suites, hotel costs,
non-clinical salaries, and on-costs
• pharmacy
• emergency departments
• prostheses
• capital — depreciation and the user cost of capital
• medical and diagnostics.

Not all of the above components are under the control of hospitals, and so care
should be exercised in interpreting differences in the total cost of an episode of care
in public and private hospitals. In particular, it should be noted that medical costs in
private hospitals are predominantly a matter between patients and their relevant
specialist(s), although private hospitals generally engage a number of salaried
doctors.

OVERVIEW XXXIX
 Box 1 Hospital cost data and their limitations
Most of the cost data used in this study were provided by the Australian Government
Department of Health and Ageing from two collections:
• National Hospital Cost Data Collection (NHCDC) — a voluntary annual survey of
hospitals, with the latest published results (2007-08) based on responses from
hospitals that accounted for 89 per cent of public acute separations and 72 per cent
of private acute separations
• Hospital Casemix Protocol (HCP) — a regular census of private health insurance
claims in public and private hospitals, collected as part of the regulation of private
health insurance. HCP data exclude both public patients, and private patients who
do not make a private health insurance claim (who comprised around 90 per cent of
separations in public hospitals and 20 per cent in private hospitals in 2007-08).
A key difference between the collections is that the NHCDC has data on hospital
expenditure (costs), whereas the HCP has data on amounts charged to patients.
The NHCDC was used as the primary data source because it is designed for cost
analysis and covers a significant share of separations in both public and private
hospitals. The HCP was only used for private-patient medical and diagnostics costs, as
these are not captured in the NHCDC.
While the NHCDC is the best available data source for the purpose of analysing costs,
it does have major limitations. For example, the NHCDC data provided to the
Commission are from an unweighted sample, and so may not be representative of all
hospitals; do not identify how the different tax treatment of for-profit and other hospitals
affects costs; and exclude the asset-value data required to calculate a user cost of
capital. Other data sources were used to fill some of these gaps, such as ABS surveys
of private hospitals and state government asset records.
The Commission also obtained data from the Department of Veterans’ Affairs (DVA) on
the cost of procuring hospital services for war veterans and their families. While DVA
data are not necessarily representative of the whole population, they may provide a
broad indication of the robustness of the Commission’s estimates. It could be argued
that DVA patients are often treated in the same hospitals and by the same clinicians as
other private patients. Furthermore, DVA patients may receive broadly comparable
treatments in public and private hospitals.

Cost per casemix-adjusted separation (all DRGs)

The Commission’s experimental cost estimates suggest that, at a national level,

public and private hospitals had a broadly similar cost per casemix-adjusted
separation in 2007-08 (table 4).

There do, however, appear to be differences between jurisdictions. In New South

Wales and Victoria, private hospitals were estimated to have a higher cost per
XL PUBLIC AND PRIVATE
HOSPITALS
casemix-adjusted separation than public hospitals. In other jurisdictions, private
hospitals were estimated to have a lower cost per casemix-adjusted separation than
public hospitals, with the gap particularly large in Western Australia. To some
extent, these differences between jurisdictions may also be attributable to
inconsistencies in how each jurisdiction measures and reports costs.

Table 4 Cost per casemix-adjusted separation by jurisdiction and

sector, 2007-08a
Dollars

NSW Vic Qld SA

Cost component Public Private Public Private Public Private Public Private

General hospitalb 2 511 1 944 2 106 2 004 2 683 1 948 2 800 1 803
Pharmacy 164 42 235 87 174 45 146 53
Emergency 205 16 251 50 211 40 135 61
Medical & diagnosticsc 733 1 497 900 1 226 794 1 404 621 1 214
Prostheses 137 620 108 527 121 491 140 495
Capitald 439 210 359 240 560 223 381 158
Totale 4 189 4 330 3 960 4 133 4 543 4 151 4 223 3 783

WA Tas, NT & ACTf Australia

Public Private Public Private Public Private

General hospitalb 3 094 1 845 3 243 2 236 2 552 1 953

Pharmacy 202 144 186 55 187 68
Emergency 147 11 238 21 208 34
Medical & diagnosticsc 1 048 1 275 725 1 391 798 1 346
Prostheses 155 555 141 540 131 542
Capitald 359 281 447 345 426 230
Totale 5 006 4 111 4 980 4 586 4 302 4 172
a Costs are casemix adjusted using DRG-level cost weights for public and private hospitals combined.
b NHCDC cost buckets for ward nursing, non-clinical salaries, allied health, critical care, operating rooms,
ward supplies and other overheads, on-costs, hotel costs, and specialist procedure suites. c Combination of
data from the NHCDC (ward medical, imaging and pathology) and HCP (medical charges, which in the HCP
includes diagnostics). d Depreciation and the user cost of capital. e Totals may not equal sum of components
due to rounding. f Data for Tasmania, the Northern Territory and the ACT are aggregated to protect the
confidentiality of the small number of hospitals in each of these jurisdictions.

There were significant differences between public and private hospitals in the
composition of costs. For general hospital costs, public hospitals were estimated to
have a higher cost per casemix-adjusted separation than private hospitals ($2552
versus $1953 at the national level). This was also the case with the experimental
estimates of capital costs ($426 versus $230). Conversely, average prostheses costs

OVERVIEW XLI
were estimated to be much lower in public hospitals ($131 versus $542). Average
medical and diagnostics costs were also estimated to be lower in public hospitals
($798 versus $1346).

A similar pattern in the cost components was evident at the jurisdiction level, and
when costs were disaggregated by hospital size and region. The estimates do
suggest, however, that private hospitals in outer regional areas had a significantly
lower cost per casemix-adjusted separation than public hospitals.

However, the Commission suggests that these estimated differences between public
and private hospitals in the composition of costs should be used cautiously. In
particular, the Commission found that a significant proportion of public-patient
medical costs in the NHCDC are embedded in the general hospital and emergency
categories (estimated to be in the order of $270 per separation nationally). Hence,
the experimental estimates overstate the cost advantage that public hospitals have in
medical and diagnostics, and the cost disadvantage that public hospitals have in
general hospital and emergency departments.

Average cost per separation (individual DRGs)

The broad similarity between public and private hospital costs was also evident at
the level of individual DRGs (as shown by clustering around the 45 degree line in
figure 2). Nevertheless, around one-fifth of DRGs had a cost per separation in
public hospitals that was at least 10 per cent lower than in private hospitals, and
nearly half of DRGs had an average cost in public hospitals that was more than
10 per cent higher than in private hospitals.

Under the DRG classification system, individual DRGs can be grouped into the
‘partitions’ of surgical, medical or other. The experimental estimates suggest that
almost three-fifths of surgical DRGs had a cost per separation in private hospitals
that was at least 10 per cent lower than in public hospitals (figure 3). Medical DRGs
were where public hospitals performed most strongly in terms of cost relative to the
private sector, with almost a quarter of medical DRGs having a cost per separation
in public hospitals that was at least 10 per cent lower than in private hospitals.

The DRG classification system also enables DRGs to be grouped into over
20 different Major Diagnostic Categories (MDCs). The Commission’s experimental
estimates suggest that, in 2007-08, cost per separation in public hospitals was:
• over 10 per cent higher than in private hospitals for almost half the MDC groups
• between 90 and 110 per cent of that in private hospitals for half the MDC groups

XLII PUBLIC AND PRIVATE

HOSPITALS
• over 10 per cent lower than in private hospitals for only one MDC (diseases and
disorders of the circulatory system).

Figure 2 Comparison of cost per separation for individual DRGs in

public and private hospitals, 2007-08a
25

45 degree
line
20
Public hospitals ($'000)

15

10

0
0 5 10 15 20 25

Private hospitals ($'000)

a A point is located above (below) the 45 degree line if the relevant DRG has a higher (lower) cost per
separation in public hospitals than in private hospitals. DRGs with a cost per separation of more than $25 000
in at least one sector are not shown on the graph. These excluded DRGs accounted for less than 2 per cent of
separations among the sampled DRGs and hospitals.

Data from the Department of Veterans’ Affairs (DVA) on the cost it incurs in
procuring hospital services for veterans and their dependants provide a useful point
of comparison with the Commission’s DRG-level cost estimates. Data were
obtained for the 20 most significant DRGs for DVA on the basis of expenditure.
The data indicate that:
• Cost per separation for DVA patients in public hospitals was within 90 to
110 per cent of the Commission’s estimates for about one-third of the 20 DRGs,
with the remainder evenly split between more than 10 per cent below and 10 per
cent above the Commission’s estimates.
• Cost per separation for DVA patients in private hospitals are more comparable
with the Commission’s estimates. Half of the 20 DRGs had a cost for DVA

OVERVIEW XLIII
 patients in private hospitals that was within 90 to 110 per cent of the
Commission’s estimates. The cost for DVA patients in private hospitals was
more than 10 per cent below the Commission’s estimate for three DRGs, and
more than 10 per cent above for the remaining seven DRGs.

Figure 3 DRG partitions by whether lower cost in public or private

hospitals, 2007-08a
100 Public cost per
separation at
Share of DRGs in partition (per cent)

least 10 per
80 cent higher
than private
cost
60 Public cost per
separation
within 10 per
40 cent of private
cost

20
Public cost per
separation at
least 10 per
0 cent lower
Surgical Medical Other All sampled than private
DRGs cost
DRG partition

a Separations are assigned to the surgical, medical or other partitions on the basis of the type of procedure
involved. A procedure is a clinical intervention that carries a procedural or anaesthetic risk, and/or requires
specialised training, facilities or equipment. A separation is classified as surgical if it involves at least one
operating-room procedure; medical if there is no procedure; and other if it involves a procedure performed
outside of an operating room, such as dental extractions and colonoscopies.

Partial productivity measures

A major determinant of cost differences between hospitals will be their productivity

(quantity of outputs relative to inputs). Ideally, this would be measured in terms of
total factor productivity (TFP), which takes account of all inputs and outputs
simultaneously. However, the necessary data to measure TFP for hospitals are not
available. Instead, the Commission has examined partial productivity measures,
which quantify output per unit of a single input, such as labour or medical supplies.

At a national level, public and private hospitals have similar separations per bed (a
measure of capital productivity). However, separations per non-medical staff
member (a measure of labour productivity) are higher in private hospitals than
public hospitals. This suggests that private hospitals have leaner staff-to-bed ratios.

XLIV PUBLIC AND PRIVATE

HOSPITALS
Since 2002-03, both public and private hospitals have increased the intensity with
which they use drug, medical and surgical supplies.

Patients in private hospitals have a shorter average length of stay per separation than
in public hospitals. This appears to be because surgical procedures in private
hospitals have shorter associated patient stays than other DRGs, and private
hospitals undertake relatively more surgical procedures than public hospitals.

These findings on hospital productivity should be interpreted with care, since they
do not control for the different characteristics of public and private hospitals, such
as whether there is an emergency department, patient-risk characteristics, and the
geographic remoteness of a hospital. In addition, partial productivity measures do
not control for differences in the use of inputs other than those included in the
measure.

Access to hospital services

Waiting times

The ability of patients to access timely medical and surgical services is an important
objective of governments and motivator for private health insurance. Under the
NHA, state and territory governments have agreed to provide hospital services
based on clinical need within a clinically-appropriate period.

Emergency department data are collected using nationally-consistent definitions.

However, care needs to be exercised when interpreting emergency department
waiting data. There are significant differences in the way states and territories
assign clinical-urgency categories. In addition, recent findings by the Victorian
Auditor-General raise questions about the accuracy of waiting-time data.

The available data suggest that the rates at which patients are seen within
emergency department triage benchmarks have improved nationally. The proportion
of patients attending emergency departments that were seen on time increased from
66 to 69 per cent between 2002-03 and 2007-08. The proportion of semi-urgent
patients that were seen on time increased from 61 to 66 per cent during the same
period. However, there were major differences between jurisdictions, with relatively
fewer patients seen on time in the ACT and Northern Territory.

The number of people seeking elective surgery grew 4.8 per cent per year between
2002-03 and 2007-08, while the number of elective surgery admissions only grew
by 1.8 per cent per year. The average number of days that the 50th percentile patient

OVERVIEW XLV
waited for elective surgery rose from 28 days in 2002-03 to 34 days in 2007-08. On
the other hand, the share of patients waiting more than a year fell in New South
Wales, Victoria, Queensland, Western Australia and Tasmania, and grew in South
Australia, the ACT and the Northern Territory. However, surgery waiting-list times
tend to underestimate the actual wait for surgery as they do not account for the time
elapsed between referral to and consultation with a surgeon, or between the surgical
consultation and being put on a waiting list.

Capacity utilisation

While it is desirable to have utilisation close to capacity in most sectors, and this is
generally viewed as an indicator of efficiency, in hospitals there is a tipping point
above which a high rate of bed occupancy can impede efficient patient flows,
especially if the hospital is subject to the uncertainty of emergency admissions.

The Australian Medical Association and Australasian College for Emergency

Medicine noted that a hospital’s bed occupancy rate is a useful indication of the
likelihood that a patient will have timely access to a hospital bed. Occupancy rates
above 85 per cent were claimed to lead to regular bed shortages and periodic bed
crises. In 2007-08, public hospitals had an average bed occupancy rate of
87 per cent, although this had declined since 2002-03. Private hospital occupancy
rates were, on average, 76 per cent in 2007-08 and had increased since 2002-03.

Improving future comparisons of costs, productivity and access

The Commission’s analysis of hospital costs, productivity and access has revealed
significant shortcomings in available data for the purpose of comparing public and
private hospitals. A foreshadowed shift to nationally-consistent activity-based
funding for public hospitals by 2014-15 should lead to more robust cost estimates
for the public sector. It would be desirable for private hospitals to report cost data
using the same methodology to ensure data consistency with public hospitals,
together with some rationalisation of existing private-hospital reporting
requirements to minimise the extent of any additional reporting burden. There is
also scope for significant improvement in the collection of data on capital costs for
both public and private hospitals.

Implementation of the NHA, under which governments have agreed to report

nationally-consistent ‘progress measures’ through the COAG Reform Council, is
expected to improve the reporting of partial productivity and access indicators. It
would be useful to also have private hospitals report such data using the same
methodologies.

XLVI PUBLIC AND PRIVATE

HOSPITALS
Hospital-acquired infections and other safety indicators
Hospital-acquired infections are the most common complication affecting hospital
patients, and in many cases are preventable. A recent study estimated that Australia
has 180 000 hospital-acquired infections annually and these occupy around two
million bed days.

Infections are typically categorised by organism, body location and/or patient

population. Organisms that have developed resistance to standard antibiotics —
such as methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-
resistant enterococci (VRE) — are often the focus of reporting regimes because
they have the greatest impacts on health systems in developed countries. Another
organism that is gaining prominence is Clostridium difficile, which usually causes
diarrhoea and is the most common cause of health care-associated gastrointestinal
infection. The body locations of greatest concern, and hence often reported, are
surgical sites and the bloodstream. It also common to report infections for high-risk
populations, such as patients in intensive-care units.

The measurement and comparison of infection rates is not straightforward. A

hospital could have a relatively high infection rate simply because its workload is
concentrated on procedures and patients that have a high risk of infection. It is
difficult to disentangle this risk effect from genuine differences in how hospitals
perform on infection rates. A common approach is to subdivide infections data into
groups with a similar risk of infection, and only compare infection rates within
those groups. Such groups could be defined by patient characteristics, procedure
and/or type of hospital. For surgical-site infections, many countries stratify data
according to a risk index which controls for a patient’s physical condition, length of
surgery and wound cleanliness. However, none of these risk-adjustment methods
fully controls for all of the sources of risk differences between hospitals.

A further problem is that Australia does not have a robust nationally-consistent data
collection for comparing hospital-acquired infections. Data from the best available
source of national data — the Australian Council on Healthcare Standards (ACHS)
— indicate that infection rates rarely differ between public and private hospitals.

ACHS data show that only four out of 47 infection indicators had a statistically
significant difference between public and private hospitals in 2007. Where
differences existed, the ACHS data suggest that private hospitals consistently have
lower infection rates than public hospitals (table 5). However, the ACHS data
collection was not designed for inter-hospital comparisons. It is limited by potential
sample-selection bias, small sample sizes, self reporting without external validation,
and no risk adjustment to reflect patient differences.

OVERVIEW XLVII
Table 5 ACHS infection indicators that differed between public and
private hospitals, 2007a
Infection No. of reporting
rateb hospitals

Indicator no. and descriptionc Units Public Private Public Private

1.2 Deep incisional SSI in hip per 100 0.99 0.63 38 96

prosthesis proceduresd procedures

1.17 Superficial incisional SSI in per 100 2.02 0.94 16 37

abdominal hysterectomy procedures

5.2 ICU-associated new MRSA per 10 000 16.70 7.18 25 23

healthcare-associated ICU overnight
infections in a nonsterile occupied bed days
site

5.4 Non ICU-associated new per 10 000 2.77 1.11 68 59

MRSA inpatient healthcare- non-ICU overnight
associated infections in a occupied bed days
nonsterile site
a The ACHS identified an indicator as differing between the public and private sectors if public/private status
explained at least 10 per cent of the variation in sampled infection rates, and statistical testing showed that the
probability of a difference between public and private rates was at least 95 per cent. However, the data may
be misleading due to sample-selection bias, small sample sizes, self reporting, and no risk adjustment to
reflect differences in patient characteristics. b Mean infection rates after applying the shrinkage estimation
method to the data. c The following abbreviations are used: ICU (intensive-care unit); MRSA (methicillin-
resistant Staphylococcus aureus); and SSI (surgical-site infection). d The ACHS (sub. 13) advised that
unpublished data for 2008 showed that indicator 1.2 was 0.68 per 100 procedures for private hospitals and
1.02 for public hospitals.

More suitable data are collected by state governments as part of their

infection-surveillance programs. Such data are not collected and reported on a
nationally-consistent basis, but public and private hospitals are included in most
cases. The data collected by state governments suggest that private hospitals have
lower infection rates than public hospitals, but this result could be misleading
because private hospitals generally undertake procedures and treat patients with a
lower risk of infection, such as planned (elective) surgery. A more definitive finding
will require the development of data collections and methods that enable risk
differences between hospitals to be more comprehensively distinguished from
genuine differences in performance.

Foreshadowed developments, such as performance reporting under the NHA, will

move Australia closer to a robust nationally-consistent data collection on
hospital-acquired infections. However, more actions will be required to enable
meaningful infection-rate comparisons between public and private hospitals. An
important step in this regard would be to include private hospitals in national
reporting arrangements. The Australian Commission on Safety and Quality in

XLVIII PUBLIC AND PRIVATE

HOSPITALS
Health Care (ACSQHC) is leading and coordinating improvements that should
improve the feasibility of future comparisons.

Other safety and quality indicators

Apart from hospital-acquired infections, the Commission examined a number of

indicators of quality and patient safety. Many of these were identified for future
reporting against the NHA.

ACHS data suggest that there is little difference between public and private
hospitals in rates of readmission to hospital within 28 days, and rates of return to
operating theatre or intensive-care unit. There are no discernible differences for
patient falls, pressure ulcers, adverse transfusion and adverse drug events,
intentional self harm, and certain obstetric indicators. ACHS data, however, must be
treated with caution, as noted for infection rates.

Data on adverse events, such as those published by the Australian Institute of

Health and Welfare (AIHW), suggest that the rate of adverse events in public
hospitals is about 5.5 per cent compared to 3.7 per cent for private hospitals.
However, these data do not account for differences in the activities undertaken by
public and private hospitals, and the extent to which some adverse events, such as
adverse reactions to medications, occurred prior to hospitalisation.

The AIHW, at the request of the ACSQHC, has proposed a set of quality and patient
safety indicators. A similar set of indicators is being proposed for the NHA. Given
the paucity of data in this area, the Commission supports this development.

Multivariate analysis
Partial indicators, such as costs and infection rates, have at least two limitations.
First, no single indicator provides an overall assessment of a hospital’s
performance, since they are by definition partial indicators. Second, there is a large
range of factors outside the control of a hospital that can influence its performance,
including patient mix and geographic location.

To address the limitations of partial indicators, the Commission also undertook a

multivariate analysis of hospital-level data. This involved the use of statistical
methods that enable the many different factors influencing a hospital’s performance
— such as types of patients treated and services provided — to be factored into an
overall assessment of performance. Such an approach has been used in many
overseas studies of hospitals.
OVERVIEW XLIX
To undertake the multivariate analysis, the Commission used 2006-07 data for:
• 368 acute public hospitals operated by governments, covering all jurisdictions
• 18 acute public hospitals operated by nongovernment bodies, covering all
mainland states
• 122 private acute hospitals that collectively account for around 50 per cent of all
private hospital separations.

The results of the multivariate analysis suggest that, after controlling for differences
in services provided and types of patients treated, the efficiency of public and
private hospitals is, on average, similar. It was estimated that the output of
individual hospitals in both sectors is, on average, about 20 per cent below best
practice among the sampled hospitals.

The multivariate results also suggest that the scope for improvement varies
somewhat with hospital size (defined by annual casemix-adjusted separations),
particularly for private hospitals. It appears that, among large and very large
hospitals (more than 10 000 annual casemix-adjusted separations), the scope to
improve efficiency is greatest for public hospitals. At the other extreme, it appears
that the scope to improve efficiency among small and very small hospitals (5000 or
fewer annual casemix-adjusted separations) is higher for private hospitals, although
this result may be partly due to a number of factors that could not be accounted for
in the analysis.

There is potential to extend the multivariate approach by also using data that is
available for the three preceding years, and to take account of differences in input
prices by also modelling a cost function. The Commission intends to undertake this
analysis in coming months and publish the results in March 2010.

Informed financial consent

The terms of reference ask the Commission to examine aspects of informed
financial consent (IFC) for privately-insured patients. IFC occurs when patients
receive relevant cost information, including notification of likely out-of-pocket
expenses (gaps), prior to agreeing to treatment. The provision of relevant cost
information prior to treatment is desirable as it allows patients to make more
informed choices.

It appears that most patients do not face a problem with a lack of IFC. The latest
available data show that around 85 per cent of hospital medical services currently
provided to privately-insured patients do not have an associated gap payment

L PUBLIC AND PRIVATE

HOSPITALS
(excluding excesses and co-payments, which are due to patient choice about
insurance cover). For the remaining 15 per cent of services that do have a gap,
4 percentage points involve known-gap agreements where there is a legal
requirement to provide IFC. Thus, it could be argued that no more than 11 per cent
of services lack IFC. A portion of this 11 per cent would be emergency cases, for
which it is not always realistic to expect IFC. Furthermore, there has been a
declining number of complaints to the Private Health Insurance Ombudsman about
a lack of IFC in recent years. Nevertheless, it is undesirable for any non-emergency
patients to incur out-of-pocket expenses without IFC, and desirable for as many
emergency patients to receive IFC as is practicable.

The Australian Government Department of Health and Ageing commissioned

surveys on IFC in 2004, 2006 and 2007. The surveys were undertaken by Ipsos
Australia and asked individuals who had recently made a private health insurance
claim for an episode of hospitalisation about the costs they incurred and the
information they received about those costs prior to treatment. This is the best
available data source to consider IFC and out-of-pocket expenses in the way
stipulated by the terms of reference.

However, the Ipsos surveys have a number of potential limitations. There might be
sample-selection bias because individuals were more likely to respond to the survey
if they did not receive IFC or incurred significant out-of-pocket expenses. This
would lead to an understatement of IFC rates and overstatement of gap payments.
Another potential limitation is that the surveys relied on patient recollections. The
possibility of sample-selection and self-reporting bias was tested by comparing the
Ipsos results with (census) data collected by the health insurance regulator (the
Private Health Insurance Administration Council, PHIAC). That comparison
indicates that the Ipsos surveys overstate the incidence and average size of out-of-
pocket expenses (table 6). Analysis of the Ipsos data should therefore be considered
in light of the aforementioned data limitations.

Table 6 Comparison of Ipsos and PHIAC data, 2007

Unit Ipsos PHIACa
Share of services with a gap paymentb % 28 18
Share of patients with a gap paymentc % 42 na
Average gap across all servicesb $ 131 23
Average gap across all services where a gap occurredb $ 465 126
a Year to June 2007. b Excludes accommodation fees. c Excludes test-related specialities. na Not available.

The 2007 Ipsos data show that the IFC rate for pre-planned admissions was lowest
in private hospitals (around 80 per cent, compared to about 90 per cent in public

OVERVIEW LI
hospitals) and for treatment by paediatricians (61 per cent). The data also show that,
for people who did not provide IFC, the average out-of-pocket gap was greatest in
private hospitals ($858, compared to $637 in public hospitals) and for treatment by
an orthopaedic surgeon ($753). There was a large range from the minimum to
maximum gap ($5 to $19 827), but this was due to a small number of very low and
high reported gaps. Of the out-of-pocket gaps where no IFC was provided, around
55 per cent were less than $500, 75 per cent less than $1000, and 90 per cent less
than $2000 (figure 4). Only 2 per cent of the gaps were greater than $4000.

Figure 4 Distribution of non-IFC out-of-pocket gap payments, 2007a

40
Per cent of observations

35
30
25
20
15
10
5
0
500 1000 1500 2000 2500 3000 3500 4000
Gap ($)

a Only patients who considered their admission to be pre-planned are included. Maternity/obstetrics
admissions are considered pre-planned. Gaps over $4 000 represented approximately 2 per cent of
observations.

The medical profession has sought to promote best practice with respect to IFC in
recent years, including through educational campaigns. In addition, some individual
specialists are using internet-based packages to inform individual patients about
their likely out-of-pocket expenses.

Medicare Levy Surcharge

The Medicare Levy Surcharge (MLS) is imposed on taxpayers who earn above a
specified income threshold and do not have private patient hospital cover for
themselves and all family members. The MLS is currently calculated at the rate of
1 per cent of taxable income. It was introduced in 1997 as part of a suite of
measures designed to arrest a decline in the share of the population with private
health insurance (PHI) policies, and to maintain the private hospital system as a
complement to the public hospital system. Other measures included a 30 per cent

LII PUBLIC AND PRIVATE

HOSPITALS
rebate on PHI premiums (introduced in 1999) and the Lifetime Health Cover
community rating scheme (introduced in 2000).

There are different MLS income thresholds for singles and couples. The thresholds
were originally set in 1997 and remained unadjusted until the 2008-09 financial
year, when there was a one-off increase in the thresholds and the introduction of
annual indexation for subsequent years. As taxable incomes generally increased
between 1997 and 2008, an increasing proportion of taxpayers became subject to
the MLS legislation. In 1997-98, around 8 per cent of single taxpayers exceeded the
singles income threshold. However with no indexation of the MLS thresholds, this
proportion increased to approximately 33 per cent of single taxpayers by 2007-08.

The Commission has been asked to consider the most appropriate indexation factor
for the MLS income thresholds. The thresholds are currently indexed by average
weekly ordinary time earnings (AWOTE). This was compared to three alternatives:
• average weekly total earnings (AWTE)
• consumer price index (CPI)
• wage price index (WPI).

The Australian Government has stated that the primary reason for indexing the
income thresholds is to ensure that the MLS remains targeted at the ‘high’ income
group for which it was intended. The suitability of the alternative indexation factors
was therefore assessed by determining how successful they would be in keeping the
share of taxpayers subject to the MLS stable over time. The assessment involved
estimating how the thresholds would have changed if there had been indexation
since 1999-2000.

If the MLS income thresholds had been indexed since 1999-2000, the proportion of
taxpayers subject to the MLS would have risen (by varying amounts) using each of
the potential indexation factors assessed by the Commission (figure 5). In theory,
this could be avoided by using an indexation factor based on the ninth decile
(90th percentile) of the income distribution, as it would specifically measure income
changes for higher income earners. However, such a measure is not currently
available in a form suitable for the regular updating of the MLS income thresholds.
The Commission has therefore concluded that AWOTE is the most appropriate
indexation factor, since among the available options it would keep the share of
taxpayers subject to the MLS most stable.

OVERVIEW LIII
Figure 5 Proportion of family taxpayers subject to the MLS under
alternative indexation options, 1999–2008a

AWOTE AWTE CPI WPI No indexation

35

30
Per cent of family taxpayers

25

20

15

10

0
99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08
Financial year

a For simplicity, the $1500 threshold increase for a second and each additional child was not taken into
account in the calculations. Therefore, the figure slightly overstates the proportion of family taxpayers who
would have been subject to the MLS.

LIV PUBLIC AND PRIVATE

HOSPITALS
Findings

Data availability
FINDING 1.1

The Commission encountered significant delays in accessing hospital-related data

beyond what could reasonably be expected to address privacy or confidentiality
concerns. There is a case for making these data more accessible to a range of users
because this could drive improvements in health care, especially as competitive
markets only have a limited role in the health sector. It could also encourage future
improvements in the data collections. Data agencies could facilitate greater data
access by:
• having established protocols allowing access to data that does not breach
privacy rules and meets certain public-interest requirements
• strengthening the mechanisms through which data users can provide ongoing
input on how data are collected and made available for analysis and research.

FINDING 1.2

Information collected by the Australian Bureau of Statistics (ABS) at considerable

public expense is a valuable resource for Australia. The Commission found that the
ABS has a number of requirements that can restrict measured use of these public
data, which the Commission suggests be reviewed by the Australian Government.
There are also barriers to accessing data held by the Australian Institute of Health
and Welfare (AIHW), due to a requirement to obtain approval from jurisdictions to
release the data. This could be addressed by the states and territories delegating the
approval of individual data requests to the AIHW, particularly where information
would only be divulged to established data users and in a form that does not identify
individual patients or hospitals.

FINDINGS LV
Australia’s public and private hospital systems
FINDING 4.1

Although there is significant diversity within and across the public and private
hospital sectors in Australia, there are a number of key similarities between public
and private hospitals that enable and encourage comparison between the sectors. It
is acknowledged that there are some differences in the activities undertaken by
public and private hospitals and that the sectors do not always service a
comparable patient population, which makes comparisons more difficult.

Costs
FINDING 5.1

Existing datasets on hospital and medical costs are limited by inconsistent

collection methods and missing information. The Commission has sought to address
these limitations by drawing on various data sources and incorporating adjustments
to make the data more comparable where possible, as well as noting data
deficiencies where they exist. The resulting estimates of hospital and medical costs
should be considered experimental.

FINDING 5.2

The Commission’s experimental cost estimates suggest that, at a national level in

2007-08, public and private hospitals had broadly similar costs per
casemix-adjusted separation. There were, however, significant differences in the
composition of estimated costs:
• the combined cost of nursing and other salaries, allied health, operating rooms
and specialist suites, critical care, hotel costs, supplies, and on-costs were on
average higher for public hospitals
• medical and diagnostics costs were higher for private hospitals, although there
are some recognised constraints with available data not separately identifying
all medical costs in public hospitals
• prostheses costs were higher in the private sector, but this is also likely to reflect
a broader range of products being available for use in private hospitals
compared to the public hospital sector
• capital costs were estimated to be somewhat higher for public hospitals, but the
extent of this result is particularly reliant on a range of data sources and
adjustments to make the data more comparable.
These differences were also evident when the estimates were disaggregated by
jurisdiction, region and hospital size.

LVI PUBLIC AND PRIVATE

HOSPITALS
 FINDING 5.3

A disaggregation of the Commission’s experimental cost estimates by

diagnosis-related groups (DRGs) suggests that in 2007-08:
• nearly one-fifth of DRGs had an average cost in public hospitals that was at
least 10 per cent lower than in private hospitals, and about half of DRGs had an
average cost in public hospitals that was more than 10 per cent higher than in
private hospitals
• almost three-fifths of surgical DRGs had a cost per separation in public
hospitals that was at least 10 per cent higher than in private hospitals, and
medical DRGs were where public hospitals performed most strongly in terms of
cost relative to the private sector.

FINDING 5.4

A foreshadowed shift to nationally-consistent activity-based funding for public

hospitals is expected to eventually lead to more robust cost data for the public
sector. However, there remains considerable scope to improve the quality and
consistency of hospital and medical cost data in Australia. In particular, there is a
need for:
• private hospitals to report cost data using the same methodology as public
hospitals, and to continue to have a high level of participation in the National
Hospital Cost Data Collection, so that the data are reliable and can be
disaggregated by sector, region, and size and type of facility
• items directly billed to private patients — such as some medical, diagnostics and
medicines — to be linked with cost data reported by hospitals so that all costs
associated with an episode of care are captured in a single collection
• reliable data on capital costs, hospital administration costs, head-office
overheads, and the cost of medicines prescribed to hospital patients
• quantification of the additional FBT liability that for-profit hospitals incur by
not having the FBT exemption that is available to other hospitals.
This may require a strengthening of data-related provisions in the National
Healthcare Agreement for public hospitals, and data-reporting requirements for
private hospitals. If this is the case, governments need to be conscious of the
regulatory burden on reporting hospitals and, where possible, seek to limit it by
avoiding duplication and inconsistency in reporting arrangements, and by utilising
cost-effective electronic reporting of data.

FINDINGS LVII
Hospital-acquired infections
FINDING 6.1

Australia does not have a robust nationally-consistent data collection on

hospital-acquired infections. The limited available evidence suggests that private
hospitals have lower infection rates than public hospitals, but this result could be
misleading because private hospitals generally treat patients who have a lower risk
of infection. A more definitive finding will require the development of data
collections that enable risk differences between hospitals to be distinguished from
genuine differences in performance.

FINDING 6.2

Foreshadowed developments, such as performance reporting under the National

Healthcare Agreement, will move Australia closer to a robust nationally-consistent
data collection on hospital-acquired infections. However, more actions will be
required to enable meaningful infection-rate comparisons between public and
private hospitals. An important step in this regard would be to include private
hospitals in national reporting arrangements. The Australian Commission on Safety
and Quality in Health Care is leading and coordinating initiatives that should
improve the feasibility of future comparisons.

Other partial indicators

FINDING 7.1

Private hospitals appear to operate relatively leaner staffing levels than public
hospitals, although it is not clear how much of this difference can be explained by
the higher provision of emergency department and outpatient clinic services by
public hospitals.

FINDING 7.2

Private hospitals exhibit shorter lengths of stay than public hospitals. This is due to
private hospitals exhibiting relatively shorter lengths of stay for surgical
procedures and undertaking relatively more surgical procedures than public
hospitals.

LVIII PUBLIC AND PRIVATE

HOSPITALS
 FINDING 7.3

Timely access to elective surgery is less likely in public hospitals than in private
hospitals. The relatively high bed occupancy rates in public hospitals restrict their
ability to manage their unpredictable workload. Equity of access is more likely in
public hospitals than private hospitals, since public hospitals provide relatively
more elective surgery to patients from poor socioeconomic areas and from more
remote areas of Australia.

FINDING 7.4

The work of the Australian Commission on Safety and Quality in Health Care and
the Australian Institute of Health and Welfare to develop a national set of safety
and quality indicators could provide a basis for future comparisons between public
and private hospitals. However, the paucity of published, comparable and reliable
hospital-level data severely limits these comparisons, and will continue to limit such
comparisons in the future. Making consistent hospital-level data available to all
interested parties would assist with future comparisons between hospital sectors
and contribute to improvements in care.

Multivariate analysis
FINDING 8.1

A multivariate analysis of Australian hospital-level data established the

best-practice benchmarks for each hospital in the sample. The benchmarks were
influenced by a number of factors. The best-practice benchmarks were lower for
hospitals that treat:
• highly morbid patients
• patients from lower socioeconomic communities
• relatively more medical cases, as these cases are more difficult to manage
• more complex cases, although this is less so for the largest hospitals.

FINDINGS LIX
FINDING 8.2

After controlling for differences in services provided and types of patients treated,
the efficiency of public and private hospitals is, on average, similar. It was
estimated that the output of individual hospitals in both sectors is, on average,
around 20 per cent below best practice among the sampled hospitals. Among large
and very large hospitals, the scope to improve technical efficiency is slightly greater
for public hospitals. At the other end of the scale, the scope to improve efficiency is
higher for small and very small private hospitals, although these results may be
partly due to a number of factors that could not be accounted for in the analysis.

Informed financial consent

FINDING 9.1

According to the Private Health Insurance Administration Council, around

90 per cent of hospital services for privately-insured patients do not have out-of-
pocket expenses that require informed financial consent. Complaints data collected
by the Private Health Insurance Ombudsman suggest that the rate of informed
financial consent has been increasing in recent years.

FINDING 9.2

The incidence and average size of out-of-pocket expenses for privately-insured

patients appear to be overstated in available survey data collected by Ipsos, due to
sample-selection and self-reporting bias. Subject to this qualification, the data
suggest that privately-insured patients have a higher rate of informed financial
consent and lower out-of-pocket expenses in public hospitals. Few conclusions can
be made about out-of-pocket expenses due to small sample sizes.

FINDING 9.3

A more robust future data source on informed financial consent (IFC) could be
created by requiring privately-insured patients to indicate on their health insurance
claim form whether they provided IFC prior to the procedure. Alternatively,
medical specialists and service providers could be required to include as part of the
billing and insurance-claim process an indication of whether documented evidence
of IFC is held for the relevant item. This information could be collected and
reported by the Private Health Insurance Administration Council.

LX PUBLIC AND PRIVATE

HOSPITALS
 FINDING 9.4

The medical profession has sought to promote best practice for informed financial
consent in recent years. This has included educational campaigns for practitioners
and internet-based packages to inform consumers of their likely expenses.

Indexation of Medicare Levy Surcharge thresholds

FINDING 10.1

Average weekly ordinary time earnings is the most appropriate indexation factor
for the Medicare Levy Surcharge income thresholds.

FINDINGS LXI
1 Introduction

Key points
• This study examines three aspects of Australia’s health care system:
– the relative performance of public and private hospitals, with particular regard to
the cost of performing clinically-similar procedures and the rate of
hospital-acquired infections
– rates of informed financial consent and out-of-pocket expenses for
privately-insured patients in public and private hospitals
– the most appropriate factor for indexing the Medicare Levy Surcharge (MLS)
income thresholds.
• The report is structured as follows:
– identification of relevant characteristics of the public and private hospital systems
(chapters 2 to 4)
– comparison of public and private hospital performance using partial indicators,
including for costs and infection rates (chapters 5 to 7)
– discussion of a more comprehensive (multivariate) approach that the
Commission has used to assess relative performance (chapter 8)
– examination of rates of informed financial consent and out-of-pocket expenses
for privately-insured patients (chapter 9)
– assessment of alternative indexation factors for the MLS income thresholds
(chapter 10).
• The Commission encountered significant delays in accessing hospital-related data
for this study that cannot be justified on privacy or confidentiality grounds.
• There is a case for making hospital data more accessible to a range of users
because this could drive improvements in health care, especially as competitive
markets have only a limited role in the health sector. It could also further encourage
future improvements in data collections.
• The Commission thanks study participants for meeting with the Commission,
participating in roundtables and teleconferences, providing data and other
assistance, and making written submissions.

This commissioned study examines issues related to Australia’s public and private
hospital systems, which are an important part of a comprehensive system of services

INTRODUCTION 1
that together contribute to the nation’s health outcomes. The Australian Government
noted in the terms of reference that it requested this study to further its commitment
to improving transparency, accountability and performance reporting within the
health system.

1.1 What the Commission has been asked to do

The terms of reference for this study are provided at the front of this report. In
summary, the Commission has been asked to undertake three distinct tasks:
• compare the relative performance of the public and private hospital systems,
with particular regard to the cost of performing clinically-similar procedures and
the rate of hospital-acquired infections
• report rates of informed financial consent and out-of-pocket expenses for
privately-insured patients
• advise the Government on the most appropriate factor for indexing the Medicare
Levy Surcharge (MLS) income thresholds.

The analysis of costs is to take into account the cost of capital, fringe-benefits tax
exemptions and other relevant factors. Hospital-acquired infections are to be
reported by type of infection. Informed financial consent and out-of-pocket
expenses are to be disaggregated by sector, region and medical specialist.

If the above tasks prove not fully possible because of conceptual problems or data
limitations, the Commission has been asked to propose developments to improve
the feasibility of future comparisons.

1.2 Report structure and study approach

The Commission has structured the analysis in this report as follows:
• identification of relevant characteristics of the public and private hospital
systems (chapters 2 to 4)
• comparison of public and private hospital performance using partial indicators,
including for costs and infection rates (chapters 5 to 7)
• discussion of a more comprehensive (multivariate) approach to assess the
relative performance of public and private hospitals (chapter 8)
• examination of rates of informed financial consent and out-of-pocket expenses
for privately-insured patients (chapter 9)

2 PUBLIC AND PRIVATE

HOSPITALS
• assessment of alternative indexation factors for the MLS income thresholds
(chapter 10).

Comparing the relative performance of hospitals has been the most challenging part
of the study, particularly in the short time available. This is because hospital
complexity and diversity make like-for-like comparisons difficult, and existing data
collections are not well suited to the task. Further elaboration of the study approach
is provided below.

Hospital complexity and diversity

Given the significance of hospital complexity and diversity to the relative

performance of public and private hospitals, a significant part of the report is
devoted to describing these characteristics (chapters 2 to 4), and this precedes the
analysis of relative performance (chapters 5 to 8).

Hospitals are generally complex organisations, with many essential services

produced by a range of health professionals in a location supported by available
technologies, with management oversight and administrative support. This makes
comparisons particularly difficult, especially to distinguish genuine differences in
performance from variation caused by differences in what hospitals do and who
they treat.

Study participants emphasised the importance of taking account of:

• variations in the types of services that hospitals provide, recognising that some
hospitals provide more complex health services that are most costly to provide
and are inherently more risky for patients
• the disruption to planned activity than can be caused by the presence of an
emergency department, and that clinical training can affect the rate of throughput
• the impact of patient characteristics on the performance of public and private
hospitals, recognising that patients with more complex conditions and those
from lower socioeconomic groups require, on average, more intensive health
treatment and are more susceptible to hospital-acquired infections
• the additional tax burden that for-profit hospitals face compared to public and
not-for-profit hospitals.

INTRODUCTION 3
Relative performance

In undertaking its assessment of relative performance, the Commission has been

mindful of the fact that the community places importance on various aspects of
hospitals, including safety, timeliness and amount of resources used. It is difficult to
capture all of these aspects in a single measure, and so it is common to report a suite
of partial indicators that each measure a particular aspect of performance.1 This
approach is reflected in the terms of reference, which specifically asks the
Commission to compare partial indicators for costs and hospital-acquired infections.
This is done in chapters 5 and 6 respectively.

Existing datasets on hospital costs are limited by inconsistent collection methods

and missing information. In chapter 5, the Commission has sought to address these
issues by drawing on various data sources and, where necessary, incorporating
adjustments to make the data more comparable. However, the Commission readily
acknowledges that a number of significant data shortcomings have limited its ability
to construct fully comparable estimates. Data on hospital-acquired infections also
have limitations for the purpose of comparing public and private hospitals, which
are discussed in chapter 6.

As requested in the terms of reference, the Commission has also considered other
indicators of relative performance. Partial indicators of hospital productivity,
access, and quality and patient safety are examined in chapter 7. Again, data
deficiencies were a constraint on this analysis.

Multivariate analysis

Apart from data deficiencies, a further limitation of individual partial indicators is

that they focus on a particular aspect of performance, such as costs, without taking
account of other aspects, such as patient safety. This makes it difficult to form an
overall assessment of performance, even when various partial indicators are
considered collectively. Furthermore, partial indicators rarely control for all
differences that are outside the control of a hospital, such as geographic location
and patient characteristics, which makes it difficult to achieve like-for-like
comparisons between hospitals.

As detailed in chapter 8, the Commission has used multivariate statistical

techniques to address the limitations of partial indicators in forming an overall

1 Examples of this multiple-indicator approach are the National Health Performance Framework
developed for the Australian Health Ministers’ Conference (AIHW 2008b; NHPC 2001), and
the ‘performance indicator framework’ that the Steering Committee for the Review of
Government Service Provision uses for public hospitals (SCRGSP 2009).
4 PUBLIC AND PRIVATE
HOSPITALS
assessment of relative performance. Multivariate analysis has been used
successfully in many overseas studies of hospital performance, but its application in
Australia has been limited to date and with identified deficiencies that the
Commission has sought to overcome.

Outcomes and outputs

Ideally, hospital performance would be measured in terms of patient outcomes.

Individuals seek hospital services in order to improve their physical and emotional
wellbeing relative to what would otherwise be the case. A wide range of measures
have been developed to measure outcomes, including changes in mortality rates, life
expectancy and quality of life.

To a limited extent, the Commission has been able to use proxies for health
outcomes such as infection rates and unplanned readmissions to hospital. However,
like many other studies, data limitations have caused the Commission to mainly
assess performance in terms of hospital outputs, such as the number of patients
treated and procedures performed.

Measuring performance in terms of outputs has the disadvantage that it does not
directly quantify the degree to which a hospital achieves its primary purpose — to
improve health outcomes. Hospital activity may lead to little improvement in health
outcomes for some individuals, or in extreme cases lead to worse outcomes.
However, outputs tend to be easier to measure than outcomes, because the latter
requires tracking of patient health after hospital discharge. As the Centre for Health
Economics (Monash University) noted, such data are not generally available:
Ideally, hospital-level data … linked up to outcomes data would be available with
associated input data on numbers/costs of staff, other inputs (drug use, technology etc)
and capital. Unfortunately this isn’t the case … (sub. 7, p. 2)

Another reason for measuring performance in terms of outputs is that outputs can be
readily attributed to how a hospital manages its resources, whereas attributing cause
and effect is far more difficult for outcomes (Hollingsworth and Peacock 2008).

Effectiveness and efficiency

The Commission has used measures of both effectiveness and efficiency to examine
the relative performance of hospitals.

Effectiveness refers to how successful a hospital is in achieving a particular

objective, such as avoiding hospital-acquired infections.

INTRODUCTION 5
Efficiency, in its broadest sense, refers to how well resources are used to benefit the
wellbeing of the community as a whole (which is determined by service quality, as
well as financial costs). This broad interpretation is known as ‘economic efficiency’
and has three components — the degree to which outputs are produced at least
possible cost (productive efficiency), how resources are allocated across different
uses so as to generate the greatest community wellbeing at a given point in time
(allocative efficiency), and to achieve the greatest possible wellbeing over time
(dynamic efficiency) (box 1.1).

Box 1.1 Components of economic efficiency

Economic efficiency is about maximising the wellbeing of the community. It requires
satisfaction of three components: productive, allocative and dynamic efficiency.
Productive efficiency is achieved when output is produced at minimum cost. It includes
technical efficiency, which refers to the extent to which, in the production of any good
or service, it is technically feasible to reduce any input without decreasing the output,
and without increasing any other input.
Allocative efficiency is about ensuring that the community gets the greatest return (very
broadly defined) from its scarce resources. A nation’s resources can be used in many
different ways. The best or ‘most efficient’ allocation of resources is the one that
contributes most to community wellbeing.
Dynamic efficiency refers to the allocation of resources over time, including allocations
designed to improve economic efficiency and to generate more resources. This can
mean finding better products and better ways of producing goods and services, which
may involve investments in education, research, development and innovation. Dynamic
efficiency can also refer to the ability to adapt efficiently to changed economic
conditions, a capacity for optimally modifying output and productivity performance in
the face of economic ‘shocks’.
Source: PC (2006b, 2008b).

In essence, the Commission has been asked to examine the (level and mix of) inputs
used by hospitals to produce their current outputs, and so the most relevant concept
is productive efficiency. The Commission has not been asked to consider the much
broader question of how well resources are allocated across different parts of the
health sector, which is measured by allocative efficiency. However, by comparing
the performance of public and private hospitals, this study could provide insights
into the potential to improve allocative efficiency.

The Commission recognises that the productive efficiency of hospitals in one sector
relative to the other could depend on the distribution of activity between the two
sectors, and so there might be an interdependence between productive and

6 PUBLIC AND PRIVATE

HOSPITALS
allocative efficiency. In this regard, the Australian Government Department of
Health and Ageing (DOHA) noted that:
Productive and allocative efficiency are often interdependent, with allocative
efficiencies allowing productive efficiencies to produce their maximum benefit. This
can account for some of the perceived differences in efficiency between public and
private hospitals. (sub. 32, p. 14)

Informed financial consent and MLS indexation

The Commission’s analysis of informed financial consent and out-of-pocket

expenses for privately-insured patients is based on survey data collected for DOHA
in 2004, 2006 and 2007 (chapter 9). In essence, these surveys provide the only
available data for this purpose. However, the Commission has had to qualify its
results because the survey data appear to be affected by sample-selection and self-
reporting biases.

In chapter 10, the suitability of four alternative available indexation factors is

assessed by estimating how successful they would have been in keeping the MLS
focused on high-income earners in recent years. The primary reason for indexing
the MLS income thresholds is to ensure that the MLS remains targeted at the high-
income group for which it was intended.

1.3 Future data improvements

Significant progress has been made since the early 1990s in developing national
data standards and datasets on health matters, and this has assisted the Commission
in undertaking this study. A key development was the 1993 National Health
Information Agreement (NHIA) between Commonwealth, state and territory health
authorities, the ABS and AIHW. The current version of the NHIA was adopted in
2004 and also included the Department of Veterans' Affairs as a signatory. More
recently, the National Healthcare Agreement (NHA) has provided added impetus
for the collection and maintenance of nationally-consistent datasets.

Governance arrangements are in place to ensure relevant agencies and jurisdictions

coordinate their efforts with regard to nationally-consistent data collections. In
particular, the National E-Health and Information Principal Committee (NEHIPC)
is responsible for overseeing implementation of the NHIA, and is advised by the
National Health Information Standards and Statistics Committee (NHISSC)
(figure 1.1). NEHIPC comprises representatives from the NHIA signatories and
reports to the Australian Health Ministers’ Advisory Council, which in turn reports

INTRODUCTION 7
to the Australian Health Ministers’ Conference. The AIHW provides a secretariat
for the NHISSC.

Figure 1.1 Governance arrangements for national health data

Source: AIHW (2009e).

Nevertheless, a common theme throughout this report is that improvements could

be made to data collections to improve the feasibility of future comparisons.
Foreshadowed changes — such as strengthened national reporting under the NHA
— will help in this regard. However, as requested in the terms of reference, the
Commission has identified other improvements that could be made.

Improvements to specific data collections are discussed in following chapters in the

context of the relevant indicators being examined. More generally, the feasibility of
future comparisons could be improved by making hospital-related data more
accessible. The Commission encountered significant delays in accessing hospital-

8 PUBLIC AND PRIVATE

HOSPITALS
related data beyond what could reasonably be expected to address privacy or
confidentiality concerns. In one instance, a jurisdiction insisted on obtaining
approval from the head of its health department before providing data that had
already been published as a chart in one of its reports.

The difficulty in accessing hospital-related data is reflected in an absence of

significant past research on the relative performance of Australian public and
private hospitals. This is despite a large amount of data being collected, and the
issue of relative performance of public and private hospitals having been debated
for many years in Australia (for example, Butler 1988b). The Commission has
previously reviewed published research on hospital performance, from which it
concluded that the gap between existing and best-practice productivity might be in
the order of 20–25 per cent for the Australian (public and private) hospitals sector
as a whole (PC 2006a). However, this conclusion was based on a combination of
the limited Australian research, particularly for the private sector, and overseas
studies which may be of limited relevance to Australia.

A number of study participants outside government shared the Commission’s

concerns about the difficulty in accessing hospital-related data (for example,
Australian Health Service Alliance, sub. DR53; Catholic Health Australia,
sub. DR62; Centre for Health Economics Research and Evaluation, sub. DR68;
National Coalition of Public Pathology, sub. DR49; Queensland Nurses Union,
sub. DR51; Rhonda Kerr and Associates, Health Planning, sub. DR44). In contrast,
government agencies responsible for collecting and/or handling data noted that they
were constrained by legislative requirements to maintain privacy and confidentiality
(for example, NSW Health, sub. DR64; SA Department of Health, sub. DR45). An
overview of existing privacy legislation is provided in box 1.2.

There is a legitimate case for privacy and confidentiality safeguards, but it would be
unfortunate if these hindered data access beyond what is necessary to maintain
privacy and confidentiality. The Centre for Health Economics Research and
Evaluation claimed that:
Researchers are often refused access to health-related administrative data with
unjustifiable claims to privacy or confidentiality. While patient privacy is an important
issue, appropriate governance procedures for de-identified data can provide the
necessary privacy protections. (sub. DR68, p. 1)

Given the public expense involved in collecting and maintaining data on health
care, and the potential gains to health outcomes from policies and processes
designed on the basis of the best evidence, there appear to be some broader public
interest aspects that also need to be considered alongside legitimate privacy and
confidentiality concerns.

INTRODUCTION 9
 Box 1.2 Privacy legislation in each jurisdiction
Australian Government — the Privacy Act 1988 (Cwlth) requires Australian
Government agencies to follow a set of eleven Information Privacy Principles, and
private health care providers to comply with a set of ten National Privacy Principles.
These are overseen by the Privacy Commissioner, who is also required under the
National Health Act 1953 (Cwlth) to issue guidelines on how Australian Government
agencies manage individuals’ Medicare and Pharmaceutical Benefits Scheme claims
information. A breach of these guidelines constitutes a violation of the Privacy Act.
The Australian Bureau of Statistics and the Australian Institute of Health and Welfare
are subject to confidentiality requirements under the Census and Statistics Act 1905
(Cwlth) and Australian Institute of Health and Welfare Act 1987 (Cwlth) respectively.
New South Wales — the Health Records and Information Privacy Act 2002 (NSW)
governs the handling of health information in the public sector, and it also seeks to
regulate the handling of health information in the private sector. Privacy NSW has
developed four statutory guidelines under this legislation, which are legally binding and
define the scope of particular exemptions in the health privacy principles.
Victoria — the Health Records Act 2001 (Vic) covers the handling of all personal
information held by health service providers in the public sector and also seeks to
govern practices in the private sector. The legislation contains a set of principles
adapted from the National Privacy Principles.
Queensland — the Information Privacy Act 2009 (Qld) contains nine principles
specifying how the Department of Health is to handle personal information. These
principles have some similarities to the National Privacy Principles in the
Commonwealth Privacy Act.
South Australia — government agencies generally have to comply with a set of
Information Privacy Principles issued under a Cabinet Administrative Instruction. There
is also a Code of Fair Information Practice, which applies to the SA Department of
Health, its funded service providers, and others with access to personal information
held by the Department.
Western Australia — government agencies do not currently have a legislative privacy
regime, but are subject to various confidentiality policies and some privacy principles
are provided for in the Freedom of Information Act 1992 (WA).
ACT — the Health Records (Privacy and Access) Act 1997 (ACT) covers health
records held in the public sector and seeks to apply to practices in the private sector
not covered by the Commonwealth Privacy Act.
Northern Territory — the Information Act 2002 (NT) covers the protection of personal
information, record keeping and archive management of information held in the public
sector. The Information Commissioner is responsible for overseeing the freedom-of-
information and privacy provisions of the legislation.
Source: AIHW (2009); Office of the Privacy Commissioner (2009a, 2009b).

10 PUBLIC AND PRIVATE

HOSPITALS
A common feature of the privacy and confidentiality arrangements is that
responsibility for handling and releasing data is assigned to one or more ‘data
custodians’ in the relevant organisation. In the Discussion Draft for this study, the
Commission expressed the concern that the term custodian could imply the purpose
is to hold data potentially from a range of users. A further concern is that the role of
data custodians could be used to censor information that may highlight deficiencies
in the provision of health services. However, the SA Department of Health noted
that:
The term data custodian is not one of restricting access but about appropriate
governance and management of databases, many of which contain significant personal
information. (sub. DR45, p. 5)

The NSW Department of Health observed that data custodians perform an important
role in the public health system, with their key responsibilities including:
• ensuring that patient privacy is maintained
• ensuring compliance with data provision legislation, probity issues and other
protocols (for example, protecting the commercial interests of private providers and
obtaining any relevant consents required for the release of data)
• ensuring due consideration of any ethical issues associated with the use and release
of data
• ensuring the completeness and accuracy of data to be released, or if necessary,
providing specific caveats regarding the data to be released where there are issues
relating to its completeness and/or accuracy. (sub. DR64, pp. 3–4)

Privacy and confidentiality requirements can be particularly problematic when the

data necessary to compare public and private hospitals are not available from a
single source. The Commission experienced this issue with its multivariate analysis,
which required access to hospital-level data held by the Australian Bureau of
Statistics (ABS) and the Australian Institute of Health and Welfare (AIHW), both of
which are subject to legislative constraints on what data can be provided to external
parties (box 1.2). An arrangement was found to allow the Commission’s
multivariate analysis to proceed, although this was not as straightforward as having
direct access to the data.

Information collected by the ABS at considerable public expense is a valuable

resource for Australia. There appears to be a case for reviewing existing
requirements to enable sensible and measured use of these public data. This could
include consideration of a protocol allowing access to data that does not breach
privacy rules and meets certain public-interest requirements.

The Commission found that a major barrier to accessing data held by the AIHW
was that the Institute had to obtain approval from jurisdictions that supplied the
INTRODUCTION 11
data, even when the information would not be released in a way that identified
individual patients or hospitals.2 The states and territories can use this power, as
both providers of hospitals and collectors of hospital-related data, to ensure
information is only provided to parties, and for purposes, they deem to be
‘appropriate’. One way to address this conflict of interest would be to have a
protocol between the AIHW and jurisdictions that placed greater onus on the AIHW
to ensure individual data requests met the legislated privacy and confidentiality
requirements of each jurisdiction. Delegating the approval of individual data
requests to the AIHW should at least occur for cases where information would only
be divulged to established data users, and in a form that does not identify individual
patients or hospitals. The protocol would be subject to the AIHW’s governance
arrangements, which include state/territory representation on the AIHW
management board.

The problem of aggregating data from more than one source, while also satisfying
privacy and confidentiality requirements, has been a barrier to developing ‘linked’
datasets that can be used to measure the impact of health interventions on outcomes.
The NSW Department of Health noted that many organisations would have to be
involved, and significant community concerns addressed:
For a range of reasons including the involvement of two levels of government as
funders and regulators, as well as the involvement of another two sectors (commercial
and not-for-profit) as providers, it is impossible to chart a patient’s journey through the
health system by analysing a ‘data trail’ because no such trail exists or can be
constructed. The databases for MBS [Medicare Benefits Schedule] and PBS
[Pharmaceutical Benefits Scheme] managed by Medicare Australia (access to which is
extremely restricted) are entirely separate from the admitted and non-admitted
databases maintained by states and territories, and also distinct from private hospital
databases.
… the community is logically uneasy about possible secondary uses of databases. This
means that if the full benefits of data linkage are to be realised, it will be necessary to
demonstrate the value to be gained by linking health activity data from different
sources, and to be very clear about the ‘rules’ that will govern this process. (sub. DR64,
p. 4)

Progress is, however, being made in this regard. At a national level, the Population
Health Research Network (PHRN) has been established to provide researchers with
access to linkable de-identified data from a diverse range of health datasets, across
jurisdictions and sectors. The PHRN has been allocated significant funding from the
Australian Government, in addition to contributions in cash and in-kind from state
and territory governments, and academic partners. A data linkage system has

2 The requirement to get approval from states and territories before releasing data derives from
s.29 of the Australian Institute of Health and Welfare Act 2007 (Cwlth).
12 PUBLIC AND PRIVATE
HOSPITALS
existed in Western Australia since the mid 1990s, with more than 600 research
projects having made use of the data since that time. The SA Department of
Health (sub. DR45) noted that it is a partner in the SA NT Data linkage Consortium
that will provide project-specific deidentified data from a number of administrative
and other datasets for research purposes. Similarly, the Centre for Health Record
Linkage was established in 2006 to create and maintain a system for linking health
and human services datasets in New South Wales and the ACT.

In summary, making hospital-related data more accessible to a broad range of users

would facilitate greater research and analysis of hospitals. This could drive
improvements in health care, especially as competitive markets have only a limited
role in the health sector. The potential gains could be significant, given the
substantial resources that Australia devotes to hospital services and the many people
treated in hospital each year. Greater data accessibility could also facilitate future
improvements in data collections by highlighting weaknesses in existing datasets.

The barriers to accessing hospital-related data are also wasteful because a

substantial amount of information is currently collected at significant cost to
governments and firms, and the potential broader public benefits from this are being
unnecessarily curtailed.

Another way in which greater benefits might be achieved is for data agencies to
strengthen mechanisms through which data users — including those outside of
government, such as academics and private health insurance funds — can provide
ongoing input on how hospital-related data are collected and made available for
analysis and research. At the Commonwealth level, DOHA (sub. DR69) noted that
it already has consultative groups for health insurers and private hospital operators.

FINDING 1.1

The Commission encountered significant delays in accessing hospital-related data

beyond what could reasonably be expected to address privacy or confidentiality
concerns. There is a case for making these data more accessible to a range of users
because this could drive improvements in health care, especially as competitive
markets only have a limited role in the health sector. It could also encourage future
improvements in the data collections. Data agencies could facilitate greater data
access by:
• having established protocols allowing access to data that does not breach
privacy rules and meets certain public-interest requirements
• strengthening the mechanisms through which data users can provide ongoing
input on how data are collected and made available for analysis and research.

INTRODUCTION 13
FINDING 1.2

Information collected by the Australian Bureau of Statistics (ABS) at considerable

public expense is a valuable resource for Australia. The Commission found that the
ABS has a number of requirements that can restrict measured use of these public
data, which the Commission suggests be reviewed by the Australian Government.
There are also barriers to accessing data held by the Australian Institute of Health
and Welfare (AIHW), due to a requirement to obtain approval from jurisdictions to
release the data. This could be addressed by the states and territories delegating the
approval of individual data requests to the AIHW, particularly where information
would only be divulged to established data users and in a form that does not identify
individual patients or hospitals.

1.4 Conduct of the study

The terms of reference for this study were received from the Assistant Treasurer on
15 May 2009. The Commission was originally to report within six months, but the
Assistant Treasurer later extended the final reporting date to early December 2009
because the Commission had encountered delays in obtaining data needed to
undertake the study, and some participants were late in lodging their submissions.

As requested in the terms of reference, the Commission consulted and invited

feedback from relevant experts and other interested parties. This was done in the
following ways:
• At the commencement of the study, a circular was mailed to people and
organisations that the Commission thought might be interested, inviting their
participation. Subsequent circulars were sent to those who had expressed an
interest in the study to keep them updated on progress.
• The study was also advertised in major national newspapers and promoted on the
Commission’s website.
• The Commission met with a cross-section of interested parties to identify
relevant issues and sources of data, including government health departments,
private hospital groups, data agencies, and private health insurers.
• An issues paper was released on 22 June 2009 to assist interested parties in
preparing submissions to the study.
• A roundtable was held with interested parties in Canberra on 30 June 2009 to
explain the study process and obtain input on what data and methodology to use.

14 PUBLIC AND PRIVATE

HOSPITALS
• A teleconference was held with technical experts from individual jurisdictions
and private hospital groups on 17 September 2009 to obtain their views on how
to interpret the cost data that they report to the Australian Government.
• The Commission released a Discussion Draft on 15 October 2009 to provide an
opportunity for interested parties to comment on draft findings prior to
completion of the Final Report.
• A further roundtable was held with interested parties in Canberra on 22 October
2009 to get feedback on the Discussion Draft.
• A teleconference was held with participants on 23 November 2009 to get their
feedback on the results of the multivariate analysis.
• Two external referees provided written comments on the multivariate analysis
(appendix G).

A total of 72 written submissions were received during the study. These were from
a variety of groups, including government health departments, academics, private
hospital groups, professional bodies, and private health insurers.

The Commission thanks study participants for meeting with the Commission,
participating in roundtables and teleconferences, providing data and other
assistance, and making written submissions. Appendix A provides details of the
individuals and organisations that participated in the study.

INTRODUCTION 15
2 Australia’s public hospital sector

Key points
• State and territory governments are assigned specific responsibilities in the delivery
of public hospital services under the National Healthcare Agreement. These include
the provision of free hospital (including emergency) services, with equitable access
to all eligible persons regardless of geographic location, and a major role in clinical
training. These responsibilities shape the volume and type of services that the
public hospital sector can deliver.
• Australia currently has 768 public hospitals, which vary widely in size and location.
However, most are small in size (with 50 beds or fewer) and most are located in
regional areas.
• The main type of service provided by public hospitals is acute care, but their services
extend to rehabilitation, palliative, geriatric, newborn and maintenance care. Around
half of all public hospital separations are same-day admissions.
• Australia’s public hospitals treated over 4 million admitted patients in 2007-08. Over
one-third of patients treated in public hospitals are aged 65 or older, and 14 per cent
of patients in public hospitals elect to be treated as private patients.
• Almost three-quarters of public hospital patients are admitted for medical treatment
while 20 per cent are admitted for surgery. The most common inpatient treatment in
public hospitals is same-day renal dialysis.
• Public hospitals delivered over 48 million occasions of service to non-admitted patients
in 2007-08, including 7.1 million accident and emergency presentations.
• Over the past four years, the number of beds per capita has remained stable, while
waiting times for elective surgery and the volume of emergency cases have risen.
• Given that the delivery of public hospital services is the responsibility of the state
and territory governments, Australia effectively has eight different public hospital
systems. The differing needs and preferences of the states and territories lead to
diversity in structure and service provision within the public hospital sector.
Variations among the states and territories are observed with respect to the:
– centralisation of governance
– mix of diagnoses treated and the share of same-day and overnight admissions
– demographic profile of patients and the proportion admitted as private patients
– number of teaching hospitals
– average salaries of public hospital staff.

AUSTRALIA'S PUBLIC 17
HOSPITAL SECTOR
The provision of public hospital services in Australia is founded on the principle
that all persons eligible for Medicare are entitled to choose to receive health and
emergency services free of charge as a public patient. The principle of public
hospital provision, as well as the responsibilities and governance arrangements of
the public hospital sector, are specified in the current National Healthcare
Agreement (NHA) (and formerly in the Australian Health Care Agreements
(AHCAs)).1

A public hospital is defined as one that is operated by, or on behalf of, the
government of the state or territory in which it is established. This includes
hospitals which are owned by private or charitable groups but are authorised or
contracted by the government to deliver public hospital services (AIHW 2009c).
Although the funding of public hospitals is shared between the Australian, state and
territory governments, the delivery of public hospital services is the responsibility
of each state and territory government. The decentralisation of management
responsibility means that Australia effectively has eight different public hospital
systems which reflect, at least partly, the different population needs, geography and
resource capacity of each state and territory.

This chapter profiles the structure and activity of Australia’s public hospital sector,
including the types of services delivered, the characteristics of the patients treated,
and the workforce of the sector. Recent developments in the public hospital sector
are also reviewed.

2.1 Role and structure of public hospitals

Role of public hospitals

Public hospitals exist in order to fulfil the government’s obligation to provide free
hospital services to all members of the community who meet Medicare eligibility
criteria. More broadly, governments have an incentive to provide public hospital
services to reap the social benefits associated with the achievement of good health
(such as higher workforce participation and productivity) and to minimise the social
costs associated with poor health (such as the costs of infection outbreaks). More
recently, the incentive for governments to provide public hospital services has been
further driven by the community’s growing expectations regarding accessibility to
health services.

1 The NHA replaced the bilateral state and territory AHCAs from July 2009.

18 PUBLIC AND PRIVATE

HOSPITALS
Under the NHA and former AHCAs, it is the responsibility of state and territory
governments to ensure all residents have equitable access to hospital services,
meaning that a broad range of hospital (including emergency) services should be
available, free of charge, throughout each state and territory. In regions where it is
not feasible for local hospitals to provide some types of acute care, patient travel
and accommodation services are often provided. Since funding for public hospitals
is sourced primarily from general taxation — rather than fees for services — the
volume of public hospital services demanded must be managed within the given
budget constraint. Excess demand for services cannot be rationed by price, but by
non-price mechanisms such as waiting lists.

The service responsibilities assigned to public hospitals have implications for their
allocation of resources. First, the obligation to provide emergency services means
that a certain volume of resources must be permanently on standby in public
hospitals which receive emergency cases. Many study participants noted that the
need to divert resources to emergencies can severely interrupt and constrain the
delivery of other hospital services such as elective surgery.

However, detailed analysis undertaken by the NSW Health Surgical Services

Taskforce suggested that this is a symptom of inadequate management rather than
an inevitable consequence of treating emergency cases:
Our Taskforce has recently published the Emergency Surgery Guidelines … In
preparation for this publication, extensive data analysis was conducted and concludes
that emergency surgical admissions were entirely predictable and could be managed
more effectively with a planned approach thereby minimising disruption to elective
surgical services. It is a failure of adequate management of the emergency load that is
the problem. (sub. DR43, p. 1)

Provision of emergency services can have resource implications for infection

control because, as noted by the Australian Healthcare and Hospitals Association
(AHHA, sub. 33), patients undergoing emergency procedures are at greater risk of
infection.

Second, the responsibility of public hospitals to invest in clinical teaching and

research — while important for the training of the future hospital workforce and the
advancement of hospital practice — imposes additional resource costs on the public
hospital sector. It is acknowledged, however, that a hospital’s engagement in
clinical teaching and research also generates potential benefits, as Bio21 Australia
Limited noted:
[T]he presence of any serious commitment to clinical research within a hospital or even
of its staff having such a commitment elsewhere is likely to enhance the quality of
professional service to patients. (sub. 35, p. 1)

AUSTRALIA'S PUBLIC 19
HOSPITAL SECTOR
Third, the responsibility assumed by governments to provide hospital services
across all geographical regions means that a number of public hospitals are
operating in regions where the degree of remoteness — and consequential small
scale of operation — may make such establishments very costly to operate relative
to the volume and type of services they can safely deliver. The Tasmanian
Department of Health and Human Services commented on this point:
Providing hospital care in rural and remote communities is almost entirely the domain
of the public sector which must absorb the scale disabilities imposed by this community
service obligation. (sub. 37, p. 3)

Additionally, public hospitals in some remote areas may take on the responsibility
of providing other public health services, such as aged care and community health
services, which would otherwise not be available in these regions.

Ownership and management

The provision of health and emergency services through the public hospital system
is the responsibility of the state and territory governments (COAG 2008d). State and
territory governments typically divide the management of public hospitals along
geographical lines (for example, metropolitan, regional and rural services), with a
separate division generally established for ambulance services. In many
jurisdictions, management structures have undergone frequent revision. For
example, the number of health service districts in Queensland has been
progressively reduced from 38 to 15 over recent years.

The degree to which the governance of public hospitals is centralised, however,

varies widely among the states and territories, as noted by the Australian
Government Department of Health and Ageing (DOHA, sub. 32). For example,
public hospitals in New South Wales lie within eight area health services which
report directly to the NSW Department of Health. In contrast, the governance of
public hospitals in Victoria is more decentralised, with hospitals, or networks of
hospitals, reporting to their own boards.

In some instances, governments facilitate the delivery of public hospital services by

contracting or authorising a private company to build or operate a hospital
establishment. In this type of arrangement, the hospital is privately owned but
classified as a public hospital because it operates on behalf of the government.
Examples of privately-owned government-contracted hospitals include Mercy
Women’s Hospital (Victoria), Noosa Hospital (Queensland) and Joondalup Hospital
(Western Australia). In some of these arrangements, public hospitals may be

20 PUBLIC AND PRIVATE

HOSPITALS
managed under the auspices of denominational groups (in many cases, Catholic
religious orders) (DOHA, sub. 32).

Public hospital services can also be delivered in partnership with the private sector
when private companies build private hospitals on public hospital campuses. These
co-location arrangements allow for the shared use of infrastructure and facilitate
teaching and research. Examples include the Jessie McPherson Private Hospital
co-located with Monash Medical Centre (Victoria), Flinders Private Hospital
co-located with Flinders Medical Centre (South Australia), and Holy Spirit
Northside Hospital co-located with Prince Charles Hospital (Queensland). A private
hospital co-located in a public establishment is not classified as a public hospital.

Funding arrangements

Although the delivery of public hospital services is the responsibility of the state
and territory governments, funding is provided by both the federal and state or
territory tiers of government.2 Non-government sources, such as insurance funds
and patients’ out-of-pocket payments, also fund a small portion of public hospital
services.

On average, the state and territory governments provide around 53 per cent of
funding for public hospital services, while the Australian Government provides
around 40 per cent and non-government sources contribute around 7 per cent
(figure 2.1). However, there are variations between jurisdictions. In particular,
funding in the Northern Territory and the ACT is more heavily sourced from the
territory government, while funding in Tasmania is more heavily sourced from the
Australian Government, relative to the average funding shares of all jurisdictions.
The contribution of non-government sources also varies widely, comprising as high
as 10 per cent of total funding in the ACT and Victoria, but no more than 4 per cent
of total funding in Queensland, South Australia and the Northern Territory.

2 Public hospital services exclude dental services, community health services, patient transport
services, public health, and health research undertaken by the hospital, but can include services
provided away from the hospital site such as dialysis (AIHW 2009a).
AUSTRALIA'S PUBLIC 21
HOSPITAL SECTOR
Figure 2.1 Funding sources for public hospital services, 2007-08

100
Per cent of total funding

80

60

40

20

0
NSW Vic Qld SA WA Tas NT ACT

Australian Government State/Territory Government Non-government

Source: AIHW (2009c).

Public hospital services constitute the largest item of total health expenditure for the
state and territory governments, and the second-largest for the Australian
Government. Between 2003-04 and 2007-08, the collective share of government
funding coming from the states and territories increased, although the share of total
public hospital funding from government sources declined (AIHW 2009c). Most of
the Australian Government’s expenditure on public hospitals is in the form of
funding conferred to the state and territories through the NHA and formerly the
AHCAs (AIHW 2009c).

In all states and territories except for the ACT, funding for acute inpatient services
is distributed at least partly on the basis of a casemix scheme. Under casemix and
similar activity-based funding schemes, each hospital is funded in relation to the
types of services it provides as well as the severity of patients’ conditions — factors
that are indicative of the hospital resources required. Among the states that adopt
casemix funding, all except for Western Australia use the AR-DRG (Australian
Refined Diagnosis-Related Group) classification system to define the casemix.
Victoria was the first jurisdiction to implement casemix funding in 1993, while
Queensland was the most recent in 2007. New South Wales uses a two-tiered
funding model, incorporating an activity-based funding component for specified
admitted activity (NSW Department of Health, sub. DR64). In some states,
particularly Victoria and South Australia, casemix funding is also applied to
sub-acute and outpatient hospital services (Hurley et al. 2009). Across most
jurisdictions, grant (or per day) funding is used for certain types of acute care where
casemix funding is deemed unsuitable, such as mental health or intensive care.
Some states supplement the casemix funding received by small hospitals in regional

22 PUBLIC AND PRIVATE

HOSPITALS
and remote areas. For example, South Australia designates 35 country hospitals as
‘minimum volume hospitals’ and provides an additional $115 million for their
budget (SA Health, sub. 4, p. 2). In the ACT, the small scale of the public hospital
sector means that budgets are largely allocated on the basis of historic costs.

Casemix and similar activity-based funding schemes are designed to promote

operational efficiency in the delivery of hospital services. Among other advantages,
casemix funding schemes:
• offer a way to link funding to the services delivered
• facilitate an assessment of hospital performance benchmarks and a comparison
of similar or peer hospitals
• enable managers and clinicians to identify inefficiencies in the system and
reallocate resources appropriately (Hurley et al. 2009; NHHRC 2009).

2.2 Characteristics of public hospitals

Traditionally, Australia’s public hospital sector has been typified by large,
metropolitan establishments, whose explicit role was to provide acute medical care,
including emergency services and complex specialist procedures, as well as to
undertake the clinical research and training needed to sustain the hospital sector. A
number of major teaching hospitals in capital cities still reflect this service
approach.

However, it is evident that the role and features of public hospitals are changing
over time, as the sector adapts to the changing needs and characteristics of the
population, and as private hospitals adopt some of the functions traditionally
reserved for the public sector (such as the provision of emergency services and
clinical training). As Australia’s hospitals continue to evolve, there are now many
public hospitals that do not typify the traditional public hospital establishment, just
as there are now many private hospitals that resemble public hospital
establishments.

The location, size and service characteristics of Australia’s public hospitals are now
driven, in part, by the obligation borne by the state and territory governments to
provide all residents with equitable access to hospital care (COAG 2008d). The
existing profile of the public hospital sector, therefore, is very much shaped by the
demographic profile of Australia’s population, as well as historical trends
underlying population growth and patterns of regional dispersion.

AUSTRALIA'S PUBLIC 23
HOSPITAL SECTOR
Number and activity of public hospitals

The most recently available data from DOHA indicate that there are currently
768 acute and psychiatric public hospitals in Australia.3 Hospital activity for
admitted patients is commonly measured in terms of separations, as explained in
box 2.1.

Box 2.1 Measures of hospital activity

Data on hospital care are generally reported in terms of separations. A separation
refers to an episode of care administered to an admitted patient, which ends by either
discharge, death, transfer to another hospital or change in the type of care. During a
hospital admission, a single patient may receive more than one episode of care — and
therefore undergo more than one separation — if they transfer to another hospital or
change their type of care (for example, from acute care to rehabilitation).
Separations only apply for patients who are admitted to hospital. Data on separations
therefore exclude patients who are treated in emergency departments or receive
outpatient services as a non-admitted patient. Care administered to patients in
emergency departments is reported in terms of presentations. Care administered to all
other non-admitted patients is reported in terms of occasions of service.

Australia’s public hospitals recorded over 4 million separations in 2007-08

(table 2.1). Half of Australia’s public hospitals are located in New South Wales and
Victoria, reflecting these states’ high population shares. Between them, these two
states recorded the majority of separations (60 per cent) in Australia’s public
hospital sector. In per capita terms, however, the Northern Territory recorded the
highest volume of admitted patient public hospital activity by a considerable
margin, recording 486 separations per 1000 residents in 2007-08. This compares to
the next highest rate of 256 separations per 1000 residents reported by the ACT.
The lowest rates of separations per capita were administered in Tasmania and
Queensland (184 and 196 separations per 1000 residents respectively).4

3 The data reported in this chapter refer to different time periods due to differences in data
availability. The most recently available data on the total number and location of public
hospitals are for September 2009 (DOHA 2009e). The most recently available data on public
hospital separations, services and expenditure are for 2007-08 (AIHW 2009a, 2009c).
Psychiatric hospitals are not included in the Commission’s analysis due to the specialised nature
and duration of psychiatric treatment and the difficulty of apportioning costs over time, but are
commonly aggregated with acute hospitals in data collections.
4 The data does not capture the extent to which hospitals treat patients outside of their
jurisdictions. Cross-border patient flows occur within all jurisdictions (AIHW 2009a). In
particular, however, ACT Health (sub DR52) noted that 25 per cent of separations reported for
ACT public hospitals are for New South Wales residents.
24 PUBLIC AND PRIVATE
HOSPITALS
Table 2.1 Number and activity of public hospitals, 2007-08a
Number of Number of Number of Proportion
hospitals separations separations per same-day
1000 residents separations b
New South Wales 228 1 466 737 203 44
Victoria 148 1 351 172 248 57
Queensland 177 831 965 196 49
South Australia 80 368 330 216 45
Western Australia 94 458 202 215 51
Tasmania 27 96 270 184 53
Northern Territory 5 90 258 486 62
ACTc 3 81 127 256 54
Australia 762 4 744 061 218 50
a Acute and psychiatric hospitals. b Measured as a per cent of total separations. The remaining share of
separations are overnight. c ACT Health (sub. DR52) advised that these data only include separations from
the ACT’s two major public hospitals.
Source: AIHW (2009a).

On average, half of all separations recorded by Australia’s public hospitals are

same-day admissions, while the remaining are overnight. Yet, some degree of
variation is evident among the states and territories. Overnight admissions are more
common in New South Wales and South Australia, while same-day admissions are
more common in Victoria, Tasmania, the ACT and, in particular, the Northern
Territory where over 60 per cent of all separations are same-day.

Size of public hospitals

Public hospitals vary widely in size: the majority have 50 beds or fewer, yet around
10 per cent have over 200 beds (table 2.2). Tasmania, South Australia, Western
Australia and Queensland have the highest concentrations of small-scale hospitals:
over 80 per cent of hospitals in each of these states have no more than 50 beds. In
contrast, the Northern Territory and the ACT have relatively higher concentrations
of large-scale hospitals, although they have fewer hospitals in total.

AUSTRALIA'S PUBLIC 25
HOSPITAL SECTOR
Table 2.2 Number of public hospitals by size, 2007-08a
0–50 beds 51–100 beds 101–200 beds Over 200 beds
New South Wales 149 29 23 27
Victoria 90 21 19 18
Queensland 142 11 10 14
South Australia 65 6 2 7
Western Australia 73 5 8 8
Tasmania 24 – 1 2
Northern Territory 2 1 1 1
ACT 1 – – 2
Australia 546 73 64 79
a Acute and psychiatric hospitals. – Nil.

Source: AIHW (2009a).

Hospital size has implications for resource efficiency and reported cost differentials.
Compared to large hospitals, small hospitals are less likely to be able to take
advantage of economies of scale or reallocate their resources when workflows vary,
thereby appearing relatively less efficient. At the same time, however, large
hospitals may be constrained in their utilisation of resources due to the requirement
that a certain level of capacity be reserved for emergencies. Large hospitals are also
likely to treat a higher share of more complex — and therefore more costly — cases
that are referred to them by smaller hospitals that are unequipped to treat the cases
themselves. Queensland Health observed that this is a particular characteristic of the
public hospital sector:
[T]here can be significant difference in the types of cases treated at different hospitals
… [T]he most complex cases are typically not undertaken in medium sized regional
hospitals or private hospitals but are referred to the major (generally public) hospitals in
large metropolitan centres. As such, the major tertiary hospitals will on average treat
high complexity (and hence higher cost) cases within any given DRG
[Diagnosis-Related Group] than regional hospitals or private hospitals. (sub. 27, p. 2)

Location of public hospitals

Public hospitals are widely dispersed geographically. Almost one-quarter of

Australia’s public hospitals are located in major cities, and a slightly smaller share
are located in remote areas, yet the majority are located in regional areas (table 2.3).
Exceptions are observed in the ACT, where all public hospitals are located in
Canberra, and in the Northern Territory, where the majority are in remote areas.
Queensland and Western Australia also have relatively high concentrations of
public hospitals in remote areas. This pattern of dispersion reflects the share of each
state or territory’s population living outside of major cities.

26 PUBLIC AND PRIVATE

HOSPITALS
Table 2.3 Number of public hospitals by location, 2009a
Major Inner Outer Remote Very All locationsb
cities regional regional remote
New South Wales 63 80 61 13 5 224
Victoria 56 57 36 2 .. 151
Queensland 18 33 66 24 53 197
South Australia 12 15 27 14 5 73
Western Australia 12 8 28 23 14 85
Tasmania .. 12 15 1 2 30
Northern Territory .. .. 1 2 2 3
ACT 3 – .. .. .. 5
Australia 164 205 234 79 81 768
a Acute and psychiatric hospitals. Location based on ABS (2001) Australian Standard Geographical
Classification. b Includes two public hospitals in New South Wales and three public hospitals in Queensland
that are unable to be geographically classified due to missing postcode data. .. Not applicable (state or
territory does not contain the respective type of area). – Nil.
Source: DOHA (2009e).

The obligation for state and territory governments to provide all residents equitable
access to public hospital services has particularly significant implications for highly
regionalised states, such as Queensland and Western Australia, and to a lesser
degree, South Australia, New South Wales and the Northern Territory. As noted by
DOHA (sub. 32), the distance of hospitals from metropolitan and regional centres
can affect estimated measures of hospital performance. For example, hospitals in
remote areas are likely to incur a higher cost of transporting hospital supplies as
well as greater difficulty attracting staff, which may necessitate higher wages.
Furthermore, some hospitals in remote areas have an added responsibility to provide
primary health and aged care services, which would otherwise not be provided in
their areas. Additionally, as noted by Queensland Health (sub. 27), many regional
and remote hospitals are very small in capacity, and therefore unable to benefit from
economies of scale. ACT Health (sub. DR52) noted that relatively higher costs are
incurred when jurisdictions provide a full range of hospital services for a small
population. The SA Department of Health acknowledged the responsibility held by
each of the state and territories governments, and the implications this bears for the
operating efficiency of public hospitals:
Public hospitals have an obligation to provide all Australians who present to them with
free public hospital care and access to services based on clinical need. Public hospital
access also needs to be provided across the state to ensure reasonable access to hospital
care by residents. This means providing the full range of specialist inpatient, outpatient,
emergency and diagnostic services at all times. For South Australia, it also means
operating minimum volume hospitals in country areas. Due to size and location, such
country hospitals are often relatively expensive to operate, but their importance to
communities cannot be underestimated. (sub. 4, p. 2)

AUSTRALIA'S PUBLIC 27
HOSPITAL SECTOR
The number of beds available per capita also appears to vary according to location.
On average, Australia’s public hospitals provide 2.7 beds per 1000 residents, but
higher bed ratios are reported in New South Wales and South Australia, while lower
bed ratios are reported in Victoria, Tasmania and the ACT (table 2.4).

Table 2.4 Number of public hospital beds per 1000 residents by location,
2007-08a
Major Inner Outer Remote Very All locations
cities regional regional remote
New South Wales 2.7 3.3 3.9 7.7 7.6 2.9
Victoria 2.4 2.7 2.9 2.9 .. 2.5
Queensland 2.3 2.5 3.4 4.0 6.4 2.6
South Australia 2.8 2.4 5.0 7.7 7.5 3.2
Western Australia 2.6 1.8 3.5 3.3 3.1 2.6
Tasmania .. 3.1 1.5 2.9 3.5 2.6
Northern Territory .. – 2.9 5.0 1.0 2.9
ACT 2.6 – .. .. .. 2.5
Australia 2.5 2.8 3.4 4.8 4.0 2.7
a Acute and psychiatric hospitals. Location based on ABS (2001) Australian Standard Geographical
Classification. .. Not applicable (state or territory does not contain the respective type of area). – Nil.
Source: AIHW (2009a).

A general feature of most states and territories is that more beds per capita are
available in remote or very remote areas than in major cities or regional areas.
However, this ratio still varies widely. In remote or very remote areas, there are at
least seven public hospital beds available per 1000 residents in New South Wales
and South Australia, while there are no more than 4 public hospital beds available
per 1000 residents in Victoria, Tasmania and Western Australia.

Patients in public hospitals

Although public hospitals primarily treat public patients, they also treat patients
who elect private status, entitling them to a choice of doctor and/or the offer of
private ward accommodation. About 14 per cent of public hospital separations in
2007-08 were for patients electing private status, most of whom were funded by
private health insurance (figure 2.2). New South Wales and Tasmania had the
highest proportion of patients being treated and billed as private patients in their
public hospitals (20 and 18 per cent respectively), while the Northern Territory and
Queensland had the lowest (5 and 8 per cent respectively).

28 PUBLIC AND PRIVATE

HOSPITALS
Figure 2.2 Share of public hospital separations by patient funding source,
2007-08a

100

80
Per cent of separations

60

40

20

0
NSW Vic Qld SA WA Tas NT ACT
Public Private health insurance Private self-funded
Dept Veterans' Affairs Compensation or other

a Acute and psychiatric hospitals. The share of self-funded patients may be underestimated as some are
unable to be identified. Compensation or other includes workers compensation, other compensation, motor
vehicle third party personal claims, other public authorities, and other funding sources. Data exclude patients
whose funding source is not reported.
Source: AIHW (2009a).

The demographic profile of patients treated in public hospitals is generally similar

across the states and territories (table 2.5). Around one-third of patients treated in
public hospitals are aged 65 and older. An exception is the Northern Territory
which has a relatively lower proportion of patients aged 65 and older, balanced by a
relatively higher proportion of patients aged 35 to 64. This point of difference
reflects the Northern Territory’s distinctly lower share of residents aged 65 and
older in its population (ABS 2008a).

The public hospital sector treats a disproportionately larger share of patients of

relatively low socioeconomic status. Fifty per cent of public hospital patients come
from the lowest 40 per cent of the population on a scale of socioeconomic
advantage (figure 2.3).

AUSTRALIA'S PUBLIC 29
HOSPITAL SECTOR
Table 2.5 Public hospital separations by patient profile, 2007-08a
Males Females
0–14 15–34 35–64 65 & over 0–14 15–34 35–64 65 & over
NSW 5.8 6.0 16.9 19.7 4.2 11.8 16.1 19.6
Vic 4.9 5.8 18.4 20.1 3.5 11.4 18.4 17.5
Qld 6.1 7.1 19.3 16.4 4.5 14.2 18.1 14.5
SA 5.6 6.0 17.6 19.5 4.1 12.1 17.1 18.1
WA 5.3 6.3 19.6 17.6 3.8 12.2 18.7 16.4
Tas 4.5 6.4 19.4 18.1 3.2 12.2 19.9 16.2
NT 5.2 6.3 26.1 7.3 3.9 12.8 31.8 6.7
ACT 4.8 7.1 20.4 19.8 3.2 11.4 16.8 16.5
Australia 5.5 6.2 18.3 18.8 4.0 12.2 17.8 17.3
a Acute and psychiatric hospitals. Per cent of total separations in each state or territory, according to patient’s
sex and age group. Each row sums to 100 per cent.
Source: AIHW (2009a).

Figure 2.3 Share of public hospital separations by socioeconomic status

of patients, 2007-08a

Most advantaged
quintile
Most disadvantaged 13%
quintile
26%

Second most
advantaged quintile
17%

Second most
disadvantaged quintile Middle quintile
23% 21%

a Quintile of socioeconomic status based on ABS (2008f) Index of Relative Socioeconomic

Advantage/Disadvantage based on the patient’s area of usual residence.
Source: AIHW (2009a).

Socioeconomically disadvantaged groups experience more ill health and have a risk
factor profile consistent with their poorer health status (Turrell et al. 2006). As the
public hospital sector treats a disproportionately large share of people of low
socioeconomic status, patients treated in public hospitals are likely to be
characterised by a relatively poor pre-existing health status and relatively more
health-related risk factors. These factors affect not only the type of treatment
30 PUBLIC AND PRIVATE
HOSPITALS
sought, but also the resources required to undertake any given procedure. For
example, patients with comorbidities or chronic conditions may be more susceptible
to infection or unplanned readmission.

2.3 Services provided by public hospitals

Admitted patient services

The main type of service provided to patients admitted to public hospitals is acute
care, which constitutes over 90 per cent of public hospital separations (table 2.6).
The next most common type of service is newborn care, followed by rehabilitation.
This distribution is generally consistent across the states and territories.

Table 2.6 Number of public hospital separations by type of care, 2007-08a

Acute Newborn Rehabilitation Palliative Geriatricb Maintenance
care care care c
NSW 1 409 636 77 326 25 954 8 273 2 813 6 065
Vic 1 305 676 55 476 13 400 5 128 13 033 870
Qld 794 041 44 600 16 853 4 266 1 037 5 778
SA 353 543 15 014 6 884 1388 460 2 341
WA 441 410 22 023 8 496 1 392 1 273 2 211
Tas 93 173 3 933 1 141 268 53 589
NT 88 197 3 341 469 311 77 404
ACT 75 465 3 955 2 249 572 561 1 283
Australia 4 561 165 225 668 75 446 21 598 19 307 19 211
a Acute and psychiatric hospitals. Excludes other and not reported care types. b Includes geriatric evaluation
and management, and psychogeriatric care. c Maintenance care refers to the provision of accommodation
and nursing care as a service itself. This can include respite care, care to patients awaiting placement, and
care to inpatients designated as nursing home type, but excludes residential aged care. – Nil.
Source: AIHW (2009a).

More specifically, the type of services provided to admitted patients can be

categorised according to patients’ diagnoses. Box 2.2 explains the common system
by which patients’ diagnoses are classified in Australia.

AUSTRALIA'S PUBLIC 31
HOSPITAL SECTOR
 Box 2.2 Australian Refined Diagnosis-Related Group (AR-DRG)
The Australian Refined Diagnosis-Related Group (AR-DRG) system categorises
separations according to the patient’s condition and the hospital resources expected to
be used. The system provides a way to record the number and type of separations
administered by a hospital in relation to the resources required.
Version 5.1 of the classification system defines 665 individual AR-DRGs. Each
separation is assigned to an AR-DRG mainly on the basis of the medical diagnosis or
surgical procedure involved, but also according to a patient’s age, length of stay, mode
of separation, the level of clinical complexity and the existence of complicating
diagnoses or procedures.
Individual AR-DRGs are grouped under 23 Major Diagnostic Categories (MDCs) which
are mostly defined by body system or disease type.
Within each MDC, individual AR-DRGs are assigned to a ‘surgical’, ‘medical’ or ‘other’
partition on the basis of the type of treatment involved. A separation is classified as
surgical if it includes an operating-room procedure, medical if it does not include any
type of procedure, and other if it includes a procedure performed outside of an
operating room (such as dental extractions and colonoscopies). In this context, a
procedure is defined as a clinical intervention that carries a procedural or anaesthetic
risk, and/or requires specialised training, facilities or equipment available only in an
acute-care setting.
Source: AIHW (2009a); DOHA (2004).

Among admitted patients, the most frequent type of diagnosis handled by

Australia’s public hospitals is a disease and disorder of the kidney or urinary tract
(figure 2.4). This type of diagnosis comprises 21 per cent of all separations
nationally, although the Northern Territory reports a considerably higher rate of
47 per cent of separations in this jurisdiction. The next most frequent types of
diagnoses treated by Australia’s public hospitals are diseases and disorders relating
to the digestive, circulatory and musculoskeletal systems and separations relating to
pregnancy and childbirth.

More broadly defined, medical cases comprise the majority (74 per cent) of
separations handled by public hospitals nationally (figure 2.5). Surgical procedures
comprise 20 per cent, while non-operating room procedures (classified as ‘other’)
comprise the remaining 6 per cent. This pattern of distribution is generally
consistent across the states and territories, with the notable exception of the
Northern Territory which handles relatively more medical and fewer surgical cases
than the national average.

32 PUBLIC AND PRIVATE

HOSPITALS
Figure 2.4 Share of public hospital separations by MDC, 2007-08a

100
Per cent of separations

80

60

40

20

0
NSW Vic Qld SA WA Tas NT ACT

Diseases and disorders of the kidney and urinary tract

Diseases and disorders of the digestive system
Diseases and disorders of the circulatory system
Pregnancy, childbirth and puerperium
Diseases and disorders of the musculoskeletal system and connective tissue
All other MDCs

a Acute and psychiatric hospitals. Per cent of total separations in each state or territory according to Major
Diagnostic Category (MDC) as defined in AR-DRG version 5.1 (box 2.2).
Source: AIHW (2009a).

Figure 2.5 Share of public hospital separations by AR-DRG partition,

2007-08a

100
Per cent of separations

80

60

40

20

0
NSW Vic Qld SA WA Tas NT ACT

Medical Surgical Other

a Acute and psychiatric hospitals. Per cent of total separations in each state or territory as defined in AR-DRG
version 5.1 (box 2.2).
Source: AIHW (2009a).

AUSTRALIA'S PUBLIC 33
HOSPITAL SECTOR
The predominance of medical cases — and the specific frequency of kidney and
urinary tract diseases and disorders — can be largely attributed to the high number
of same-day renal dialysis admissions handled by the public hospital sector. This
type of admission is the most frequent type of separation treated by public hospitals,
constituting one-third of same-day separations and 18 per cent of separations in
total (figures 2.6 and 2.7).5 The predominance of medical cases may also be
explained by the high number of obstetric separations handled by the public hospital
sector, which are also largely classified as medical cases.

Figure 2.6 Most frequent same-day public hospital separations by

AR-DRG, 2007-08a

Red blood cell disorders (w/o cscc)

Other skin, subcutaneous tissue

and breast procedures

Other gastroscopy for non-major

digestive disease

Chest pain

Other factors influencing health status

Antenatal and other obstetric admission

Lens procedure

Colonscopy (not complex)

Chemotherapy

Renal dialysis

0 100 200 300 400 500 600 700 800 900

Number of separations ('000)

a Acute and psychiatric hospitals. Ten most frequent same-day separations, as defined in AR-DRG version
5.1 (box 2.2). w/o: without. cc: complications and comorbidities. cs: catastrophic or severe.
Source: AIHW (2009a).

5 The Australian Private Hospitals Association (APHA sub. DR65) suggested that the high
number of same-day renal dialysis cases handled in the public sector may be due to some health
insurance funds capping the benefits payable for renal dialysis. It is also acknowledged that a
renal dialysis patient typically undergoes multiple same-day admissions as part of their ongoing
treatment (NSW Department of Health, sub. 40).
34 PUBLIC AND PRIVATE
HOSPITALS
Figure 2.7 Most frequent overnight public hospital separations by
AR-DRG, 2007-08a

Chronic obstructive airways disease

(w/o cscc)

Abdominal pain or mesenteric adenitis

(w/o cc)

Chronic obstructive airways disease

(with cscc)

Vaginal delivery
(single uncomplicated with other condition)

Antenatal and other obstetric admission

Cellulitis
(age>59 w/o cscc or age<60)

Caesarean delivery (w/o cscc)

Oseophagitis, gastroent and misc. digestive

system disorders (w/o cscc)

Chest pain

Vaginal delivery (w/o cscc)

0 20 40 60 80 100 120
Number of separations ('000)

a Acute and psychiatric hospitals. Ten most frequent overnight separations, as defined in AR-DRG version 5.1
(box 2.2). w/o: without. cc: complications and comorbidities. cs: catastrophic or severe. misc.: miscellaneous.
Source: AIHW (2009a).

Another feature of Australia’s public hospitals is their range of specialist units. The
most common specialist unit in acute public hospitals is domiciliary care (assisting
people with reduced ability to care for themselves in their own homes), followed by
obstetrics and maternity facilities, and nursing home care (table 2.7). At the state or
territory level, domiciliary care constitutes the most common specialist service in
New South Wales, Victoria, Western Australia and South Australia, whereas
obstetrics and maternity facilities constitute the most common specialist service in
Queensland, Tasmania, the Northern Territory and the ACT. New South Wales
provides a relatively large share of less common specialist services, having 75 of

AUSTRALIA'S PUBLIC 35
HOSPITAL SECTOR
Australia’s 108 alcohol and drug units, 46 of Australia’s 109 coronary care units,
and 36 of Australia’s 75 level-three intensive care units (AIHW 2009a).

Table 2.7 Number of specialist service units in public hospitals, 2007-08a

Dom Obs Nurs Ren Ger Rehab Psych Hspice Paed Oncon

NSW 158 78 73 56 71 58 45 45 46 43
Vic 96 58 77 59 35 32 35 24 30 36
Qld 39 40 13 23 10 18 18 9 18 11
SA 47 31 43 14 12 9 9 16 8 8
WA 58 32 39 12 22 19 18 29 9 10
Tas – 2 – 2 3 3 3 1 3 3
NT 1 5 – 4 – 2 3 1 2 –
ACT – 2 – 1 2 2 2 1 2 2
Australia 399 248 245 171 155 143 133 126 118 113
a Ten most common specialist service units in acute public hospitals in Australia. Dom: Domiciliary care
service. Obs: Obstetrics and maternity facility. Nurs: Nursing home care unit. Ren: Maintenance renal dialysis
centre. Ger: Geriatric assessment unit. Rehab: Rehabilitation unit. Psych: Psychiatric ward or unit.
Hspice: Hospice care unit. Paed: Specialist paediatric unit. Oncon: Oncology unit. – Nil.
Source: AIHW (2009a).

Non-admitted patient services

In addition to treating admitted patients, public hospitals provide a considerable

volume of services to non-admitted patients. The AIHW (2009a) reported that
public hospitals administered over 48 million occasions of service to non-admitted
patients in 2007-08, including 7.1 million accident and emergency presentations
(table 2.8). The provision of emergency services fulfils the responsibility assigned
to public hospitals under the NHA and former AHCAs.

Reported data on non-admitted patient services, however, are subject to wide

variation in methods of collection and classification among the states and territories.
For example, data reported by New South Wales under ‘other individual patient care’
include a large volume of pharmacy and district nursing services administered by
Justice Health that would not typically occur in other hospitals. As the AIHW (2009a)
cautioned, such variations preclude direct comparisons from being made.

36 PUBLIC AND PRIVATE

HOSPITALS
Table 2.8 Services to non-admitted patients in public hospitals, 2007-08a
Accident and Outpatient care Other individual Group sessions
emergency patient care b
New South Wales 2 417 721 6 400 364 12 414 382 242 174
Victoria 1 522 573 2 864 208 3 115 414 23 016
Queensland 1 471 377 3 324 742 5 867 454 17 853
South Australia 544 439 1 203 133 456 785 80 296
Western Australia 778 119 1 697 777 2 287 313 63 456
Tasmania 142 633 459 539 399 480 ..
Northern Territory 125 315 296 259 194 087 300
ACT 98 441 122 694 150 878 1 756
Australia 7 100 618 16 368 716 24 885 793 428 851
a Reported number of occasions of service to non-admitted patients in acute public hospitals. Due to
widespread differences in the collection and coverage of non-admitted patient data among the states and
territories, the AIHW advises that the reported data should be interpreted and compared with caution. b Other
types of care include: mental health, alcohol and drug, pharmacy, community health, district nursing,
pathology, radiology and organ imaging and other outreach services. There are large variations in the types of
care reported in this category by each of the states and territories. .. Not applicable.
Source: AIHW (2009a).

As another broader point of difference in data reporting, a patient receiving a given

type of treatment may be classified as an admitted patient in one jurisdiction, but as
an non-admitted outpatient in another jurisdiction. For example, most chemotherapy
patients are classified as non-admitted outpatients in New South Wales, South
Australia and the ACT, but classified as admitted patients in other jurisdictions
(AIHW 2009a). This point of difference must be considered when evaluating not
only the volume of services reported for non-admitted patients, but also the volume
of separations reported for admitted patients.

Data on hospital services delivered to non-admitted patients are further complicated

by differences among the states and territories in the delivery of services to
non-admitted patients outside of hospital settings. For example, outpatient hospital
services delivered in community health centres by some states and territories are
effectively excluded from reported hospital statistics (AIHW 2009a).

Research and training

Medical and health services research has traditionally been undertaken in public
hospitals, and research remains an important function of the public sector. Likewise,
under the NHA and former AHCAs, it is the responsibility of state and territory
governments to provide clinical training for undergraduate students and
specialists-in-training (COAG 2008d). DOHA commented on the value of this
training role:

AUSTRALIA'S PUBLIC 37
HOSPITAL SECTOR
 Public hospitals play a vital role in all health professional training programs by
providing clinical placements and supervision. For medical education, on completion of
university undergraduate or graduate education programs, graduates enter
pre-vocational training [for one year] at a major public teaching hospital to become
registered to practice. Most registered doctors then also complete another one to two
years pre-vocational training, gaining experience in different clinical departments and
in different hospital settings such as in rural hospitals. Most doctors then enter a four to
six year vocational training toward becoming independent practitioners accredited by
specialty colleges. For most specialties (other than general practice), this vocational
training takes place largely in public hospital settings. (sub. 32, p. 8)

Australia’s public hospital system includes 69 public teaching hospitals (excluding

psychiatric hospitals). The majority of public teaching hospitals are located in
Queensland and New South Wales, which have 22 and 20 respectively
(AIHW 2009a).

The teaching role of hospitals has implications for their costs and resource
allocation, as noted by study participants:
The cost structures in the large teaching hospitals in particular reflect the teaching and
research components of the work that is done. Both these activities are inseparable
from the provision of care. A doctor can be treating his patients, teaching his registrars
and gathering material for research all at the same time, using all the same facilities and
drawing on the same support staff and services … It is universally understood and
accepted that surgery will take longer if doctors in training are being taught during it.
(Australian Medical Association, sub. 28, p. 1)

This point was also made by Access Economics, in their submission prepared for
Medibank Private, Australian Unity, Bupa Australia and Ramsay Health Care:

Many public hospitals have a research and teaching role, which adds costs to the public
sector …. Significant numbers of staff specialists in a teaching hospital devote a
substantial proportion of their time to teaching, research, clinical management and
service to specialist societies and professional colleges. These costs are … not
distinguished from the general budget of public hospitals. However, research and
teaching do provide additional benefits to a hospital by granting access to students,
collaborative research facilities, and personnel to undertake some tasks (at a small
internalised cost) (sub. DR60, p. 12)

Although clinical training is increasingly being undertaken in the private hospital

sector, the public hospital sector retains primary responsibility for this role (DOHA,
sub. 32; Queensland Health, sub. 27; SA Department of Health, sub. 4). State and
territory governments continue to direct the majority of funding for clinical training
to public hospitals, although some states fund clinical training places in other
healthcare settings.

38 PUBLIC AND PRIVATE

HOSPITALS
2.4 Workforce characteristics
Nurses represent the largest share (45 per cent) of the public hospital sector
workforce nationally (table 2.9). Diagnostic and allied health professionals, and
administrative and clerical staff, each represent approximately 15 per cent of the
workforce, medical officers represent 11 per cent, and domestic and other staff
constitute the remaining 13 per cent.

Table 2.9 Number of staff in public hospitals, by occupation, 2007-08a

Salaried Nurses Diagnostic Domestic Administrative Other
medical and allied and other and clerical personal
officers health staff staff care staff b
professionals

NSW 8 353 36 726 12 470 9 627 11 099 na

Vic 6 783 27 024 12 412 6 578 10 802 ..
Qld 5 622 19 219 4 860 7 462 6 177 923
SA 2 190 9 152 2 043 1 848 3 071 787
WA 2 667 9 593 2 906 4 194 4 053 na
Tas 512 2 222 527 995 660 ..
NT 342 1 212 321 558 432 15
ACT 526 1 945 474 173 614 181
Australia 26 996 107 089 36 013 31 434 36 909 na
a Number of full-time equivalent staff in acute and psychiatric hospitals. Data exclude one hospital in Victoria,
two hospitals in Tasmania, and pathology staff in Queensland. b For New South Wales and Victoria, Other
personal care staff are classified as Diagnostic and allied health professions or Domestic and other staff. .. Not
applicable. na Not available.
Source: AIHW (2009a).

Some variations from the national averages are apparent. For example, medical
officers constitute a relatively larger share of the public hospital workforce in the
ACT, and a relatively smaller share in New South Wales, Victoria and Tasmania.
Nurses constitute a relatively larger share of the public hospital workforce in New
South Wales, South Australia and the ACT, and a relatively small share in Victoria,
Western Australia and the Northern Territory. Victoria is characterised by a
relatively high share of diagnostic and allied health professionals.

Some notable differentials in the average salaries of public hospital staff are also
evident between the states and territories. Although the average annual salaries of
nursing staff are generally similar among the states and territories, the Northern
Territory stands out for offering almost $15 000 more than the national average
(table 2.10). The average salaries of medical officers show relatively more
variation. Comparatively high annual salaries are offered to medical officers in the
Northern Territory and Western Australia (up to $27 000 more than the national

AUSTRALIA'S PUBLIC 39
HOSPITAL SECTOR
average), while medical officers receive the lowest annual average salaries in
New South Wales, South Australia and the ACT.6

Table 2.10 Average salaries of staff in public hospitals, 2007-08a

Salaried Nurses Diagnostic Domestic Administrative Other
medical and allied and other and clerical personal
officers health staff staff care staff b
professionals

NSW 137 766 73 702 64 367 51 188 59 834 na

Vic 152 284 75 503 53 505 62 953 44 809 na
Qld 159 069 72 044 67 317 46 252 51 662 50 635
SA 141 196 72 152 72 019 35 312 52 568 41 722
WA 178 762 77 422 71 333 53 021 56 719 na
Tas 158 685 71 293 64 432 41 612 49 529 na
NT 181 065 89 656 76 490 53 233 58 693 67 279
ACT 142 171 76 261 74 449 45 284 58 995 49 758
Australia 151 211 74 237 62 259 51 491 52 910 47 020
a Average salaries (in dollars) of full-time equivalent staff in acute and psychiatric hospitals. b For New South
Wales and Victoria, Other personal care staff are classified as Diagnostic and allied health professions or
Domestic and other staff. na Not available.
Source: AIHW (2009a).

2.5 Recent developments in public hospitals

From 2003-04 to 2007-08:
• The number of beds available in acute public hospitals increased, but only
matched population growth. The system has maintained a ratio of around
2.6 beds per 1000 residents.
• The number of separations has increased by around 13 per cent, equivalent to
10 more separations per 1000 residents. The relative shares of same-day and
overnight separations have stayed roughly constant.
• An increase in the annual number of separations per capita, alongside the fact
that the number of beds per capita has remained stable, suggests that the average
length of a separation per bed has shortened. This is supported by data showing
that the average length of stay and the relative stay index have fallen slightly
over these four years.

6 Salary comparisons in the other staff categories are not computed due to differences in data
reporting.

40 PUBLIC AND PRIVATE

HOSPITALS
• The waiting time for elective surgery has lengthened, although proportionally
fewer patients are waiting more than one year.
• The number of occasions of service delivered to non-admitted patients has
increased by 4.7 million, equivalent to more than 10 per cent over this four-year
period (table 2.11).

Table 2.11 Changes in acute public hospitals, 2003-04 to 2007-08a

2003-04 2004-05 2005-06 2006-07 2007-08

Number of hospitals 741 739 736 739 742

Number of beds available 51 038 52 806 52 236 53 563 53 137
Number of beds available per 1000 residents 2.55 2.61 2.54 2.57 2.55
Number of separations (‘000) 4 183 4 261 4 451 4 646 4 729
Number of separations per 1000 residents 206.9 207.3 212.8 218.0 216.9
Per cent of same-day separations 49.1 49.2 49.7 50.2 49.9
Average length of stay for overnight separations 6.4 6.4 6.3 6.2 6.2
(days)
Relative stay indexb 1.02 0.99 0.99 0.98 0.98

Waiting times for elective surgeryc

Days waited at 50th percentile 28 29 32 32 34
Days waited at 90th percentile 193 217 237 226 235
Per cent patients waiting more than 365 days 3.9 4.8 4.9 3.1 3.0
Number of non-admitted occasions of service 43 660 42 759 44 750 46 141 48 355
(‘000)
a Data refer to acute hospitals only, whereas data reported in previous tables refer to acute and psychiatric
hospitals. b Directly standardised relative stay index. A value greater than 1 indicates that an average
patient’s length of stay is greater than expected for their type of separation, while a value less than 1 indicates
that the length of stay is shorter than expected. c Data for reporting hospitals only, and include some private
hospitals that are contracted to provide elective surgery to public patients.
Source: AIHW (2009a).

Growth in public hospital activity over this four-year period is mainly attributed to
growth in medical cases, rather than surgical or other types of treatment
(AIHW 2009a). Same-day admissions for renal dialysis (classified as medical) is
the fastest growing type of separation in Australia’s public hospital sector,
increasing by over 30 per cent in the four-year period up to 2007-08 (figure 2.8).
The next fastest growing types of separations in public hospitals are admissions for
chest pain, childbirth and other obstetrics, and digestive disorders. The ten most
rapidly growing types of separations in the public hospital sector were all among the
ten most frequent types of overnight or same-day separations in public hospitals in
2007-08 (as illustrated earlier in figures 2.6 and 2.7).

AUSTRALIA'S PUBLIC 41
HOSPITAL SECTOR
Figure 2.8 Ten fastest increasing public hospital separations by AR-DRG,
2003-04 to 2007-08a

Complex gastroscopy (same-day)

Total separations 2003-04
Cellulitis (age>59 w/o cscc or age<60)
(overnight) Increase in separations from
2003-04 to 2007-08
Caesarean delivery (w/o cscc)
(overnight)

Lens procedures (same-day)

Antental and other obstetric admission

(same-day)

Other factors influencing health status

(same-day)

Oesophagitis, gastroent and misc.

digestive system disorders

Vaginal delivery (w/o cscc)(overnight)

Chest pain (overnight or same-day)

Admit for renal dialysis (same-day)

0 100 200 300 400 500 600 700 800 900

Number of separations ('000)
a Acute and psychiatric hospitals. Data for 2003-04 are defined according to AR-DRG version 5.0. Data for
2007-08 are defined according to AR-DRG version 5.1 (see box 2.2). Data classifications are subject to minor
revision between years. w/o: without. cc: complications and comorbidities. cs: catastrophic or severe.
misc.: miscellaneous.
Source: AIHW (2009a).

Emergency department services are one of the fastest growing type of services
provided by public hospitals (AIHW 2009a). In the four-year period up to 2007-08,
emergency department presentations in public hospitals grew by almost 30 per cent
(table 2.12). Despite the higher volume of activity, the proportion of presentations
treated in the clinically appropriate time and median waiting times have remained
stable (although care needs to be exercised when interpreting waiting-time data for
emergency departments, as there appears to be significant variation between
hospitals in how waiting times are measured and in the assignment of clinical
urgency categories).

42 PUBLIC AND PRIVATE

HOSPITALS
Table 2.12 Public hospital emergency department activity, 2003-04 to
2007-08a
2003-04 2004-05 2005-06 2006-07 2007-08

Number of emergency department 4 308 4 529 4 915 5 287 5 537

presentations (‘000)
Per cent of presentations seen on timeb
Resuscitation 99 100 99 99 100
Emergency 76 76 77 78 76
Urgent 63 64 64 65 63
Semi-urgent 65 65 65 66 66
Non-urgent 87 88 87 88 87
Median waiting time (minutes)
Resuscitation – – – – –
Emergency 5 5 5 5 6
Urgent 22 21 21 20 21
Semi-urgent 38 37 37 36 36
Non-urgent 28 28 29 28 28
a Data for reporting hospitals only. Not all hospitals include an emergency department. b Time period is
specified according to the triage category. – Nil.
Source: AIHW (2009a).

The future governance, funding arrangements and delivery of services in the public
hospital sector are set for reform under COAG’s NHA and National Partnership
Agreement on Hospital and Health Workforce Reform (COAG 2008b, 2008d,
2008e). The anticipated impact of these reforms, alongside the suite of
recommendations recently proposed by the National Health and Hospital Reform
Commission (NHHRC 2009), are further considered with the future direction of the
private hospital sector in chapter 4.

AUSTRALIA'S PUBLIC 43
HOSPITAL SECTOR
3 Australia’s private hospital sector

Key points
• Private hospitals differ greatly in size, function and management. Of the 556 private
hospitals in Australia, there are large organisations operating many hospitals, as
well as smaller bodies running single or only a few facilities. A substantial number of
private hospitals are run as not-for-profit entities while others are operated on a
for-profit basis.
• Around 80 per cent of patients in private hospitals are privately insured, and the
majority of private hospital funding is received from private health insurers for
treating their members. Indeed, patient revenue (including from self-funded patients)
accounted for around 96 per cent of private acute and psychiatric hospital income in
2006-07.
• Private hospitals treated 40 per cent of all hospital inpatients and performed
64 per cent of elective surgeries in Australia in 2007-08. About 17 per cent of
separations from private hospitals in 2007-08 were for chemotherapy, renal dialysis
and same-day colonoscopies. Fewer than 10 per cent of private acute and
psychiatric hospitals had emergency departments in 2006-07.
• Rights of private practice for medical specialists are an important feature of
workforce arrangements in private hospitals, and there is evidence suggesting that
medical specialists are generally able to earn higher incomes in private hospitals
than in public hospitals. There is little publicly available data about the wages and
conditions of nursing staff in the private hospital sector.
• Private hospitals have recently experienced significant increases in the number of
separations and some changes to the composition of services provided. There also
appears to be some increase in the extent of clinical teaching by private hospitals.

Private hospitals are privately owned and operated institutions, catering for patients
who are treated by a doctor of their own choice.1 Patients are charged fees for
accommodation and other services provided by private hospitals and relevant
medical and paramedical practitioners (AIHW 2009a). Private hospitals exist in
response to patients’ willingness-to-pay for a choice of doctor, private ward

1 Some hospitals which deliver public hospital services are privately owned. Such hospitals are
classified as public as they operate on behalf of, and are funded by, a government.
AUSTRALIA’S 45
PRIVATE HOSPITAL
SECTOR
facilities and relatively faster access to hospital services. The Australian Private
Hospitals Association (APHA, sub. 25, p. 2) commented that the private hospital
sector exists in ‘explicit recognition that individuals should be able to exercise
choice in health care’.

Recent data show that there are 556 private hospitals in Australia, of which 285 are
acute or psychiatric hospitals and 271 are freestanding day hospitals
(DOHA 2009c). Acute hospitals provide at least some medical, surgical or obstetric
care for admitted patients and provide round-the-clock comprehensive qualified
nursing services, as well as other necessary professional services. Freestanding day
hospital facilities provide investigation and treatment for acute conditions on a
day-only basis (ABS 2008e).

This chapter profiles the structure and activity of Australia’s private hospitals,
including the types of services delivered, the characteristics of the patients treated,
and the workforce. Recent developments in the public hospital sector are also
examined. While focus is placed on the activity of acute and psychiatric hospitals,
private freestanding day hospitals are separately profiled, given their role in the
wider private hospital system.

3.1 Structure of private hospitals

Ownership and management

Of the 289 private acute and psychiatric hospitals in Australia in 2006-07, 165 were
run on a for-profit basis and 124 were not-for-profit (table 3.1). Not-for-profit
hospitals are those which qualify as a non-profit organisation with either the
Australian Taxation Office or the Australian Securities and Investments
Commission. These are further categorised as ‘religious or charitable’ and ‘other’
(ABS 2008e).

Both for-profit and not-for-profit entities are among the largest providers of private
hospital services in Australia. The for-profit companies Ramsay Health Care and
Healthscope are among the ten largest enterprises — by market capitalisation — in
the Australian Securities Exchange’s listed healthcare sector (ASX 2009). Ramsay
Health Care operates over 65 hospitals and day surgery units across Australia, while
Healthscope owns or manages 44 medical and surgical, rehabilitation and
psychiatric hospitals (Ramsay Health Care Limited 2009a; Healthscope Limited,
sub. 42). In the not-for-profit sector, Catholic services represent the largest grouping
of health, community and aged care services in Australia, providing 9500 beds in 75

46 PUBLIC AND PRIVATE

HOSPITALS
(private and public) healthcare facilities including seven teaching hospitals (CHA
2009b). In relation to the services provided by Catholic hospitals, Catholic Health
Australia (CHA) noted that:
Catholic hospitals also have a mission focus which is often reflected in providing a
wider range of treatments, such as palliative care, than might be the case than if the
hospital was purely focused on profit maximisation. It also means that some Catholic
hospitals are located in geographic regions which might not necessarily be attractive to
for-profit operators. (sub. 20, p. 2)

Table 3.1 Number of private acute and psychiatric hospitals, 2006-07

For-profit Not-for-profit
All
Religious or charitable Othera
New South Wales np 18 np 85
Victoria 50 18 14 82
Queensland 27 24 6 57
South Australia 6 7 17 30
Western Australia np 10 np 23
Tas, NT and ACTb np 6 np 12
Australia 165 83 41 289
a Comprises bush nursing, community and memorial hospitals. b Data for Tasmania, the Northern Territory
and the ACT are aggregated to protect the confidentiality of the small number of hospitals in each of these
jurisdictions. np Not published but included in totals where applicable, unless otherwise indicated.
Source: ABS (Private Hospitals, Cat. no. 4390.0).

Funding arrangements

Private hospitals source their revenue largely from use of their operating theatres
and bed facilities, and a number provide a broad range of services such as
diagnostics, chemotherapy and sub-acute care that also generate revenue from
patients. Private hospitals generally order and pay for prostheses and then recoup
the cost from health insurance funds and, in some instances, patients. Medical fees
are usually billed separately and direct to the patient from the medical provider, as
opposed to being directed via hospital accounts.

Private hospitals operate under fee-for-service funding models that reward

additional activity. Therefore, private hospitals generally have an incentive to
maximise throughput. Private hospitals, particularly the for-profit sector, seek to
maximise returns on their capital investment and labour force, for the benefit of
owners/shareholders. While the revenue-generation motive is likely to be less strong
for not-for-profit private hospitals, these hospitals also aim to avoid making losses.

AUSTRALIA’S 47
PRIVATE HOSPITAL
SECTOR
Across most private hospital structures, patient revenue is the dominant source of
hospital income. In 2006-07, this ranged from 95 per cent of hospital income for
religious or charitable private acute and psychiatric hospitals to 98 per cent for
for-profit providers (table 3.2). DOHA advised that:
Revenue for private hospitals and day hospital facilities can come from a number of
sources (e.g. Department of Veterans’ Affairs, state/territory health authorities’
contracts, self-funding by patients and compensable patients), but the majority of
funding is received from private health insurers for treating their members. It is
therefore in the interest of facilities to negotiate comprehensive contracts with
individual insurers. (sub. 32, p. 6)

Table 3.2 Income and expenditure of private hospitals, 2006-07

For-profit Not-for-profit

Religious or charitable Otherb

$’000 $’000 $’000

Income
Patient revenuec 3 543 450 3 012 142 254 890
Recoveriesd 53 238 74 009 6 650
Othere 39 336 92 335 6 427
Total income 3 636 025 3 178 486 267 967
Recurrent expendituref
Wages and salaries including on-costs 1 700 724 1 556 302 144 770
Drug, medical and surgical suppliesg 873 026 810 497 68 778
Food supplies 45 478 47 084 4 951
Other domestic services 45 630 49 801 4 179
Administrative expenses 208 042 239 737 17 022
Repairs and maintenance 46 237 39 177 3 537
Otherh 297 164 353 347 26 230
Total recurrent expenditure 3 216 301 3 095 945 269 464
Gross capital expenditurei 207 984 220 453 11 282
a Acute and psychiatric hospitals (excludes freestanding day hospitals). b Comprises bush nursing,
community and memorial hospitals. c Includes revenue received by, and due to, the hospital in respect of
patient liability for accommodation and other fees. d Recoveries includes income received from items such as
staff meals and accommodation, and facility fees paid by medical practitioners. e Includes investment income,
income from charities, bequests and visitors’ expenses. f Expenditure on goods and services which does not
result in the creation or acquisition of fixed assets. g Includes surgically implanted prostheses and homograft
items. h Includes interest, depreciation, contract services and transport. i Expenditure on the acquisition or
enhancement of assets (excluding financial assets).
Source: ABS (Private Hospitals, Cat. no. 4390.0).

Wages and salaries, and drug, medical and surgical supplies are the biggest
recurrent expenditure items for private acute and psychiatric hospitals. Together
these represented 78 per cent of total recurrent expenditure in 2006-07
(ABS 2008e). Wages and salaries (including on-costs) constituted around or just

48 PUBLIC AND PRIVATE

HOSPITALS
above 50 per cent of total recurrent expenditure for both for-profit and
not-for-profit private hospitals (ABS 2008e). These expenditure figures do not
include the costs of non-salaried medical staff (such as independent practitioners).

Capital expenditure varies year-to-year in the private health sector due to the
irregular nature of such expenditure. Significant purchases or construction
undertaken in a given year are unlikely to be repeated for some time (ABS 2008e).

Service costs

The cost structures for services in the private hospital sector are very different than
those in the public sector. The SA Department of Health commented that:
Medical services in private hospitals are provided on a fee-for-service basis rather than
by the hospital. One result is that it is in the doctors’ best interests to ensure as many
theatre cases as possible are done in each set of booked theatre time. (sub. 4, p. 4)

Medical costs for the private sector are difficult to ascertain, as doctors usually bill
patients directly. Private hospitals are not made aware of, and so do not record,
these costs (APHA, sub. DR65). The Australian Health Service Alliance noted that:
Doctor costs in the private sector are in general a matter between the patient and the
doctors involved in the care. This applies to doctors involved in such care whether they
are the primary treating physician or surgeon, or other medical practitioners involved in
care such as anaesthetists, pathologists and radiologists … Prostheses costs have a
different basis in the public and private sector. In the public sector they are included in
hospital funding. In the private sector they are in effect negotiated separately at the
industry level and the hospital is simply the conduit by which prostheses are supplied to
patients by their treating doctor. (sub. 1, pp. 4–5)

Tax regimes differ between for-profit and not-for-profit hospitals. Fringe benefits of
up to $17 000 per employee are exempt from fringe-benefits tax for not-for-profit
hospitals (and public hospitals). Not-for-profit private hospitals (and public
hospitals) are also exempt from payroll tax. The tax arrangements for private and
public hospitals are discussed further in chapter 5 and appendix D.

The average cost per patient day tends to increase as hospital size increases, which
the ABS noted ‘is a reflection of the greater complexity of procedures undertaken at
the larger hospitals’ (ABS 2008e, p. 18).2 More complex procedures necessitate
greater use of highly trained staff, expensive equipment, drugs and medical
supplies. It is also noted that religious and charitable hospitals have relatively higher

2 Patient days are the aggregate number of days of stay for all overnight-stay patients who were
separated from hospital during the year. Same-day patients are each counted as having a stay of
one day (ABS 2008e).
AUSTRALIA’S 49
PRIVATE HOSPITAL
SECTOR
average costs per patient day than for-profit and other not-for-profit hospitals.
(ABS 2008e). The fact that religious and charitable hospitals constitute over half of
the largest sized private acute and psychiatric hospitals (more than 200 beds) —
while comprising less than 30 per cent of all private acute and psychiatric hospitals
— may help to explain the higher average costs of larger-sized hospitals.

Licensing

State and territory health authorities are responsible for licensing private hospitals
and private day hospitals, and mandate a range of operational and quality
requirements. Licensing requires these facilities to meet a range of criteria, such as
building regulations, provision of speciality services, as well as safety and quality.
Licensing requirements vary from one jurisdiction to another and, in some
jurisdictions, differ for private hospitals and day hospital facilities.

New South Wales currently has separate regulations for private hospitals and
private day hospitals. The regulations for private hospitals are somewhat more
prescriptive around furnishings, staffing and quality assurance processes than for
private day hospitals. New South Wales’ proposed Private Health Facilities
Regulation 2009 will remove the distinction between private hospitals and private
day hospitals and impose compliance burdens based on services offered, rather than
facility type, size or location.

In Victoria, private hospitals and private day hospitals are subject to the same
regulatory requirements (Health Services (Private Hospitals and Day Procedure
Centres) Regulations 2002). These regulations contain minimum nursing
staff-to-patient ratios and the mix of nursing staff. These staffing requirements are
more specific than for other jurisdictions.

Among the jurisdictions, South Australia, Tasmania and the Northern Territory do
not have specific licensing criteria for day hospital facilities but inspect new
facilities and provide assurances that the facilities are suitable for Australian
Government declaration as private hospitals. The Australian Department of Health
and Ageing (DOHA) noted that the Commonwealth Minister for Health and Ageing
has the power to declare private hospitals for health insurance purposes, Medicare
benefits and the Pharmaceutical Benefits Scheme (DOHA, sub. 32).

50 PUBLIC AND PRIVATE

HOSPITALS
3.2 Characteristics of private hospitals
Private hospitals exhibit great diversity in the choice of where, how and what
services they offer, and may specialise in particular procedures or types of
patients. For example, some private hospitals are women’s or children’s hospitals,
or have units or departments specialising in women’s and/or children’s health.
Examples include the Mater Children’s Private Hospital and the Mater Mothers’
Private Hospital (Queensland), Woodvale Private Hospital for Women (Western
Australia), and the Allowah Presbyterian Children’s Hospital (New South Wales).

Number and activity of private hospitals

The majority of private acute and psychiatric hospitals are located in New South
Wales and Victoria (table 3.3). New South Wales had a lower number of private
acute and psychiatric hospital patient separations per 1000 residents than the
national average in 2007-08, while Queensland’s number was comparatively higher.
On average, 57 per cent of separations in private acute and psychiatric hospitals
were same-day admissions.

Table 3.3 Number and activity of private hospitals, 2007-08a

Number of Number of Number of Proportion
hospitals separations separations per same-day
b
1000 residents separations
New South Wales 84 662 743 90.8 59.7
Victoria 75 663 465 114.9 55.6
Queensland 55 596 730 138.8 56.9
South Australia 31 196 865 112.0 53.1
Western Australia 24 259 807 120.5 55.9
Tas, NT and ACTc 11 np np np

Australia 280 2 461 852 115.1 56.8

a Acute and psychiatric hospitals (excludes freestanding day hospitals).

Source: AIHW (2009a).

Size of private hospitals

A significant proportion of private acute and psychiatric hospitals are relatively

small: 43 per cent have fewer than 50 beds, while less than 10 per cent have more
than 200 beds (table 3.4).

AUSTRALIA’S 51
PRIVATE HOSPITAL
SECTOR
Table 3.4 Number of private hospitals by size, 2006-07a
0–50 bedsb 51–100 beds 101–200 beds Over 200 beds Total
New South Wales 35 32 15 3 85
Victoria 37 24 16 5 82
Queensland 22 16 12 7 57
South Australia 8 7 np np 30
Western Australia 10 3 6 4 23
Tas, NT and ACTc 3 5 np np 12
Australia 125 87 55 22 289
a Acute and psychiatric hospitals (excludes freestanding day hospitals). Data refers to different time period to
previous table due to differences in data availability. b Number refers to hospitals with 26 to 50 beds for South
Australia, Tasmania, the Northern Territory and the ACT. The number of hospitals with fewer than 26 beds is
not published for these states and territories. c Data for Tasmania, the Northern Territory and the ACT are
aggregated to protect the confidentiality of the small number of hospitals in each of these jurisdictions np Not
published but included in totals where applicable.
Source: ABS (Private Hospitals, Cat. no. 4390.0).

In terms of the number of beds, acute and psychiatric hospitals operated by religious
and charitable institutions were generally larger than other private hospitals. They
accounted for around 20 per cent of the smallest private hospitals (up to 50 beds),
but for 40 per cent of hospitals with 101 to 200 beds, and around 60 per cent of
those with more than 200 beds (ABS 2008e).

Location of private hospitals

About 75 per cent of all available private acute and psychiatric hospital beds in
2006-07 were located in capital cities, even though only 64 per cent of Australia’s
population lived in these areas (ABS 2008e). There are more private acute and
psychiatric hospitals in metropolitan than regional areas nationally (table 3.5). The
one exception at a state level is Queensland, which has 63 per cent of its private
acute and psychiatric hospitals, and 50 per cent of its private acute and psychiatric
hospital beds, outside of Brisbane. For other jurisdictions (where data are available),
at least 70 per cent of all beds are located within capital cities. There are relatively
fewer private hospital beds outside of capital cities than there are private hospitals,
suggesting that private hospitals outside of capital cities are, on average, smaller
than their capital city counterparts.

Acute and psychiatric hospitals operated by religious and charitable institutions

provided 10 246 beds during 2006-07, equivalent to 42 per cent of the total number
of available beds in all private acute and psychiatric hospitals. Around 75 per cent
of these beds were located in capital cities while the remaining share were outside
capital cities (ABS 2008e).

52 PUBLIC AND PRIVATE

HOSPITALS
Table 3.5 Number of private hospitals and beds by location, 2006-07a
Number of hospitalsb Number of beds

Capital city Rest of Total Capital city Rest of Total

state/territory state/territory
New South Wales 57 28 85 4 573 1 823 6 396
Victoria 61 21 82 5 605 1 005 6 610
Queensland 21 36 57 2 863 2 824 5 687
South Australia 21 9 30 1 748 130 1 878
Western Australia 17 6 23 np np 2 795
Tas, NT and ACTc 8 4 12 np np 1 061
Australia 185 104 289 18 095 6 332 24 427
a Acute and psychiatric hospitals (excludes freestanding day hospitals). b These data relate to a different
reporting period (July 2006 to June 2007) than the DOHA figures cited earlier (September 2009). c Data for
Tasmania, the Northern Territory and the ACT are aggregated to protect the confidentiality of the small
number of hospitals in each of these jurisdictions. np Not published but included in totals where applicable.
Source: ABS (Private Hospitals, Cat. no. 4390.0).

Patients in private hospitals

In 2007-08, private hospitals administered 3.1 million separations, of which nearly

2.5 million took place in acute and psychiatric hospitals. Female patients accounted
for around 55 per cent of all private hospital separations, while patients aged 35 to
64 accounted for around 45 per cent of all private hospital separations. Of note is
the lower incidence of private hospital separations for females aged 65 and over in
Western Australia (table 3.6).

Table 3.6 Share of private hospital separations by patient profile,

2007-08a
Males Females

0–14 15–34 35–64 65 & over 0–14 15–34 35–64 65 & over
New South Wales 2.3 4.6 19.5 18.4 1.7 9.7 25.1 18.5
Victoria 1.7 4.4 18.7 18.7 1.3 10.5 26.6 18.1
Queensland 1.9 3.9 20.1 20.2 1.4 9.7 24.4 18.4
South Australia 2.0 4.5 19.3 20.5 1.4 7.8 24.9 19.5
Western Australia 2.6 5.4 21.4 17.0 1.9 11.1 26.5 13.9
Tas, NT and ACTb 2.3 5.1 19.9 16.8 1.7 11.5 26.6 15.7
Australia 2.1 4.5 19.7 18.9 1.5 10.0 25.5 17.9
a All private hospitals. Per cent of total separations in each state or territory, according to patient’s sex and
age group. Each row sums to 100 per cent. b Data for Tasmania, the Northern Territory and the ACT are
aggregated to protect the confidentiality of the small number of hospitals in each of these jurisdictions.
Source: AIHW (2009a).

AUSTRALIA’S 53
PRIVATE HOSPITAL
SECTOR
The majority of patients in private hospitals are funded by private health insurance
or self-funded. Nationally, the proportion of private hospital separations covered by
private hospital insurance was 80 per cent in 2007-08 (figure 3.1). For Queensland,
a larger proportion of private hospital patients are self-funded or funded by the
Department of Veterans’ Affairs than is the case for other jurisdictions.

Figure 3.1 Share of private hospital separations by patient funding source,

2007-08a

100
Per cent of separations

80

60

40

20

0
NSW Vic Qld SA WA
Private health insurance Private self-funded
Public Dept Veterans' Affairs
Compensation or other

a The share of self-funded patients may be underestimated as some are unable to be identified. Data exclude
patients whose funding source is not reported. Compensation or other includes workers compensation, other
compensation, motor vehicle third party personal claim, other public authorities, and other funding sources.
Data for Tasmania, the Northern Territory and the ACT are not published.
Source: AIHW (2009a).

These rates of private hospital usage do not appear to be fully explained by private
health insurance participation rates of the states and territories (figure 3.2). Among
the jurisdictions, Western Australia and the grouping of New South Wales and the
ACT have the highest rates of private health insurance, while the Northern Territory
and Queensland have the lowest.

54 PUBLIC AND PRIVATE

HOSPITALS
Figure 3.2 Share of population with private health insurance, 2008

60
Per cent of population

50

40

30

20

10

0
NSW Vic Qld SA WA Tas NT Australia
and ACT

Source: PHIAC (2009c).

The share of private hospital separations according to patients’ socioeconomic

status is presented in figure 3.3. In 2007-08, nearly 30 per cent of patients in private
hospitals were from areas of the highest quintile of socioeconomic advantage. The
top two quintiles of socioeconomic advantage accounted for around 50 per cent of
all private hospital separations. Comparatively, 14 per cent of patients in private
hospitals were from the lowest quintile.

Figure 3.3 Share of private hospital separations by socioeconomic status

of patients, 2007-08a
Most
disadvantaged
quintile
14% Most advantaged
quintile
28%
Second most
disadvantaged
quintile
17%

Middle quintile Second most

19% advantaged quintile
22%
a Quintileof socioeconomic status based on ABS (2008f) Index of Relative Socioeconomic
Advantage/Disadvantage score based on the patient’s area of usual residence.
Source: AIHW (2009a).

AUSTRALIA’S 55
PRIVATE HOSPITAL
SECTOR
3.3 Services provided by private hospitals
Private hospitals often specialise in a limited range of surgical procedures, although
there are also a number of full-service private hospitals that offer a comparable
range of services to those provided by the large public teaching hospitals. Private
hospitals tend to provide more elective procedures than public hospitals, accounting
for approximately 64 per cent of all elective surgery separations in Australia in
2007-08 (AIHW 2009a).

Patient services

Nationally, 41 per cent of separations from private hospitals were classified as

surgical in 2007-08. A further 38 per cent of private hospital separations were
classified as medical (figure 3.4).

Figure 3.4 Share of private hospital separations by AR-DRG partition,

2007-08a

100
Per cent of separations

80

60

40

20

0
NSW Vic Qld SA WA Australia

Medical Surgical Other

a As defined in AR-DRG version 5.1 (see box 2.2). Data for Tasmania, the Northern Territory and the ACT are
not published, but are included in totals.
Source: AIHW (2009a).

The most frequent private hospital same-day and overnight separations in 2007-08
are listed in figures 3.5 and 3.6. The ten most frequent same-day separations made
up around 37 per cent of all private hospital separations.

56 PUBLIC AND PRIVATE

HOSPITALS
Figure 3.5 Most frequent same-day private hospital separations by
AR-DRG, 2007-08a

Follow up with endoscopy

Mental health treatment without ECT

Other factors influencing health status

Complex gastroscopy

Dental extractions and restorations

Other gastroscopy for non-major

digestive disease

Lens procedures

Admit for renal dialysis

Other colonoscopy

Chemotherapy

0 50 100 150 200

Number of separations ('000)

a Acute and psychiatric hospitals (excludes freestanding day hospitals). Ten most frequent same-day
separations, as defined in AR-DRG version 5.1 (box 2.2). ECT: electroconvulsive therapy.
Source: AIHW (2009a).

AUSTRALIA’S 57
PRIVATE HOSPITAL
SECTOR
Figure 3.6 Most frequent overnight private hospital separations by
AR-DRG, 2007-08a

Hysterectomy for non-malignancy

b
Circulatory disorders

c
Laparoscopic cholecystectomy (w/o cscc)

Tonsillectomy and/or adenoidectomy

Inguinal and femoral hernia procedures

Knee replacement and reattachment

Other shoulder procedures

Caesarean delivery (w/o cscc)

Sleep apnoea

Vaginal delivery (w/o cscc)

0 50 100 150 200 250 300 350 400

Number of separations ('000)

a Acute and psychiatric hospitals (excludes freestanding day hospitals). Ten most frequent overnight
separations, as defined in AR-DRG version 5.1 (box 2.2). b Without acute myocardial infarction; with invasive
cardiac investigation procedure; without complex diagnosis or procedure. c Without common bile duct
exploration. w/o: without. cc: complications and comorbidities. cs: catastrophic or severe.
Source: AIHW (2009a).

The type of care offered by private hospitals differed across jurisdictions (table 3.7).
There are a comparatively high number of private geriatric separations in Victoria, a
low number of private newborn separations in South Australia, a high number of
private maintenance care separations in Queensland, and a high number of private
rehabilitation separations in New South Wales.

58 PUBLIC AND PRIVATE

HOSPITALS
Table 3.7 Number of private hospital separations by care type, 2007-08a
Acute Rehabilitation Newborn Geriatricb Palliative Maintenance
care care care c
New South Wales 783 374 68 039 22 917 .. 441 105
Victoria 777 176 13 717 4 057d 6 778 511 63
Queensland 748 685 25 036 18 716 66 2 433 1 208
South Australia 235 971 6 511 920 35 199 10
Western Australia 319 665 1 159 10 177 64 2 098 258
Tas, NT and ACTe 118 464 1 197 4 640 1 84 55
Australia 2 983 335 115 659 61 427 6 944 5 766 1 699
a Acute, psychiatric and freestanding day hospitals. Excludes other and not reported care types. b Includes
geriatric evaluation and management, and psychogeriatric care. c Maintenance care refers to the provision of
accommodation and nursing care as a service itself. This can include respite care, care to patients awaiting
placement, and care to inpatients designated as nursing home type, but excludes residential aged care. d The
reporting of newborns with unqualified days only is not compulsory for the Victorian private sector, resulting in
a low number of separations in this category. e Data for Tasmania, the Northern Territory and the ACT are
aggregated to protect the confidentiality of the small number of hospitals in each of these jurisdictions. .. Not
applicable.
Source: AIHW (2009a).

Research and training

Research and training are important and growing activities of private hospitals.
Historically, these services have been more typically provided by the public sector.
However, 47 private hospitals provided teaching to medical staff and
undergraduates in 2006-07 (ABS 2008e).

The interim results of a recent CHA clinical placements survey suggest that
midwifery, nursing and medical placements are the most common clinical
placements in CHA facilities (both public and private) (CHA 2009a).3 In its
submission to this study, CHA (sub. 20) noted the significant clinical teaching and
research that is undertaken by Catholic private hospitals. For-profit private hospital
operators also offer clinical training opportunities for health care students. For
example, Ramsay Health Care provided over 2 million clinical placement hours to
undergraduate nursing and medical students in 2007-08 (Ramsay Health Care
Limited 2009c). Further, APHA advised that:
Australia’s private hospitals invest $35 000 000 a year in the education and training of
surgeons, doctors, nurses and other healthcare professionals. (sub. 25, p. 5)

3 The CHA clinical placements survey reported 40 responses received from CHA facilities, and
includes aged care facilities as well as public and private hospitals (CHA 2009a).
AUSTRALIA’S 59
PRIVATE HOSPITAL
SECTOR
There are a number of government programs facilitating greater training in private
hospitals. For example, under the Expanded Specialist Training Program, medical
registrars on training programs undertake rotations through a range of settings,
including: private hospitals; specialists’ rooms; clinics; day surgeries; and
Aboriginal medical services (DOHA 2009g). The private hospital sector’s
increasing role in training is discussed further in section 3.6.

Emergency care in private hospitals

Comparatively few private hospitals provide emergency department services.

APHA noted that:
Some of the large acute medical/surgical private hospitals provide similar services to
their public sector counterparts, including accident and emergency services. However,
this applies largely in the densely-populated metropolitan areas. (sub. 25, p. 3)

In 2006-07, 24 private hospitals had emergency departments and 47 private

hospitals treated accident and emergency cases. Among the states, Victoria had the
largest number of private hospitals with accident and emergency services (15),
followed by Queensland (13) (table 3.8).

Table 3.8 Accident and emergency treatment in private hospitals, 2006-07a

Number of hospitals Number of hospitals Number of accident and
treating accident and with an emergency emergency patients
emergency cases department treated
New South Wales 5 4 54 829
Victoria 15 4 106 095
Queensland 13 9 162 758
South Australia 8 np 35 345
Western Australia np 3 np
Tas, NT and ACTb np np np
Australia 47 24 453 572
a Acute and psychiatric hospitals (excludes freestanding day hospitals). b Data for Tasmania, the Northern
Territory and the ACT are aggregated to protect the confidentiality of the small number of hospitals in each of
these jurisdictions. np Not published but included in totals where applicable.
Source: ABS (Private Hospitals, Cat. no. 4390.0).

Without an emergency department, private hospitals arguably have more

predictable workflows and uninterrupted throughput of patients. However, the NSW
Health Surgical Services Taskforce (sub. DR43) observed that disruptions to
elective surgical services can be minimised through effective management of
emergency surgical admissions.

60 PUBLIC AND PRIVATE

HOSPITALS
The Australasian College for Emergency Medicine noted that some private
hospitals have established emergency departments for a number of reasons:
a. Emergency departments are the safest and most efficient way to provide acute,
unscheduled care whether in public or private hospitals.
b. As private hospitals began to expand their range and complexity of services, a
mechanism was needed to provide acute care whether to new patients or to
patients post-discharge.
c. Private hospitals needed to respond to the expectation from insured patients that
they should be able to choose private hospital care in the event of an emergency
and not have to attend a public hospital. Some private hospitals, in particular the
church and charitable organisations, felt they had an obligation to their
communities to provide an alternative to public hospital emergency
departments, albeit at a cost to the patient.
d. An emergency department gave the hospital a source of patients that was
independent of their attending specialists. It also attracted patients at times and
on days when elective admissions were less frequent such as weekends thereby
increasing the productivity of hospital facilities.
e. In general, health insurance rates were declining during the 1990s. Therefore,
ensuring 24-hour availability and expanding services beyond elective or
scheduled care gave private hospitals with an emergency department a
competitive edge.
f. Patients who were admitted via an emergency department represent a casemix
and complexity that helps attract new specialists to a private hospital.
g. For some private hospitals, an emergency department was the only way they
could fund a medical staff presence on the campus 24 hours a day.
(sub. 14, p. 2)

3.4 Workforce characteristics

There were 46 718 salaried staff in private acute and psychiatric hospitals in
2006-07 (table 3.9), although this figure does not include medical practitioners with
rights of private practice. In these hospitals, nursing staff accounted for 60 per cent
of total salaried staff, and there were an average of 1.4 nurses per occupied bed.
Salaried medical officers and other diagnostic professionals accounted for close to
7 per cent of total salaried staff in these hospitals, and administrative and clerical
staff accounted for 16 per cent (ABS 2008e). Within the category of administrative
and clerical staff are occupations including: accounting; engineering; information
technology; and communications. The number of hospital managers and their roles
are difficult to discern from the data, as hospital managers may or may not also be
clinicians (Centre for Health Economics, Monash University, sub. 7).

AUSTRALIA’S 61
PRIVATE HOSPITAL
SECTOR
Table 3.9 Number of staff in private hospitals, 2006-07a
NSW Vic Qld SA WA Tas, NT Australia
and ACT b
All salaried staff 12 066 12 152 10 882 3 672 6 093 1 854 46 718
Salaried medical officers and other
diagnostic health professionals 881 988 589 129 419 100 3 106
Nursing 7 244 7 247 6 454 2 244 3 398 1 214 27 801
Administrative and clerical 1 815 1 885 2 099 589 959 297 7 645
Domestic and other 2 126 2 031 1 740 710 1 317 242 8 166
Staff per bedc 2.5 2.4 2.3 2.4 2.7 2.3 2.4
Nursing staff 1.5 1.4 1.4 1.5 1.5 1.5 1.4
Otherd 1.0 1.0 0.9 0.9 1.2 0.8 1.0
a Full-time equivalent staff in acute and psychiatric hospitals (excludes freestanding day hospitals).
b Tasmania, the Northern Territory and the ACT are aggregated to protect the confidentiality of the small
number of hospitals in these states/territories. c Average number of staff per occupied bed. d Includes
salaried medical officers and other diagnostic health professionals, administrative, domestic and other staff.
Source: ABS (Private Hospitals, Cat. no. 4390.0).

In the case of nursing staff, agreements at the enterprise or jurisdiction level

between employers and employee unions are common means of setting wages.
Available data suggest that pay rates for nurses in the private hospital sector are
close to those in the public hospital sector (NSW Nurses’ Association 2009). In
Victoria, by 31 January 2010, the most experienced Grade 2 registered nurse will be
paid $1201.40 per week in public and Healthscope hospitals and $1207.83 per week
at the Epworth hospitals (Workplace Authority 2008a, 2008b; Department of Health
Victoria 2009). In occupations such as nursing, employment conditions (for
example, the length of breaks between shifts) are also a key focus in collective
agreement negotiations.

Rights of private practice

In the hospital system, many medical specialists have rights of private practice as
well as having an established relationship with one or more private and/or public
hospitals. As CHA noted:
Many private hospitals are co-located with a public hospital. Many doctors work in
both sectors — as a salaried or sessional medical officer in the public sector and as an
independent practitioner in the private sector. Many doctors view their work time spent
across both types of hospitals as complementary and contributing to their overall work
and remuneration package. Remuneration rates are lower in the public system
compared with the private sector and many doctors who work in the private sector see
it as part of their professional duties to work for part of a week in a public hospital —
including undertaking teaching responsibilities. (sub. 20, p. 7)

62 PUBLIC AND PRIVATE

HOSPITALS
Furthermore, CHA noted that it is not only doctors who move across both sectors —
nurses, allied health professionals and medical students are among those who may
also have this mobility.

In the private sector, medical specialists are in non-salaried positions and work
independently of the hospital. Indicative requirements for such positions include:
fellowship from an Australian specialty college or recognised equivalent; eligibility
to be registered as a specialist with the relevant Australian medical board; and
appropriate indemnity insurance (Healthscope Limited 2009b; Ramsay Health Care
Limited 2009b).

Consultation with a number of study participants has suggested that there are no
data collections to indicate the number of specialists who have the right to admit
patients to private hospitals, the specific nature of these arrangements, or whether
there are specialties for which such arrangements are more common. One possible
reason may be that granting admission rights to medical practitioners is a decision
for individual hospitals.

3.5 Private freestanding day hospitals

Private freestanding day hospitals make up around half of all private hospitals. The
number of private freestanding day hospitals has been increasing over time, since
many procedures which used to require overnight hospital stay can now be
performed on a day-only basis due to advances in technology and treatment
methods.

Private freestanding day hospitals are fundamentally different to private acute and
psychiatric hospitals and public hospitals, making comparison with these entities
difficult. Private freestanding day hospitals often focus on a small number of
procedures at the exclusion of many other activities undertaken by larger acute
hospitals. This is a key reason why these facilities have been excluded from the later
comparative analysis. While not part of this study’s direct comparison of public and
private hospitals, these facilities are important for understanding the hospital sector
overall and the sector’s development over time.

Most private freestanding day hospitals are in metropolitan areas, and there are
more in New South Wales than in any other state or territory (table 3.10). Private
freestanding day hospital facilities accounted for almost 670 000 or around
20 per cent of total private sector separations in 2007-08 (AIHW 2009a). New
South Wales has a lower ratio of separations to beds than Queensland and Victoria.

AUSTRALIA’S 63
PRIVATE HOSPITAL
SECTOR
This may, in part, reflect differences in the activities undertaken by private
freestanding day hospitals among the jurisdictions (table 3.10).

Table 3.10 Private freestanding day hospital facilities, 2007-08

Number of Number of Number of Separations per
facilities beds separations 1000 residents
New South Wales 88 722 195 177 27
Victoria 73 558 168 826 31
Queensland 51 340 183 569 43
South Australia 24 130 46 732 26
Western Australia 28 352 65 611 30
Tasmania 2 9 np np
Northern Territory – – np np
ACT 6 40 np np
Australia 272 2 151 668 033 32
np Not published but included in totals where applicable. – Nil.
Source: AIHW (2009a).

The output of private freestanding day hospitals is mainly surgical. Of the 30

AR-DRGs with the largest number of separations from private freestanding day
hospitals in 2007-08, 16 were classified as surgical, seven were medical and seven
were classified as ‘other’ (AIHW 2009a). In 2006-07, specialist endoscopy,
ophthalmic, plastic/cosmetic and general surgery facilities accounted for around
65 per cent of private freestanding day hospitals (table 3.11). In 2006-07, New
South Wales accounted for 53 per cent of general surgery day facilities and around
44 per cent of ophthalmic day facilities in Australia.

Table 3.11 Number of private freestanding day hospital facilities by type of

centre, 2006-07
General Specialist Ophthalmic Plastic/ Othera Totalb
surgery endoscopy cosmetic
NSW 8 21 25 7 29 90
Vic np 31 9 np 23 73
Qld 4 15 14 3 16 52
SA .. 4 3 9 8 24
WA .. np np .. 12 17
Tas, NT and ACTc np np np np 4 12
Australia 15 76 57 28 92 268
a Includes fertility and sleep disorders clinics. b These data (2006-2007) are for a different reporting period
than the AIHW data in table 3.10 (2007-08). The total number of hospitals changed between reporting periods.
c Data for Tasmania, the Northern Territory and the ACT are aggregated to protect the confidentiality of the
small number of hospitals in each of these jurisdictions. np Not published but included in totals where
applicable. .. Not applicable.
Source: ABS (Private Hospitals, Cat. no. 4390.0).

64 PUBLIC AND PRIVATE

HOSPITALS
Nursing staff accounted for around 58 per cent of total staff in private freestanding
day hospitals in 2006-07 (table 3.12). This is a similar proportion to nursing staff in
private acute and psychiatric hospitals. Administrative and clerical staff accounted
for about 32 per cent of staffing in private freestanding day hospitals, which is
around double the proportion of these staff in private acute and psychiatric
hospitals. This may reflect an increased administrative burden from higher patient
turnover in these facilities.

Table 3.12 Number of staff in private freestanding day hospitals, 2006-07

General Specialist Ophthalmic Plastic/ Othera Totalb
surgery endoscopy cosmetic
Nursing staff 122 344 330 101 477 1 373
Administrative and clerical 33 240 173 52 263 761
Otherc 14 51 74 8 105 251
Total staff 169 634 577 160 845 2 385
a Including fertility and sleep disorders clinics. b Full-time equivalent staff. c Includes salaried medical officers
and other diagnostic health professionals, administrative, domestic and other staff.
Source: ABS (Private Hospitals, Cat. no. 4390.0).

3.6 Recent developments in private hospitals

Increasing numbers of patient separations

Total private hospital separations per 1000 residents have increased by more than
12 per cent since 2003-04. Within this, separations from private acute and
psychiatric hospitals have risen by close to 9 per cent over the same period
(figure 3.7). This highlights the rapid growth of private freestanding day hospitals
(close to 27 per cent over the period), albeit from a lower base. There was also an
increase in the average number of beds in private acute and psychiatric hospitals in
the capital cities by 410 beds, and a decrease in the average number of beds in
regional Australia by 96 beds, between 2005-06 and 2006-07 (ABS 2008e).

The proportion of private hospital separations funded by private hospital insurance

increased slightly between 2003-04 and 2007-08, from around 76 per cent to
80 per cent (figure 3.8). The proportions of private hospital patient separations
accounted for by public, Department of Veterans’ Affairs and self-funded patients
each declined by around 1 per cent over the period.

AUSTRALIA’S 65
PRIVATE HOSPITAL
SECTOR
Figure 3.7 Private hospital separations per 1000 residents,
2003-04 to 2007-08
150
Separations per 1000

120
residents

90

60

30

0
2003-04 2004-05 2005-06 2006-07 2007-08

All private hospitals

Private acute and psychiatric hospitals
Private freestanding day hospital facilities

Source: AIHW (2009a).

Figure 3.8 Private hospital patient separations by funding source,

2003-04 to 2007-08a

100
Per cent of separations

80

60

40

20

0
2003-04 2004-05 2005-06 2006-07 2007-08
Private health insurance Private self-funded
Public Dept Veterans' Affairs
Compensation or other

a Data exclude patients whose funding source is not reported.

Source: AIHW (2009a).

Changing private hospital services

The most common AR-DRGs among private hospital patients have changed little in
recent years. Of the 30 most common private hospital AR-DRGs in 2007-08,

66 PUBLIC AND PRIVATE

HOSPITALS
28 were among the 30 most common in 2003-04.4 However, there have been
significant increases in the number of private hospital separations for the most
common AR-DRGs. Table 3.13 sets out recent changes in the number of
separations for the 15 most common private hospital AR-DRGs in 2007-08. All but
two reported an increased number of separations between 2003-04 and 2007-08.

Table 3.13 Number of separations for the most common private hospital
AR-DRGs, 2003-04 to 2007-08a
AR-DRG Description 2007-08 Change
from
2003-04
R63Z Chemotherapy 176 372 32 228
G44C Other colonoscopy, same-day 169 234 29 529
L61Z Admit for renal dialysis 164 480 30 862
C16B Lens procedures, same-day 121 181 23 934
G45B Other gastroscopy for non-major digestive disease, same-day 97 758 - 5 061
D40Z Dental extractions and restorations 93 575 14 826
G46C Complex gastroscopy, same-day 89 533 28 689
Z64B Other factors influencing health status, same-day 77 046 34 772
U60Z Mental health treatment, same-day, without electroconvulsive 75 018 9 624
therapy
Z40Z Follow up with endoscopy 64 058 8 901
I18Z Other knee procedures 64 026 8 477
J11Z Other skin, subcutaneous tissue and breast procedures 53 625 3 280
O05Z Abortion with operating room procedure 51 114 4 305
N07Z Other uterine and adnexa procedures for non-malignancy 49 167 11 483
O60B Vaginal delivery without catastrophic or severe complications or 34 498 - 675
comorbidities
a Data for 2003-04 are defined according to AR-DRG version 5.0. Data for 2007-08 are defined according to
AR-DRG version 5.1 (see box 2.2). Data classifications are subject to minor revision between years.
Source: AIHW (2009a).

Some study participants noted recent changes in the breadth and composition of
private hospital services. CHA observed:
The private hospital sector is providing an increasing proportion of total hospital
services in many different specialty groups, particularly in the areas of cardiac medical,
cardiac interventional, oncology, obstetrics, orthopaedics and gastroenterology.
(sub. 20, p. 2)

APHA (sub. 25) commented that a number of complex procedures and treatments
traditionally associated with public hospitals are now performed more often in

4 The following AR-DRGs were not among the 30 most common private hospital separations in
2003-04: retinal procedures and other female reproductive system operating room procedures
(patients aged under 65) without malignancy, complications or comorbidities.
AUSTRALIA’S 67
PRIVATE HOSPITAL
SECTOR
private hospitals, including knee replacements, procedures of the digestive system,
prostatectomies, chemotherapy and major malignant breast conditions. It is also
noteworthy that private hospital separations increased between 2003-04 and
2007-08 across 22 of the 23 Major Diagnostic Categories, with the exception of a
small decrease of 62 separations in the burns category (AIHW 2009a).

Increasing teaching role

It appears that private hospitals are increasingly involved in teaching, although data
to support this development is limited. The number of private hospitals teaching
medical staff, nursing staff and allied health professionals increased between
2005-06 and 2006-07 (table 3.14).

Table 3.14 Private hospitals with teaching roles, 2005-06 to 2006-07a

2005-06 2006-07 Percentage
change
Number of private hospitals teaching:
Medical staff/undergraduates 42 47 12
Nursing staff/undergraduates 163 171 5
Allied health professionals 58 61 5
Number of hospitals with affiliated teaching status 67 64 -5
Nursing staff providing nurse education 272 319 17
a Acute and psychiatric hospitals (excludes freestanding day hospitals). Measured in full-time equivalent staff.

Source: ABS (Private Hospitals, Cat. no. 4390.0).

On the private sector’s take-up of teaching responsibilities, Australasian College for

Emergency Medicine noted that:
Whilst most private emergency departments would embrace and welcome involvement
in medicine training, it must be recognised that if students are placed without adequate
resourcing and process re-engineering, it is likely that the performance (both financial
and throughput) of private emergency departments will deteriorate. (sub. 14, p. 5)

68 PUBLIC AND PRIVATE

HOSPITALS
4 Public and private hospitals in the
health system

Key points
• Public and private hospitals are similar in a number of ways. A comparison of the
types of diagnoses most frequently treated by public and private hospitals indicates
that the two sectors offer many of the same hospital services, particularly
chemotherapy, renal dialysis and medical obstetrics. More broadly, a number of
large metropolitan private hospitals offer a range of services on par with large public
hospitals including, in some instances, accident and emergency treatment and
clinical training.
• Differences between public and private hospitals in terms of hospital size, location
and services are, in part, a function of their business models, government
requirements and community expectations.
– The public hospital sector handles the majority of acute care separations and
accounts for most regional and remote hospitals, while private hospitals are more
concentrated in metropolitan areas and are more likely to treat patients of higher
socioeconomic advantage.
– The public sector’s activity is concentrated on medical cases (including those
typically admitted through emergency departments) while the private sector’s
activity is more concentrated on surgical (typically elective) procedures.
• The overall relationship between the two sectors is not clear cut, especially as the
sectors do not operate in isolation, as exemplified by co-located hospitals sharing
resources and medical staff working across both sectors. Although differences
between and within the sectors make valid comparison difficult, Australia’s robust and
well-established system of public and private hospitals — and the overlap in their
services — enables a comparison of their respective performance to be considered.
• The respective roles, responsibilities and accountabilities of public and private
hospitals may change with the new National Healthcare Agreement and the future
response of governments to the recommendations of the National Health and
Hospitals Reform Commission.

Australia’s hospitals are an integral part of the nation’s health system, providing an
extensive range of services — including acute, emergency, newborn, geriatric,
diagnostic, rehabilitation, palliative and outpatient care — designed to fulfil the

PUBLIC AND PRIVATE 69

HOSPITALS IN THE
HEALTH SYSTEM
health needs of all residents across the nation. As part of Australia’s comprehensive
system of health services, hospitals contribute to the nation’s health outcomes in
combination with other key components, including health-promotion strategies,
disease-prevention strategies, chronic disease management, other primary care
services and accessible aged care. The performances of each of these components
are interdependent, such that deficiencies in one part of the health system are likely
to place greater demand on another. Shortfalls in the provision of primary care and
aged care in some more remote areas of Australia, for example, necessitate that
hospitals deliver these essential services to these communities. Hospitals must
therefore be evaluated in the context of their roles and responsibilities.

It is clearly recognised that public and private hospitals are driven by different
operational motives, typically treat different types of patients, and typically deliver
different suites of services. As such, these differences need to be taken into account
if comparisons between the public and private hospital sectors are to be valid and
useful. Given the broad scope of differences between public and private hospitals,
one role of this study is to highlight the complexities and limitations of conducting a
comparative assessment. At the same time, this study aims to identify the factors by
which the two sectors can be appropriately compared, allowing for the possibility
that discernible differences in the services offered by the two sectors may be partly
indicative of their respective fields of specialty and relative efficiency.

This chapter looks at the role of public and private hospitals as part of Australia’s
overall health system. The similarities and differences between the two sectors, and
their degree of complementarity and competitiveness, are examined. The possible
future directions of public and private hospitals in Australia, including the wider
pressures facing the hospital system, are discussed.

4.1 Similarities and differences

The different operational motives of the public and private hospital sectors
contribute to differences in the size and location of public and private hospital
establishments, as well as in the services they offer and patients they treat. This
section elaborates on some key differences between the public and private sectors in
the context of the hospital system as a whole, while also recognising their
similarities.

Core differences in the operational incentives, roles and responsibilities of the

public and private hospital sectors highlight the complexity of directly comparing
the sectors on the basis of efficiency. Indeed, study participants have confirmed that
efficiency may not be the primary objective of all hospital providers, and that some

70 PUBLIC AND PRIVATE

HOSPITALS
explicit objectives of the public sector — the delivery of emergency care, equity of
access, and clinical research and training — also contribute to the overall quality of
the health system:
[H]ospital performance may not be the greatest priority in the objectives of certain
groups. State administrators have different priorities to hospital managers (who may or
may not also be clinicians). There are also equity considerations, and many other issues
such as the capacity which must be held back when running an emergency department,
as well as staff commitments to teaching and medical research — all critical
components to a high quality healthcare system, and again reducing comparability of
providers. (Centre for Health Economics, Monash University, sub. 7, p. 2)

Given the complexities involved in assessing the comparative performance of the

two hospital sectors, this study acknowledges that points of similarity will help
identify the factors by which the sectors can be appropriately compared. For
example, large private hospitals that treat a similar casemix to the public sector,
conduct teaching and research and run emergency departments are likely to be
sufficiently similar to many public hospital establishments to merit an evaluative
comparison. On the other hand, points of difference between the sectors should not
necessarily preclude a comparative assessment, but serve to highlight the factors
that must be taken into account in the assessment, and may potentially signal areas
of relative efficiency.

Similarities and differences in hospital establishments

The majority of very small hospitals (with 50 or fewer beds) and very large
hospitals (more than 200 beds) are public. Moderately sized hospitals (between
51 and 200 beds) are more evenly distributed across the two sectors (table 4.1).

Table 4.1 Share of public and private hospitals by sizea

Public hospitalsb Private hospitalsc Australia

% % %
0–50 beds 81 19 100
51–100 beds 46 54 100
101–200 beds 54 46 100
201 beds or more 78 22 100
All hospitals 73 27 100
a Acute and psychiatric hospitals. b Data for public hospitals are for 2007-08. c Data for private hospitals are
for 2006-07.
Source: ABS (Private Hospitals, Cat. no. 4390.0); AIHW (2009a).

PUBLIC AND PRIVATE 71

HOSPITALS IN THE
HEALTH SYSTEM
Public and private hospitals operate in broadly similar numbers in metropolitan
areas. Given public hospitals’ service obligations in rural and regional areas, as well
as the private hospital sector’s commercial considerations, all hospitals in remote
and very remote areas of Australia are provided by the public sector (table 4.2).

Table 4.2 Public and private hospitals by location, 2009a

Public hospitals Private hospitals Australia

% % %
Major cities 45 55 100
Inner regional 76 24 100
Outer regional 93 7 100
Remote 100 — 100
Very remote 100 — 100
a Acute and psychiatric hospitals. Remoteness area based on ABS (2005) Australian Standard Geographical
Classification. — Nil.
Source: DOHA (2009c, 2009e).

The need for the public hospital sector to ensure service delivery in rural and
regional areas, while also undertaking teaching in the large metropolitan areas,
helps to explain the diversity of hospital establishments in the public hospital sector.
On the other hand, the different operational objectives of the private hospital sector
help to explain why few establishments are small enough to be adversely affected
by a lack of economies of scale, and none are located in remote areas where
distance and isolation contribute to higher resource costs.

Similarities and differences in services

Patient services

In 2007-08, public hospitals accounted for around 61 per cent of total hospital
separations, while private acute and psychiatric hospitals accounted for around
31 per cent. The remainder were separations in private freestanding day hospitals.
Public hospitals provided the majority of all types of hospital care with the
exception of rehabilitation services, and also accounted for the majority of medical
cases handled by Australia’s hospital system. In contrast, private hospitals
accounted for the majority of surgical and other procedures, performing around
two-thirds of all elective surgery (table 4.3). New South Wales Health (sub. 40)
commented that private hospitals may be, in effect, more selective in their
admissions than public hospitals, because they are less likely to be equipped to treat
long-stay or highly complex medical cases. These types of cases are likely to be

72 PUBLIC AND PRIVATE

HOSPITALS
redirected to the public hospital sector instead. However, the Australian Private
Hospitals Association (APHA) (sub. 25) commented that a number of complex
procedures and treatments traditionally associated with public hospitals are now
commonly performed in private hospitals.

Table 4.3 Share of hospital separations by sector, 2007-08

Public hospitals Private hospitals Australia

% % %
Separations by type of care
Acute care 60 40 100
Rehabilitation 39 61 100
Palliative care 79 21 100
Geriatric 74 26 100
Maintenance care 92 8 100
Newborn 79 21 100
Separations by type of diagnosis
Surgical 43 57 100
Medical 75 25 100
Other 31 69 100
Elective procedures 36 64 100
a Defined by AR-DRG partition (box 2.2).

Source: AIHW (2009a).

Outside of admitted patient services, public hospitals handled over 90 per cent of
the total number of accident and emergency presentations reported in 2007-08
(AIHW 2009a).

A detailed comparison of the types of services provided by the two hospital sectors is
presented in the following lists of the 30 most frequent overnight and same-day
separations that were treated in each sector in 2007-08, categorised according to
Australian Refined Diagnosis-Related Groups (AR-DRGs) (box 2.2).

In a comparison of the 30 most frequent overnight AR-DRGs treated in each sector

(which represent 33 per cent of overnight separations in the public sector and
42 per cent of overnight separations in the private sector), only eight AR-DRGs
were common to both sectors (table 4.4). These common AR-DRGs mainly relate to
obstetrics. Of the 22 overnight AR-DRGs that were distinct to the public sector in
this sample, many were the type of separation commonly admitted through
emergency departments, such as respiratory and cardiac-related cases, injuries,
seizures, and treatment for poisoning and the toxic effects of drugs. All but one of
these treatments were medical cases, although the one exception — appendicectomy
— is a surgery usually performed in emergency circumstances to treat appendicitis.
Also distinct to the public sector’s most frequent types of separations in this sample
PUBLIC AND PRIVATE 73
HOSPITALS IN THE
HEALTH SYSTEM
was cellulitis — an infection commonly caused by the Staphylococcus aureus
bacteria (RCPA 2009), which is a condition used in the monitoring of hospital
quality. Of the 22 overnight AR-DRGs that were distinct to the private sector in this
sample, many were elective procedures such as knee or hip replacements. All but
two of these AR-DRGs were surgical cases.

There appears to be considerably more overlap between the sectors in their

same-day separations. In a comparison of the 30 most frequent same-day AR-DRGs
treated in each sector (which represented 70 per cent of same-day separations in the
public sector and 81 per cent of same-day separations in the private sector in
2007-08), 22 AR-DRGs were common to both sectors, including the four most
frequent same-day separations overall (table 4.5). Although differing in order of
frequency, the top four activities in both sectors were: renal dialysis, chemotherapy,
non-complex colonoscopy and lens procedures. The greater degree of overlap in
same-day separations between the two sectors may be partly explained by the fact
that same-day separations in both hospital sectors appear to be concentrated among
a smaller number of AR-DRGs.

Although the public and private hospital sectors displayed more similarity in their
same-day separations than in their overnight separations, the concentration of
medical cases in the public sector and surgical cases in the private sector was again
apparent. Of the eight most frequent treatments distinct to the private sector in this
sample, all but two were surgical cases. All eight of the treatments distinct to the
public sector in this sample were medical cases.

74 PUBLIC AND PRIVATE

HOSPITALS
Table 4.4 Thirty most frequent overnight separations in public and private hospitals by AR-DRG, 2007-08a
Public hospitals Private hospitals

AR-DRG Partitionb Descriptionc Rank Number of Per centd Rank Number of Per cente
separations separations

O60B Med Vaginal delivery (w/o cscc) 1 101 245 4.47 1 34 421 3.39
F74Z Med Chest pain 2 52 326 2.31 26 8 427 0.83
G67B Med Oesophagitis, gastroent and misc. digestive system disorders (age>9; w/o cscc) 3 42 082 1.86 23 9 212 0.91
O01C Surg Caesarean delivery (w/o cscc) 4 41 510 1.83 3 28 324 2.79
J64B Med Cellulitis (age>59 (w/o cscc)) or age<60) 5 35 070 1.55 na
O66A Med Antenatal and other obstetric admission 6 33 277 1.47 25 8 504 0.84
O60C Med Vaginal delivery (single uncomplicated w/o other condition) 7 24 183 1.07 na
E65A Med Chronic obstructive airways disease (with cscc) 8 22 370 0.99 na
G66B Med Abdominal pain or mesenteric adenitis (w/o cc) 9 22 032 0.97 na
E65B Med Chronic obstructive airways disease (w/o cscc) 10 21 571 0.95 na
E62C Med Respiratory infections/inflammations (w/o cc) 11 21 547 0.95 na
E69C Med Bronchitis and asthma (age<50 w/o cc) 12 21 328 0.94 na
F62B Med Heart failure and shock (w/o catastrophic cc) 13 21 228 0.94 na
E62B Med Respiratory infections/inflammations (with severe or moderate cc) 14 20 407 0.90 na
U67Z Med Personality disorders and acute reactions 15 20 369 0.90 na
F71B Med Non-major arrhythmia and conduction disorders (w/o cscc) 16 20 295 0.90 na
D63B Med Otitis media and upper respiratory tract infection (w/o cc) 17 19 677 0.87 na
H08B Surg Laparoscopic cholecystectomy (w/o common bile duct exploration; w/o cscc) 18 18 558 0.82 8 16 815 1.65
G07B Surg Appendicectomy (w/o cscc) 19 18 388 0.81 na
U63B Med Major affective disorders (age<70 w/o cscc) 20 17 111 0.76 12 12 217 1.20
I68B Med Non-surgical spinal disorders (w/o cc) 21 16 198 0.71 27 8 088 0.80
U61A Med Schizophrenia disorders (with mental health legal status) 22 16 002 0.71 na
X60C Med Injuries (age<65) 23 15 743 0.69 na
B76B Med Seizure (w/o cscc) 24 15 450 0.68 na
K60B Med Diabetes (w/o cscc) 25 14 981 0.66 na

(Continued next page)

75
Table 4.4 (continued)

Public hospitals Private hospitals

AR-DRG Partitionb Descriptionc Rank Number of Per centd Rank Number of Per cente
separations separations

L63B Med Kidney and urinary tract infections (age>69 or with severe cc) 26 14 831 0.65 na
X62B Med Poisoning or toxic effects of drugs and other substances (age<60 w/o cc) 27 14 672 0.65 na
L63C Med Kidney and urinary tract infections (age<70 w/o cscc) 28 14 364 0.63 na
O60A Med Vaginal delivery (with cscc) 29 14 282 0.63 na
F73B Med Syncope and collapse (w/o cscc) 30 14 040 0.62 na
E63Z Med Sleep apnoea na 2 34 109 3.36
I16Z Surg Other shoulder procedures na 4 26 536 2.61
I04Z Surg Knee replacement and reattachment na 5 22 184 2.18
G09Z Surg Inguinal and femoral hernia procedures (age>0) na 6 18 605 1.83
D11Z Surg Tonsillectomy and/or adenoidectomy na 7 17 619 1.73
F42B Other Circulatory disordersf na 9 15 506 1.53
N04Z Surg Hysterectomy for non-malignancy na 10 14 168 1.39
I03C Surg Hip replacement (w/o cscc) na 11 12 697 1.25
M02B Surg Transurethral prostatectomy (w/o cscc) na 13 11 721 1.15
K04Z Surg Major procedures for obesity na 14 11 718 1.15
I18Z Surg Other knee procedures na 15 11 396 1.12
F15Z Surg Percutaneous coronary interventiong na 16 10 183 1.00
I20Z Surg Other foot procedures na 17 10 104 0.99
I10B Surg Other back and neck procedures (w/o cscc) na 18 9 974 0.98
D10Z Surg Nasal procedures na 19 9 929 0.98
N06Z Surg Female reproductive system reconstructive procedures na 20 9 771 0.96
D06Z Surg Sinus, mastoid and complex middle ear procedures na 21 9 574 0.94
J06B Surg Major procedures for non-malignant breast conditions na 22 9 284 0.91
I29Z Surg Knee reconstruction or revision na 24 8 723 0.86

(Continued next page)

76
Table 4.4 (continued)

Public hospitals Private hospitals

AR-DRG Partitionb Descriptionc Rank Number of Per centd Rank Number of Per cente
separations separations

G11B Surg Anal and stomal procedures (w/o cscc) na 28 7 897 0.78
J06A Surg Major procedures for malignant breast conditions na 29 7 610 0.75
I30Z Surg Hand procedures na 30 7 527 0.74

Total 30 most frequent overnight AR-DRGs 745 137 32.89 422 843 41.60
a Number of separations ranked according to frequency in each sector, as defined in AR-DRG version 5.1. b Med: medical partition. Surg: surgical partition (see box 2.2).
c w/o: without. cc: complications and comorbidities. cs: catastrophic or severe. misc: miscellaneous. d Per cent of total public hospital overnight separations. e Per cent of
total private hospital overnight separations. f without acute myocardial infarction; with invasive cardiac investigation procedure; without complex diagnosis or procedure.
g without acute myocardial infarction; with stent implantation. na: AR-DRG is not among the 30 most frequent in the sector (data unavailable).

Source: AIHW (2009a).

77
Table 4.5 Thirty most frequent same-day separations in public and private hospitals by AR-DRG, 2007-08a
Public hospitals Private hospitals

AR-DRG Partitionb Descriptionc Rank Number of Per centd Rank Number of Per cente
separations separations

L61Z Med Admit for renal dialysis 1 815 622 34.82 3 164 469 8.29
R63Z Med Chemotherapy 2 121 703 5.20 1 176 290 8.89
G44C Other Other (non-complex) colonoscopy 3 53 385 2.28 2 169 234 8.53
C16B Surg Lens procedures 4 51 907 2.22 4 121 181 6.11
O66B Med Antenatal and other obstetric admission 5 45 835 1.96 na
Z64B Med Other factors influencing health status 6 45 378 1.94 8 77 046 3.88
F74Z Med Chest pain 7 36 115 1.54 na
G45B Other Other gastroscopy for non-major digestive disease 8 34 160 1.46 5 97 758 4.93
J11Z Surg Other skin, subcutaneous tissue and breast procedures 9 33 188 1.42 13 50 106 2.53
Q61C Med Red blood cell disorders (w/o cscc) 10 30 008 1.28 24 16 961 0.86
G67B Med Oesophagitis, gastroent and misc. digestive system disorders (age>9; w/o cscc) 11 28 628 1.22 na
Z40Z Other Follow up with endoscopy 12 26 873 1.15 10 62 510 3.15
G46C Other Complex gastroscopy 13 23 513 1.00 7 89 533 4.51
O05Z Surg Abortion (with operating-room procedure) 14 23 431 1.00 12 50 165 2.53
D40Z Other Dental extractions and restorations 15 22 983 0.98 6 91 399 4.61
U60Z Med Mental health treatment (w/o electroconvulsive therapy) 16 21 734 0.93 9 75 018 3.78
R61C Med Lymphoma and non-acute leukaemia 17 21 578 0.92 27 14 061 0.71
L41Z Other Cystourethroscopy 18 20 801 0.89 15 26 074 1.31
X60C Med Injuries (age<65) 19 20 723 0.88 na
G66B Med Abdominal pain or mesenteric adenitis (w/o cc) 20 18 412 0.79 na
I68C Med Non-surgical spinal disorders 21 17 057 0.73 18 19 575 0.99
N10Z Surg Diagnostic curettage or diagnostic hysteroscopy 22 16 288 0.70 22 17 291 0.87
N09Z Surg Conisation, vagina, cervix and vulva procedures 23 16 174 0.69 29 12 754 0.64
L67C Med Other kidney and urinary tract diagnoses (w/o cscc) 24 15 231 0.65 na
I30Z Surg Hand procedures 25 14 286 0.61 23 17 288 0.87
I18Z Surg Other knee procedures 26 14 115 0.60 11 52 630 2.65

(Continued next page)

78
Table 4.5 (continued)

Public hospitals Private hospitals

AR-DRG Partitionb Descriptionc Rank Number of Per centd Rank Number of Per cente
separations separations

Q60C Med Reticuloendothelial and immunity disorders (w/o malignancy; w/o cscc) 27 13 921 0.59 na
N07Z Surg Other uterine and adnexa procedures for non-malignancy 28 13 787 0.59 14 43 463 2.19
I74C Med Injury to forearm, wrist, hand or foot (age<75; w/o cc) 29 13 517 0.58 na
G11B Surg Anal and stomal procedures (w/o cscc) 30 11 873 0.51 20 18 525 0.92
N11B Surg Other female reproductive system operating-room procedures (age<65) (w/o cc)e na 16 22 819 1.15
J08B Surg Other skin graft and/or debridement procedures (w/o cscc) na 17 22 770 1.15
C03Z Surg Retinal procedures na 19 19 401 0.98
J10Z Surg Skin, subcutaneous tissue and breast plastic operating-room procedure na 21 18 323 0.92
F42B Other Circulatory disordersf na 25 16 862 0.85
B05Z Surg Carpal tunnel release na 26 15 269 0.77
V62B Med Alcohol use disorder and dependence na 28 12 972 0.65
D13Z Surg Myringotomy with tube insertion na 30 12 043 0.61

Total 30 most frequent same-day AR-DRGs 1 642 226 70.11 1 603 790 80.87
a Number of separations ranked according to frequency in each sector, as defined in AR-DRG version 5.1. b Med: medical partition. Surg: surgical partition (see box
2.2). c w/o: without. cc: complications and comorbidities. cs: catastrophic or severe. misc.: miscellaneous. d Per cent of total public hospital same-day separations.
e Per cent of total private hospital same-day separations. e without malignancy. f without acute myocardial infarction; with invasive cardiac investigation procedure;
without complex diagnosis or procedure. na AR-DRG is not among the 30 most frequent in the sector (data unavailable).
Source: AIHW (2009a).

79
Teaching and training

An historical difference between the public and private hospital sectors has been
their teaching and training responsibilities. As noted in earlier chapters, the public
sector delivers the vast bulk of teaching and training, although private sector
involvement is increasing, and recent government programs are facilitating this
expansion (DOHA 2009g). Catholic Health Australia (CHA) (sub. 20) noted that
many large Catholic hospitals have much in common with large public hospitals
through their involvement in teaching and research, and the APHA (sub. 25) noted
that the private hospital sector provides a significant volume of training and
education to undergraduate medical and nursing students that is not funded by
government. It has been argued that greater responsibility for teaching should be
assigned to private hospitals, in order to offer more training opportunities in areas of
private sector specialisation (Crotty 2005).

Similarities and differences in patients

Patient demographics and socioeconomic status differ between the public and
private hospital sectors. Although the public hospital sector handles the majority of
separations for patients of all age groups, some variation is apparent (figure 4.1).
Children and young people comprise a larger share of the case load of public
hospitals, with 28 per cent of patients admitted to public hospitals being under
35 years of age. In private hospitals, the age group with the highest proportion of
patients admitted is 50 to 64 years (27 per cent). Patients aged between 75 to
84 years comprise 15 per cent of those admitted to both public and private hospitals,
while those aged 85 and over comprise 6 per cent of those admitted to public
hospitals and 5 per cent of those admitted to private hospitals.

A higher proportion of patients in public hospitals are from disadvantaged

socioeconomic areas, compared to patients in private hospitals. Even so, around
26 per cent of patients from the most disadvantaged areas were treated in private
hospitals treated in 2007-08 (figure 4.2). Around 40 per cent of patients from the
most advantaged socioeconomic areas were treated in public hospitals.

80 PUBLIC AND PRIVATE

HOSPITALS
Figure 4.1 Share of separations by sector and patient age, 2007-08

100
Per cent of separations

80

60

40

20

0
0-14 15-34 35-44 45-54 55-64 65 & over
Age groups
Public hospitals Private hospitals

Source: AIHW (2009a).

Figure 4.2 Share of separations by sector and socioeconomic status of

patients, 2007-08a

100
Per cent of separations

80

60

40

20

0
Most Second most Middle Second most Most
disadvantaged disadvantaged advantaged advantaged

Quintiles of socioeconomic advantage/disadvantage

Public hospitals Private hospitals

a Quintiles of socioeconomic status based on ABS (2006) Index of Relative Socioeconomic

Advantage/Disadvantage score based on the patient’s area of usual residence.
Source: AIHW (2009a).

In 2007-08, the majority of patients electing to be treated as private patients were

treated by private hospitals, while public hospitals treated nearly all public patients
(table 4.6). Public hospitals treated 14 per cent of patients using private health
insurance to fund their hospital stay. Private hospitals treated 62 per cent of
Department of Veterans’ Affairs patients in 2007-08.

PUBLIC AND PRIVATE 81

HOSPITALS IN THE
HEALTH SYSTEM
Table 4.6 Share of patient separations by sector and funding source,
2007-08
Public hospitals Private hospitals Australia

% % %
Public patient 98 2 100
Private health insurance 14 86 100
Self-funded 17 83 100
Department of Veterans’ Affairs 38 62 100
Compensation or othera 47 53 100
All funding sources 60 40 100
a Includes workers compensation, other compensation, motor vehicle third party personal claim, other public
authorities and other funding sources.
Source: AIHW (2009a).

4.2 Relationship between the two sectors

Based on the differences in the business structures, responsibilities, services and
patient profiles of public and private hospitals, there is strong support from study
participants that the two sectors are fundamentally different, as noted:
[I]t is critical to emphasise that the public and private hospital sectors are driven by
different requirements. The core function may be to assemble infrastructure, workforce
and knowledge around the care of patients to improve their health, but the associated
‘imperatives’ have created two very different structures. (Royal Australasian College of
Surgeons, sub. 30, pp. 1-2)
The two sectors have fundamentally different roles in that the public health system is
required under the National Health Care Agreement to treat all patients that present for
care, while the private sector provides preferential access to care for those who can
afford it. (Tasmanian Department of Health and Human Services, sub. 37, p. 3)
[P]ublic and private hospitals serve different populations, do different things, operate in
very different ways and are funded differently. (Australian Healthcare and Hospitals
Association, sub. 33, p. 2)
For the most part, private hospitals are quite different from public hospitals in size and
types of services offered. For example, private facilities in the mental health sector
provide treatment for quite distinct conditions to those treated in the public sector. The
majority of chemotherapy treatment for people with cancer is delivered in the private
sector. Most of the rehabilitation for people who have had accidents, injuries or falls is
provided in private hospitals … And nearly all in-hospital specialist palliative care
services for the dying are provided in private hospitals, especially in regional Australia.
(APHA, sub. 25, p. 3)

A key difference between the sectors is the type of product they deliver. For certain
types of care offered by both sectors, private hospital services can be accessed
82 PUBLIC AND PRIVATE
HOSPITALS
relatively sooner, and with the added benefits of private ward accommodation and a
choice of doctor, subject to a patient’s willingness-to-pay. Based on this distinction,
the two sectors are effectively delivering differentiated products. As CHA
commented, the private sector has to distinguish itself from the public sector in
order to attract demand:
Given that privately insured patients have already paid for public insurance under
Medicare, the private health sector, in order to attract additional funding from
individuals, needs to provide a different patient experience to the public sector —
particularly in areas where seemingly similar clinical services are offered. (sub. 20, p. 6)

These fundamental differences in public and private hospitals suggest that the two
sectors may be described as complementary, in the sense that they provide a
different range of services which supplement that of the other. Many interested
parties held this view, at least with respect to some services (ACSQHC, sub. 24;
Rhonda Kerr and Associates, sub. 34). Queensland Health noted:
In general terms, … the provision of public and private health care [is] predominantly
complementary rather than competitive. Therefore any competitive analysis will need
to carefully address the complexities of the environment in which health care takes
place across the public and private systems. (sub. 27, p. 1)

On the other hand, some study participants commented that there is little evidence
that the two sectors are complementary (Centre for Health Economics, Monash
University, sub. 7), while other participants pointed towards the sectors’
similarities. As well as offering many of the same services, the two sectors
effectively compete for the same resources, such as trained medical staff
(Rhonda Kerr and Associates, sub. 34). The noted similarities between public and
private hospitals suggest that, at least in some aspects, the two sectors may be
described as competitive markets, sharing common functions and offering
substitutable services.

Many study participants, however, concluded that the relationship between the
public and private sectors is not clear cut. While some public and private hospital
establishments share common features and may be seen to operate in competition,
others are sufficiently distinct that they may be viewed as complementary
components of the total hospital system. On this point, APHA noted:
Private hospitals are in some aspects similar to public ones. Some of the large acute
medical/surgical private hospitals provide similar services to their public sector
counterparts, including accident and emergency services. However, this applies largely
in the densely populated metropolitan areas. For the most part, private hospitals are
quite different from public hospitals in size and types of services offered. (sub. 25, p. 3)

Likewise, the Australian Medical Association (AMA) observed that the similarities
between the sectors are restricted to a select number of features:
PUBLIC AND PRIVATE 83
HOSPITALS IN THE
HEALTH SYSTEM
 There are some limited areas where the two hospital sectors can be seen as being in
competition with each other. However, the two sectors do have quite markedly
differing casemix. (sub. 28, p. 2)

Defining the profiles of the public and private sectors is further complicated by
signs of increasing overlap and interaction between them, as the Australian
Government Department of Health and Ageing observed:
The boundaries between public hospitals and private hospitals and the services
provided within each sector are becoming increasingly blurred. Examples include:
public hospitals provide services to private patients, while private hospitals provide
services to public patients; … emergency care is provided by public hospitals and also
by some private hospitals; the education and training of healthcare professionals is now
occurring in some private facilities. (sub. 32, p. 12)

Similarly, the Royal Australasian College of Surgeons observed that the private
hospital sector is increasingly adopting some of the traditional characteristics of the
public sector:
Historically [in the private sector] there has been an emphasis on elective or
semi-elective presentations. However, there has also been a growing recognition that
patients present as emergencies, and this has seen the development of emergency
departments in the larger private hospitals. Patients are also referred by medical
practitioners, usually specialists. Consequently another focus of the private hospital
sector has been on securing a steady referral of patients, which has driven its
development. (sub. 30, p. 2)

Furthermore, defining the responsibilities and incentives of each sector is

complicated by the fact that a single provider can deliver services in both sectors, as
CHA commented:
In the case of the Catholic hospitals, the delineation between public and private is
further confounded by the fact that a number of Catholic healthcare providers operate
significant hospitals in both the public and private sectors including the St Vincent’s
Health Australia, Little Company of Mary Health Service and the Mater Health Service
in Brisbane. (sub. 20, p. 5)

The strength of interaction between the two sectors has led some to describe them
as ‘interdependent’ (Queensland Health, sub. 27, p. 2), reflecting the fact that the
services and functions of each sector help to support the other. The interaction
between the sectors may be demonstrated by, for example: the sharing of resources
in co-located establishments; medical staff working across both sectors and the
exchange of spillover benefits (the private sector benefiting from the public sector’s
investment in medical training and research). The AMA identified the benefits of
Australia’s dual hospital system:
[T]he plural nature of the public and private hospital system is one of the strengths of
Australia’s health system. It is readily apparent that taxpayer-sourced funding cannot
84 PUBLIC AND PRIVATE
HOSPITALS
 bear the whole load of financing health care … Australia gets its best results when the
two sectors have a strong symbiotic relationship … It makes no economic sense to have
duplicated and underutilised resources in both sectors if there is scope for resources to
be shared. (sub. 28, pp. 1–2)

4.3 Possible directions for hospitals

Australia’s hospital system is under pressure, largely due to the impact of
population ageing and the associated fiscal burden, but also due to technological
advancements, heightened community expectations regarding access to health care,
and trends in the health status of the population. These factors are changing not only
the type of hospital services sought by patients, but also how and where treatment is
delivered.

Under these pressures, healthcare expenditure currently consumes around 9 per cent
of Australia’s GDP, of which the equivalent of 3.5 per cent of GDP is allocated to
hospitals (AIHW 2009c). In 2007-08, Australian Government expenditure on
hospitals amounted to $14.9 billion, most of which was allocated to public hospital
services (figure 4.3). Around 13 per cent of the Australian Government’s hospital
expenditure was used to fund private health insurance premium rebates. In the same
year, state and local government expenditure on hospitals amounted to
$16.8 billion, almost all of which was directed to public hospitals.
Non-governmental expenditure on hospitals amounted to $6.9 billion, most of
which was directed to private hospitals. From 2005-06 to 2007-08, total government
expenditure on public and private hospitals increased by 12 per cent and 9 per cent
respectively. Over the same timeframe, non-governmental expenditure on public
and private hospitals increased by 17 per cent and 7 per cent respectively (AIHW
2009c).

Accounting for growth in both demographic and non-demographic factors (such as

medical developments), the Australian Treasury estimated that from 2006-07 to
2046-47, Australian Government expenditure on hospitals (including private health
insurance rebates) will increase from 1.2 to 2.3 per cent of GDP, equivalent to a
three-fold increase in real expenditure per person (Treasury 2007). Consistent with
these figures, the Commission estimated that Australian Government expenditure
on hospitals (excluding private health insurance rebates) will increase to 2.25 per
cent of GDP by 2044-45 (PC 2005). It estimated that state and territory government
expenditure on hospitals will increase to 1.9 per cent of GDP by 2044-45, up from
1.2 per cent of GDP in 2002-03 (PC 2005).

PUBLIC AND PRIVATE 85

HOSPITALS IN THE
HEALTH SYSTEM
Figure 4.3 Funding sources of public and private hospitals, 2007-08a

Non-governmental
– private hospitals
($4.7 billion)
Non-governmental Australian
– public hospitals Government
($2.2 billion) – public hospitals
($12.1 billion)
State/Territory
Governments
– private hospitals
($0.3 billion)

Australian
State/Territory Government
Governments – private hospitals
– public hospitals ($2.8 billion)
($16.6 billion)

a Expenditure on public hospital services and private hospitals.

Source: AIHW (2009c).

Given the intensified demand for hospital services, matched with increasing
budgetary pressures, current policy reforms are focused on generating greater
efficiency, accountability and performance quality within Australia’s hospital
system, and across both public and private sectors. This is especially the case with
the new National Health Agreement (NHA) agreed to by COAG in late 2008. Other
reform proposals, such as the recommendations of the National Health and
Hospitals Reform Commission (NHHRC), are consistent with these directions.

COAG reforms

The NHA aims to change the future delivery of public hospital services. It was
agreed to by COAG in December 2008 and took effect from July 2009. The
Australian Government and all state and territory governments are signatories to the
agreement, which declares as one of its objectives that Australians ‘receive
appropriate high quality and affordable hospital and hospital-related care’ (COAG
2008d, p. A-4). The agreement is intended to improve the efficiency of service
delivery in the public hospital sector, while also reflecting the community service
obligations of small and regional hospitals.

86 PUBLIC AND PRIVATE

HOSPITALS
Among the changes, the agreement proposes a range of measures designed to:
• move towards activity-based funding in public hospitals
• reduce average waiting times for elective surgery
• improve access to rehabilitation, post-acute and transition care services
• improve the quality of data on services to non-admitted patients
• improve comparative assessment of the public and private hospital sectors
• increase levels of informed financial consent for private patients in both public
and private hospitals
• increase hospital quality and safety by building on the priorities of the
ACSQHC.

Although the NHA reforms are primarily applicable to the public hospital sector,
measures to monitor the progress of the NHA objectives will apply to the private
hospital sector where relevant (COAG 2008d). Further details on the NHA are
provided in appendix B.

The National Partnership Agreement on Hospital and Health Workforce Reform,

agreed to by COAG in 2008, sets specific measures for the public hospital sector to:
• introduce a nationally consistent approach to activity-based funding (starting
development in 2009-10 and implementation in 2013-14)
• improve health workforce capability and supply
• enhance the provision of sub-acute services
• relieve pressure on emergency departments (COAG 2008e).

Among the key budgetary measures to support the two agreements, COAG agreed
to the following funding allocations:
• $60.5 billion as a Specific Purpose Payment to the states to facilitate the
objectives of the NHA
• $750 million as a National Partnership (NP) payment in 2008-09 to relieve
pressure on public hospital emergency departments
• $500 million in 2008-09 to provide 1600 more sub-acute care beds
• $1.75 billion as an NP payment over five years from 2008-09 to expand training
programs — especially in regional hospitals — in addition to funding by state
governments (COAG 2008b).

The total funding package delivered to the states as part of the NHA represents an
increase of $22.4 billion compared to the previous Australian Health Care
PUBLIC AND PRIVATE 87
HOSPITALS IN THE
HEALTH SYSTEM
Agreements. The funding is targeted to facilitate higher throughput in public
hospitals, with public statements suggesting the extra funding will enable 370 000
more separations, 350 000 more emergency department presentations and
2.5 million more outpatient services over four years from 2008-09 (COAG 2008a).

National Health and Hospitals Reform Commission

In June 2009, the NHHRC presented a suite of recommendations to the Australian

Government proposing, among many changes, a redesign of the funding and
governance arrangements of the health system, higher standards of service delivery,
expanded clinical training programs and improvements in the reporting of hospital
performance indicators (NHHRC 2009). Among an extensive list, the NHHRC
specifically recommended:
• the use of activity-based funding for both public and private hospitals, covering
inpatient, outpatient and emergency services
• the use of National Access Targets to ensure timely access and safe care in
hospitals
• data reporting requirements for both public and private hospitals
• that a greater share of responsibility for the funding and delivery of health
services be transferred from the states and territories to the Australian
Government.

Although the recommendations of the NHHRC are largely consistent with the NHA
— including the move towards activity-based funding to drive higher efficiency —
some distinct changes in funding and governance arrangements are proposed. The
NHHRC’s recommendation that the Australian Government assume full funding
responsibility for public hospitals over time would introduce a more centralised
approach to health services, compared to the NHA’s more collaborative
Commonwealth–State approach.

Also distinct from the NHA, the NHHRC proposed a new model for health care in
Australia more generally, called Medicare Select, which would build on the existing
Medicare scheme. Under this new model, all patients would be covered by a ‘health
and hospital’ plan to access their universal service Medicare entitlements, but given
greater flexibility to switch to a private provider without forgoing their Medicare
entitlements. The proposed new model is intended to promote greater competition
between public and private health service providers, thereby improving resource
efficiency.

88 PUBLIC AND PRIVATE

HOSPITALS
The recommendations of the NHHRC report are currently under consideration by
the Australian Government.

Private health insurance

Currently, around 45 per cent of Australians have private health insurance, which is
a relatively high rate compared to most other OECD countries (OECD 2003;
PHIAC 2009c). There are a number of possible reasons for Australia’s high rate of
coverage, including the desire for security and peace of mind expressed by those
who choose to take out insurance coverage (ABS 2009b). Australia’s high rate of
private health insurance coverage can also be attributed, in part, to three policy
measures: the Medicare Levy Surcharge, the Private Health Insurance Rebate and
Lifetime Health Cover. These measures are intended to encourage those with the
capacity to pay for their own private health insurance to do so, particularly at a
younger age, and to maintain their membership throughout their lifetime. These
policies are based on the principle that healthcare fees should be aligned, to some
degree, to patients’ capacity to pay. By encouraging the take-up of private health
insurance, these policies are also intended to help relieve demand on the public
hospital sector.

As part of the 2009-10 Budget, the Australian Government proposed a reduction in

the rebate for high income earners, on the rationale that the existing scheme
disproportionately favours high income earners and demands unsustainably high
levels of government funding1 (Australian Senate Community Affairs Legislation
Committee 2009). In 2007-08, Australian Government expenditure on private health
insurance rebates across all health services amounted to $3.6 billion, representative
of a 16 per cent increase from the preceding year (AIHW 2009c). If the proposed
rebate reduction does affect private health insurance membership rates, it may
redirect some patients from private to public election status and consequently shift
demand for certain hospital services such as elective procedures. However, the
Australian Treasury estimated that the proposed changes would not significantly
reduce private health insurance membership (cited by Australian Senate Community
Affairs Legislation Committee 2009). Other analysts have commented that that
Treasury’s estimates may be understated (Access Economics, commissioned by
CHA, 2009).

The NHHRC proposed that the role of private health insurance should be examined
alongside its proposed healthcare model, focusing on the way in which private

1 This includes rebates claimed through the taxation system and rebates paid directly to health
insurance funds which allow them to reduce premiums.
PUBLIC AND PRIVATE 89
HOSPITALS IN THE
HEALTH SYSTEM
health insurance would complement the new system of ‘health and hospital plans’,
and the potential impact of the new model on the viability of private health
insurance membership and expected premiums (NHHRC 2009).

FINDING 4.1

Although there is significant diversity within and across the public and private
hospital sectors in Australia, there are a number of key similarities between public
and private hospitals that enable and encourage comparison between the sectors. It
is acknowledged that there are some differences in the activities undertaken by
public and private hospitals and that the sectors do not always service a
comparable patient population, which makes comparisons more difficult.

90 PUBLIC AND PRIVATE

HOSPITALS
5 Hospital and medical costs
Key points
• Comparing the costs of public and private hospitals has been one of the most
challenging parts of this study because:
– existing data collections are limited by inconsistent collection methods and
missing information
– differences between hospitals in the types of patients treated and services
provided make like-for-like comparisons difficult.
• The Commission has sought to address these problems by drawing on various data
sources and incorporating adjustments to make the data more comparable, but the
resulting estimates should be considered experimental.
• The Commission’s experimental cost estimates suggest that, at a national level,
public and private hospitals had a broadly similar cost per casemix-adjusted
separation in 2007-08. However, significant differences were found in the
composition of costs:
– medical and diagnostics costs and prostheses costs were higher in private
hospitals
– capital costs were estimated to be somewhat higher in public hospitals, but this
result is particularly reliant on a range of data sources and adjustments to make
the data more comparable
– the combined cost of nursing and other salaries, allied health, operating rooms
and specialist suites, critical care, hotel costs, supplies, and on-costs were on
average higher for public hospitals.
• A disaggregation of the experimental cost estimates by diagnosis-related groups
(DRGs) suggests that in 2007-08:
– half of DRGs had an average cost in public hospitals that was more than 10 per
cent higher than in private hospitals, and one-fifth of DRGs had an average cost
in public hospitals that was at least 10 per cent lower than in private hospitals
– almost three-fifths of surgical DRGs had an average cost in public hospitals that
was at least 10 per cent higher than in private hospitals, and medical DRGs were
where public hospitals performed most strongly in terms of costs.
• To some extent, the results reflect differences between sectors in types of patients
treated and services provided, as there was a limit to which such differences could
be controlled for without advanced statistical methods. Multivariate techniques are
used in chapter 8 to take account of the many factors influencing performance.
• A foreshadowed shift to nationally-consistent activity-based funding for public
hospitals is expected to lead to more robust cost data. However, there remains
considerable scope to improve the quality and consistency of information on
hospital and medical costs for both public and private hospitals.

HOSPITAL AND 91
MEDICAL COSTS
The terms of reference direct the Commission to report comparative hospital and
medical costs for clinically-similar procedures performed by public and private
hospitals. This has been one of the most challenging parts of the study, particularly
in the short time available, because:
• existing data collections are limited by inconsistent collection methods and
missing information
• differences between hospitals in the types of patients treated and services
provided make like-for-like comparisons difficult.

Many study participants raised doubts about whether meaningful cost comparisons
were possible, given these difficulties (for example, Australian Healthcare and
Hospitals Association, sub. 33; Australian Medical Association, sub. 28; Catholic
Health Australia, sub. 20; NSW Department of Health, sub. 40; Royal Australasian
College of Surgeons, sub. 30; SA Department of Health, sub. 4; Tasmanian
Department of Health and Human Services, sub. 37).

Nevertheless, a number of participants acknowledged that the cost estimates were

consistent with their expectations (for example, NSW Department of Health,
sub. 41; Dr. John Deeble, sub. DR56; Catholic Health Australia, sub DR62).

It should also be noted that costs are a partial indicator of hospital performance,
since they do not include information on other aspects of performance, such as
quality and patient safety (Australian Medical Association, sub. 28). Nevertheless,
there is a strong case for monitoring and comparing hospital costs, given that
hospital services account for a large proportion of Australia’s health spending, and
competitive markets only have a limited role in driving efficiency improvements in
the health sector. Indeed, governments already participate in a number of
performance indicator frameworks that include the reporting of hospital costs,
particularly for public hospitals (AIHW 2009a; DOHA 2009a; NHPC 2004;
SCRGSP 2009). Unfortunately, these initiatives have yet to lead to comprehensive
and nationally-consistent reporting of hospital costs.

The Commission has sought to address data limitations, and take account of the
diversity and complexity of hospitals, by drawing on various data sources and,
where necessary, incorporating adjustments to make the data more comparable.
However, the Commission readily acknowledges that a number of significant data
shortcomings have limited its ability to construct fully-comparable costs. The
Commission therefore stresses that the cost estimates presented in this chapter
should be treated as experimental.

The next section describes the cost indicators used in this report. This is followed by
an overview of data sources and estimation methods. The resulting estimates are
92 PUBLIC AND PRIVATE
HOSPITALS
then presented. The chapter concludes with a discussion of data developments that
could improve the feasibility of future cost comparisons.

5.1 Cost indicators

Two commonly-used measures of hospital costs were estimated for this study:
• cost per casemix-adjusted separation — the average cost of treating a range of
different diagnoses, after controlling for differences in the complexity of
required treatments (casemix adjustment)
• cost per separation — the average cost of treating a group of diagnoses that are
clinically similar.

Clinically-similar diagnoses were defined according to the widely-accepted system

of Australian Refined Diagnosis-Related Groups (AR-DRGs).1 This classification
system provides a clinically-meaningful way of relating types of patients treated to
required resources (DOHA 2004). Individual DRGs represent a class of patients
with similar clinical conditions that require similar hospital services (AIHW 2009a;
Department of Health and Ageing, sub. 32).

The AR-DRG system only applies to acute-care admitted-patient services, and so it

was not possible to compare costs for other hospital services. Admitted-patient
services accounted for 71 per cent of the costs incurred by overnight acute-care
hospitals in 2007-08 (AIHW 2009a).2

Casemix-adjusted separations were calculated by weighting the number of

separations for each DRG by its relative complexity. In line with established
practice (for example, AIHW 2009a; DOHA 2009a), the relative complexity of each
DRG was measured by its cost weight — the average cost of the DRG across all
relevant hospitals divided by the average cost for all DRGs.

The grouping of similar outputs by DRG, and casemix adjustment when comparing
costs for more than one DRG, is an important step in making cost comparisons

1 The AR-DRG system is used by governments across Australia to measure and fund health
services, with its origin dating back to the early 1990s. It is managed by the Department of
Health and Ageing in consultation with state and territory health authorities, the Clinical
Casemix Committee of Australia, Clinical Classification and Coding Groups, and National
Centre for Classification in Health (DOHA 2004).
2 Victoria admits patients for treatments that other jurisdictions may administer as nonadmitted
(outpatient) services, such as chemotherapy and dialysis, and so a disproportionate share of
Victorian separations may be categorised as admitted-patient services (Victorian Department of
Health, pers. comm., 30 September 2009).
HOSPITAL AND 93
MEDICAL COSTS
more meaningful. As noted by study participants, the range and type of patients
treated by a hospital (casemix) will have a major influence on its costs (Australian
Nursing Federation, sub. 17; Catholic Health Australia, sub. 20; NSW Department
of Health, sub. 40; SA Department of Health, sub. 4; Tasmanian Department of
Health and Human Services, sub. 37).

Some participants were concerned that individual DRGs are not sufficiently
homogeneous to enable like-for-like comparisons (for example, Medical
Technology Association of Australia, sub. DR48; Queensland Health, sub. 27;
Tasmanian Department of Health and Human Services, sub. 37; Women’s and
Children’s Hospitals Australasia, sub. 21).

It is inevitable that any patient classification system will have some heterogeneity
within individual categories, as no single patient is precisely identical to another,
and so the question is whether such heterogeneity is significant and likely to
prejudice any cost comparison. The Commission notes that DRGs are sometimes
categorised by factors such as patient age and whether there are comorbidities, and
so it appears that these factors are to some extent controlled for. In addition, the
AR-DRG system has been refined over a period of more than a decade with input
from national, state and territory health departments so that only patients with
similar clinical conditions and resource requirements are grouped into the same
DRG (DOHA 2004).

5.2 Data sources and estimation methods

The terms of reference direct the Commission to use cost data that the states and
territories will provide to the Australian Government under the National Health
Agreement, and private hospitals already provide to the Government. However,
recent policy developments — such as a foreshadowed move to nationally-
consistent activity-based funding — have yet to lead to the reporting of all costs on
a comprehensive and consistent basis. The Commission therefore had to draw on
various existing data sources and, where necessary, incorporate adjustments to
make the data more comparable. These data sources and adjustments are
summarised below, with further details provided in appendix D.

Data sources

Most of the cost data were sourced from two data collections managed by the
Australian Government Department of Health and Ageing (DOHA) as part of its
regulatory and oversight functions:

94 PUBLIC AND PRIVATE

HOSPITALS
• National Hospital Cost Data Collection (NHCDC) — a voluntary annual survey
of hospitals, with the latest published results (2007-08) based on responses from
hospitals which accounted for 89 per cent of public acute separations and
72 per cent of private acute separations.
• Hospital Casemix Protocol (HCP) — a regular census of private health insurance
claims in public and private hospitals, collected as part of the regulation of
private health insurance. HCP data excludes both public patients and private
patients who do not make a private health insurance claim (who together
comprised around 90 per cent of separations in public hospitals and 20 per cent
in private hospitals in 2007-08).

A key difference between the collections is that the NHCDC has data on hospital
expenditure (costs), whereas the HCP has data on amounts charged to patients.

The NHCDC was used as the primary data source because it is designed for cost
analysis and covers a significant share of separations in both public and private
hospitals. The HCP was only used for private-patient medical and diagnostics costs,
as these are not captured in the NHCDC. Other data sources — such as surveys of
private hospitals by the Australian Bureau of Statistics and the National Hospital
Morbidity Database — were also used where NHCDC data are incomplete.

Study participants noted that the NHCDC is the best available data source for the
purpose of analysing costs, but also cautioned that it has major limitations (for
example, Australian Health Service Alliance, sub. 1; Australian Unity, sub. 31;
Catholic Health Australia, sub. 20). Some of the deficiencies of available cost data
are outlined in box 5.1.

Around 11 per cent of DRGs (less than 2 per cent of separations) in the 2007-08
NHCDC sample were excluded from the analysis because they had few separations
(episodes of care) in at least one sector, and/or involved less than three hospitals.

The Commission also obtained data from the Department of Veterans’ Affairs
(DVA) on the cost of treating a selected number of DRGs for veterans and their
dependants. While DVA data are not necessarily representative of the whole
population (Repatriation Commission, sub. 39), as a client of both the public and
private hospital sectors across Australia, DVA’s experience provides useful insights
into the relative performance of the two sectors. In 2007-08, DVA-funded patients
accounted for 4.1 per cent of all hospital separations (2.6 per cent of separations in
public hospitals and 6.4 per cent in private hospitals) (AIHW 2009a).

It should be noted, however, that the DVA data are based on prices negotiated
between DVA and the providers of hospital services, rather than the cost incurred

HOSPITAL AND 95
MEDICAL COSTS
by hospitals in providing those services. The extent to which there is a mark up over
costs could vary across jurisdictions for public hospitals and between different
operators of private hospitals. DVA cost results are presented in appendix D.

Box 5.1 Some of the deficiencies of existing cost data

There are inconsistencies within and between the public and private sectors in the
reporting of individual cost items in the National Hospital Cost Data Collection
(NHCDC) because data are provided without any auditing or reconciliation controls
(Department of Health and Ageing (DOHA), sub. 32). For example:
• depreciation is not reported for Victorian public hospitals
• building depreciation is not reported for public hospitals in Queensland
• there are differences in the extent to which jurisdictions and hospitals report
head-office overheads.
Another weakness of the NHCDC is that the user cost of capital is not explicitly
included.
There are also key differences in the way hospitals measure costs for the NHCDC:
• ‘Patient-costed’ sites measure costs at the patient level. This approach is used for
75 per cent of public-hospital cost data (DOHA, sub. 32).
• ‘Cost-modelled’ sites allocate aggregate costs across different items using
pre-determined service weights, which can be more than ten years old (Catholic
Health Australia, sub. 20). As a result, reported costs are probably not as accurate
as those from patient-costed sites. More than 90 per cent of private hospitals in the
NHCDC report cost-modelled data (DOHA 2009b).
The NHCDC data that DOHA provided to the Commission are based on an unweighted
sample, and so may not be representative of all hospitals. The data do cover a high
proportion of separations in both sectors, but some types of hospitals and regions —
such as large metropolitan hospitals — may be over-represented relative to others. For
example, the NSW Government has noted that the majority of NSW data in the
NHCDC are drawn from principal referral and major teaching hospitals, which on
average are more costly due to higher infrastructure costs and a more complex
casemix (DOHA 2009a). The SA Department of Health (sub. 4) estimated that teaching
accounted for over 5 per cent of total costs in SA metropolitan hospitals.
Items directly billed to patients — especially medical and diagnostics costs for private
patients — are not collected for the NHCDC. In such cases, charges data had to be
used from the HCP, which is a census of private health insurance claims covering all
hospitals. The Commission requested HCP data for only the hospitals in the NHCDC
sample, but received averages based on the full HCP census population. Thus, HCP
and NHCDC averages for different populations had to be used.

96 PUBLIC AND PRIVATE

HOSPITALS
Estimation methods

Overall costs were estimated by summing the various items that contribute to an
episode of care. Cost data on these items have varying degrees of accuracy and
comparability, and so the Commission distinguished between them using the six
categories listed in table 5.1.

Table 5.1 Cost componentsa

1. General hospital 2. Pharmacy 5. Capital

Ward nursingb Pharmacyc Depreciationd

Ward supplies & other overheads User cost of capitale
Allied health 3. Emergency

Critical caref Emergency departmentg 6. Medical & diagnostics

Operating rooms Incurred by the hospital
Specialist procedure suites 4. Prostheses Ward medicalh
Hotel costs Prosthesesi Imagingj
Non-clinical salariesk Pathologyl
On-costsm Billed directly to the patient
Medical chargesn
a Individual items are NHCDC cost buckets unless otherwise indicated. Details provided in appendix D.
b Excludes nursing salaries and wages reported in imaging, pathology, critical care, operating rooms,
emergency departments, specialist procedure suites, allied health, and pharmacy. c Excludes pharmacy costs
reported in critical care, operating rooms, specialist procedure suites, emergency departments, pathology, and
imaging. d Sourced from the NHCDC, except for Queensland and Victorian public hospitals, which were
derived from data published in SCRGSP (2009). e Derived from data published by the ABS (2008e) and
SCRGSP (2009). f Critical care covers intensive-care units and coronary-care units. g Emergency department
cost associated with patients who are subsequently admitted. h Excludes medical salaries and wages
reported in imaging, pathology, critical care, operating rooms, emergency departments, specialist procedure
suites, allied health, and pharmacy. i Excludes prostheses acquired directly by patients or their doctors, rather
than the hospital. j Excludes imaging costs reported in critical care, operating rooms, emergency departments,
specialist procedure suites, pharmacy, and pathology. k All other costs of service provision, but primarily other
salaries and wages, such as for patient-care assistants. l Excludes pathology costs reported in critical care,
operating rooms, emergency departments, specialist procedure suites, pharmacy, and imaging. m Includes
superannuation, termination payments, lump-sum payments, fringe-benefits tax, long-service leave, workers
compensation, and recruitment costs. Excludes items paid as part of a salary package, such as salaries and
wages, leave, allowances, and hotel costs. n Sourced from the HCP. Includes both medical and diagnostics
charges.

In order to compare hospital and medical costs for similar procedures performed by
public and private hospital systems it is necessary to combine a number of cost
components. In particular, to build a measure of the cost of an episode of care in the
private sector that is comparable to the cost of an episode of care in the public
system it is necessary to combine costs from different sources — private hospitals
are generally not responsible for the bulk of medical or diagnostics costs incurred in
a private episode of care, rather they are billed directly to the patients by the
specialists involved (Australian Health Services Alliance, sub. DR53; Australian
HOSPITAL AND 97
MEDICAL COSTS
Private Hospitals Association, sub. DR65). In this sense, this is a comparison of
costs associated with the public and private hospital systems.

The footnotes to table 5.1 are based on how costs are meant to be allocated
according to the NHCDC Hospital Reference Manual (DOHA 2008c). In practice,
jurisdictions do not always follow the NHCDC data specifications for public
hospitals, and compliance in the private sector can also be inconsistent. For
example, Victorian public hospitals record costs according to the methods of the
Clinical Costing Standards Association of Australia, and these are subsequently
mapped to the NHCDC cost structure. One of the consequences of this is that
NHCDC data for ‘ward nursing’ in Victorian public hospitals include other ward
costs, such as consumables, lighting and cleaning, and non-ward costs for admitted
patients, such as hospital-in-the-home and maternity post-domiciliary nursing care.

FBT exemptions

The terms of reference require the Commission to take account of fringe-benefits

tax (FBT) exemptions when comparing costs. Public and not-for-profit private
hospitals can provide ‘capped’ fringe benefits up to a value of $17 000 per
employee without incurring FBT. In addition to the capped exemption, public and
private not-for-profit hospitals are also eligible for the uncapped meal-entertainment
exemption. The meal-entertainment exemption has the potential to confer large
benefits to individuals employed by eligible institutions, and is likely to affect
resource allocation and employee behaviour (PC 2009).

In order to match the post-tax salary package a given employee receives in a public
or not-for-profit private hospital, it is therefore necessary for for-profit hospitals to
incur an FBT liability that other hospitals are exempted from (Australian Health
Insurance Association, sub. 18).

To facilitate like-for-like comparisons, the Commission estimated the cost that for-
profit hospitals incur by not having access to the capped FBT exemption, and
reduced their reported costs accordingly. Where the labour costs of for-profit
private hospitals could be identified in the NHCDC, they were reduced by around
1.4 per cent. It was not possible to estimate the cost of the uncapped meal-
entertainment exemption due to a lack of information about its use.

Capital costs

The Commission was also directed to take account of the capital costs of hospitals.
This has two elements — depreciation and the user cost of capital (UCC).

98 PUBLIC AND PRIVATE

HOSPITALS
Depreciation is the reduction in an asset’s value due to usage and obsolescence. The
NHCDC has DRG-level data on depreciation, except for public hospitals in Victoria
and Queensland.3 To address this data deficiency, aggregate public hospital
depreciation reported by the Victorian and Queensland governments (and published
by SCRGSP 2009) was apportioned across DRGs according to the pattern observed
in other jurisdictions. It was not possible to apply a similar adjustment to cost
estimates by region and hospital size, and so public hospital depreciation is
understated for these disaggregations.

The UCC is the opportunity cost of the capital used to deliver services. That is, the
return that could be generated if the funds tied up in the capital were employed in
their next best use. The UCC for public hospitals was based on the methodology
and asset data that jurisdictions have for some years contributed to for national
reporting of public hospital costs under the auspices of the Steering Committee for
the Review of Government Service Provision (SCRGSP 2009).4 A similar method
was applied to private hospitals, using asset values estimated from investment and
depreciation data collected by the ABS (2008e).

As detailed in appendix D, the capital cost estimates are particularly reliant on a

range of data sources and adjustments to make the data comparable. They could
underestimate capital costs in the public sector if asset data exclude public-private
partnership arrangements and contracting out of public-patient services to private
operators. It appears that assets are reported inconsistently between jurisdictions.
Capital costs could be underestimated in the private sector to the extent available
asset data exclude leased hospitals.

Administration and corporate overheads

In recent years, concerns have been expressed about a rapid increase in hospital
administrative staff relative to numbers of beds and treated patients (for example,
Sammut 2009). It is difficult to fully quantify the extent of this issue, but available
data do suggest that there has been some growth in administrative staff. Between
2001 and 2006, the number of medical administrators and nursing directors
employed across all areas of the health system grew by 69 per cent, compared to
23 per cent for all health workers (AIHW 2008c). Data presented in chapter 7
indicate that hospital administration and clerical workers per bed increased by

3 Queensland public hospital data include non-building depreciation, but does not cover
depreciation relating to buildings (DOHA 2009a).
4 The public-hospital asset data used were at a jurisdiction level. The Commission requested
hospital-level asset data from each jurisdiction. Such data was not provided by all jurisdictions,
and so it was not possible to use hospital-level data and maintain a consistent approach.
HOSPITAL AND 99
MEDICAL COSTS
8 per cent in public hospitals and 19 per cent in private hospitals between 2002-03
and 2007-08. As a further comparison, the number of available or licensed hospital
beds in Australia grew by 3 per cent, and the number of separations grew by
19 per cent, between 2000-01 and 2005-06 (AIHW 2006, 2009a).

It is not currently possible to separately identify the wages and salaries of

administrative staff in the NHCDC data, because administrative staff are often
included in the costs of their relevant work area, such as operating rooms,
pathology, and emergency departments. There may also be inconsistencies between
jurisdictions in how the cost of hospital administrative staff are allocated.

Furthermore, there are inconsistencies in the extent to which administrative and

head-office overheads are included in the NHCDC (section 5.5). For example, costs
associated with financial, payroll and human resource management services are not
included in public hospital costs in Queensland (DOHA 2009a). The Commission
understands that head-office costs are similarly excluded in public hospital costs
from New South Wales, Victoria, South Australia, Western Australia and the ACT.
In contrast, Tasmania includes a number of head office and administration costs, but
has reduced the extent to which they are included over recent NHCDC collections
in order to be more consistent with other jurisdictions (DOHA 2009a). The
Commission also understands that insurance costs for public hospitals are often
treated differently across jurisdictions — while all jurisdictions include costs
associated with workers’ compensation insurance, it is unclear whether other costs
such as medical indemnity, public liability and building and contents insurance
costs are accounted for in the NHCDC.

The extent to which hospital administration and head-office costs are included in
private hospital data is unclear. While it is expected that NHCDC cost data should
include shared costs where a hospital is part of a larger group, whether this extends
to head-office costs is not clear, beyond the exclusion of executive costs
(DOHA 2009a).

Pharmaceuticals

Accurate estimation of pharmaceutical costs incurred in public and private hospitals

is hampered both by incomplete coverage of pharmaceutical expenditure and the
attribution of pharmaceutical cost data to other sources. Ideally, pharmaceutical data
would provide a complete picture of the cost of all pharmaceuticals routinely
provided by hospitals in areas such as wards and operating theatres, the more
expensive highly-specialised drugs prescribed for treatments such as chemotherapy,
and other medicines obtained through prescriptions for individual hospital patients.
NSW Department of Health (sub. DR64) noted that all public-patient
100 PUBLIC AND PRIVATE
HOSPITALS
pharmaceuticals are covered in the NHCDC. The Commission understands that this
includes expensive highly-specialised drugs, where applicable.

However, the cost of medicines used to treat private hospital patients is not fully
captured in either the NHCDC or HCP (Dr. John Deeble, sub. DR56), and it is
unclear whether or not high-cost drugs used by patients in private hospitals would
be included in the NHCDC. Data published by the AIHW (2009d) suggest that
private hospitals have either substantially lower pharmaceutical costs, or up to 40
per cent of the pharmaceutical costs for patients in private hospitals are met by
external arrangements, such as the (publicly-funded) Pharmaceutical Benefits
Scheme.

Public hospital pharmaceutical costs are reported differently across jurisdictions.

The NHCDC cost bucket for ‘pharmacy’ is meant to include the costs of purchase,
production, distribution, supply and storage of drug products and clinical pharmacy
services, including salaries and wages in the pharmacy cost centre (DOHA 2008c).
Pharmacy costs reported in critical care, operating rooms, emergency departments,
pathology, imaging, and specialist procedure suites are not reported separately as
pharmacy costs. The Commission understands that Victoria is an exception in this
respect, and includes all pharmacy costs under the pharmacy cost bucket.

It is not currently possible to separately identify the total cost of pharmaceuticals

from the NHCDC data, given the allocation of pharmaceutical costs to other cost
areas and variations in jurisdictional approaches. However, the Commission has
obtained estimates of the amount of pharmaceutical costs included in other cost
buckets, for public hospitals. For those jurisdictions that allocate costs across
various different buckets, around 24 per cent of total pharmacy costs were included
in operating rooms and critical care costs in 2007-08 (appendix D).

FINDING 5.1

Existing datasets on hospital and medical costs are limited by inconsistent

collection methods and missing information. The Commission has sought to address
these limitations by drawing on various data sources and incorporating adjustments
to make the data more comparable where possible, as well as noting data
deficiencies where they exist. The resulting estimates of hospital and medical costs
should be considered experimental.

HOSPITAL AND 101

MEDICAL COSTS
5.3 Cost per casemix-adjusted separation

National and jurisdiction-level estimates

The Commission’s experimental cost estimates suggest that, at a national level,

public and private hospitals had a broadly similar cost per casemix-adjusted
separation in 2007-08 (table 5.2). There do, however, appear to be differences
between jurisdictions. In New South Wales and Victoria, private hospitals were
estimated to have a higher cost per casemix-adjusted separation than public
hospitals. In other jurisdictions, private hospitals were estimated to have a lower
cost per casemix-adjusted separation than public hospitals, with the gap particularly
large in Western Australia. To some extent, these differences between jurisdictions
may be attributable to inconsistencies in how each jurisdiction measures and reports
costs for the NHCDC.

There were significant differences between public and private hospitals in the
composition of costs. For general hospital costs, public hospitals were estimated to
be more costly than private hospitals ($2552 versus $1953 at the national level).
This was also the case with the experimental estimates of capital costs ($426 versus
$230). Conversely, average prostheses costs were estimated to be much lower in
public hospitals ($131 versus $542). Average medical and diagnostics costs were
also lower in public hospitals ($798 versus $1346). A similar pattern in the cost
components was evident at the jurisdiction level.

102 PUBLIC AND PRIVATE

HOSPITALS
Table 5.2 Cost per casemix-adjusted separation by jurisdiction and
sector, 2007-08a
Dollars
NSW Vic Qld SA
Cost component Public Private Public Private Public Private Public Private

General hospitalb 2 511 1 944 2 106 2 004 2 683 1 948 2 800 1 803
Pharmacy 164 42 235 87 174 45 146 53
Emergency 205 16 251 50 211 40 135 61
Medical & diagnosticsc 733 1 497 900 1 226 794 1 404 621 1 214
Prostheses 137 620 108 527 121 491 140 495
Capitald 439 210 359 240 560 223 381 158
Totale 4 189 4 330 3 960 4 133 4 543 4 151 4 223 3 783

WA Tas, NT & ACTf Australia

Public Private Public Private Public Private

General hospitalb 3 094 1 845 3 243 2 236 2 552 1 953

Pharmacy 202 144 186 55 187 68
Emergency 147 11 238 21 208 34
Medical & diagnosticsc 1 048 1 275 725 1 391 798 1 346
Prostheses 155 555 141 540 131 542
Capitald 359 281 447 345 426 230
Totale 5 006 4 111 4 980 4 586 4 302 4 172
a Costs are casemix adjusted using DRG-level cost weights for public and private hospitals combined.
b NHCDC cost buckets for ward nursing, non-clinical salaries, allied health, critical care, operating rooms,
ward supplies and other overheads, on-costs, hotel costs, and specialist procedure suites. c Combination of
data from the NHCDC (ward medical, imaging and pathology) and HCP (medical charges, which in the HCP
includes diagnostics). d Depreciation and the user cost of capital. e Totals may not equal sum of components
due to rounding. f Data for Tasmania, the Northern Territory and the ACT are aggregated to protect the
confidentiality of the small number of hospitals in each of these jurisdictions.
Source: Productivity Commission estimates.

General hospital costs

The relative significance of individual items within the general hospital category
differed between the public and private sectors (figure 5.1). To some degree, this
could reflect differences in how costs are allocated between different items, rather
than genuine variation in the composition of costs. The extent of this issue is
unknown. As noted later, some medical costs for public hospitals are captured in
this general hospital category, rather than the medical cost bucket.

HOSPITAL AND 103

MEDICAL COSTS
Figure 5.1 Composition of general hospital costs by sector, 2007-08a

Public hospitals Private hospitals

Other Ward nursing

19% 24%
Other
Ward nursing
27%
36%

Allied health
3%

Non-clinical Allied health

salaries 2%
7%

Non-clinical
salaries
Critical care 6%
11%
Critical care
9%
Operating rooms & Operating rooms &
specialist procedure specialist procedure
suites suites
24% 32%

a Care should be taken in interpreting this figure due to differences between jurisdictions in how public
hospital costs are reported. The differences may cause public-hospital nursing costs to be overstated, and
supplies and on-costs to be understated. Costs were casemix adjusted using a ‘general hospital’ cost weight
for this figure. ‘Other' comprises NHCDC cost buckets for ward supplies and other overheads, hotel costs and
on-costs.
Source: Productivity Commission estimates.

Nevertheless, general hospital costs in aggregate differed significantly between the

sectors. It was estimated that, in 2007-08, the general hospital cost per casemix-
adjusted separation was about 30 per cent higher in public hospitals compared to
private hospitals. Based on the reported data, it appears that this was largely due to
greater expenditure on ward nursing per separation in public hospitals. There was
also a sizeable difference in the average cost of critical care. Critical care covers
intensive-care and coronary-care units, and so the estimated difference in average
costs might reflect the fact that most of these units are in public hospitals.

Medical and diagnostics

Medical and diagnostics costs are incurred differently in the public and private
hospital systems. In the public sector, such costs generally relate to the wages and
salaries of doctors and specialists, whereas in the private sector they largely (though
not exclusively) consist of fees charged to patients by doctors and are not under the
control of the hospital in which treatment is performed (Australian Health Service
Alliance, sub. DR53; Australian Private Hospitals Association, sub. DR65).

104 PUBLIC AND PRIVATE

HOSPITALS
Medical and diagnostics costs were estimated to make up a greater proportion of the
total cost associated with the treatment of patients in private hospitals than they do
in public hospitals (32 per cent compared to 19 per cent), with the difference
estimated to be $548 at the national level. At a jurisdictional level, the gap between
public and private medical and diagnostics costs was estimated to range from as
little as $227 in Western Australia to $764 in New South Wales (table 5.2).

However, the estimated differences between public and private hospitals in medical
and diagnostics costs should be interpreted with care. Around one-third of public-
patient medical costs in the NHCDC are embedded in the general hospital and
emergency categories (estimated to be around $268 per separation nationally, as
detailed in appendix D). Hence, the experimental estimates overstate the cost
advantage that public hospitals have in medical and diagnostics, and the cost
disadvantage that public hospitals have in general hospital and emergency
departments.

Catholic Health Australia (sub. DR62) noted that the estimated medical and
diagnostics costs for public hospitals in New South Wales presented in the
Discussion Draft for this study was inconsistent with Catholic Health Australia’s
experience and unexpectedly low. Catholic Health Australia referred to an
interjurisdictional comparison of payments made to Visiting Medical Officers and
salaried and sessional staff (AIHW 2009a). It is worth noting that medical and
diagnostics costs associated with ‘ungroupable’ HCP separations have been
included in the Commission’s estimates since the publication of the Discussion
Draft, and this has increased the estimated medical and diagnostics costs for public
hospitals in New South Wales (appendix D).

It should also be noted that the Commission’s experimental estimates for medical
and diagnostics costs are broadly consistent with comments from other study
participants, who observed that:
• doctors in private practice tend to charge higher fees to private patients to
compensate for the lower earnings they receive from treating public patients in
public hospitals (often described as cross subsidisation) (Australian Health
Service Alliance, sub. 1; Australian Medical Association, sub. 28)
• there may be less incentive to limit medical costs in the private sector
(Prof. Richard Harper, sub. 6).

HOSPITAL AND 105

MEDICAL COSTS
Prostheses

As with medical and diagnostics costs, prostheses costs are actively managed and
borne by public hospitals, while private hospitals are limited in their ability to
control them:
[M]ost prostheses are actually purchased by the hospital and supplied to the patient by
the hospital — although the choice of prosthetic devices is made by the treating
doctors. (Catholic Health Australia, sub. DR62, p. 6)

At a national level, the cost of prostheses per casemix-adjusted separation was much
greater for procedures performed in private hospitals than in public hospitals. A
similar disparity was evident in all jurisdictions, but was greatest in New South
Wales and Victoria. The relatively high estimated cost of prostheses in private
hospitals is consistent with the views of participants (for example, Australian Health
Services Alliance, sub. 1). However, comments by CHA suggest that the estimated
disparity between sectors may be at least partly attributable to data deficiencies:
… the arrangements for the management and purchasing of prostheses in both sectors
are quite different and should be excluded from this particular study. In particular,
whilst the private sector has detailed prostheses billing data (a requirement for
reimbursement), this does not apply in the public sector where prostheses tracking is
less detailed and usually modelled using weights rather than actual utilisation. To put
this into perspective, prostheses can be over 20 per cent of costs in some hospitals,
depending on the casemix. (sub. 20, p. 9)

Nevertheless, the Commission considered it important to include prostheses costs in

this study, given that they can account for a large share of costs for some types of
treatment. In addition, the Commission has reported prostheses costs separately
from other items so that readers can see how they affect total cost per
casemix-adjusted separation.

Differences in prostheses costs should be considered carefully as they reflect not

only differential pricing of prostheses between the public and private sectors, but
also the effect of bulk-purchasing arrangements in the public sector, and a wider
range of often higher-priced products being available in the private sector
(Australian Private Hospitals Association, sub DR65; Catholic Health Australia,
sub. DR62; Medical Technology Association of Australia, sub. DR48).

At an aggregate level, participants suggested that the Commission’s method for

estimating prostheses costs may overstate the relative difference between public and
private sectors, due to there being proportionally fewer cases with significant
prostheses costs in the public sector (for example, Australian Health Services
Alliance, sub. DR54). This means that there are a large number of separations over
which the total prostheses expenditure is spread.

106 PUBLIC AND PRIVATE

HOSPITALS
In order to provide an indication of prostheses costs for procedures where they are
commonly used, the Australian Health Service Alliance (sub. DR53) suggested that
the prostheses cost per casemix-adjusted separation be calculated using only DRGs
with an average prostheses cost over $30 per separation. Using this approach, the
Commission estimated that the prostheses cost per casemix-adjusted separation —
using this constrained segment of separations and DRGs — was $790 and $1514 in
public and private hospitals respectively in 2007-08.5

Capital costs

The Commission’s experimental estimates suggest that capital costs account for a
notable proportion of total costs in both sectors, and so should be included in any
cost comparison. This is consistent with broader data published by the Australian
Institute of Health and Welfare (AIHW 2009c), which show that capital expenditure
on health facilities and investments accounted for 5.4 per cent of Australia’s total
health spending in 2007-08.

At a national level, the Commission’s experimental estimates suggest that capital

cost per casemix-adjusted separation in public hospitals was over 80 per cent higher
than in private hospitals. This is largely due to a significant difference between
sectors in the estimated UCC. The Commission’s experimental estimates suggest
that, in 2007-08, the UCC per casemix-adjusted separation was around $280 in
public hospitals, compared to around $100 in the private sector.

The Commission acknowledges that this result is particularly reliant on a range of

data sources and adjustments to make the data comparable, with the experimental
UCC estimates for both private and public hospital thought to be at the lower range
of the spectrum. This is particularly apparent with the private sector UCC estimates,
which rely on asset values estimated using a perpetual inventory model
(appendix D).

NSW Department of Health (sub. 41; sub. DR64) and Dr. John Deeble (sub. DR56)
favoured a different approach in which profits were used to measure the UCC for
private hospitals. Such an approach is likely to be misleading because many private
hospitals are run on a not-for-profit basis. A further problem is that it would confuse
profits recorded for accounting purposes with the economic concept of the UCC
(appendix D). Nevertheless, despite using a different methodology, both NSW

5 Using the Commission’s DRG-level cost estimates for the Discussion Draft, the Australian
Health Service Alliance (sub. DR53) estimated the prostheses costs per casemix-adjusted
separation for public and private hospitals to be $782 and $1512 respectively. These figures are
slightly different to those presented above due to revision of the cost estimates since the
Discussion Draft.
HOSPITAL AND 107
MEDICAL COSTS
Department of Health and Dr. Deeble appear to have reached a similar conclusion
to that found by the Commission. NSW Department of Health (sub. 41) estimated
that the average amount of capital used per bed in public hospitals is much higher
than in private hospitals ($388 000 versus $244 000 per bed). Dr. Deeble (sub.
DR56) agreed that capital costs would be significantly higher in public hospitals
compared to private hospitals.

The Commission’s results are also consistent with comments from other study
participants:
A presentation given by the Queensland Department of Health in 2008 … on the
development of the new Queensland Children’s Hospital indicates that the cost per bed
for the 360 public hospital beds is in the order of $3.055 million per bed or $14 763 per
square metre. This compares with current costs from the acute private hospital sector in
Queensland of around $5000 per square metre for high-cost areas such as operating
theatres and $3500 per square metre for areas such as patient wards/rooms and
administrative offices. The differential is in the order of 250–300 per cent. (Australian
Private Hospitals Association, sub. 25, p. 8)
Whilst it is always difficult to directly compare construction costs in the healthcare
environment, the UCH [UnitingCare Health] experience over the last two years has
shown that construction costs per bed (excluding equipment and professional fees) is
around $450 000. (UnitingCare Health, sub. 15, p. 2)

The Commission undertook a sensitivity analysis to test the robustness of its

conclusion on the relative size of capital costs in the public and private sectors. This
showed that capital costs were higher in public hospitals than private hospitals
under a range of plausible assumptions (appendix D).

Costs by region and hospital size

The previously noted pattern in the cost components was also evident by region and
hospital size (tables 5.3 and 5.4). That is, general hospital and capital costs were
lower in the private sector, whereas prostheses, and medical and diagnostics costs
were lower in the public sector.

Based on the estimates, it appears that total cost per casemix-adjusted separation
was essentially the same for public and private hospitals in major cities. Total cost
was also estimated to be similar across sectors in inner regional areas, and for very
large and large hospitals. However, private hospitals were estimated to have lower
costs than public hospitals when located in outer regional areas, and when of
medium to very small size.

108 PUBLIC AND PRIVATE

HOSPITALS
Table 5.3 Cost per casemix-adjusted separation by region and sector,
2007-08a
Dollars
Major cities Inner regional Outer regional
Cost component Public Private Public Private Public Private

General hospitalb 2 552 1 948 2 373 2 013 2 802 1 718

Pharmacy 192 71 181 49 155 33
Emergency 215 37 202 14 142 –
Medical & diagnosticsc 792 1 360 866 1 246 813 1 406
Prostheses 136 557 123 468 104 308
Capitald 362 229 227 193 254 143
Totale 4 249 4 204 3 972 3 983 4 270 3 609

Remotef Very remotef

Public Private Public Private

General hospitalb 3 712 .. 4 947 ..

Pharmacy 124 .. 209 ..
Emergency 316 .. 292 ..
Medical & diagnosticsc 707 .. 776 ..
Prostheses 11 .. 1 ..
Capitald 343 .. 381 ..
Totale 5 212 .. 6 607 ..
a Regions are based on ABS Australian Standard Geographical Classification (ASGC), Cat. no. 1216.0. Costs
are casemix adjusted using DRG-level cost weights for public and private hospitals combined. b NHCDC cost
buckets for ward nursing, non-clinical salaries, allied health, critical care, operating rooms, ward supplies and
other overheads, on-costs, hotel costs, and specialist procedure suites. c Combination of data from the
NHCDC (ward medical, imaging and pathology) and HCP (medical charges, which in the HCP includes
diagnostics). d Depreciation and the user cost of capital. e Totals may not equal sum of components due to
rounding. f There are no private hospitals in remote and very remote regions. .. Not applicable. – Nil or
rounded to zero.
Source: Productivity Commission estimates.

HOSPITAL AND 109

MEDICAL COSTS
Table 5.4 Cost per casemix-adjusted separation by hospital size and
sector, 2007-08a
Dollars

Very large Large Medium

Cost component Public Private Public Private Public Private

General hospitalb 2 591 2 007 2 408 1 922 2 361 1 836

Pharmacy 203 82 165 48 127 45
Emergency 209 48 241 21 183 5
Medical & diagnosticsc 816 1 393 762 1 324 840 1 310
Prostheses 139 614 110 500 124 405
Capitald 358 222 373 267 402 227
Totale 4 317 4 365 4 060 4 082 4 037 3 827

Small Very small

Public Private Public Private

General hospitalb 2 629 1 774 3 150 1 929

Pharmacy 105 42 102 87
Emergency 115 – 56 –
Medical & diagnosticsc 769 1 346 711 629
Prostheses 102 299 22 16
Capitald 436 212 465 224
Totale 4 157 3 673 4 507 2 885
a Hospital size defined by annual casemix-adjusted separations as follows: very large (more than 20 001),
large (10 001 to 20 001), medium (5001 to 10 000), small (2001 to 5000), and very small (up to 2000).
Casemix adjustment for the purpose of allocating hospitals to a size group was undertaken by DOHA using
separate cost weights for public and private hospitals. The casemix adjustment used by the Productivity
Commission to calculate costs was based on DRG-level cost weights for public and private hospitals
combined. b NHCDC cost buckets for ward nursing, non-clinical salaries, allied health, critical care, operating
rooms, ward supplies and other overheads, on-costs, hotel costs, and specialist procedure suites.
c Combination of data from the NHCDC (ward medical, imaging and pathology) and HCP (medical charges,
which in the HCP includes diagnostics). d Depreciation and the user cost of capital. e Totals may not equal
sum of components due to rounding. – Nil or rounded to zero.
Source: Productivity Commission estimates.

Hospitals that are very small, or located in a remote or very remote region, were
estimated to have relatively high costs per separation, even after casemix
adjustment. This is consistent with the view that remote and small hospitals face
additional costs because of their remoteness and/or inability to achieve the scale
economies of larger establishments in more densely populated regions. The
Tasmanian Department of Health and Human Services (sub. 37) cautioned that the
scale inefficiencies of small hospitals also tend to increase the cost of larger public

110 PUBLIC AND PRIVATE

HOSPITALS
hospitals, because the larger hospitals in a health network often provide
administrative support for smaller hospitals.

Some participants were concerned that inclusion of the relatively costly remote and
very remote hospitals in other comparisons — national, jurisdiction and hospital
size — biased the results in favour of private hospitals, since all remote and very
remote hospitals are in the public sector. However, this effect is likely to be
relatively minor and would not change the broad conclusions in this chapter. In the
2007-08 NHCDC sample provided by DOHA, only 5.4 per cent of hospitals and
1.6 per cent of separations in the public sector were in remote and very remote
regions (appendix D). For very small hospitals these proportions were larger —
15 per cent of hospitals and 21 per cent of separations for very small public
hospitals were in remote or very remote regions — but remote and very remote
establishments still accounted for a minority of the public-sector sample.

Excluding remote and very remote hospitals from the cost analysis would has a
limited impact on the overall cost per casemix-adjusted separation. It was estimated
that the cost per casemix-adjusted separation for public hospitals would decrease by
around $10 at a national level, and around $1 for Queensland, $16 for South
Australia, $4 for Western Australia, and $52 for Tasmania, the ACT and NT
combined.6 It is estimated that the cost per casemix-adjusted separation for small
and very small public hospitals would fall by $116 (2.8 per cent) and $224 (5 per
cent) respectively if remote and very remote hospitals were excluded from the
analysis.

Very large hospitals were estimated to have among the highest cost per
casemix-adjusted separation. This probably reflects the tendency of the largest
hospitals to treat patients with the most complex clinical conditions, maintain a
capability for major trauma events that is rarely used, and provide a large proportion
of clinical training.

6 Excluding remote and very remote hospitals does not change reported estimates for public
hospitals in New South Wales or Victoria, as the NHCDC sample had no hospitals in these
jurisdictions within the remote or very remote category.
HOSPITAL AND 111
MEDICAL COSTS
FINDING 5.2

The Commission’s experimental cost estimates suggest that, at a national level in

2007-08, public and private hospitals had broadly similar costs per
casemix-adjusted separation. There were, however, significant differences in the
composition of estimated costs:
• the combined cost of nursing and other salaries, allied health, operating rooms
and specialist suites, critical care, hotel costs, supplies, and on-costs were on
average higher for public hospitals
• medical and diagnostics costs were higher for private hospitals, although there
are some recognised constraints with available data not separately identifying
all medical costs in public hospitals
• prostheses costs were higher in the private sector, but this is also likely to reflect
a broader range of products being available for use in private hospitals
compared to the public hospital sector
• capital costs were estimated to be somewhat higher for public hospitals, but the
extent of this result is particularly reliant on a range of data sources and
adjustments to make the data more comparable.
These differences were also evident when the estimates were disaggregated by
jurisdiction, region and hospital size.

5.4 Average cost of individual DRGs

The experimental cost estimates for individual DRGs can be downloaded from the
Commission’s website at http://www.pc.gov.au/projects/study/hospitals. These
cover 592 DRGs and are available by jurisdiction, region and hospital size.

In summary, the DRG-level cost estimates suggest that, in 2007-08, many DRGs
had broadly similar costs in public and private hospitals. This is evident from
clustering around the 45 degree line in figure 5.2, which compares the cost per
separation for individual DRGs in the public and private sectors.

Nevertheless, it is also apparent that the cost in one sector relative to the other
varies between DRGs. This variation can be examined by using a measure of
relative cost — the ratio of cost per separation in public hospitals relative to that in
private hospitals. If a DRG has a public-private cost ratio of one, it indicates that
public and private hospitals have the same cost per separation. A ratio of less than
(more than) one indicates that the cost per separation is lower (higher) in public
hospitals relative to private hospitals.

112 PUBLIC AND PRIVATE

HOSPITALS
Figure 5.2 Comparison of cost per separation for individual DRGs in
public and private hospitals, 2007-08a

25

45 degree
line
20
Public hospitals ($'000)

15

10

0
0 5 10 15 20 25

Private hospitals ($'000)

a A point is located above (below) the 45 degree line if the relevant DRG has a higher (lower) cost per
separation in public hospitals than in private hospitals. DRGs with a cost per separation of more than $25 000
in at least one sector are not shown on the graph. These excluded DRGs accounted for less than 1.5 per cent
of separations among the sampled DRGs and hospitals.
Source: Productivity Commission estimates.

In figure 5.3, DRGs are ranked from the lowest public-private cost ratio to the
highest, and graphed against their cumulative share of all DRGs.

Given the experimental nature of the estimates, the Commission suggests that there
is no discernable difference in cost between sectors if the estimated cost of a DRG
in public hospitals is within 90 to 110 per cent of the cost in private hospitals (a
public-private cost ratio in the range of 0.9 to 1.1). Around 32 per cent of the
analysed DRGs were estimated to have an average cost per separation in public
hospitals that was within this range. These DRGs accounted for 29 per cent of
separations and 40 per cent of aggregate costs among the sampled DRGs and
hospitals.

HOSPITAL AND 113

MEDICAL COSTS
Figure 5.3 Cumulative distribution of DRGs ranked by public-private cost
ratio, 2007-08a

100
Cumulative percentage of DRGs

80

60

40

20

0
0.0 0.5 1.0 1.5 2.0 2.5
Public-private cost ratio
a The public-private cost ratio measures cost per separation for a DRG in public hospitals relative to that in
private hospitals. A ratio of one indicates that, for the relevant DRG, public and private hospitals have the
same cost per separation. A ratio of less than (more than) one indicates that the cost per separation is lower
(higher) in public hospitals. Three DRGs with a public-private cost ratio of more than 2.5 are not shown on the
graph. These excluded DRGs accounted for about 0.04 per cent of separations among the sampled DRGs
and hospitals.
Source: Productivity Commission estimates.

Around 18 per cent of the analysed DRGs were estimated to have an average cost
per separation in public hospitals that was at least 10 per cent lower than in private
hospitals (a public-private cost ratio of less than 0.9). These DRGs accounted for
22 per cent of separations and 20 per cent of aggregate costs among the sampled
DRGs and hospitals.

Around 50 per cent of the analysed DRGs were estimated to have an average cost
per separation in public hospitals that was more than 10 per cent higher than in
private hospitals. These DRGs accounted for 48 per cent of separations and
40 per cent of aggregate costs among the sampled DRGs and hospitals.

DRG partitions and Major Diagnostic Categories

Under the DRG classification system, individual DRGs can be grouped into the
‘partitions’ of surgical, medical or other. The experimental estimates suggest that
almost three-fifths of surgical DRGs had a cost per separation in public hospitals
that was at least 10 per cent higher than in private hospitals (figure 5.4). Medical
DRGs were where public hospitals performed most strongly in terms of cost relative
to the private sector, with 22 per cent of medical DRGs having a cost per separation

114 PUBLIC AND PRIVATE

HOSPITALS
in public hospitals that was at least 10 per cent lower than in private hospitals.
Nearly one third of medical separations occur in these DRGs.

Figure 5.4 DRG partitions by public cost relative to private cost, 2007-08a
100 Public cost per
separation at
Share of DRGs in partition (per cent)

least 10 per
80 cent higher
than private
cost
60
Public cost per
separation
within 10 per
40 cent of private
cost

20
Public cost per
separation at
least 10 per
0
cent lower
Surgical Medical Other All sampled than private
DRG partition DRGs cost

a Separations are assigned to the surgical, medical or other partitions on the basis of whether the separation
involves a procedure, and whether that procedure requires an operating room. A procedure is a clinical
intervention that carries a procedural or anaesthetic risk, and/or requires specialised training, facilities or
equipment. A separation is classified as surgical if it involves at least one operating-room procedure; medical if
there is no procedure; and other if it involves a procedure performed outside of an operating room, such as
dental extractions and colonoscopies.
Source: Productivity Commission estimates.

Around 55 per cent of the DRGs classified as ‘other’ were found to have no clear
cost difference (a cost per separation in public hospitals between 90 and
110 per cent of the cost in private hospitals).7 The strong performance of private
hospitals in surgical and other DRGs could reflect their tendency to specialise in
relatively routine procedures, whereas public hospitals have to provide a broader
range of services and manage the potentially disruptive effects of emergency
admissions (Queensland Health, sub. 27; SA Department of Health, sub. 4;
Tasmanian Department of Health and Human Services, sub. 37).8

7 A separation is classified as ‘other’ if it involves a procedure performed outside of an operating

room, such as dental extractions and colonoscopies.
8 Chapters 2 and 3 of this report discuss emergency departments in public and private hospitals
respectively.
HOSPITAL AND 115
MEDICAL COSTS
The DRG classification system also enables DRGs to be grouped into over
20 different Major Diagnostic Categories (MDCs). The Commission’s experimental
estimates suggest that, in 2007-08, cost per separation in public hospitals was:
• over 10 per cent higher than in private hospitals for almost half the MDC groups
• between 90 and 110 per cent of that in private hospitals for half the MDC groups
• over 10 per cent lower than in private hospitals for only one MDC (diseases and
disorders of the circulatory system).9

FINDING 5.3

A disaggregation of the Commission’s experimental cost estimates by

diagnosis-related groups (DRGs) suggests that in 2007-08:
• nearly one-fifth of DRGs had an average cost in public hospitals that was at
least 10 per cent lower than in private hospitals, and about half of DRGs had an
average cost in public hospitals that was more than 10 per cent higher than in
private hospitals
• almost three-fifths of surgical DRGs had a cost per separation in public
hospitals that was at least 10 per cent higher than in private hospitals, and
medical DRGs were where public hospitals performed most strongly in terms of
cost relative to the private sector.

Casemix complexity

A number of participants asserted that, where public hospitals have a higher cost per
casemix-adjusted separation than private hospitals, this is because public hospitals
have a more complex casemix (for example, Queensland Health, sub 27; Tasmanian
Department of Human Services, sub. 37).

To investigate this issue, the Commission examined the relative importance (in
terms of separations) of individual DRGs in each sector, and their associated cost
weight. As noted above, a cost weight measures a DRG’s average cost across all
hospitals, relative to the average cost for all DRGs across all hospitals. Cost weights
are commonly used as an indicator of the relative complexity of a DRG, and to
calculate an overall casemix-adjusted cost per separation.

The Commission estimate that the average cost weight for DRGs in public hospitals
was 0.96 in 2007-08, suggesting that the overall casemix of public hospitals is

9 Cost per casemix-adjusted separation was estimated for each MDC by dividing the total cost
over all separations in that MDC by the total number of cost-weighted separations in that MDC.
116 PUBLIC AND PRIVATE
HOSPITALS
slightly less complex than that of private hospitals.10 The average cost weight for
DRGs in private hospitals was estimated to be 1.09.

In figure 5.5, DRGs are ranked from the lowest cost weight to the highest in each
sector, and graphed against their cumulative share of separations. It can be seen
that, in 2007-08, public hospitals had a greater share of their workload in DRGs
with a relatively low cost weight. Around 25 per cent of separations in public
hospitals had a cost weight of less than 0.25, in comparison to 14 per cent in private
hospitals.

Figure 5.5 Cumulative distribution of separations in each sector ranked by

DRG cost weight, 2007-08a
Cumulative public separations Cumulative private separations
100
Cumulative separations (per cent)

80

60

40

20

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
DRG cost weight
a DRG cost weight is the average cost for each DRG relative to the average cost for all DRGs. DRGs with a
cost weight of more than four are not shown on the graph. There are 86 DRGs not shown, which accounted
for 3.1 per cent of public separations and 4.7 per cent of private separations in the NHCDC sample.
Source: Productivity Commission estimates.

Differences in average cost weights are largely driven by the different number of
separations in low-cost and high-volume DRGs, such as renal dialysis and
chemotherapy, that are experienced in the public and private sectors. When the
renal dialysis and chemotherapy DRGs are excluded from the calculations, the
average cost weights converge to equal 1.00 in both the public and private sectors.

10 This was calculated as a weighted average, where the weights were the number of separations
for each DRG.
HOSPITAL AND 117
MEDICAL COSTS
5.5 Improving future cost comparisons
As noted previously, the cost estimates in this report should be considered
experimental, since they are based on datasets with inconsistent collection methods
and missing information. In addition, there was a limit to which differences between
the public and private sectors — such as types of patients treated and services
provided — could be controlled for without using advanced statistical methods.
Multivariate techniques are used in chapter 8 to take account of the many factors
influencing hospital performance.

The remainder of this chapter outlines foreshadowed improvements to the collection

and reporting of cost data, and considers what other improvements could be made to
improve the feasibility of future comparisons of public and private hospitals.

Foreshadowed improvements to data collections

In late 2008, the Council of Australian Governments (COAG) made the National
Partnership Agreement on Hospital and Health Workforce Reform, in which
jurisdictions agreed to introduce a nationally-consistent approach to activity-based
funding (ABF) for public hospitals. (COAG 2008e). This will involve development
of a nationally-consistent costing and funding model for all care types and all
non-clinical hospital services, including teaching and research. The costing model is
to be built on the NHCDC.

The Australian Government committed $154 million from 2008-09 to 2012-13 to

fund the ABF initiative, with $133 million of this to be allocated to the states and
territories. It was agreed that the Australian Government would provide national
leadership and coordination for the development of classification systems and
costing models.

Specific performance targets were outlined for the development of ABF

infrastructure in the COAG partnership agreement:
• By 30 June 2011, 100 per cent of admitted-patient episodes are to be classified
and costed using a nationally-consistent model.
• By 30 June 2015, 100 per cent of emergency-department services, sub-acute,
outpatient services and hospital-auspiced community health services are to be
classified and costed using a nationally-consistent model.
• If agreed by COAG, by 30 June 2016, 100 per cent of admitted-patient episodes,
emergency department, sub-acute outpatient services and hospital-auspiced

118 PUBLIC AND PRIVATE

HOSPITALS
 community health services are to be funded through a nationally-consistent ABF
model.

These developments would go a long way towards addressing the problems that the
Commission has encountered with existing cost data for public hospitals. It would,
however, be desirable for all private hospitals to report cost data using the same
methodology as public hospitals. It is notable therefore that the COAG partnership
agreement gives the Australian Government responsibility, in collaboration with
states and territories, to engage with the private sector to improve the comparability
of performance between the public and private sectors. DOHA (sub. 32) noted that
steps have already been undertaken in this regard.

However, at this stage it appears that participation in the NHCDC will remain
voluntary for private hospitals. It is unlikely that this will ensure that the quality and
comparability of private hospital data improves. Catholic Health Australia,
representing a large proportion of private hospitals, noted that it:
… supports compulsory participation of the private sector in contributing to the cost
data collections and for data input into these collections to be made consistent across all
jurisdictions and between the public and private sectors. (sub. DR62, p. 2)

The Australian Health Service Alliance (sub. 1) suggested that consideration be

given to making it mandatory for all hospitals, or at least all hospitals other than the
very small ones, to provide data to the NHCDC.

Participants suggest that there may be scope to rationalise and improve consistency
with other reporting requirements — including to the HCP as part of the regulation
of private health insurance — so that there is not a major additional reporting
burden on private hospitals from being required to participate in the NHCDC
(Australian Private Hospitals Association, sub. 25; Catholic Health Australia,
sub. DR62; DOHA, sub. 32). DOHA noted:
… anecdotal evidence suggests that some private facilities may contribute to as many
as nine different collections with different formats and requirements, and those that
have facilities in different jurisdictions may have up to eight different reporting regimes
to comply with for very similar information. (sub. 32, p. 28)

Similarly, the Australian Private Hospitals Association (sub. 25, p. 6) called for a
rationalisation of the ‘existing plethora of regulation and reporting requirements
imposed on private hospitals’.

Catholic Health Australia (sub. 20; sub. DR62) advocated the establishment of an
Office of Hospital Cost Data within DOHA to oversee a nationally-consistent data
collection for both public and private hospitals. It also recommended that, with the
implementation of ABF for public hospitals, Commonwealth funding to the states
HOSPITAL AND 119
MEDICAL COSTS
should be made contingent on them providing data that are consistent and high
quality. In addition, an independent data audit agency was proposed to ensure the
quality of submitted data.

Capital costs

The Commission’s experimental estimates for capital costs are particularly reliant
on a range of data sources and adjustments to make the data comparable. While the
precise value of capital costs remains open to debate, the Commission considers that
this item is likely to account for a notable share of total costs and so should be
included in data collections.

The shift to a nationally-consistent data collection for ABF purposes may partially
address problems with the reporting of depreciation, as this item is already covered
by the NHCDC, but it is unclear what will be done to measure the UCC. This will
require consistent approaches to measuring asset values, which is currently not the
case, particularly for public hospitals. For private hospitals, the total value of assets
is not currently available, and so the Commission had to estimate it from investment
and depreciation data collected by the ABS (2008e). Thus, there is a strong case for
including asset values, reported on a consistent basis, for both public and private
hospitals in the NHCDC. This would need to include assets subject to public-private
partnership arrangements, contracting out of public-patient services to the private
sector, and the use of leased assets.

Items billed directly to patients

It is unclear how the move to a nationally-consistent data collection for ABF

purposes would address problems with identifying medical and diagnostics costs for
private patients. The Commission has had to rely on HCP data for these items, but
public hospitals often fail to assign private-patient costs to DRGs when coding HCP
data. To address this problem, it was necessary to apportion unassigned medical
costs across DRGs by using separation data for private patients in public hospitals
and private hospital cost weights (appendix D).

It would be desirable for all costs associated with an episode of care — including
those directly billed to patients, such as medical, diagnostic and pharmaceutical
costs — to be captured in the new national cost collection.

120 PUBLIC AND PRIVATE

HOSPITALS
Prostheses

The Commission’s experimental estimates suggest that the cost of prostheses in

public hospitals is considerably lower than in private hospitals. As noted previously,
this is consistent with the expectations of study participants, but there may also be
problems with how prostheses costs are currently reported. The NHCDC data
obtained by the Commission seem to have noticeably different prostheses costs
across separations within a particular DRG and sector. The Commission
understands that changes are being considered for the next round of the NHCDC to
address some of the problems with the reporting of prostheses costs.

Hospital administration costs and head-office overheads

As noted previously, concerns have been expressed in recent years about a rapid
increase in hospital administrative staff relative to numbers of beds and treated
patients. As noted above, it was not possible to separately identify the wages and
salaries of administrative staff in the NHCDC data, because administrative staff are
often included in the costs of their relevant work area, such as operating rooms,
pathology, and emergency departments. There are also inconsistencies between
jurisdictions in how the costs of administrative staff are allocated. There are
opportunities for improvements in data to respond to these identified deficiencies.

Study participants noted that head-office overheads — such as for centralised

procurement of supplies and provision of information-technology services —
should also be included in cost comparisons as they are part of the cost of supplying
hospital services (for example, Australian Health Insurance Association, sub. 18;
Australian Medical Association, sub. 28; Australian Private Hospitals Association,
sub. 25). There are currently inconsistencies between jurisdictions/hospitals in the
extent to which head-office overheads are reported in the NHCDC, and this is not
clearly documented. The Commission understands that the Technical Reference
Group for the NHCDC — comprising representatives from the jurisdictions and
private hospital groups — has been considering an approach to collect head-office
overheads consistently in future cost data collections.

FBT exemption

As noted above, the Commission has attempted to achieve like-for-like comparisons

by removing the FBT liability that for-profit private hospitals incur due to not
having access to the FBT exemption available to public and not-for-profit hospitals.
This involved adjusting the NHCDC cost data for ward medical, ward nursing, and
non-clinical salaries. It is possible that this adjustment only partially accounts for
HOSPITAL AND 121
MEDICAL COSTS
the additional FBT burden faced by for-profit hospitals, since some labour costs are
incorporated into other NHCDC cost buckets. Ideally, the adjustment would be
applied to the NHCDC cost bucket for on-costs, as this is where FBT is supposed to
be included by reporting hospitals (DOHA, sub. 32). However, the Commission did
not have access to data on the proportion of on-costs that were FBT payments.

Pharmaceuticals

As noted above, the cost of medicines used to treat hospital patients is not fully
captured by the NHCDC. Ideally, the data would include the cost of medicines
routinely provided by hospitals in areas such as wards and operating theatres, the
more expensive highly-specialised drugs prescribed for treatments such as
chemotherapy, and other medicines obtained through prescriptions for individual
hospital patients (including those currently subsidised under the Pharmaceutical
Benefits Scheme and provided through community pharmacies).

FINDING 5.4

A foreshadowed shift to nationally-consistent activity-based funding for public

hospitals is expected to eventually lead to more robust cost data for the public
sector. However, there remains considerable scope to improve the quality and
consistency of hospital and medical cost data in Australia. In particular, there is a
need for:
• private hospitals to report cost data using the same methodology as public
hospitals, and to continue to have a high level of participation in the National
Hospital Cost Data Collection, so that the data are reliable and can be
disaggregated by sector, region, and size and type of facility
• items directly billed to private patients — such as some medical, diagnostics and
medicines — to be linked with cost data reported by hospitals so that all costs
associated with an episode of care are captured in a single collection
• reliable data on capital costs, hospital administration costs, head-office
overheads, and the cost of medicines prescribed to hospital patients
• quantification of the additional FBT liability that for-profit hospitals incur by
not having the FBT exemption that is available to other hospitals.
This may require a strengthening of data-related provisions in the National
Healthcare Agreement for public hospitals, and data-reporting requirements for
private hospitals. If this is the case, governments need to be conscious of the
regulatory burden on reporting hospitals and, where possible, seek to limit it by
avoiding duplication and inconsistency in reporting arrangements, and by utilising
cost-effective electronic reporting of data.
122 PUBLIC AND PRIVATE
HOSPITALS
6 Hospital-acquired infections

Key points
• Infections are the most common complication affecting hospital patients, and in
many cases are preventable. A recent study estimated that Australia has 180 000
hospital-acquired infections annually and these occupy almost two million bed days.
• A simple comparison of infection rates across hospitals may not provide an accurate
indication of the potential for performance improvements and associated benefits
because:
– the risk of infection depends on patient characteristics (such as age) and types of
treatments provided (such as surgery), and these vary between hospitals
– there are many different types of infections, with varying degrees of prevalence
and potential harm.
• In order to take account of this diversity, it is common to limit comparisons to groups
with a similar risk of infection (such as patients in intensive-care units) and
distinguish infections by organism (such as Staphylococcus aureus) and body
location (such as surgical sites). However, these methods do not remove all of the
factors outside the control of a hospital that can cause its infection rate to differ from
other establishments.
• A further problem is that Australia does not have a robust nationally-consistent data
collection for comparing hospital-acquired infections. The currently available
national data were not designed for cross-hospital comparisons, and may be
affected by issues such as sample-selection bias and unaudited self-reporting.
• Data collected by state governments as part of their infection-surveillance programs
suggest that private hospitals have lower infection rates than public hospitals.
However, this result could be misleading because private hospitals generally treat
patients who have a lower risk of infection, and the data do not fully control for this.
• Foreshadowed developments — such as performance reporting under the National
Healthcare Agreement — will move Australia closer to a robust nationally-consistent
data collection on hospital-acquired infections. However, there is scope for further
reforms, such as including private hospitals in national reporting arrangements.
• The Australian Commission on Safety and Quality in Health Care is leading and
coordinating initiatives that should improve the feasibility of future infection-rate
comparisons.

HOSPITAL-ACQUIRED 123
INFECTIONS
The terms of reference ask the Commission to compare the rate of hospital-acquired
infections in public and private hospitals, disaggregated by type of infection. This is
an important indicator of service quality because infections are the most common
complication affecting hospital patients, and in many cases are preventable.
Hospital-acquired infections also place a significant burden on the health system,
with an estimated 180 000 cases in Australia each year that occupy almost two
million bed days (Graves, Halton and Robertus 2008).

In this chapter, the Commission identifies the key types of hospital-acquired

infections; considers how infection rates should be compared between hospitals;
undertakes a comparison between public and private hospitals using available data;
and outlines developments to make future comparisons more feasible and robust.

6.1 Types of hospital-acquired infections

In 2008, the Australian Commission on Safety and Quality in Health Care
(ACSQHC) published a detailed report on healthcare-associated infections
(Cruickshank and Ferguson 2008). The study observed that, in order to make
monitoring tractable, the approach has generally been to focus on infections of
greatest concern, based on prevalence and/or harm to individual patients. The key
types of infection are usually defined by organism, physical location and/or patient
population:
• Organisms that have developed resistance to standard antibiotics — such as
methicillin-resistant Staphylococcus aureus (MRSA), vancomycin-resistant
enterococci (VRE), and multiresistant gram-negative bacteria — are often the
focus of reporting regimes because they have the greatest impacts on health
systems in developed countries. Another organism that is gaining prominence is
Clostridium difficile.1
• Physical locations of greatest concern, and hence often reported, are surgical
sites and the bloodstream.
• Patient populations considered to have a high risk of acquiring and/or
experiencing significant harm from an infection — such as newly-born infants
and patients in intensive-care units (ICUs) — are also often a priority for
monitoring programs.

The ACSQHC report recommended that all hospitals should monitor

hospital-acquired infections because this, combined with timely feedback, has been
proven to reduce infection rates. The report identified major deficiencies in

1 Further details about these organisms are provided in the following parts of section 6.1.

124 PUBLIC AND PRIVATE

HOSPITALS
Australia’s existing monitoring arrangements and recommended strengthened
surveillance of high-priority organisms, locations and populations (detailed later in
this chapter).

Issues associated with high-priority organisms, locations and populations are

outlined below because they have implications for how infection rates should be
compared across hospitals.

Infection organisms

Staphylococcus aureus

Staphylococcus aureus — sometimes termed golden staph — is a type of bacteria

often found on the skin and in the nose with no apparent ill effects. People with this
condition are said to be colonised (bacteria present, but not causing an infection).
However, if Staphylococcus aureus enters the body through broken skin, such as
surgical wounds, it can cause an infection and require treatment with antibiotics.
Staphylococcus aureus is responsible for the largest proportion of
healthcare-associated bacterial infections (Cruickshank and Ferguson 2008).

Staphylococcus aureus is usually spread by direct skin contact (typically via hands)
with a person who is infected or colonised, or through contact with shared items,
such as towels and shared surfaces like door handles, taps and benches.

A distinction is often made between methicillin-sensitive Staphylococcus aureus

(MSSA) and MRSA.2 In most Australian hospitals, MSSA remains a more common
cause of healthcare-acquired infections, and is likely to have a larger impact on the
health system than MRSA (Christiansen et al. 2008). However, MRSA is not
uncommon and can be more serious for individual patients because it has developed
resistance to many antibiotics.

MRSA was traditionally associated with hospital admissions. However, in recent

years community-acquired MRSA — involving people who have not recently been
hospitalised or had a medical procedure — has been a growing problem (Nimmo et
al. 2006). MRSA infections in the community are usually manifested as skin
infections, such as pimples and boils, and occur in otherwise healthy people (SA
Department of Health 2008).

2 Resistance to methicillin is traditionally used as an indicator of antibiotic resistance because it

used to be the drug of choice for treating Staphylococcus aureus infections. MRSA can be
subdivided into organisms that are multiresistant (resistant to multiple antibiotics) and not
multiresistant (only resistant to methicillin).
HOSPITAL-ACQUIRED 125
INFECTIONS
Research undertaken for the ACSQHC found that MRSA infections were the
second-most costly adverse event per patient (after post-procedure endocrine and
metabolic disorders), adding an extra $19 892 to the cost of an episode of care
(based on 2005-06 data for Victoria and 2006-07 data for Queensland) (Jackson,
Ngheim, Rowell, Jorm and Wakefield 2009). The next most costly adverse event
was Clostridium difficile that caused enterocolitis (inflammation of the colon and
small intestine), costing $19 745 per episode.

Clostridium difficile

Clostridium difficile is an anaerobic toxin-producing bacterium that has an ability to

form spores, which enable it to survive in the environment for extended periods of
time (Thomas and Riley 2003). The organism usually causes diarrhoea and is the
most common cause of healthcare-associated gastrointestinal infection
(TIPCU 2009). A new virulent strain emerging in North America and Europe has
led to epidemics and extensive mortality (Cruickshank and Ferguson 2008).

The main risk factor for Clostridium difficile colonisation and infection is prior
exposure to antibiotics, possibly because antibiotics disrupt the normal balance of
bacteria and other micro-organisms in the gut, allowing Clostridium difficile to
spread (Thomas and Riley 2003).

Most cases of Clostridium difficile occur in hospitals or long-term care facilities.

Transmission usually occurs through shared equipment, a contaminated
environment or the hands of healthcare workers. The organism can be readily
cultured from inanimate environmental sources such as beds, cupboards, floors and
walls, as well as from the hands of healthcare workers caring for patients with a
Clostridium difficile infection. The impact of Clostridium difficile on the health-care
system is considerable, with patients requiring additional infection-control
precautions and specific treatment, and can spend an extra 1 to 3 weeks in hospital
(McGregor, Riley and Van Gessel 2008).

The ACSQHC has recommended that surveillance systems for Clostridium difficile
be established nationally (Cruickshank and Ferguson 2008; TIPCU 2009).

Vancomycin-resistant enterococci

Vancomycin is an antibiotic used to treat infections caused by enterococci, which

are bacteria normally residing in the bowel without causing any illness. However,
enterococci can invade other parts of the body and cause an infection. The most

126 PUBLIC AND PRIVATE

HOSPITALS
common sites of enterococcal infection are the urinary tract, wounds, blood and
heart lining (SA Department of Health 2009b).

Some enterococci have become resistant to vancomycin. These vancomycin-

resistant enterococci (VRE) infections are harder to treat because of their antibiotic
resistance. VRE infections are dangerous for people with a weakened immune
system, but most recover with appropriate antibiotic treatment. However, the cost of
containing VRE outbreaks in health facilities can be considerable. A 2003 outbreak
in a large WA teaching hospital was brought under control over a three-month
period at a cost of $2.7 million (Christiansen et al. 2004).

VRE infections are typically spread by physical contact with faeces, or skin or
objects that have been contaminated with VRE. This includes contact with
contaminated hands, hospital equipment, bathroom taps and door handles. Hand
washing is one of the best ways to prevent the spread of VRE. Regular cleaning of
frequently-touched surfaces is also important because VRE can survive in the
environment for a long time (SA Department of Health 2009b).

VRE colonisation of patients is more common than infection. It is estimated that for
every patient detected as having VRE, there will be at least ten others in an
institution who are colonised. Colonised patients and their immediate physical
environment act as reservoirs for the ongoing transmission of VRE within hospitals
(Mascinie and Bonten 2005; TIPCU 2009).

Multiresistant gram-negative bacteria

Gram-negative bacteria are a group of organisms that can be identified by using a

‘gram-staining’ test. These bacteria are of concern because they can have a
highly-transmissible resistance to antibiotics (Christiansen et al. 2008).

Mortality rates of up to 100 per cent can occur from gram-negative bacteria
infections if the bacteria are resistant to multiple antibiotics (that is, multiresistant
gram-negative bacteria) and are not treated with an antibiotic that is active against
the organism. The risk of experiencing a multiresistant gram-negative bacteria
infection is greater for people admitted to a teaching hospital, treated in an ICU,
having other medical conditions, being treated with a central intravenous line or
urinary catheter, having a longer stay in hospital, and being previously treated with
antibiotics (Christiansen et al. 2008).

HOSPITAL-ACQUIRED 127
INFECTIONS
Infection locations

Surgical-site infections

Surgical-site infections (SSIs) result from a range of organisms. Skin flora — such
as Staphylococcus aureus and coagulase-negative staphylococci — are most often
responsible for SSIs that follow ‘clean procedures’. SSIs from ‘contaminated
procedures’ can be associated with polymicrobial infection and flora normally
found in the viscus that is opened, such as gram-negative infections following rectal
surgery. The range of organisms causing SSIs is also influenced by the choice and
timing of prophylactic antibiotics prior to surgery (Bull et al. 2008).

SSIs can be difficult to monitor because more than 50 per cent become apparent
after discharge from hospital, and any associated readmission may not be to the
establishment where the surgery occurred. For example, SSIs can occur up to four
weeks after deep-incisional surgery, and up to 12 months after joint-replacement
surgery (HQCC 2009).

The risk of experiencing an SSI, and the associated adverse impacts, differ between
surgical procedures and according to patient characteristics. For example,
Bull et al. (2008) noted that infection rates tend to be low for major-joint prosthesis
replacements, but the consequences of infection are significant. The patient may
require further surgery, removal of the prosthetic joint, replacement with another
joint, and months of intravenous antibiotic therapy, followed by oral antibiotic
therapy. In other procedures, such as caesarean sections, infection rates tend to be
much higher, but the consequences are less severe and may not even require
readmission to a hospital.

The ACSQHC recommended routine local surveillance of SSIs, including coronary

artery bypass graft surgery, major-joint prosthesis insertion, and other procedures
that have higher-than-expected SSI rates at the local level (Cruickshank and
Ferguson 2008).

Bloodstream infections

Bloodstream infections (BSIs) occur when the blood contains bacteria (in which
case the infection is termed a bacteraemia) or fungi (fungaemia).
Collignon et al. (2008, p. 53) noted that BSIs can have significant adverse impacts:
Studies in Australia document that 17–29 per cent of patients with hospital-acquired
BSIs die while still in hospital. Patients who develop BSIs are also more likely to suffer
complications during their hospital stay that result in a longer hospital stay and an
increased cost of hospitalisation.
128 PUBLIC AND PRIVATE
HOSPITALS
Staphylococcus aureus bacteraemia (SAB) is the most common type of
healthcare-associated BSI (Collignon et al. 2008). SAB cases are detected when
Staphylococcus aureus is isolated in a blood culture. An often-cited study by
Collignon et al. (2005) estimated that Australia has around 7000 SAB cases per
year. The authors concluded that approximately one-half of all SAB cases were
hospital acquired, and a further one-sixth were linked to healthcare procedures in
other settings. The remaining one-third were deemed to be community-acquired
BSIs.

MSSA strains are responsible for the largest share of hospital-acquired SAB cases,
but MRSA incidence is significant. Collignon et al. (2005) estimated that MRSA
accounted for around 40 per cent of Australian hospital-onset episodes of SAB.
Overseas evidence indicates that the median death rate for MSSA SAB infections is
25 per cent, and for MRSA SAB infections is 34 per cent (Cosgrove et al. 2003).

Most healthcare-related SAB cases are attributed to intravascular catheters. SAB

rates are therefore important markers of intravenous catheter management and the
effectiveness of hand hygiene within a hospital or institution. Catheter-associated
BSIs are a particular problem in intensive-care patients and immunocompromised
patients who depend on artificial vascular access.

BSIs used to be mainly acquired during hospitalisation, but this situation is

changing as increasing numbers of people are managed at home with intravascular
catheters, have medical procedures performed as outpatients, or are discharged early
from hospitals with percutaneous (through-the-skin) invasive medical devices in
place.

The ACSQHC recommended mandatory reporting by hospitals of SAB BSIs,

central-line associated BSIs in ICUs, and haemodialysis-access-associated BSIs
(Cruickshank and Ferguson 2008).

Patient populations

As noted above, certain populations are more likely to acquire a particular type of
infection. For example, Staphylococcus aureus is a greater concern for those who
have undergone major surgery. Other patient populations that are often the focus of
infection indicators are newly-born infants and ICU patients. They have a relatively
high risk of acquiring an infection and experiencing significant harm, including
possibly death. The ACSQHC recommended routine monitoring of bacterial sepsis
for babies during the first week of life (Cruickshank and Ferguson 2008).

HOSPITAL-ACQUIRED 129
INFECTIONS
6.2 How should infection rates be measured and
compared?
The measurement and comparison of infection rates is not straightforward. A
hospital could have a relatively high infection rate simply because its workload is
concentrated on services and patients with a high risk of infection. It is also
desirable to take account of heterogeneity between different infection organisms
and their location in the body. A hospital could have the same rate of total
infections as its peers, but the infection organisms and locations could be far more
serious.

Data presented in earlier chapters show that the types of services provided and
patients treated can differ markedly between hospitals, both within and between the
public and private sectors. It therefore follows that a comparison of infection rates
in public and private hospitals could encounter difficulties in separating the effects
of casemix differences from genuine differences in the performance of hospitals in
reducing and managing the rate of infections. The difficulty of the task was noted
by several study participants:
Meaningful comparison of rates of HAIs [hospital-acquired infections] will be difficult,
and needs to test whether there are fundamental differences between the public and
private hospital casemix. (ACSQHC, sub. 24, p. 5)
While hospital-acquired infections are an important indicator of quality they must be
compiled and assessed carefully lest they mislead. Some common problems include
comparing hospitals with significantly different casemix, considering so many infection
indicators that any analysis is too granular and drawing conclusions that are not
statistically robust. On top of this there [is] what is potentially the most confounding
influence of all — is the relevant data complete and accurate? (Australian Health
Service Alliance, sub. 1, p. 6)

A significant amount of work has been done overseas to implement methodologies

that enable hospitals to compare their infection rates from one period to the next,
and against best practice in other hospitals. This has been prompted by research that
shows that the monitoring of hospital-acquired infections, combined with a
prevention program that uses the data to drive improvements, can lead to lower
infection rates (Clezy et al. 2008).

The US National Nosocomial Infections Surveillance (NNIS) system, which began

in 1970, has played a leading role in establishing methodologies for comparable
infection rates. The methods developed by the NNIS have underpinned surveillance
programs in many other countries. Advanced national surveillance programs now
exist in countries such as the Netherlands, Belgium, Brazil, Canada, Denmark,
France, Germany, Norway, Sweden and the United Kingdom. Participation in the

130 PUBLIC AND PRIVATE

HOSPITALS
NNIS has been shown to have significantly reduced BSIs, SSIs, urinary-tract
infections and pneumonia in ICU patients in the United States. In Europe, studies
have found that German and Dutch infection surveillance programs have led to
reductions of 24–57 per cent for SSIs and 20–29 per cent for ICU infections (Clezy
et al. 2008).

Australia has yet to implement a national surveillance regime for hospital-acquired

infections, but studies of infection monitoring in individual hospitals and states have
confirmed that monitoring can reduce infections (Collignon et al. 2006). In addition,
multi-hospital surveillance regimes do exist in varying forms within individual
Australian jurisdictions, and these are in part based on NNIS methodologies
(detailed in section 6.3).

The ACSQHC (sub. 24) stressed that it is desirable to use a statistical technique that
accounts for risk differences when comparing infection rates across hospitals. A
common way of doing this is to subdivide infections data into groups with a similar
risk of infection, and only compare infection rates within those groups. Such groups
could be defined by patient characteristics, procedure and/or type of hospital. The
ACSQHC (sub. 24) suggested that such groupings would preferably be defined by
patient population characteristics (including procedure profile), but should at least
be defined by the peer groups (identified by scale and services provided) used for
national reporting on public hospitals (detailed in AIHW 2009a). However, the
ACSQHC noted that the peer group classification system would need to be revised
because it does not currently include private hospitals.

A popular approach for grouping SSIs with similar risks is to stratify the data using
a risk index developed by the US National Healthcare Safety Network (NHSN)
(formerly the NNIS) (box 6.1). However, some recent studies have found that the
NHSN/NNIS risk index does not always provide an accurate measure of risk
(including studies of Victorian and Queensland infections by Friedman et al. 2007
and Clements et al. 2007 respectively). This has caused some SSI monitoring
programs to use a modified version of the NHSN/NNIS methodology, including
those in Queensland and Victoria (CHRISP 2003; Victorian Department of Human
Services 2008a).

HOSPITAL-ACQUIRED 131
INFECTIONS
 Box 6.1 NHSN/NNIS risk index for SSIs
The NHSN/NNIS (National Healthcare Safety Network/ National Nosocomial Infections
Surveillance System) risk index is widely used internationally, including in Australia, to
enable the comparison of groups with similar risks of experiencing an SSI. The risk
index is calculated by summing scores for three characteristics:
• physical status
• length of surgery
• wound class.
The resulting index has four possible values — zero, one, two or three — with a higher
value indicating there is a greater risk of having an SSI. The scoring system for each
characteristic is outlined below.

ASA classification of physical status

The physical status of patients undergoing general anaesthesia is categorised using a
system developed by the American Society of Anaesthesiology (ASA). Patients
categorised under the ASA system as having severe systemic disease, or being
moribund and not expected to survive without the operation, are given a score of one
for physical status. Other patients are given a physical-status score of zero.
Length of surgery
If the surgery has a longer duration than 75 per cent of comparable procedures, based
on a database of past procedures, then a score of one is assigned for length of
surgery. Other patients are given a length-of-surgery score of zero.
Wound class
If the surgical team classifies a surgical wound as being ‘contaminated’ or ‘dirty
infected’, using a classification system adapted from guidelines developed by the US
Centres for Disease Control and Prevention, then a score of one is assigned for wound
class. Other patients are given a wound-class score of zero.
Source: CHRISP (2003); Cruickshank and Ferguson (2008); WA Department of Health (2008).

The WA Government’s infection-surveillance unit cautioned that the NHSN/NNIS

risk index does not control for all risk differences between public and private
hospitals:
… there are almost certainly key patient-risk factors that are not incorporated into this
[NHSN/NNIS risk stratification] method, and that systematically differ between
patients having surgery at private and public hospitals.
For example, diabetic control, obesity and cigarette smoking are among known risk
factors for SSIs that are not modifiable by hospitals and are not specifically
incorporated into NHSN risk stratification. (Health Care Associated Infection Unit,
Communicable Disease Control Directorate, Department of Health WA, sub. 38, p. 4)

132 PUBLIC AND PRIVATE

HOSPITALS
Another factor that is not controlled for in the NHSN/NNIS risk index is the volume
of each procedure performed by a hospital. Units performing more joint
replacements have lower infection rates for the same patient risk group (Dailey, Van
Geesel and Peterson 2009; Katz et al. 2001). However, NSW Health (sub. DR64)
suggested that ‘SSI is not related to volume but is clearly related to the frequency
and application of best practice “bundles” of care’.

Another method for dealing with risk differences is the Standardised Infection Ratio
(SIR). This is calculated as the number of observed infections divided by the
number of expected infections. Expected infections can be based on an historical
database of infection rates across multiple hospitals that is stratified into different
risk groups (such as procedures). The usefulness of the SIR depends on the
relevance and accuracy of that benchmark data. An SIR of more than one indicates
that there are more infections than expected, and a SIR less than one indicates fewer
infections than expected.

6.3 Available evidence on hospital-acquired infections

National data

The Commission identified two potential sources of national data on

hospital-acquired infections that cover both public and private hospitals:
• the Clinical Indicator Program (CIP), which is managed by the Australian
Council on Healthcare Standards (ACHS)
• the National Hospital Morbidity Database (NHMD), which is managed by the
Australian Institute of Health and Welfare (AIHW).

These sources are discussed below.

ACHS Clinical Indicators Program

The CIP collects a large number clinical indicators from hospitals, including 47 of
which measure healthcare-associated infections linked to specific procedures.3
These are grouped into five broad categories:
• SSIs (18 indicators)

3 Health care-associated infections are measured by 47 out of 49 indicators collected under the
ACHS Infection Control Indicators Version 3 (two indicators measure staff exposure to blood
and bodily fluids, which may not necessarily result in an infection).
HOSPITAL-ACQUIRED 133
INFECTIONS
• central-line associated BSIs (14 indicators)
• BSIs associated with dialysis (5 indicators)
• neonatal infections (6 indicators)
• MRSA cases (4 indicators).

Published data for these indicators suggest that infection rates rarely differ between
the public and private hospital systems. In 2007 (the latest published data), only
four of the 47 CIP indicators of healthcare-associated infections had a statistically
significant difference between public and private hospitals (table 6.1).4 In the few
cases where such differences were evident, the data suggest that the private sector
consistently outperformed the public sector. However, this could be misleading
because the CIP is not designed to monitor the relative performance of the public
and private sectors. It is a service offered to individual healthcare providers to help
them improve their service quality (ACHS, sub. 13). As a result, the CIP data have a
number of limitations:
• participation in the CIP is voluntary, and so the sample may not be
representative of either the public or private sectors (sample-selection bias)
• the number of reporting hospitals is often small, and so sample sizes may not be
sufficient to reach robust conclusions about the relative performance of a
particular sector
• hospitals self-report data without external validation, and have the option of only
reporting indicators of interest to them
• there is no risk adjustment to reflect differences in patient characteristics.5

For 2007 (the latest published data), a total of 284 hospitals reported
infection-control indicators to the ACHS.6 But CIP participants are not obliged to
report every indicator because some may not be relevant to services provided by
their organisation. As a result, individual infection-control indicators were based on

4 This was also the case in 2006 (based on data published in ACHS 2007) and in 2008 (based on
unpublished data the ACHS provided to the Commission).
5 The CIP indicators do to some extent stratify the data according to risk, since in many cases the
indicators are specific to a certain type of procedure/treatment and/or whether an infection
occurred in an ICU.
6 The ACHS (sub. 13) advised that its (yet to be published) 2008 infections data will be based on
responses from 292 hospitals, with 128 of these in the public sector and 164 in the private
sector.
134 PUBLIC AND PRIVATE
HOSPITALS
samples that ranged from 1 to 142 hospitals in 2007, with the median sample being
only 13 hospitals (ACHS 2008).7

Table 6.1 ACHS infection indicators that differed between public and
private hospitals, 2007a
Infection No. of reporting
rateb hospitals

Indicator no. and descriptionc Units Public Private Public Private

1.2 Deep incisional SSI in hip per 100 0.99 0.63 38 96

prosthesis proceduresd procedures

1.17 Superficial incisional SSI in per 100 2.02 0.94 16 37

abdominal hysterectomy procedures

5.2 ICU-associated new MRSA per 10 000 16.70 7.18 25 23

healthcare-associated ICU overnight
infections in a nonsterile occupied bed days
site

5.4 Non ICU-associated new per 10 000 2.77 1.11 68 59

MRSA inpatient healthcare- non-ICU overnight
associated infections in a occupied bed days
nonsterile site
a The ACHS identified an indicator as differing between the public and private sectors if public/private status
explained at least 10 per cent of the variation in sampled infection rates, and statistical testing showed that the
probability of a difference between public and private rates was at least 95 per cent. However, as noted in the
main text, the data may not be suitable for public-private comparisons due to sample-selection bias, small
sample sizes, self-reporting, and no risk adjustment to reflect differences in patient characteristics. b Mean
infection rates after applying the shrinkage estimation method to the data. c The following abbreviations are
used: ICU (intensive-care unit); MRSA (methicillin-resistant Staphylococcus aureus); and SSI (surgical-site
infection). d The ACHS (sub. 13) advised that unpublished data for 2008 showed that indicator 1.2 was 0.68
per 100 procedures for private hospitals and 1.02 for public hospitals.
Source: ACHS (2008).

In addition, Clezy et al. (2008) claimed that the methods used to collect the CIP
indicators could vary widely between facilities because they are not precisely
specified, and there is limited training on applying indicator definitions and on
best-practice methods for detecting infections.

National Hospital Morbidity Database

The National Hospital Morbidity Database (NHMD) contains patient-level data

from almost all hospitals in Australia on diagnoses, procedures and external causes
of injury. It is compiled by the AIHW from data supplied by state and territory

7 Excluding the two infection-control indicators that measure staff exposure to blood and bodily
fluids, which were based on samples of 198 and 202 hospitals respectively.
HOSPITAL-ACQUIRED 135
INFECTIONS
health authorities. The Australia Health Insurance Association (AHIA, sub. DR58)
encouraged the Commission to use data from the NHMD as a source of information
on hospital-acquired infections. However, the AIHW advised that NHMD data from
2007-08 and previous years cannot reliably identify whether a health condition
arose during care. These data have not been used by the Commission for reporting
on hospital-acquired infections.

For 2008-09, the NHMD coding standards were revised to include a

‘condition-onset flag’ that identifies whether a health condition arose during an
episode of care (Australian Government Department of Health and Ageing,
sub. 32).8 This, in combination with codes for specific infection organisms and
locations, could be used to identify different types of hospital-acquired infections,
such as VRE urinary-tract infections.

However, the AIHW advised the Commission that infections data from the 2008-09
NHMD will not be available until after this study is completed. Similar data were,
however, obtained from the Victorian Government, which has for many years
required hospitals to identify conditions that arose during an episode of care. These
data are detailed in appendix F, and were used in the Commission’s analysis of
state-level data below.

State-level data

Given the limitations of national infections data, the Commission decided to also
draw on evidence collected by state governments. Government monitoring of
hospital-acquired infections is largely undertaken by state governments, reflecting
their role as providers of public hospitals and regulators of private hospitals.

Details about the data collected by individual jurisdictions and what they show
about infection rates in public and private hospitals are provided in appendix F. In
summary, the data are not collected and reported on a nationally-consistent basis,
but public and private hospitals are included in most cases (the exceptions are New
South Wales and the Northern Territory). However, government infection
surveillance programs in Victoria, Queensland and Tasmania only began collecting
information from private hospitals recently, and so limited data are available. The
Commission was able to obtain additional data for Victoria, derived from that
state’s morbidity database. Western Australia has the most transparent reporting of
results for public and private hospitals, and provided the most comprehensive data
from its surveillance regime for this study.

8 NHMD data are coded according to the International Statistical Classification of Diseases and
Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM).
136 PUBLIC AND PRIVATE
HOSPITALS
The different collection and reporting methods across jurisdictions mean that it is
not possible to construct national estimates of hospital-acquired infections from
state-level data, or to generally compare across jurisdictions. Nevertheless, the data
show a consistent pattern within jurisdictions — for a given type of infection,
private hospitals have a lower rate of hospital-acquired infection than public
hospitals. However, this result could be misleading because private hospitals
generally treat patients who have a lower risk of infection, and the data do not fully
control for this. The pattern evident in the state-level data should therefore be
interpreted with caution.

It is also evident from the state-level data that infection rates can vary over time.
This could be due to a range of factors, including the occurrence of a pandemic and
a change in adherence to infection-control procedures in hospitals.

FINDING 6.1

Australia does not have a robust nationally-consistent data collection on

hospital-acquired infections. The limited available evidence suggests that private
hospitals have lower infection rates than public hospitals, but this result could be
misleading because private hospitals generally treat patients who have a lower risk
of infection. A more definitive finding will require the development of data
collections that enable risk differences between hospitals to be distinguished from
genuine differences in performance.

6.4 Developments to improve future comparisons

It is evident from the preceding analysis that Australia’s inconsistent and sometimes
less-than-comprehensive approaches to monitoring hospital-acquired infections
hinders comparisons across hospitals. This is of concern, given that studies have
shown that monitoring and benchmarking, combined with a prevention program that
uses the data to drive improvements, can reduce the rate of infections. Consumers
Health Forum of Australia (sub. DR59) shared this concern, advocating mandatory
reporting of hospital-acquired infections and calling for infection data to be made
available to consumers.

The problems with existing monitoring arrangements prompted the recent

ACSQHC study of healthcare-associated infections to recommend
nationally-consistent approaches, including strengthened surveillance for:
• BSIs classified as being SAB, central-line associated in ICUs, and
haemodialysis-access associated

HOSPITAL-ACQUIRED 137
INFECTIONS
• SSIs, including those linked to coronary artery bypass graft surgery, major-joint
prosthesis insertion, and other procedures that have higher-than-expected SSI
rates at the local level
• bacterial sepsis in the first week of life, including meningitis (Cruickshank and
Ferguson 2008).

Some progress has been made in this regard. In December 2008, the Australian
Health Ministers’ Conference (AHMC) decided that all public hospitals will have to
report two infection indicators — SAB BSIs and Clostridium difficile — on a
nationally-consistent basis (ACSQHC, sub. 24). In November 2009, AHMC
included SAB BSIs and Clostridium difficile in a national set of safety and quality
indicators (DOHA, sub. DR 69, p. 9). Reporting of these two indicators is being
facilitated by the ACSQHC, which is also considering other infection indicators at
the request of the AHMC. The states and territories have implemented, or are in the
process of implementing, the required surveillance and reporting arrangements.

Health Ministers also noted at their December 2008 meeting that hand hygiene is a
key element in the prevention of hospital-acquired infections. They therefore
supported a National Hand Hygiene Initiative, which commenced in early 2009 and
is managed by Hand Hygiene Australia with oversight by the ACSQHC. The
Ministers called for nationally-consistent measurement of hand hygiene compliance
using approaches facilitated by the ACSQHC. The Royal Australasian College of
Surgeons supported consistent reporting of hand hygiene and infection rates:
The College has a longstanding interest in hospital-based infections, infection control,
the use of antibiotics, approaches to hand hygiene and the impact of these on individual
patient care. Substantial work over the past decade has again highlighted the
importance of systemwide approaches to hand hygiene and its impact on key infections
like MRSA bacteraemias and surgical-site infections. The College would certainly
support the introduction of nationwide reporting on some of these key measures. To our
knowledge, however, there is no current methodology for this at the individual state
and territory level for public or private hospitals. (sub. 30, p. 3)

The Health Care Associated Infection Unit, Communicable Disease Control

Directorate, Department of Health WA (sub. 38) noted that process indicators —
such as hand-hygiene compliance rates, compliance with surgical antibiotic
prophylaxis, influenza staff vaccination rates and central-line insertion and care
protocols — have the advantage that they do not need to be adjusted for patient-risk
factors.

States and territories will also have to report indicators for some hospital-acquired
infections in public hospitals under the National Healthcare Agreement (NHA). The

138 PUBLIC AND PRIVATE

HOSPITALS
Australian Government Department of Health and Ageing advised that a
performance benchmark had been set under the NHA for SAB BSIs:
Sound comparisons between the public and private sectors on the basis of
hospital-acquired infections is necessary with a clear need to set benchmarks and
identify best practice. One of the performance benchmarks adopted under the NHA is
that the rate of Staphylococcus aureus (including MRSA) bacteraemia be no more than
2.0 per 10 000 occupied bed days for acute care public hospitals by 2011-12 in each
state and territory. (sub. 32, p. 21)

This has strengthened the acceptance by jurisdictions of nationally-consistent

definitions for SAB BSIs under the abovementioned AHMC decision.

While the abovementioned developments are welcome, more actions will be needed
to establish a comprehensive and nationally-consistent approach to infection
monitoring. For example, private hospitals are not subject to the recent AHMC
decision and NHA reporting requirements. The ACSQHC also noted the need for
further reforms and recommended:
• the eventual incorporation of private hospitals in national health care reporting, such
as those currently managed through National Health Information Agreements
• the development of national hospital peer groupings which include and classify
private hospitals
• the national development of standard measures of safety and quality which are
applied across all Australian hospitals
• the promotion of routine review of safety and quality indicators by all hospitals in
Australia. (sub. 24, p. 10)

The Productivity Commission supports the above recommendations made by the

ACSQHC, which was established to lead and coordinate national improvements in
the safety and quality of health care.

FINDING 6.2

Foreshadowed developments, such as performance reporting under the National

Healthcare Agreement, will move Australia closer to a robust nationally-consistent
data collection on hospital-acquired infections. However, more actions will be
required to enable meaningful infection-rate comparisons between public and
private hospitals. An important step in this regard would be to include private
hospitals in national reporting arrangements. The Australian Commission on Safety
and Quality in Health Care is leading and coordinating initiatives that should
improve the feasibility of future comparisons.

HOSPITAL-ACQUIRED 139
INFECTIONS
7 Other partial indicators

Key points
• The complexity of hospital services means that indicators additional to those
reported in chapters 5 and 6 are necessary to compare the performance of public
and private hospital systems. In broad terms, these indicators cover partial
productivity, access and quality and patient safety.
• The Commission observed a paucity of reliable published data with which to
compare the hospital sectors.
• Best available partial productivity data suggest that private hospitals operate leaner
staffing levels and have shorter average lengths of stay per episode of care.
• The shorter average lengths of stay are due to:
– private hospitals (on average) having relatively shorter lengths of stay for the
same surgical procedures
– private hospitals undertaking relatively more surgical procedures.
• In terms of access, the number of accident and emergency department visits grew
from 5.8 to 7.1 million between 2002-03 and 2007-08. The proportion of patients
being seen within their triage benchmarks is reported to have improved. However,
there are some questions about the veracity of available data.
• More people are on public hospital elective surgery waiting lists and they are waiting
longer to be admitted into public hospitals than in 2002-03.
– Yet, public hospitals were observed to be meeting their hospital waiting time
targets. Data manipulation is alleged to have occurred with elective surgery
waiting lists in Victoria, and data need to be collected on a consistent basis
across Australia.
• Bed occupancy rates are a more comparable measure of timely access to hospitals.
Public hospitals have average bed occupancy rates that are above or near the
levels where regular bed shortages can occur.
• The reporting of national quality and patient safety indicators is still at a formative
stage, and few data can provide a comparison between public and private hospitals.
• The best available data on the percentage of hospital separations that involve an
adverse event have many shortcomings, but suggest a lower incidence of adverse
events in private hospitals.

The terms of reference ask the Commission to examine and report on the relative
performance of the public and private hospital systems. In doing so, the
Commission is to consider other performance indicators, including the ability of

OTHER PARTIAL 141

INDICATORS
such indicators to inform comparisons of hospital performance and efficiency, and
to propose any developments that would improve the feasibility of future
comparisons.

Many partial indicators of hospital performance have been reported in other

publications (for example, DOHA 2009f; SCRGSP 2009; AIHW 2008c, 2009a).
The Commission has included such indicators to provide as complete a picture as
possible of the performance and efficiency of the public and private hospital
systems, and the accessibility and quality of care each system provides. Some well-
established indicators of the performance of the public hospital system, such as
waiting times, are not available for the private sector. The Commission nevertheless
considers that such indicators provide useful information on the performance of the
public hospital system.

Chapters 5 and 6 reported on average costs per separation and hospital-acquired

infection rates. To identify additional indicators of system performance, the
Commission reviewed a number of existing performance indicator frameworks, as
well as frameworks under development such as the proposed National Healthcare
Agreement indicator set. The partial indicators covered in this chapter (as well as
those reported earlier) are summarised in box 7.1.

In choosing the indicators, the Commission was guided by three overarching

principles. The indicators should be:
• outcome and output based — they should reflect how hospitals are improving
health outcomes, or if that is not possible, how well they are providing services
• hospital-wide — they should reveal the overall performance of hospitals and
hospital sectors rather than specific aspects of clinical care
• based on the best available data — except for those data that are manifestly
inadequate, the best available data should be reported and their known
deficiencies documented.

There are four implications that arise from the choice of indicators and data. First, a
number of input or process-oriented indicators are not reported here, even though
they are found in other frameworks. For example, indicators on the workforce
sustainability of hospitals are excluded because their interpretation is ambiguous. A
high proportion of an older nursing workforce, for example, may indicate
insufficient recruitment of junior nurses, or it can indicate a more effective,
experienced workforce.

142 PUBLIC AND PRIVATE

HOSPITALS
 Box 7.1 Summary of partial indicators
• Costs (chapter 5)
– cost per casemix-adjusted separation
– average cost of individual DRGs
• Productivity
– labour productivity
– bed productivity
– drug, medical and surgical supplies productivity
– relative stay index
• Access to hospital services
– emergency department waiting times
– elective surgery waiting lists and times
– occupancy rates
– private hospital insurance costs
– rates of separation for elective surgery
• Quality and patient safety
– hospital-acquired infections (chapter 6)
– accreditation of hospitals and hospital beds
– unplanned readmissions and returns
– adverse events
– obstetric indicators.

Second, much of the data presented in this chapter are publicly available and have
well-known deficiencies, such as the data published by the Australian Council on
Healthcare Standards (ACHS). The Commission has also made a number of
suggestions that would improve the availability and comparability of data in the
future.

Third, as with average costs and hospital-acquired infection rates, each of the
indicators in box 7.1 are partial measures of hospital performance. This means that
it is potentially misleading to interpret the results of any one indicator without
regard to others.

Finally, most of the partial indicators presented in this chapter are drawn from
published sources and are averages for each sector. The Australian Health Service
Alliance (AHSA) cautioned that:
While overall sector information is useful when comprehensive and robust, such
aggregate information may conceal the existence of a small number of hospitals with
particularly low quality. (sub. DR53, p. 6)
OTHER PARTIAL 143
INDICATORS
As noted in chapters 2, 3 and 4, public hospitals differ considerably from each other
in size and location and private hospitals have a range of operational motives. These
differences are not captured in the partial indicators presented in this chapter.

7.1 Productivity
An important policy objective is whether hospitals are making economical use of
their resources. Wasted resources can mean a lost opportunity to improve health
outcomes. A measure of the efficiency of resource use is technical efficiency
(chapter 1). A hospital is said to be technically efficient if, in the provision of a
service, it is not feasible to reduce any input without also decreasing the service and
without increasing any other input.

The preferred measure of technical efficiency is total factor productivity (TFP). TFP
indicates how effectively hospitals are able to transform all of their inputs (labour,
capital, and pharmaceutical, medical and surgical supplies) to provide all of their
services (inpatient and outpatient services and emergency department visits).

TFP, however, is difficult to measure for hospitals. This is because some of the
information needed to calculate TFP, such as measures of capital and revenue, are
not available for hospitals. Instead, this section reports four partial measures of
productivity:
• labour productivity
• bed productivity
• drug, medical and surgical supplies productivity
• relative stay index.

Care must be exercised when interpreting these measures. First, each of the partial
productivity measures need to be read together. Second, except for casemix
adjustment, none of them account for a range of factors that can affect a hospital’s
performance, such as:
• the range of non-inpatient services hospitals provide (such as accident and
emergency departments)
• the patient risks hospitals manage
• other factors outside the control of hospitals.

144 PUBLIC AND PRIVATE

HOSPITALS
Labour productivity

Labour productivity is a measure of the extent to which non-medical staff (all

hospital staff except doctors and surgeons) contribute to a hospital’s services.1
Labour costs are a significant component of hospital costs, and hospitals with high
labour productivity are likely to be more technically efficient in their use of labour.
Two measures of labour productivity of inpatient services were used:
• patient days per non-medical staff — defined as the number of days for which a
hospital provides lodging to patients divided by the number of non-medical staff
• casemix-adjusted separations per non-medical staff — defined as
casemix-adjusted separations divided by the number of non-medical staff.

It is difficult to draw accurate comparisons of labour productivity between public

and private hospitals because, as noted earlier, no account was made for teaching
and research functions or for the non-inpatient services provided by many public
hospitals. Valid comparisons can be made over time, however. These trends suggest
that both patient days per non-medical staff member and separations per
non-medical staff member have been declining in public hospitals and increasing
for private hospitals between 2002-03 and 2007-08 (table 7.1).

There are three other limitations to this indicator. First, hospital establishment data
do not count the number of doctors and surgeons exercising their rights of private
practice in private and public hospitals (AIHW 2009a; ABS 2008f). As a result, to
improve comparability between sectors, private practice and salaried doctors and
surgeons were excluded from the count of staff for both private and public
hospitals.2

Second, high labour productivity may not always be desirable. Catholic Health
Australia said that measuring the numbers of separations per doctor or nurse:
… could send signals that use of fewer than clinically appropriate numbers of clinical
staff is to be encouraged. (sub. 20, p. 13)

A relatively high ratio of medical and nursing staff to patients may provide a higher
level of personal care to patients at levels that are clinically appropriate.

1 The reasons for excluding medical staff from this calculation of labour productivity are detailed
below.
2 Despite this general exclusion, it was not possible to fully exclude from the data the few salaried
medical officers employed in private hospitals. Their inclusion does not materially affect the
results.
OTHER PARTIAL 145
INDICATORS
Finally, labour productivity estimates include psychiatric hospitals. Data for private
acute and psychiatric hospitals were not available separately, so for comparability,
psychiatric hospitals were included in the public hospital dataset.

Table 7.1 Labour productivity, 2002-03 and 2007-08a

Patient days per non-medical staff Casemix-adjusted separations per
member non-medical staff memberb

2002-03 2007-08c 2002-03 2007-08c

Public hospitals
NSW 92 89 22 23
Vic 91 78 24 23
Qld 94 77 23 22
SA 106 96 26 24
WA 86 79 21 22
Tas 100 87 24 23
NT 105 103 26 25
ACT 84 82 23 25
Australia 93 84 23 22
Private hospitals
NSW 154 149 49 55
Vic 145 151 44 52
Qld 149 159 44 53
SA 150 150 49 56
WA 145 138 42 40
Tas, NT and ACTd 138 156 np np
Australia 148 151 44 52
a ABS data for private hospitals do not exclude psychiatric hospitals. For comparability, psychiatric hospitals
have been included in both hospital samples. bExcludes newborns with no qualified days, hospital boarders
and posthumous organ donors. c2006-07 data for private hospitals. Includes a small number of salaried
medical officers. dData on private hospitals in Tasmania, the Northern Territory and the ACT are aggregated
to protect the confidentiality of the small number of hospitals in each of these jurisdictions. np Not published.
Source: AIHW (2004, 2009a); ABS (Private Hospitals, Australia, Cat. no. 4390.0); Productivity Commission
estimates.

Bed productivity

Capital productivity is a measure of the extent to which a hospital’s capital stock

(buildings, plant and equipment) contributes to the provision of hospital services.
There are various difficulties associated with quantifying the capital stock of
hospitals in dollar terms (chapter 5), and so the number of licensed or available beds
is used here as a proxy. Two measures of bed productivity were used:
• patient days per bed — the number of days for which a hospital provides lodging
to patients divided by the number of licensed or available beds
146 PUBLIC AND PRIVATE
HOSPITALS
• casemix-adjusted separations per bed — casemix-adjusted separations divided
by the number of licensed or available beds.

Hospitals with high rates of bed productivity are likely to be more technically
efficient in their use of capital.

Hospital bed productivity measures are relatively more comparable measures of

inpatient services than labour productivity. This is because beds, unlike labour, are
used exclusively for inpatient services. Public hospitals recorded more patient days
per bed (316) than private hospitals (279) in 2007-08, although there was
considerable variability for both public and private hospitals in each state and
territory (table 7.2). Nationally, the rate of bed utilisation declined slightly for
public hospitals and increased for private hospitals between 2002-03 and 2007-08.
Separations per bed were similar between public and private hospitals in 2002-03,
but by 2007-08, private hospital bed productivity had grown relatively larger.

Table 7.2 Bed productivity, 2002-03 and 2007-08

Patient-days per bed Casemix-adjusted separations
per beda
2002-03 2007-08 2002-03 2007-08

Public hospitals
NSW 337 311 79 79
Vic 357 351 93 102
Qld 295 281 73 79
SA 330 324 81 81
WA 301 302 73 83
Tas 336 302 82 78
NT 362 423b 90 104
ACT 322 326 88 98
Australia 330 316 81 85
Private hospitals
NSW 264 308 86 115
Vic 281 279 87 95
Qld 280 298 83 101
SA 282 302 92 115
WA 260 215 79 79
Tas, NT and ACTc np 199 np np
Australia 264 279 82 99
a Excludes newborns with no qualified days, hospital boarders and posthumous organ donors. bPatient days
per bed may exceed 365 since beds may have more than one same-day patient. cData on private hospitals in
Tasmania, the Northern Territory and the ACT are aggregated to protect the confidentiality of the small
number of hospitals in each of these jurisdictions. np Not published.
Source: AIHW (2004, 2009a); ABS (Private Hospitals, Australia, Cat. no. 4390.0); Productivity Commission
estimates.

OTHER PARTIAL 147

INDICATORS
The use of beds as a proxy for capital costs is not without its limitations. While bed
productivity is acceptable for comparing across hospitals of similar peer groups in a
given year, it is less appropriate for comparing across different types of hospitals (as
beds do not reflect differences in specialisations) or for considering trends over time
(as it does not capture the gradual uptake of new technologies).

Another limitation of the bed productivity indicator is that high bed productivity
(utilisation) is not unambiguously desirable. Some spare capacity is necessary to
manage the unpredictable workload associated with emergency admissions and to
provide timely access to elective surgery (see section 7.2).

Drug, medical and surgical supplies productivity

Drug, medical and surgical supplies represent a significant component of hospital

operating costs. The Steering Committee for the Review of Government Service
Provision (SCRGSP 2009) reported that about 15 per cent of public hospital
recurrent (non-capital related) expenditure was attributable to drug, medical and
surgical supplies. Hospitals with high productivity in the use of drug, medical and
surgical supplies are likely to be more technically efficient in that area.

Two measures of drug, medical and surgical supplies productivity are used in this
chapter:
• patient days per $1000 of expenditure on drug, medical and surgical supplies —
the number of days for which a hospital provides lodging to patients divided by
$1000 of constant price expenditure on drug, medical and surgical supplies
• casemix-adjusted separations per $1000 of expenditure on drug, medical and
surgical supplies — casemix-adjusted separations divided by $1000 of constant
price expenditure on drug, medical and surgical supplies.

Drug, medical and surgical supply productivity is not easily comparable between
public and private hospitals due to the different prices paid by the sectors for their
supplies. A given sector’s productivity, however, is more comparable over time.
Both the patient-days and separations per $1000 spent on drug, medical and surgical
supplies declined for both public and private hospitals (table 7.3). Given that prices
are held constant, the ‘productivity decline’ represents an increase in the intensity of
the use of drug, medical and surgical supplies in the delivery of services.

To the extent that public hospitals pay less for their drug, medical and surgical
supplies than private hospitals, actual public hospital productivity will be lower than
indicated by the rates reported in table 7.3.

148 PUBLIC AND PRIVATE

HOSPITALS
Table 7.3 Drug, medical and surgical supplies productivity, 2002-03 and
2007-08a
Patient-days per $1000 drug, Separations per $1000 drug,
medical and surgical supplies medical and surgical suppliesb

2002-03 2007-08c 2002-03 2007-08c

Public hospitals
NSW 6.4 4.0 1.4 0.9
Vic 7.3 3.9 2.0 1.2
Qld 6.5 3.4 1.6 0.9
SA 10.0 5.7 2.4 1.3
WA 6.8 4.2 1.7 1.2
Tas 6.7 3.2 1.5 0.8
NT 6.6 5.4 2.2 1.9
ACT 5.8 4.0 1.7 1.2
Australia 6.9 4.0 1.7 1.1
Private hospitals
NSW 4.6 3.4 1.4 1.2
Vic 5.7 4.0 1.8 1.4
Qld 6.2 4.2 1.8 1.4
SA 5.5 3.9 1.7 1.3
WA 6.0 3.5 2.0 1.2
Tas, NT and ACTd 2.0 4.0 0.7 1.5
Australia 5.3 3.8 1.7 1.3
a Deflator is constructed from drug (imported and domestic) indices, and medical and surgical supply
(imported and domestic) indices. ABS data for private hospitals do not exclude psychiatric hospitals. For
comparability, psychiatric hospitals have been included in both hospital samples.b Excludes newborns with no
qualified days, hospital boarders and posthumous organ donors. C 2006-07 data for private hospitals. d Data
on private hospitals in Tasmania, the Northern Territory and the ACT are aggregated to protect the
confidentiality of the small number of hospitals in each of these jurisdictions.
Source: AIHW (2004, 2009a); ABS (Private Hospitals, Australia, Cat. no. 4390.0); Productivity Commission
estimates.

Relative stay index

Hospital length of stay is sometimes used as a measure of hospital efficiency. It is

an indirect measure of technical efficiency because hospitals that have shorter
lengths of stay for a patient are presumed to be able to spread the total cost of fixed
overheads and capital assets across a greater number of patients.

The average length of stay (ALOS) is a useful indicator in this regard, and is
commonly used to compare hospital performance for individual procedures
(AIHW 2009a). However, it is a poor indicator when comparing the performance of
hospitals across all of their inpatient activity. This is because a hospital’s reported
ALOS does not adjust for its composition of services.

OTHER PARTIAL 149

INDICATORS
A more appropriate measure of a hospital’s ALOS is its relative stay index (RSI).
The RSI is defined as the actual number of acute care patient days divided by the
expected number of acute care patient days, adjusted for casemix. RSI differs from
ALOS in two respects:
• the length of stay is standardised using a reference composition of procedures (in
much the same way as age standardisation)
• each DRG is adjusted for its relative cost weight (AIHW 2009a; SCRGSP 2009).

Public hospitals exhibited relatively shorter lengths of stay than private hospitals for
medical DRGs (with an RSI of 0.94 compared to 1.20) in 2007-08 (table 7.4).
Private hospitals exhibited shorter lengths of stay for surgical DRGs than public
hospitals (0.98 compared to 1.03) in 2007-08.

Table 7.4 Relative stay index, public and private hospitals, by DRG
procedure partitions, 2002-03 and 2007-08a
Medical Surgical Other

2002-03 2007-08 2002-03 2007-08 2002-03 2007-08

Public hospitals
NSW 1.03 1.01 1.08 1.05 1.19 1.16
Vic 0.91 0.86 0.99 1.01 1.00 1.01
Qld 0.92 0.89 0.99 0.98 1.04 1.09
SA 0.94 0.97 1.01 1.04 1.00 1.03
WA 1.03 0.97 1.04 1.04 1.00 0.99
Tas 1.01 0.95 1.07 1.02 1.09 1.10
NT np 1.08 np 1.44 np 1.4
ACT np 0.88 1.07 0.94 1.10 0.91
Australia 0.96 0.94 1.03 1.03 1.07 1.07
Private hospitals
NSW 1.26 1.31 0.93 0.96 0.87 0.96
Vic 1.13 1.17 0.97 0.99 0.95 0.97
Qld 1.16 1.18 0.98 0.95 0.98 0.98
SA 1.13 1.16 0.96 0.97 0.91 0.97
WA 1.19 1.18 1.04 1.07 0.98 0.98
Tas, NT and ACTb np np np np np np
Australia 1.17 1.2 0.97 0.98 0.93 0.97
a Under the direct standardisation method, RSI is calculated by multiplying the ALOS for each sub-group of
hospital by total number of separations undertaken by all hospitals, divided by the total patient days for all
hospitals. b Owing to commercial-in-confidence restrictions on ABS data, it was not possible to estimate the
RSI of private hospitals in Tasmania, the ACT and the Northern Territory, though their contributions are
included in the Australian total. np Not published.
Source: AIHW (2004, 2009a).

150 PUBLIC AND PRIVATE

HOSPITALS
It follows that the relatively low ALOS observed for private hospitals is due to a
combination of their lower RSI for surgical and other DRGs (compared to public
hospitals), and their tendency to undertake relatively more separations in the
surgical and other DRG partitions than public hospitals (chapter 4).

Even though the RSI represents an improvement on simply reporting hospital-level

ALOS, both indicators have their limits. Neither reports the actual resources used in
the delivery of an episode of care — TFP is a better measure in this regard. Two
hospitals, for example, that exhibit the same RSI or ALOS might still differ in
technical efficiency if one hospital employs more staff and other resources than the
other.

Labour intensity of hospitals

Labour intensity is a measure of how many staffing resources hospitals employ

relative to other inputs. Labour intensity is a useful descriptor of the way a hospital
manages its workload. A high (or low) labour intensity is not necessarily desirable,
but it can illustrate some of the drivers behind the productivity measures described
above.

Public hospitals employed more allied health workers, nursing and other personal
care staff per bed than did private hospitals over the period 2002-03 to 2007-08
(table 7.5). This is understandable, given that public hospitals are more likely to
provide accident and emergency departments and outpatient clinics than private
hospitals.

The number of administration and clerical workers per bed is likely to be more
comparable, although this depends on the extent to which clerical and
administration staff are part of a hospital’s ‘overheads’ and do not vary substantially
with its activities. In broad terms, public hospitals employed twice as many
administration and clerical staff than did private hospitals over the period 2002-03
to 2007-08 (table 7.5). The administration and clerical staff reported here do not
include off-campus staff (such as head-office staff), and no account has been made
of differences in the provision of emergency departments and outpatient services.

Public hospitals reduced the number of allied health, nursing and other personal
care workers per bed between 2002-03 and 2007-08, while both hospital sectors
increased the number of administration and clerical staff per bed during the same
period.

OTHER PARTIAL 151

INDICATORS
Table 7.5 Labour intensity of public and private hospitals, 2002-03 and
2007-08a
Allied health, nursing and Administration and clerical staff
other care workers per bedb per bed

2002-03 2007-08c 2002-03 2007-08c

Public hospitals
NSW 3 2.93 0.65 0.56
Vic 3.2 3.63 0.73 0.85
Qld 2.63 3.05 0.5 0.58
SA 2.53 2.77 0.59 0.62
WA 2.84 3.09 0.65 0.75
Tas 2.87 2.93 0.47 0.52
NT 2.81 3.42 0.63 0.68
ACT 3.05 3.26 0.78 0.72
Australia 3.05 2.84 0.6 0.65
Private hospitals
NSW 1.53 1.67 0.22 0.3
Vic 1.72 1.67 0.23 0.31
Qld 1.55 1.46 0.34 0.35
SA 1.59 1.43 0.28 0.27
WA 1.6 1.79 0.26 0.34
Tas, NT and ACTd np 1.08 0.25 0.21
Australia 1.6 1.58 0.26 0.31
a All staff measured in full-time equivalents. b Includes a small number of salaried medical officers for private
hospitals because these could not be separated from the data. c 2006-07 data for private hospitals. d Data on
private hospitals in Tasmania, the Northern Territory and the ACT are aggregated to protect the confidentiality
of the small number of hospitals in each of these jurisdictions. np Not published.
Source: AIHW (2004, 2009a); ABS (Private Hospitals, Australia, Cat. no. 4390.0); Productivity Commission
estimates.

FINDING 7.1

Private hospitals appear to operate relatively leaner staffing levels than public
hospitals, although it is not clear how much of this difference can be explained by
the higher provision of emergency department and outpatient clinic services by
public hospitals.

FINDING 7.2

Private hospitals exhibit shorter lengths of stay than public hospitals. This is due to
private hospitals exhibiting relatively shorter lengths of stay for surgical
procedures and undertaking relatively more surgical procedures than public
hospitals.

152 PUBLIC AND PRIVATE

HOSPITALS
7.2 Access to hospital services
The timeliness of access to hospital care can have a profound influence on clinical
outcomes. As the Victorian Auditor-General remarked:
Timely access to hospital care is important. For patients requiring emergency care or
elective surgery, the time taken to receive services can significantly affect clinical
outcomes. (Victorian Auditor-General 2009, p. 1)

The ability of patients to access medical and surgical services is thus an important
policy objective of governments and an important motivator for private hospital
insurance.

Under the National Healthcare Agreement (NHA), state and territory governments
have agreed to provide free public hospital services based on:
• clinical need within a clinically appropriate time period
• equitable access regardless of a patient’s geographic location.

The NHA, like the Australian Health Care Agreements which preceded it, requires
states and territories to collect data on waiting times for emergency treatment and
elective surgery in public hospitals. These access measures are widely available and
well-established indicators of public hospital performance.

In contrast to the regular reporting of public sector waiting times, few measures of
timely access to private hospitals are available. The Commission has only been able
to directly compare the accessibility of care between the public and private systems
for a small number of indicators relating to equitable access to elective surgery.

Timely access to emergency treatment

Emergency medicine is concerned with addressing the injury or illness that poses an
immediate risk to a patient’s life or long-term health. The Australian Triage Scale
recognises that the most urgent (resuscitation) patients need to be seen immediately
and the least urgent emergency patients need to be seen within two hours of
presentation to a hospital emergency department (ACEM 2000).

Public hospitals

Public hospitals (including privately-owned hospitals that provide public hospital

services) are required by each state and territory health department to collect and
report data on the number of persons presenting to emergency departments, the

OTHER PARTIAL 153

INDICATORS
assignment of clinical urgency, the time at which the person was seen by a medical
officer, and the outcomes of the emergency care.

Data are reported to the Australian Government Department of Health and Ageing
(DOHA) and the Australian Institute of Health and Welfare (AIHW) under the
National Non-Admitted Emergency Department Care Data Collection and are
widely published (AIHW 2009a; SCRGSP 2009; DOHA 2009f).

Between 2002-03 and 2007-08, the number of recorded visits to public hospital
accident and emergency departments grew from 5.8 million to 7.1 million, or at
about 4.1 per cent per year. In most states, the proportion of patients that were seen
within their prescribed benchmark times improved in recent years (table 7.6). The
greatest improvements were observed for urgent and semi-urgent cases. For
example, the proportion of semi-urgent patients that were seen within their
benchmark times increased from 61 to 66 per cent between 2002-03 and 2007-08.
Public hospitals in the ACT and the Northern Territory, however, experienced
significant declines in the proportion of non-resuscitation patients that were seen
within their benchmark times (AIHW 2009a).

Care needs to be exercised when interpreting waiting-time data for emergency

departments. There appears to be a significant variation between hospitals in how
waiting times are measured, including the assignment of clinical urgency categories
(SCRGSP 2009).

Potentially more serious, however, is the suggestion that data manipulation has
occurred in at least one and possibly more Victorian public hospitals. The Victorian
Auditor-General found that the quality of Victorian emergency department
waiting-time data had ‘fundamental flaws both with data accuracy and the rigour of
data capture processes’ (Victorian Auditor-General 2009, p. v). In particular:
… hospitals inconsistently interpreted reporting rules, data capture methods were
susceptible to error, and the accuracy of some data was impossible to check. This
means incorrect data can go undetected. In one hospital, data manipulation had
occurred. (Victorian Auditor-General 2009, p. 3)

Dr Stephen Parnis, an emergency physician at St Vincent’s Hospital and vice

president of the Victorian Branch of the Australian Medical Association (AMA), in
giving evidence to a Victorian parliamentary inquiry, expressed doubts over the
credibility that all Victorian resuscitation patients were seen on time (Medew 2009).

154 PUBLIC AND PRIVATE

HOSPITALS
Table 7.6 Public hospital emergency department waiting times, 2002-03
and 2007-08
NSW Vic Qld SA WA Tas NT ACT Aust

Number of accident and emergency department occasions of service (thousands of visits)

2002-03 1 982 1 261 1 223 472 571 97 94 96 5 796
2007-08 2 418 1 523 1 471 544 778 143 125 98 7101
Rate of change 4.1 3.8 3.8 3.8 -0.9 8.1 5.9 0.5 4.1
Proportion of resuscitation patients seen on time
2002-03 100 100 99 99 94 92 100 100 99
2007-08 100 100 98 100 99 99 100 100 100
Rate of change 0.0 0.0 -0.3 0.3 1.0 1.6 0.0 0.1 0.2
Proportion of emergency patients seen on time
2002-03 77 84 73 66 73 55 60 82 75
2007-08 81 79 69 72 69 74 59 81 76
Rate of change 1.1 -1.2 -1.0 1.9 -1.2 6.1 -0.5 -0.3 0.2
Proportion of urgent patients seen on time
2002-03 57 76 55 47 64 61 64 74 61
2007-08 69 71 56 54 56 54 47 52 63
Rate of change 3.7 -1.4 0.3 2.9 -2.8 -2.4 -6.0 -6.8 0.7
Proportion of semi-urgent patients seen on time
2002-03 62 65 55 49 68 60 59 67 61
2007-08 75 65 61 60 59 58 47 51 66
Rate of change 3.7 -0.1 1.9 4.0 -2.9 -0.5 -4.3 -5.4 1.6
Proportion of non-urgent patients seen on time
2002-03 87 86 80 85 87 90 88 79 85
2007-08 90 86 87 80 86 86 86 78 87
Rate of change 0.8 0.1 1.7 -1.1 -0.3 -0.9 -0.4 -0.3 0.4
Proportion of all patients seen on time
2002-03 65 73 60 53 73 64 65 74 66
2007-08 76 71 63 61 61 60 52 58 69
Rate of change 3.1 -0.5 1.1 3.0 -3.6 -1.4 -4.5 -4.8 0.9
Source: AIHW (2004, 2009a); Productivity Commission estimates.

Private hospitals

A number of privately operated hospitals provide emergency departments, with

some serving public patients. The Australasian College for Emergency Medicine
(ACEM, sub. 14) argued that waiting times at privately-operated emergency
departments were generally less than their public hospital counterparts. According
to statistics supplied by the ACEM for three private hospitals in different states (one
of which was contracted to provide public hospital emergency department services)
in 2007-08:

OTHER PARTIAL 155

INDICATORS
• 96 per cent of resuscitation patients were seen on time (though this was based on
a very small sample of patients)
• 79 per cent of emergency patients were seen on time
• 70 per cent of urgent patients were seen on time
• 72 per cent of semi-urgent patients were seen on time (though there was a large
range, 46 to 99 per cent)
• 93 per cent of non-urgent patients were seen on time (sub. 14; pers. comm.).3

The ACEM (sub. 14) claimed that some of the main reasons why private hospital
emergency departments were able to achieve shorter waiting times were because
they had:
• relatively more senior staff in decision-making positions
• incentive-based payments that are aligned with patient throughput
• efficient department sizes
• processes and systems designed to reduce waiting times and improve efficiency.

Timely access to elective surgery

Elective surgery is any form of surgery that a patient’s doctor or health professional
considers to be necessary but which can be delayed by at least 24 hours
(DOHA 2009f).

The two hospital systems have different approaches to providing access to elective
surgery. In the public hospital system, the decision to grant a patient access is made
by the hospital, reflecting its judgement about the patient’s clinical needs as well as
its resourcing and performance targets. The decision to access private hospital
services, by comparison, is usually made by the patient and their consulting
physician. The decision reflects the patient’s willingness to pay for the expected
benefits of the hospital care, where there is a know out-of-pocket cost.

The AHSA observed that these different approaches have implications for
comparing timely access between public and private hospitals:
[Timely access] is an area where one can anticipate the private sector will produce
better results than the public sector. This is because in the private sector funding is
uncapped which means there is an incentive and additional resources to facilitate the
treatment of additional patients. In the public sector funding and throughput are capped

3 Pers. comm., Dr Yusuf Nagree, Chair Private Practice Committee, Australasian College for
Emergency Medicine, 28 August and 14 September 2009.
156 PUBLIC AND PRIVATE
HOSPITALS
 by finite levels of funding and this will reduce access to public hospitals. The
comparison and anticipated differences are thus primarily driven not by the hospitals
but the differing funding conditions in the two sectors. (sub. 1, p. 8)

Despite uncapped funding to the private sector, some patients may experience
difficulties gaining access to private hospitals. The Doctors Reform Society of
Australia (sub. DR50) referred to a survey conducted by the Victorian Branch of the
AMA. The survey found that private hospitals sometimes refuse to admit patients,
with those aged 75 and over most commonly refused admission (AMA Victoria
2001).

Many study participants noted that the need to divert resources to emergencies
can interrupt and constrain the delivery of elective surgery. In contrast, the NSW
Health Surgical Services Taskforce submitted that:
… emergency surgical admissions [are] entirely predictable and could be managed
more effectively with a planned approach thereby minimising disruption to elective
surgical services.(sub. DR43, p. 1)

The Commission examined two measures of timely access to elective surgery:

• public hospital waiting times
• occupancy rates.

Consideration of waiting times was limited to the public sector because private
hospital operators do not maintain elective surgery waiting list data. Even though
hospital operators maintain detailed morbidity data for each admitted patient (which
include a record of the date of admission), such data do not typically record the date
of the consultation during which it was decided to admit the patient, so it is not
possible to calculate the time taken to receive treatment at a private hospital.

Public hospital waiting times

Public hospitals are required by each state and territory health department to collect
data on the number of patients on their waiting lists, the clinical categories assigned
to those patients, and the time it has taken from listing to admission for elective
surgery. State and territory health departments routinely publish statistics about the
performance of public hospitals against performance targets (for example, NSW
Department of Health nd; DHS (Victoria) 2009).

Elective surgery waiting times data are collected under the National Elective
Surgery Waiting Times Data Collection, and are widely published (AIHW 2008e;
SCRGSP 2009; DOHA 2009f). Three elective surgery waiting list indicators are
presented:
OTHER PARTIAL 157
INDICATORS
• net growth (change) in the number of people on waiting lists
• average number of days waited for an admission
• proportion of people being admitted from a waiting list within the benchmark
times, by clinical urgency.

The number of people seeking elective surgery grew by 4.8 per cent per year
between 2002-03 and 2007-08, with particularly significant growth in 2006-07 and
2007-08 (table 7.7). In contrast, the number of elective surgery admissions only
grew by 1.8 per cent per year. After allowing for ‘removals’, there was a net
increase in the number of people waiting for elective surgery between 2002-03 and
2007-08.

The reasons why people were removed from a waiting list is also instructive. The
number of people who were taken off the waiting list because they were admitted as
an emergency patient, sought treatment outside the public hospital system or could
not be contacted (or were presumed to have died) grew more quickly than the
number of patients being admitted to public hospitals for elective surgery
(table 7.7).

Table 7.7 Additions and removals from public hospital elective surgery
waiting lists, 2002-03 to 2007-08
2002-03 2003-04 2004-05 2005-06 2006-07 2007-08 Rate of
change

no. no. no. no. no. no. %

Total additions 586 744 608 680 621 015 638 904 734 715 740 952 4.8

Total removals 601 972 618 180 645 340 657 401 650 973 661 275 1.9

Elective admission 517 503 528 949 549 746 556 953 556 770 565 501 1.8

Emergency admission 3 541 3 985 6 757 6 154 5 909 5 650 9.8

Not contactable/died 7 142 6 832 8 488 9 792 9 036 9 514 5.9

Treated elsewhere 14 217 15 842 22 537 26 565 21 015 22 520 9.6

Other removals 59 569 62 572 57 812 57 937 58 243 58 090 -0.5

Net change to list -15 228 -9 500 -24 325 -18 497 83 742 79 677 ..

.. Not applicable

Source: AIHW (2004, 2005, 2006, 2007a, 2008b, 2009a); Productivity Commission estimates.

The second indicator reports how long people waited before being admitted for
elective surgery. The average number of days waited between 2002-03 and 2007-08
increased from 28 to 34 days for the 50th percentile patient and from 197 days to
158 PUBLIC AND PRIVATE
HOSPITALS
235 days for the 90th percentile patient (table 7.8). The proportion of admitted
patients that had to wait longer than a year declined from 4 per cent to 3 per cent,
though it increased in South Australia, the Northern Territory and the ACT
(AIHW 2004, 2009a). Study participants suggested that policy decisions to focus
efforts on reducing the number of people waiting longer than a year may have led to
the observed increases in average waiting times.

Table 7.8 Elective surgery waiting times, public hospitals, 2002-03 and
2007-08
NSW Vic Qld SA WA Tas NT ACT Aust

Days waited at 50th percentilea

2002-03 29 28 21 34 27 42 45 48 28
2007-08 39 33 27 42 30 36 43 72 34
Average growth rateb 6.1 3.3 5.2 4.3 2.1 -3.0 -0.9 8.4 4.0
Days waited at 90th percentilea
2002-03 227 197 113 181 207 389 305 300 197
2007-08 278 221 137 208 206 369 337 372 235
Average growth rate b 4.1 2.3 3.9 2.8 -0.1 -1.1 2.0 4.4 3.6
Percentage waited more than 365 days
2002-03 4.2 4.2 2.6 3.0 3.9 10.9 7.0 7.1 4.0
2007-08 1.8 3.6 2.3 3.9 3.0 10.1 8.6 10.3 3.0
Average growth rateb -15.6 -3.0 -2.4 5.4 -5.1 -1.5 4.2 7.7 -5.6
a Average wait in days. b Average annual rate of growth between 2002-03 and 2007-08.

Source: AIHW (2004, 2005, 2006, 2007a, 2008b, 2009a); Productivity Commission estimates.

The length of waiting lists, however, does not provide an indication of how long a
patient actually waited for elective surgery. Waiting times do not take into account
the time waited between the referral to the surgeon and the appointment with the
surgeon, or the time between the appointment and being placed on the waiting list
(AIHW 2008e; Victorian Auditor-General 2009).

The third indicator is a measure of how well hospitals met their respective targets to
reduce waiting lists for each of three clinical urgency categories. The three
generally accepted urgency categories for elective surgery are:
• category 1 (urgent) — admission is desirable within 30 days
• category 2 (semi-urgent) — admission is desirable within 90 days
• category 3 (non-urgent) — admission at some time in the future is acceptable.

There is no national benchmark for admitting non-urgent cases but the term
‘extended wait’ is used for patients waiting longer than 12 months and some
jurisdictions do set targets for non-urgent cases (SCRGSP 2009).
OTHER PARTIAL 159
INDICATORS
The data suggest that the proportion of:
• patients still on waiting lists who have waited longer than their recommended
times in all clinical urgency categories declined (improved) between 2002-03
and 2006-07 for all jurisdictions (except the Northern Territory which increased,
and Queensland and Tasmania for which there was little change)
• patients admitted after having waited longer than their recommended times
increased in most jurisdictions (except for NSW, for which this proportion
decreased, and Tasmania, for which there was little change) (table 7.9).

Table 7.9 Elective surgery waiting times by clinical category, public

hospitals, 2002-03 and 2006-07a
NSW Vic Qld SA WA Tas NT ACT

Percentage of patients on waiting lists with extended waits

2002-03
Category 1 – 30 day 38.9b – 2.3 17.0 39.2 52.0c 57.8 6.6
Category 2 – 90 day 40.2b 39.1 5.3 22.1 48.0 66.0c 52.0 55.8
Category 3 10.6b 27.1 38.2 18.3 29.7 31.0c 26.5 32.5
Total 22.7b 31.1 26.0 18.8 34.8 49.0c 35.8 41.5
2006-07
Category 1 – 30 day 5.1 – 6.4 21.6 26.2 39.7 53.7 6.8
Category 2 – 90 day 28.9 34.0 20.5 16.8 46.2 64.8 51.7 54.0
Category 3 0.2 10.5 32.5 11.3 6.5 32.0 39.3 24.3
Total 8.5 20.5 25.6 13.5 21.9 48.8 45.9 38.7
Percentage of admitted patients with extended waits
2002-03
Category 1 – 30 day 38.9b – 9.3 13.5 15.3 28.0c 14.5 8.8
Category 2 – 90 day 40.2b 20.5 11.8 15.6 23.5 43.0c 24.0 47.1
Category 3 10.6b 8.8 13.0 4.9 6.4 23.0c 14.6 18.1
Total 22.7b 12.4 11.1 10.1 13.6 32.0c 17.9 26.6
2006-07
Category 1 – 30 day 12.9 – 13.2 22.5 28.8 25.0 19.2 7.2
Category 2 – 90 day 25.5 25.3 17.7 22.1 44.0 46.1 43.0 49.1
Category 3 4.4 8.5 11.7 9.5 24.3 22.6 39.9 30.4
Total 14.2 14.5 14.9 17.4 31.6 32.4 31.1 32.4
a Care must be taken when comparing between jurisdictions because of differences in how patients are
assigned to clinical categories. b 2004-05 data for NSW. c 2005-06 data for Tasmania. – Nil or rounded to
zero.
Source: SCRGSP (2005, 2009).

The AIHW commented that there ‘is evidence of considerable variation in the
assignment of clinical urgency categories’ (AIHW 2008e, p. 3), and that the lack of

160 PUBLIC AND PRIVATE

HOSPITALS
comparability of clinical urgency categories means that indicators are ‘not
meaningful or comparable across jurisdictions’ (AIHW 2008e, p. 4).

The Special Commission of Inquiry into Acute Care Services in NSW Public
Hospitals (the Garling Review) was mindful of the quality of waiting list data, and
recommended:
NSW Health should institute an audit program of waiting lists kept for each hospital in
NSW, conducted by staff who are not associated with the relevant area health service
or the hospital. The audits should examine all paperwork that the hospital is required to
maintain for the waiting lists including correspondence with referring doctor, and
should include the auditing of any reclassification of patients’ clinical urgency
category. (emphasis added) (Garling Review 2008, Recommendation 82, p. 54)

Occupancy rates

The demand for hospital services — like fire, ambulance and police services — is
highly variable and is often difficult to predict. In the presence of this
unpredictability, a common strategy among these services is to ensure that the
service has sufficient spare capacity. Even though increases in capacity utilisation
are sometimes desirable from a productivity perspective, too high a capacity
utilisation for a hospital can lead to problems with timeliness of access.

A measure of the capacity of a hospital to provide timely access is its occupancy

rate, which is the number of patient days per bed divided by 365 times 100. An
occupancy rate of 85 per cent has been suggested as an optimal target (for example,
Baghurst, Place and Posnett 1999; Sprivulis et. al 2006). Forero and Hillman (2008)
observed:
There is clear evidence that occupancy rates in most urban public hospitals are greater
than 85 per cent. When occupancy rates exceed 85 per cent, regular bed shortages and
periodic bed crises are expected. If average bed occupancy rises to 90 per cent or more,
access block crises are routinely expected. Spare bed capacity is essential for the
effective management of emergency admissions and to have surge capacity. (Forero
and Hillman 2008, p. 1)

Occupancy rates are greater than 85 per cent for public hospitals in most
jurisdictions, although these have been declining in recent years. Conversely,
relative occupancy rates for private hospitals were less than 85 per cent and have
been increasing in recent years (table 7.10).

OTHER PARTIAL 161

INDICATORS
Table 7.10 Occupancy rates, 2002-03 to 2007-08a
2002-03 2003-04 2004-05 2005-06 2006-07 2007-08

Public hospitals
NSW 92 87 82 88 88 85
Vic 98 99 100 98 98 96
Qld 81 77 81 78 80 77
SA 83 82 81 83 82 83
WA 91 93 91 94 95 89
Tas 92 90 85 90 87 83
NT 88 94 92 94 91 89
ACT 99 103 108 118 118 116
Australia 90 88 87 89 89 87
Private hospitalsb
NSW 72 71 73 76 80 84
Vic 77 76 75 78 84 76
Qld 77 78 80 78 79 82
SA 71 73 74 65 67 59
WA 77 75 77 68 70 83
Tas, NT and ACT np np np 53 53 55
Australia 72 71 72 74 76 76
a It is possible for a hospital to have an occupancy rate of more than 100 per cent because beds may have
more than one same-day separation, and same-day separations are counted as having a length of stay of one
day. b Includes private acute and psychiatric hospitals. np Not published.
Source: Productivity Commission estimates based on AIHW (2004, 2005, 2006, 2007a, 2008b, 2009a).

Care, however, needs to be exercised when interpreting occupancy rates. Hospitals

that specialise in same-day separations will be observed to have higher rates since
beds may have more than one same-day occupant, and same-day separations are
usually counted as having a length of stay of one day. As a result, the occupancy
rates for the private sector as a whole are likely to be relatively high. Similarly,
occupancy rates are likely to be higher in jurisdictions that classify people receiving
same-day treatments (such as chemotherapy and renal dialysis) as admitted patients
rather than outpatients.4

Also, a benchmark of 85 per cent occupancy rate was applied for all hospitals,
including those that did not operate emergency departments. Hospitals that have a
more predictable patient flow can, arguably, operate to higher capacity levels.
However, effective management of emergency surgical admissions can minimise

4 Victoria admits patients for treatments that other jurisdictions may administer as non-admitted
(outpatient) services, such as chemotherapy and dialysis, and so a disproportionate share of
Victorian separations may be categorised as admitted-patient services (Victorian Department of
Health, pers. comm., 30 September 2009).
162 PUBLIC AND PRIVATE
HOSPITALS
disruption to elective surgical services (NSW Health Surgical Services Taskforce,
sub. DR43, p. 1).

Equitable access to elective surgery

The Commission examined two measures of equity of access to elective surgery:

• the affordability of private hospital insurance
• elective surgery separation rates of public and private hospitals.

Affordability of private hospital insurance

Two measures of the affordability of private hospital services are:

• the costs of private hospital insurance
• the proportion of the population with private hospital insurance.

In the case of the first indicator, the higher is the out-of-pocket cost of private
hospital insurance the lower is the financial accessibility of private hospital
services. The out-of-pocket cost for private hospital insurance rose from $957 to
$1311 per policy or $451 to $631 per insured person between 2002-03 and 2007-08,
after deducting for the private health insurance rebate (table 7.11). Since the
average price of private hospital insurance rose slightly more quickly than average
weekly earnings, the relative cost of private hospital insurance also rose slightly.

Drawing on data published by the Private Health Insurance Administration Council,

the proportion of the Australian population that was covered by private hospital
insurance (with or without ancillary cover) grew from 43.4 per cent of the
population in 2002-03 to 44.7 per cent in 2007-08 (table 7.11).

The data indicate that even though financial accessibility slightly worsened during
this period, the accessibility of private hospital services improved with the increased
take-up of private hospital insurance.

OTHER PARTIAL 163

INDICATORS
Table 7.11 Average cost and population coverage of private hospital
insurance, 2002-03 to 2007-08a
Average
annual
2002-03 2003-04 2006-07 2007-08 change

Cost per policy %

Current pricesb $ 957 1 053 1 274 1 311 6.5
Share of earningsc % 2.6 2.7 2.9 2.8 2.2
Cost per insured person
Current pricesb $ 451 497 608 631 6.9
Share of earningsc % 1.2 1.3 1.4 1.4 2.6
Population coveraged
No. of persons covered ‘000 8 639 8 627 9 145 9 534 2.0
Share of population % 43.4 42.9 43.5 44.7 ..
a Hospital-only insurance, excludes ambulance and general (ancillary) care. b After deducting for private
health insurance rebate. The deduction for the rebate is allocated on a pro-rata basis between hospital and
general (ancillary) health insurance components. c Average annual earnings, all persons, full time and non-
full-time workers. d Includes hospital-only and hospital and ancillary cover. .. Not applicable
Source: PHIAC (2003, 2004, 2007a, 2008); ABS (Average weekly earnings, Cat. no. 6302.0); Productivity
Commission estimates.

Elective surgery separation rates

Separation rates for public and private hospitals is an indirect measure of hospital
usage and therefore equity of access. Its advantage is that it provides a consistent
basis for comparing levels of equity of access across sub-populations such as
socioeconomic status, Indigenous status, gender, and remoteness status of the
patient, as well as the hospital.

Separation rates are defined as the number of separations per 1000 population and
are standardised for age. This approach ‘incorporates an assumption that levels of
“need” are the same, on average, for different populations, or that variation in need
can be accounted for using data analysis (such as age standardisation)’
(AIHW 2008e, p. 6).

Australian hospitals provided approximately 1.6 million elective surgery separations

in 2004-05 (the latest year for which the AIHW published these statistics). Private
hospitals undertook more elective surgery separations (48.3 separations per 1000
population, age-standardised) than public hospitals (31.0 separations per 1000
population) (table 7.12).

164 PUBLIC AND PRIVATE

HOSPITALS
Table 7.12 Elective surgery separation statistics, 2007-08a
Public elective Private elective All elective
surgical surgical surgical
separations separations separations
By patient’s remoteness categoryb
Major cities 26.0 54.1 80.1
Inner regional 34.3 49.1 83.4
Outer regional 36.7 40.5 77.2
Remote 35.3 32.5 67.8
Very remote 30.3 20.5 50.8
All patients 29.0 52.0 81.0
By patient’s socioeconomic statusc
Most disadvantaged 37.9 37.9 75.7
Second most disadvantaged 34.0 45.0 79.0
Middle quintile 30.6 51.0 81.6
Second most advantaged 24.9 55.7 80.6
Most advantaged 16.9 69.1 86.0
All patients 29.0 52.0 81.0
By patient’s Indigenous status
Indigenous Australians 38.1 9.9 47.9
Other Australians 27.7 50.2 77.9
All persons 27.8 49.6 77.4
By patient’s genderd
Male 27.3 42.8 70.1
Female 34.9 54.6 89.4
All patients 31.0 48.3 79.2
a Rates of separation per 1000 people, age-standardised to the estimated resident population 30 June 2001.
b Australian Standard Geographical Classification, Remoteness Areas. Coverage of the linked data by
remoteness areas ranged from 60 per cent in Remote areas to 100 per cent in Major cities. c Socio-Economic
Indexes for Areas (SEIFA) classification. Coverage by quintile of socioeconomic advantage/disadvantage
ranged from 85 per cent for the Middle quintile to 100 per cent for the Most advantaged quintile. d 2004-05.
Source: AIHW (2008e, 2009a).

As a broad generalisation, patients from the most disadvantaged socio-economic

areas and regional and remote areas were more likely to use public hospitals.
Conversely, patients who were from more advantaged socio-economic areas,
non-Indigenous or from the major cities were more likely to use private hospitals
(table 7.12).

Consequences of hospital congestion

Hospitals operate a number of units and wards. The ability of a unit or ward to treat
patients depends in part on its ability to refer them from one area of the hospital to
another. For example, the ability of an emergency department to admit a patient

OTHER PARTIAL 165

INDICATORS
depends on the availability of a ward bed. Similarly, the ability of a patient to be
transferred from an emergency department or general ward to an intensive-care unit
(ICU) bed depends on the availability of an ICU bed.

The ACHS reports several statistics that describe the extent of congestion (or ‘bed
blocking’) within a hospital. As noted in chapter 6, the ACHS data are based on
relatively few hospitals (between 30 and 60 hospitals for these statistics). The
collection is voluntary, so there is a risk of sample selection bias. Finally, the
statistics do not account for differences in casemix. The ACHS (sub. 13) noted that
its data are not intended for benchmarking purposes but rather internal review
processes.

The ACHS (2008) reported that the percentage of emergency department patients
who waited longer than eight hours to be admitted increased from 25 per cent to
33 per cent between 2002 and 2007.5 Except for 2006, there were no statistically
significant differences between public and private hospitals.

The ACHS collects and reports three statistics on the accessibility to and from ICU
and high-dependency unit (HDU) beds. These are the proportion of:
• patients who could not be admitted to an ICU because of a lack of ICU
resourcing
• elective surgery deferred or cancelled due to lack of ICU or HDU bed
• patients who were transferred to another facility or area due to unavailability of
ICU or HDU bed (ACHS 2008).

Even with the relatively small sample, there were statistically significant differences
between public and private hospitals. For example, the percentage of patients that
were not admitted to an ICU because of a lack of resourcing in public hospitals was
8.6 per cent compared to 1.2 per cent in private hospitals.

The ACHS also reported on the extent of delays on discharge from the ICU or HDU
of more than 12 hours (ACHS 2008). Approximately 16 per cent of patient transfers
from ICU or HDU to hospital beds were delayed in 2007, across the 38 reporting
public and private hospitals. There was a statistically significant difference between
public and private hospitals. The rate for private hospitals was 3.5 per cent
compared to 18.4 per cent for public hospitals.

5 Including patients who waited longer than eight hours and were planned for admission but
discharged from an emergency department without reaching an inpatient bed, were transferred to
another hospital for admission, or died in the emergency department.
166 PUBLIC AND PRIVATE
HOSPITALS
 FINDING 7.3

Timely access to elective surgery is less likely in public hospitals than in private
hospitals. The relatively high bed occupancy rates in public hospitals restrict their
ability to manage their unpredictable workload. Equity of access is more likely in
public hospitals than private hospitals, since public hospitals provide relatively
more elective surgery to patients from poor socioeconomic areas and from more
remote areas of Australia.

7.3 Quality and patient safety

Hospital-acquired infections, an important aspect of quality and patient safety, were
discussed in chapter 6. This section examines a selection of other quality and patient
safety indicators of hospital care. The Melbourne Institute of Applied Economic and
Social Research noted that measuring quality and patient safety is complex and that
there is no concise measure:
… hospital quality is a multifaceted concept that covers aspects such as effectiveness of
treatment, timeliness of service delivery, quality of amenities, technological
sophistication, incidences of in-hospital adverse events and so on. Constructing,
comparing and synthesizing measures across different quality dimensions are a
challenging task …
The difficulty is compounded by the fact that often within a given quality dimension
there exist multiple measures and/or outcomes. For example, in the case of in-hospital
adverse events, there are more than 20 common measures covering four different
aspects: hospital-acquired infections, operative and post-operative complications,
sentinel events and obstetrics … (sub. 16, p. 6)

The Australian Commission on Safety and Quality in Health Care (ACSQHC) has
worked with the AIHW to develop a set of indicators that can be used to measure
hospital quality and patient safety (AIHW 2009g). Such a framework is intended to
be useful for public and for private hospitals. In addition, there are indicator
frameworks used for the NHA (appendix B); by the National Health Performance
Committee and the SCRGSP (appendix C); and by the Women’s and Children’s
Hospitals associations, the Australian Healthcare and Hospitals Association, the
Australian Private Hospitals Association, and state and territory health departments.

Given the potentially very large number of possible indicators, the Commission has
selected those indicators that best indicate whole-of-hospital performance. That is,
the indicators that can be widely applied and are not disease or injury specific. In
addition, the Commission has reported only those indicators for which data are
published for both public and private hospitals (box 7.2).

OTHER PARTIAL 167

INDICATORS
 Box 7.2 Quality and patient safety indicators
Indicators for which the Commission has reported publicly available data include:
• accreditation
– the proportion of hospitals and beds that are accredited
• readmission and returns
– the rates of unplanned readmissions to hospital within 28 days of a surgical
admission
– the rates of unplanned return to theatre or operating room during an admission
– the rates of unplanned return to an ICU within 72 hours of discharge
• adverse events
– patient falls
– pressure ulcers
– complications of blood transfusion
– adverse drug events
• intentional self-harm (and suicide)
• obstetrics
– foetal, neonatal and perinatal mortality rates
– caesarean section rates

Accreditation

Accreditation indicates that a hospital regularly reviews its programs, services and
organisation to ensure processes are in place to support quality of care to patients
(ACSQHC 2008). Accreditation does not mean that errors do not occur, but that
processes are in place to support quality care and that those processes are checked
regularly.

Hospital accreditation is currently available through a number of providers, with

most public and private hospitals seeking accreditation as members of the ACHS. In
January 2008, 95 per cent of public hospitals and day procedure facilities and 97 per
cent of private hospitals and day procedure facilities were accredited (ACSQHC
2008). Since June 2008, the Private Health Insurance (Accreditation) Rules 2008
(Cwlth) have required hospitals that provide privately insured services to be
accredited.

The ACSQHC noted that it is currently developing an alternative model for safety
and quality accreditation (ACSQHC 2008).

168 PUBLIC AND PRIVATE

HOSPITALS
Unplanned readmissions and returns

The three indicators of unplanned readmissions and returns include:

• unplanned readmissions after 28 days of a surgical admission
• unplanned return to ICU (or HDU) within 72 hours of discharge
• unplanned return to operating room during an admission.

An indicator of whether a hospital’s care is clinically effective is the extent to which

a discharged patient is subsequently readmitted for the same or substantially similar
clinical condition in a relatively short space of time (SCRGSP 2009). One such
measure is the unplanned (and unexpected) readmission to hospital within 28 days
of a surgical admission. Unplanned/unexpected readmissions are included in the
reporting structure for the NHA (COAG 2008d).

There is currently only one national collection that reports this indicator. The ACHS
reported that unplanned and unexpected hospital readmissions within 28 days
declined from 2 per cent in 2003 to 1.2 per cent in 2007 (ACHS 2008). From a
sample of between 310 and 334 hospitals, of which approximately 170 were private,
there were no statistically significant differences observed between public and
private hospitals between 2003 and 2005, though private hospitals reported lower
rates of readmission in 2006 and 2007.

A high rate of unplanned readmission into an ICU may reflect less than optimal
management of a patient (including ward management), or a patient’s early
discharge to accommodate other ICU admissions (ACHS 2008). The rate of
unplanned readmission to ICU is defined as the number of unplanned readmissions
into an ICU within 72 hours of discharge from an ICU divided by the number of
admissions into an intensive care unit.

The rate of unplanned readmissions to an ICU was 1.7 per cent in 2007
(ACHS 2008). This rate has been relatively constant since 2001. Though some
jurisdictions exhibited statistically significant different rates relative to the group
average, there were no statistically significant differences reported for public and
private hospitals.

A related indicator is the extent to which patients are unexpectedly returned to an

operating room (theatre) after surgery during the same admission. Unplanned return
to theatre has been proposed by the AIHW for inclusion in a national set of safety
and quality indicators (AIHW 2009g).

As with unplanned readmission to hospital, there is only one national data collection
that reports this indicator. The ACHS reported that 0.41 per cent of patients
OTHER PARTIAL 169
INDICATORS
experienced an unplanned return to theatre in 2007 (ACHS 2008). From a sample of
between 274 and 291 hospitals, the rate remained largely unchanged between 2003
and 2007, and there were no statistically significant differences between public and
private hospitals during this period.

Limitations of indicators

There are four limitations to the reported statistics on readmissions and returns to
theatre.

First, no account has been made for the considerable differences between hospitals
in the policy environments in which they operate, the diseases they treat, and their
geographic locations. To ensure that indicators meaningfully reveal the underlying
characteristics of a hospital rather than those of its external operating environment,
it is important that there is some process for standardising or adjusting for such
differences. The meaningfulness of these statistics would be improved if they were
adjusted for:
• hospital casemix (for example, specific diagnostic categories such as chronic
heart failure and chronic obstructive pulmonary disease are believed to be at
higher risk of readmission)
• patient-risk characteristics, such as age, gender and comorbidities
• the extent to which the patient receives ongoing support outside of hospital after
discharge (such as from outpatient clinics, community and family support
services)
• the extent to which the patient complies with the prescribed self-management
strategies (Hasan 2001; SCRGSP 2009).

Second, while 300 or so hospitals represent a relatively large sample of public and
private hospitals, given the voluntary nature of the ACHS Clinical Indicator
Program, it is likely that there is an element of self-selection in the data. Statistics
based on a census would provide a more accurate measure of the differences
between public and private hospitals.

Third, for readmission rates to be meaningful, care must be taken to distinguish

between a readmission and the recurrence of a chronic condition (such as asthma)
(AHSA, sub. 1).

Fourth, the reported readmission rates are likely to be understated, since a number
of patients are admitted to another hospital and those admissions are not counted
towards the readmission rates. The Commission sees merit in using linked datasets,

170 PUBLIC AND PRIVATE

HOSPITALS
such as the Western Australian Data Linkage System and the NSW and ACT Centre
for Health Record Linkage, to establish more reliable estimates of readmissions to
hospital.

Adverse events

According to the Royal Australasian College of Surgeons (RACS 2008, p. 13), ‘an
adverse event is defined as the unintentional harm arising from an episode of
healthcare and not due to the disease process itself’. According to the AIHW
(2009a, p. 53), adverse events:
… include infections, falls resulting in injuries, and medication and medical device
problems. Some of these adverse events may be preventable.

There are several data collections that provide data on the prevalence, causes and
consequences of adverse events. These include:
• National Hospital Morbidity Data Collection (NHMD) — a national data
collection of episodes of hospital care that include information on the diagnoses
and treatments for adverse events
• state- and territory-based incident reporting systems — which cover the extent,
seriousness, causes and consequences of a variety of incidents, as reported by
healthcare staff. A well-known example is the Australian Incident Monitoring
System (AIMS)
• sentinel event reporting — which is the reporting of a very limited range of
serious adverse events, in which death or serious harm to a patient has occurred
• mortality reviews — such as those undertaken by medical peer committees (such
as by surgeons and anaesthetists), in specific contexts (such as maternal
mortality review committees) and coronial inquiries (which establish the cause
of death).

A fifth source of data is the detailed clinical information compiled by physicians

during the course of surgical and medical care for use in detailed audit reports. An
example is the Australian Orthopaedic Association’s National Joint Replacement
Registry.

National Hospital Morbidity Data Collection

The NHMD is one of two national collections of adverse events that can provide
insights into public and private hospitals. The AIHW regularly publishes data on the

OTHER PARTIAL 171

INDICATORS
percentage of hospital separations for which there were associated adverse events
(AIHW 2009a) (table 7.13).

Table 7.13 Hospital separations with an adverse event, 2002-03 to

2007-08a, b
Per cent
2002-03 2003-04 2004-05 2005-06 2006-07 2007-08

Public hospitals
External cause codesc
Adverse drug, medicament or 1.4 1.6 1.7 1.8 1.8 1.8
biological substance effects
Misadventures to patients 0.2 0.2 0.2 0.2 0.2 0.2
Procedures causing abnormal
3.1 3.3 3.4 3.4 3.4 3.2
reactions/complications
Other external causes 0.1 0.1 0.1 0.1 0.1 0.1
Place of occurrence codesd 4.4 4.7 5.0 5.3 5.4 5.2

Diagnosis codese 5.1 5.4 5.6 5.8 5.9 5.6

Private hospitals

External cause codesc

Adverse drug, medicament or
0.6 0.7 0.7 0.7 0.7 0.6
biological substance effects
Misadventures to patients 0.1 0.1 0.1 0.1 0.1 0.1
Procedures causing abnormal
2.6 2.6 2.7 2.8 2.8 2.7
reactions/complications
Other external causes – – – – – –
Place of occurrence codesd 3.1 3.3 3.4 3.6 3.6 3.5

Diagnosis codese 3.6 3.6 3.7 3.9 3.8 3.7

a Separations for which the care type was reported as Newborn with no qualified days, and records for
Hospital boarders and Posthumous organ procurement have been excluded. b Percentages are not equal
because multiple diagnoses and external causes can be recorded for each separation and external cause
codes can be used together to describe an adverse event. c Includes ICD-10-AM categories Y40 to Y59,
Y60 Y82, Y83-Y84, Y88 and Y95. d Includes ICD-10-AM codes Y92.22. e Includes ICD-10-AM codes E89,
G97, H59, H95, I97, K91, M96, N99, T81.0, T81.4, T82-T85, and others included above (T80 to T88 and
T98.3). – Nil or rounded to zero.
Source: AIHW (2004, 2005, 2006, 2007a, 2008b, 2009a).

The data are categorised by ICD-10-AM codes for diagnoses, places of occurrence,
and external causes of injury and poisoning. These codes indicate that an adverse
event was treated and may have occurred during hospitalisation (AIHW 2009a). The
data do not, however, provide any indication of the seriousness and therefore
consequences of the events.

An adverse event may be recorded for one classification but not for another, so the
aggregates for the external cause, place of occurrence and diagnosis codes may not

172 PUBLIC AND PRIVATE

HOSPITALS
be equal. For example, the rates for place of occurrence tend to be lower than the
diagnosis codes, indicating that not all diagnosed adverse events occurred in a
health-care setting.

It would appear from these data that the incidence of adverse events was higher in
public hospitals than for private hospitals. For example, adverse events were coded
as having occurred in over 5 per cent of separations for public hospitals and
between 3.4 and 3.7 per cent of separations for private hospitals in 2007-08
(table 7.13).

However, there are three reasons to treat these estimates with great care. First, the
AIHW noted that the estimates probably under-represent the incidence of adverse
events because other ICD-10-AM codes may indicate the occurrence of an adverse
event or its treatment (AIHW 2009a). Second, the data do not distinguish between
events that occurred in hospitals and those that did not. Third, as noted, the data do
not directly measure the consequences of the adverse events.

Wilson et al. (1995) and Jackson (2008) have each estimated a higher prevalence of
adverse events than those reported above. Wilson et al. (1995) reviewed the medical
records of over 14 000 admissions to 28 hospitals in NSW and South Australia.
They estimated that approximately 8.3 per cent of all separations involved
in-hospital adverse events. About 4.9 per cent of adverse events resulted in
mortality and 13.7 per cent resulted in a permanent disability.

Jackson (2008) used Victorian and Queensland data (which contained a flag to
indicate if a condition arose during hospitalisation) to develop a method for
quantifying and classifying hospital-acquired diagnoses, for the purpose of
estimating the cost of adverse events. Jackson (2008) found that adverse events
occurred in 12.3 per cent of separations. The introduction of a condition-onset flag
in the NHMD from 2008-09 (DOHA sub. 32) should greatly improve the
availability and reliability of the coded data on adverse events and hospital-acquired
diagnoses.

Incident reporting data

State and territory governments have adopted a variety of incident monitoring

systems for public hospitals (see for example, NSW Department of Health 2005b;
DHS 2008; LAOS Victoria 2008; SA Department of Health 2006; OSQHC 2006),
and many private hospitals have adopted their own systems. Data reported in ACHS
(2008) draw on incident monitoring systems. Reported indicators include:
• adverse drug events

OTHER PARTIAL 173

INDICATORS
• adverse transfusion events
• patient falls
• pressure ulcers
• intentional self-harm (including attempted and actual suicide).

An examination of ACHS (2008) data suggests that there are no statistical

differences between public and private hospitals for any of these indicators.
However, there are also four reasons to be cautious about using incident reporting
data such as those published by the ACHS. First, these data are self-reported by
doctors and nurses. Although doctors and nurses are encouraged to record adverse
events, the self-reporting nature of these collections means that there is likely to be
a degree of under-reporting and differences in reporting rates between professions.
For example, in a report on adverse events, the Office of Safety and Quality in
Health Care (Western Australia) (2006) noted that 91 per cent of AIMS events were
reported by nurses. The majority of these were patient falls, medication errors and
behaviour-related incidents. The Auditor-General for Western Australia (2007)
estimated that only about one-third of all adverse events are reported.

Second, the ACHS data are based on a voluntary survey of between 150 and
350 hospitals. Third, the data are not adjusted for possible differences arising out of
casemix. Finally, adverse event data rely on the self-reporting of incidents as part of
a hospital’s incident reporting system.

Mortality ratios

Mortality statistics are regularly published by the AIHW (2007b). However,

in-hospital mortality rates, with the exception of foetal, neonatal and perinatal
mortality rates (SCRGSP 2009), are not generally reported.

Mortality statistics are potentially useful partial indicators of the quality and patient
safety of hospitals. This is primarily because death is ‘unequivocal and generally
accurately reported’ (Ben-Tovim et al. 2009, p. 1).

The Commission explored the possibility of calculating and reporting it own

estimates of hospital standardised mortality ratios (HSMRs) based on data extracted
from the NHMD. However, the absence of reliable hospital-level establishment
identifiers in the NHMD meant that HSMRs could not be calculated.6 Inclusion of

6 The majority of states do not use the private hospital identifier field to identify actual hospitals.
In some states, each private hospital identifier denotes several private hospitals within some kind
174 PUBLIC AND PRIVATE
HOSPITALS
individual hospital establishment identifiers in all data submitted by jurisdictions to
the NHMD would greatly enhance the utility of these data.

Obstetric indicators

Obstetric services represent a significant aspect of hospital services. There were

over 500 000 pregnancy, childbirth and puerperium separations in Australia in
2006-07, of which about 350 000 took place in public hospitals and the remainder in
private hospitals. Yet, unlike other hospital separations, pregnancy and childbirth
are not injuries or diseases, and thus many of the quality and patient safety
indicators described above are not relevant for this area of hospital activity.

The Commission examined the reporting frameworks of the SCRGSP (2009) and
Women’s Hospitals Australasia (2007), and concluded that there were some
indicators that would be useful to report, including:
• episiotomy rates for all first births — the rate at which there is an incision of the
perineum
• third and fourth degree tears for all first births
• foetal, neonatal and perinatal death rates
• rates of caesarean section for selected first births
• rates of significant blood loss within 24 hours following a vaginal birth.

The ACHS (2008) published some of these data for public and private hospitals. Its
data suggest that the proportion of primiparous patients with an intact lower genital
tract declined between 2001 and 2007 from 30.5 per cent to 26.3 per cent. While the
incidence of second and third degree tears among primiparous patients was
observed to have increased during this period, there is no consistent evidence to
suggest a statistically significant difference between public and private hospitals.

Robson, Laws and Sullivan (2009) used 2001–2004 data from the National
Perinatal Data Collection to study the outcomes of women delivering a single baby.
They found that the rate of third and fourth degree perineal injury was 1.4 per cent
in public hospitals and 0.8 per cent in private hospitals. Likewise, after adjusting for
risk factors — such as maternal age, smoking status and number of previous
pregnancies — babies born in public hospitals were more likely to have low Apgar
scores, need high-level resuscitation or be admitted to a special care or neonatal
intensive care nursery.

of health administrative or organizational boundary, while in other states the aggregation is at a

much higher level, with very many hospitals merged together.
OTHER PARTIAL 175
INDICATORS
However, Pesce (2009) cautioned that it is not possible to extrapolate the findings
made by Robson, Laws and Sullivan to conclude that private care provided by
obstetricians, with its increased intervention rates, prevents perinatal mortality and
morbidity in Australia. Tracy et al. (2009) and Chambers (2009) questioned the
findings of the Robson, Laws and Sullivan study, considering that failure to account
for the effects of low birthweight undermines its conclusions. Likewise, Evans,
Malcolm and Gordon (2009) noted that less than 10 per cent of stillbirths occur
during delivery, suggesting that differences in perinatal death rates are not related to
hospital care or interventions performed during labour. Nevertheless, Coory et al.
(2009) considered that:
… the results from [Robson, Laws and Sullivan’s] article might be due to confounding,
but they should not be dismissed and should be investigated with more detailed clinical
data.

The AIHW (2008d) reports foetal, neonatal and perinatal mortality rates for public
and private hospitals. Each of these rates is lower in the private sector (table 7.14).
However, caution should be exercised in interpreting these rates, as they have not
been adjusted for different patient-risk characteristics, such as age, gender and
comorbidities. The lack of risk adjustment is particularly problematic as there are
few level III neonatal intensive care units located in private hospitals, suggesting
that most of the sickest infants are treated in the public hospital sector.

Table 7.14 Rates of foetal, neonatal and perinatal deaths by hospital

sector, rate per 1000 birthsa, 2006
Foetal deaths Neonatal deathsb Perinatal deathsb
Public 8.5 3.2 11.8
Private 6.7 2.7 9.5
a Foetal and perinatal death rates were calculated using all births (live births and foetal deaths). Neonatal
death rates were calculated using all live births. b Numerators exclude neonatal deaths in NT. Denominators
exclude live births in NT. Except in WA, these may exclude neonatal deaths within 28 days of birth for babies
transferred to another hospital or readmitted to hospital, and those dying at home.
Source: AIHW (2008d).

There are no national data on rates of caesarean section for selected first births in
public and private hospitals. The AIHW reports caesarean section rates for all births
in public and private hospitals, although these rates are not adjusted for different
patient-risk characteristics, such as age, gender, previous pregnancies and
comorbidities. There has been a steady growth in the proportion of caesarean births
in both public and private hospitals (figure 7.1). In 1991, 16 per cent of births in
public hospitals and 22 per cent of births in private hospitals were caesarean births.
By 2006, 28 per cent of births in public hospitals and 41 per cent of births in private
hospitals were by caesarean section. Although the appropriate caesarean section rate

176 PUBLIC AND PRIVATE

HOSPITALS
is not known, the rise in caesarean section rates has prompted considerable concern
(for example, Dietz and Peek 2004).

Figure 7.1 Rates of caesarean section by hospital sector, 1991–2006

45

40
Private hospitals
35
Per cent

Australia
30

25
Public hospitals
20

15
1991 1993 1995 1997 1999 2001 2003 2005

Source: AIHW (2008d and previous issues).

The Commission approached the states and territories about obtaining unpublished
data on episiotomy rates for all first births, third and fourth degree tears for all first
births, and rates of significant blood loss. The data were not able to be provided in
time for this study.

7.4 Developments to improve future comparisons

There are a large number of potential indicators that could be used to assess the
performance of public and private hospital sectors. Most of these indicators are
specific to particular illnesses, injuries and medical care, so they are not generally
applicable for assessing the performance of hospital sectors. There remain,
nonetheless, a large number of indicators that can be used for this purpose, as the
Commission has found.

The Commission, however, encountered a paucity of comparative public and

private hospital-level statistics. Many that are published have significant limitations,
such as relatively small samples and problems surrounding the self-reporting of
data. Where the Commission has sought to obtain unpublished hospital-level data, it
faced considerable obstacles to accessing them in a timely manner. Many of these

OTHER PARTIAL 177

INDICATORS
obstacles did not appear reasonable on privacy or confidentiality grounds, given the
Commission’s intended use and reporting of the data.

There are some initiatives underway that will lead to wider availability of hospital
performance indicators. For example, implementation of the NHA, under which
governments have agreed to report nationally-consistent ‘progress measures’
through the COAG Reform Council, should improve the reporting of partial
productivity and access indicators for the public sector. It would be useful if the
same methodologies were used when such data are collected from private hospitals.

There has also been progress towards national safety and quality indicators, with the
Australian Health Ministers’ Conference agreeing on 13 November 2009 to
fast-track the implementation and reporting of a core set of nine national indicators
of safety and quality for public hospitals (DOHA sub. DR69). Another set of
hospital performance indicators is being prepared for the National Healthcare
Agreement (appendix B). Additionally, the recent introduction of the condition-
onset flag in the NHMD should greatly improve the availability and reliability of
data on adverse events and hospital-acquired diagnoses.

Some of the quality and patient safety indicators proposed to the Commission are
comparatively new for Australia, such as the open disclosure of adverse events.
Open disclosure refers to the open discussion of incidents that result in harm to a
patient while receiving health care. A successful open disclosure regime, as
promoted by the Centre for Health Communication (University of Technology,
Sydney), could yield significant benefits to the community, because the absence of
such disclosure has been identified ‘as a major reason for patients and family
members to file complaints and pursue legal action’ (Centre for Health
Communication, sub. 3, pp. 1-2).

Yet developing quality and patient safety indicators alone is not sufficient.
Long-term improvements to health outcomes need comprehensive public reporting
of quality and patient safety by all hospitals. The ACSQHC (sub. 24), for example,
argued for the reporting of safety and quality by both public and private hospitals.

The reporting of data should not be confined to jurisdictional or sectoral level data,
as is the case with this report, but should be principally at the hospital level.
Hospitals vary significantly, and reporting broad statistics masks the major variation
that can occur between hospitals, as observed by the ACSQHC (sub. 24). It is
hospital-level data, not jurisdictional, that health care consumers, providers, funders
(private health insurers and governments), regulators and policy makers need to
inform their decisions.

178 PUBLIC AND PRIVATE

HOSPITALS
The Commission notes that the introduction of market mechanisms, like
activity-based funding, may provide an incentive to expand the role for quality and
patient safety data — particularly where quality and patient safety outcomes are tied
to funding. Yet the hospital sector is likely to remain highly regulated, so there will
remain a role for governments to facilitate this information provision.

FINDING 7.4

The work of the Australian Commission on Safety and Quality in Health Care and
the Australian Institute of Health and Welfare to develop a national set of safety
and quality indicators could provide a basis for future comparisons between public
and private hospitals. However, the paucity of published, comparable and reliable
hospital-level data severely limits these comparisons, and will continue to limit such
comparisons in the future. Making consistent hospital-level data available to all
interested parties would assist with future comparisons between hospital sectors
and contribute to improvements in care.

OTHER PARTIAL 179

INDICATORS
8 Multivariate analysis

Key points
• Multivariate analysis can overcome some of the shortcomings of reporting individual
partial indicators by generating a single measure of performance that
simultaneously accounts for the diversity of hospital activity and the range of factors
outside the control of hospitals.
• The Commission obtained the permission of 122 for-profit and not-for-profit private
hospitals to use their hospital-level morbidity data. This group accounts for 60 per
cent of private hospital sector separations.
• Together with data for 368 public acute hospitals and 18 non-government hospitals
that provide public hospital services, the Commission had access to a unique
dataset of 508 public and private acute hospitals.
• The Commission modelled a number of factors using stochastic frontier analysis.
The factors considered included:
– outputs — admitted patient, emergency department and outpatient activities
– inputs — labour inputs, drug, medical and surgical supplies, and beds
– quality and patient safety — in-hospital mortality
– patient risk characteristics — comorbidity scores and socioeconomic status of
patients
– roles and functions — whether a hospital had a teaching role, and the complexity
of services of the hospital.
• Factors that were found to influence hospital performance include higher
comorbidity rates among patients, patients from lower socioeconomic areas and a
greater proportion of medical cases and complex cases.
• Using data for 2006-07, it was estimated that, on average, the technical efficiencies
of the hospitals within the sample were about 20 per cent below best practice.
• The efficiencies of public and private hospitals were broadly similar, except that:
– large and very large private hospitals were slightly more technically efficient than
public hospitals
– very small and small public hospitals were more technically efficient than private
hospitals, although this may in part reflect the way such hospitals were modelled
in the analysis.
• The Commission will undertake further analysis using data for 2003-04 to 2006-07,
where it will examine the cost performance of hospitals as well as the performance
of individual peer groups of hospitals.

MULTIVARIATE 181
ANALYSIS
The partial indicators discussed in the previous chapters are readily computable and
well understood by practitioners, but they suffer from at least two limitations. First,
since they are by definition partial, no one indicator provides an overall assessment
of a hospital’s performance. Instead, a large number of indicators that cover costs,
quality and patient safety need to be read in conjunction to infer an overall
assessment of hospital performance.

Second, there are a large range of factors outside the control of a hospital that can
influence its performance. These include the characteristics of its patients (such as
the patient’s Indigenous status, socioeconomic status, gender, age and
comorbidities), and the roles and functions of the hospital (such as whether it
provides teaching services). Many of the partial indicators presented in chapters 5 to
7 do not take account of these factors.

The multivariate analysis undertaken here has the potential to advance our
understanding of the performance of public and private hospital systems. This
chapter provides an overview of the data, methods employed in the multivariate
analysis and the findings. A more detailed discussion is provided in appendix E.

The Commission examined the scope for improving technical efficiency for
hospitals in 2006-07. The Commission intends to undertake further analysis into
hospital costs, and to use additional years of data (2003-04 to 2005-06). The results
are to be published in March 2010.

8.1 About the Commission’s multivariate analysis

To measure ‘what hospitals do’ and ‘how well they do their tasks’, the Commission
treated hospital establishments (and in some instances, campuses) as the principal
object of measurement. It is generally at the hospital level that decisions are made to
use a variety of ‘inputs’ (such as nurses, administration and clerical staff,
medications, and technologies) to produce a range of ‘outputs’ (such as surgical
medical and surgical procedures, emergency department episodes of care).

Measuring the relationship between inputs and outputs, known as the production
function approach, is a well-established method for analysing the performance of
hospitals (Hollingsworth and Peacock 2008) and establishments in other industries.
In the case of hospitals, its chief advantage is that it permits public, for-profit and
not-for-profit hospitals to be directly compared because the approach does not,
unlike the cost function approach, depend upon hospitals behaving in a particular
manner (such as minimising costs).

182 PUBLIC AND PRIVATE

HOSPITALS
Technical efficiency

The production function approach permits the calculation of each hospital’s

technical efficiency. Technical efficiency is one of several measures of efficiency,
and is the extent to which a hospital’s outputs can be increased without increasing
input use, or conversely, the extent to which inputs can be reduced without
decreasing outputs. One of the first steps in measuring technical efficiency is to
identify the factors that determine the frontiers (or best-practice benchmarks),
against which each hospital will be compared. It is recognised that a hospital’s
range of services vary markedly due to a number of reasons, including the relative
comorbidities of their patient population and the community’s expectations about
the types of services that hospitals should provide.

While technical efficiency is a measure of hospital performance, it is not a measure

of cost efficiency. Technical efficiency represents whether the most is being made
out of the hospital’s scarce resources. Although technical efficiency contributes to a
hospital’s cost efficiency, a hospital’s cost efficiency can be influenced by many
other factors, including its price and input mix. It is entirely feasible that a hospital
might be less technically efficient than another hospital, but because of superior
buying power, might still be reported to have lower costs.

Stochastic frontier analysis

The Commission used stochastic frontier analysis (SFA) to estimate the production
function and each hospital’s technical efficiency. Like any other multivariate
regression analysis, SFA allows for statistical relationships to be established
between the dependent variable and the independent variables. In addition to
establishing the ‘slope’ of the regression equation, the coefficients can also be
interpreted as establishing the best-practice benchmarks faced by each hospital. This
is the subject of section 8.4.

SFA also simultaneously determines the technical efficiency of each hospital after
controlling for:
• factors that affect hospital performance
• random variations between hospitals reflecting
– the effect of measurement error in the variables and other random factors that
affect hospital outputs, such as disease outbreaks
– the combined effects of other omitted factors, many of which cannot be
measured.

MULTIVARIATE 183
ANALYSIS
Compared with alternative methods, such as data envelopment analysis, SFA yields
more conservative estimates of the scope to increase output, because the estimated
potential to raise output is determined after controlling for the identified factors,
random events and differences among hospitals. The technical efficiency of public
and private hospitals is reported in section 8.5.

A detailed description of SFA is provided in appendix E.

Scope of the analysis

The analysis covers hospitals only. It does not include the medical workforce —
except to the extent that diagnostic and allied health professionals are also
employed by each hospital. This scope differs from that of the cost analysis in
chapter 5, which explicitly examined all of the costs incurred in supplying hospital
services.

While the analysis considers the majority of public acute hospitals, only 122 private
sector acute (overnight) hospitals are represented in the sample. The majority of
these are for-profit hospitals. While this does not affect any conclusions drawn
about the hospitals in the sample, care must be taken when inferring about the
performance of the private hospital sector as a whole, and of not-for-profit hospitals
in particular.

8.2 Profile of hospitals in the sample

The Commission obtained permission to access the morbidity and hospital
establishment data for 703 public hospitals. The financial variables for a number of
Victorian hospital observations were reported at the network level. Tasmania
provided a single observation for all public hospitals. In both cases, the respective
observations were disaggregated into hospital-level observations on the basis of
casemix-adjusted separations or the number of beds.

The Commission also obtained the permission of 130 privately owned and operated
hospitals to use their data and unique hospital identifiers in the study. The state and
territory health departments in all jurisdictions agreed to supply their private
hospital data and identifiers to the Commission except for Tasmania, which was
unable to release the private hospital identifiers in time for this study.

From the total sample of 833 observations, 325 hospital observations were excluded
as these were classified as non-acute, sub-acute, psychiatric or free-standing day
hospital facilities. The remaining dataset comprises 508 public and private acute
184 PUBLIC AND PRIVATE
HOSPITALS
hospital observations for 2006-07 (table 8.1). One private hospital observation is an
aggregate of all Tasmanian private hospitals. In total, there were observations for
368 public acute hospitals, 122 private acute hospitals and 18 hospitals that were
classified ‘public contract’ because they were managed by non-government entities
and offered public hospital services.

Table 8.1 Profile of sample hospitals by location and size, 2006-07a

Major cities Outside major cities All
hospitals
Public Private Public Public Private Public
contract contract
Very large 53 np np 15 np – 98
Large 21 16 np 16 6 np 70
Medium 14 26 – 31 12 – 83
Very small & small 8 np – 210 np np 257
All hospitals 96 93 15 272 29 3 508
a Sample refers to all the acute hospitals included in the Commission’s multivariate analysis. Hospital location
is defined by the Australia Standard Geographical Classification (ABS 2001). Hospital size is defined by
number of casemix-adjusted separations per year, where Very large refers to 20 001 or more casemix-
adjusted separations; Large is defined as 10 001 to 20 000 casemix-adjusted separations per year; Medium is
defined as 5001 to 10 000 casemix-adjusted separations per year; Small is defined as 2001 to 5000 casemix-
adjusted separations per year; and Very small is defined as 2000 or fewer casemix-adjusted separations per
year. np Not published due to confidentiality requirements but included in totals where applicable. – Nil.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates.

The remaining acute private hospital in the sample account for around 60 per cent
of all private hospital separations in Australia (excluding same-day facilities), with
a higher representation of for-profit private hospitals compared to not-for-profit
private hospitals (AIHW unpublished data).

Measures of output and productivity for the sample of hospitals included in the
analysis are reported in table 8.2. These data indicate the volume and type of
activity undertaken by hospitals, as well as the productivity of the inputs used.
Observable differences between the public and private sectors can be used to
identify the factors driving hospital efficiency and explain comparative differentials.

Service and patient characteristics of the hospitals in the analysis are profiled in
table 8.3. Observable variations between public and private hospitals — in terms of
the functions they undertake and the patients they treat — draw attention to the need
to control for these factors when calculating and assessing their technical efficiency.

MULTIVARIATE 185
ANALYSIS
Table 8.2 Profile of sample hospitals, output and partial productivity measures, 2006-07a
All hospital sizes Very large Large Medium Small & very small

Public Private Private Public Public Private Public Private Public Private Public Private
for-profit not-for- contract
profit
Output measures (average, per hospital)
All separations (not casemix-adjusted) 11 245 np 15 686 19 186 40 550 28 066 16 509 13 331 8 405 7 898 1 797 3 070
All separations b 11 571 np 18 173 20 216 45 032 32 910 15 297 14 591 7 166 7 229 1 410 2 840
Acute separationsb 9 043 np 15 296 16 459 35 616 28 942 11 623 12 106 5 373 4 924 1 074 1 775
Pregnancy/neonatal separations b 1 245 np 1 446 2 077 4 819 1 663 2 003 1 499 606 679 134 79
Mental/alcohol services separationsb 562 np 233 790 2 121 365 617 191 450 519 89 501
Other separationsb 458 np 778 470 1 440 1 106 741 410 589 1 001 77 447
Emergency dept occasions of service 15 035 np – 564 41 556 9 603 26 310 1 363 15 363 170 4 781 –
Outpatient occasions of service 34 371 np – 42 409 136 634 1 826 45 265 251 17 725 1 953 4 059 2 642
Partial productivity measures
Casemix-adjusted separations per staffc 22.9 np 63.2 32.2 25.3 61.7 27.3 66.8 26.1 71.4 20.6 60.9
Casemix-adjusted separations per bed 70.9 np 114.0 106.6 110.3 141.9 103.8 125.7 87.0 102.4 49.6 62.3
Patient days per bed 246.0 np 335.5 309.5 352.2 363.9 317.6 323.0 296.1 316.3 189.8 244.5
Patient days per staffc 84.8 np 196.0 94.2 82.2 157.6 86.3 175.9 95.3 224.1 83.1 234.5
Non-medical staff per bed 3.0 np 1.9 3.6 4.4 2.4 4.0 2.0 3.4 1.5 2.3 1.2
Occupancy rate 67.4 np 91.9 84.8 96.5 99.7 87.0 88.5 81.1 86.7 52.0 67.0
Average length of stay (days) 3.2 np 4.0 3.1 3.6 2.8 3.0 2.5 3.4 3.0 3.1 6.5
Number of observations 368 94 28 18 68 24 37 22 45 38 218 38
a Sample refers to all the acute hospitals included in the Commission’s multivariate analysis. Data disaggregated by size excludes public contract hospitals due to
confidentiality requirements. Private hospital data disaggregated by size refers to both for-profit and not-for-profit hospitals. b Casemix-adjusted. c Per non-medical staff
member. np Not published due to confidentiality requirements but included in totals where applicable. – Nil or rounded to zero.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates

186
Table 8.3 Profile of sample hospitals, by service and patient characteristics, 2006-07a
All hospital sizes Very large Large Medium Small & very small

Public Private Private Public Public Private Public Private Public Private Public Private
for-profit not-for- contract
profit
Services
Medical DRG (% of separations) 75.7 np 53.5 63.8 72.7 45.7 68.3 37.3 66.3 48.0 79.8 60.4
Surgical/other DRG (% separations) 24.3 np 46.5 36.2 27.3 54.3 31.7 62.7 33.7 52.0 20.2 39.6
Same-day separations (% separations) 52.4 np 44.5 51.5 45.7 43.4 55.7 52.9 62.6 68.2 51.9 65.1
Cost weight (ratio) 0.84 np 1.01 1.05 1.10 1.15 0.94 1.03 0.91 0.93 0.73 1.08
E&W 1 – Complexity (ratio)b 0.55 np 0.68 0.94 1.45 1.13 0.91 0.74 0.63 0.36 0.20 0.12
E&W 2 – Compl. adj. for size (ratio)b 0.49 np 0.58 0.81 1.17 0.95 0.81 0.65 0.60 0.34 0.20 0.12
Teaching hospital (%) 17.1 np 75.0 44.4 75.0 95.8 24.3 15.6 4.4 68.4 0.5 60.5
Network (%) 6.8 np – – 23.5 – 13.5 – 20.0 – 5.5 –
Patient characteristics (average %)
Funding election status
Public 77.9 np 16.7 77.4 83.4 7.0 81.5 7.7 82.5 4.5 74.6 5.8
Private or other non-public 21.9 np 83.3 22.5 16.4 93.0 18.3 92.0 17.4 95.2 25.1 93.8
Residence
From major city 26.4 np 66.0 78.4 71.2 80.4 54.1 71.3 34.3 67.8 6.0 66.5
From inner regional 36.2 np 23.7 14.1 20.2 13.3 36.5 18.3 31.1 24.3 42.2 25.6
From outer regional 26.5 np 9.1 7.0 7.2 5.5 6.1 9.6 30.9 7.2 35.2 7.3
From remote 5.0 np 0.9 0.3 0.8 0.7 1.6 0.6 3.0 0.4 7.3 0.5
From very remote 5.9 np 0.3 0.1 0.5 0.2 1.7 0.2 0.8 0.2 9.4 0.1
Index of socio-economic advantage
SEIFA 1 (most disadvantaged) 40.3 12.4 24.3 19.9 21.4 14.3 28.4 15.6 36.5 14.0 48.9 16.6
SEIFA 2 26.5 15.2 11.1 18.9 22.4 12.5 22.3 14.8 26.7 17.5 28.4 11.9
SEIFA 3 16.6 26.0 13.7 17.8 22.7 18.9 17.7 28.0 17.1 18.6 14.4 27.5
SEIFA 4 9.9 23.5 20.7 24.7 18.3 25.4 15.8 24.6 13.7 23.6 5.5 19.4
SEIFA 5 (most advantaged) 6.8 22.9 30.2 18.8 15.1 29.0 15.9 17.0 6.0 26.3 2.8 24.6
(Continued next page)

187
Table 8.3 (continued)

All hospital sizes Very large Large Medium Small & very small

Public Private Private Public Public Private Public Private Public Private Public Private
for-profit not-for- contract
profit
Charlson comorbidity score
Score 0 (no comorbidities) 74.1 82.2 70.9 71.2 69.4 72.2 75.0 80.5 71.3 81.6 76.0 81.8
Score 1 (fewest comorbidities) 9.1 5.2 6.6 7.2 7.6 5.9 7.0 5.4 8.1 5.4 10.2 5.4
Score 2 10.4 7.2 10.5 14.2 15.1 11.8 11.0 7.9 12.7 8.0 8.4 5.4
Score 3 1.6 1.1 1.5 1.6 1.9 1.1 1.4 0.8 1.7 1.3 1.6 1.3
Score 4 1.5 0.5 1.2 1.3 1.7 1.0 1.4 0.7 2.6 0.6 1.3 0.6
Score 5 2.6 3.6 8.2 3.6 3.6 7.6 3.7 4.3 3.0 2.8 2.1 4.9
Score 6 or higher (most comorbid.) 0.5 0.2 1.0 0.9 0.7 0.4 0.5 0.3 0.6 0.2 0.4 0.6
Average score 0.55 0.42 0.72 0.70 0.73 0.72 0.60 0.43 0.67 0.39 0.46 0.47
Age
<1yr 2.2 np 2.0 2.8 3.4 0.7 3.2 1.1 1.6 1.4 1.7 0.9
1-4yrs 2.9 np 2.5 3.1 3.6 1.2 4.6 1.1 2.0 1.2 2.6 1.4
5-14yrs 3.9 np 3.2 4.0 4.8 1.6 5.4 1.9 2.9 2.0 3.6 2.0
15-49yrs 36.1 np 30.2 38.7 35.8 30.2 35.9 36.1 34.7 34.1 34.5 33.4
50-59yrs 12.2 np 15.6 12.7 11.9 17.6 11.6 17.3 13.1 14.8 12.2 18.3
60-69yrs 13.8 np 17.1 13.6 13.8 19.0 13.3 17.6 15.6 16.3 13.5 15.3
>70yrs 28.8 np 29.4 25.1 26.8 29.7 25.9 24.8 30.0 30.2 29.7 28.8
Female, share 0.53 np 0.55 0.55 0.53 0.52 0.55 0.55 0.54 0.58 0.53 0.56
Indigenous, percentage 9.1 np 0.7 2.1 4.2 0.3 5.0 0.8 6.2 0.3 12.0 0.2
Quality indicator
Mortality rates 1.36 0.42 3.97c 1.30 1.14 0.70 0.98 0.57 1.15 0.46 1.54 2.74
Number of observations 368 94 28 18 68 24 37 22 45 38 218 38
a Sample refers to all the acute hospitals included in the Commission’s multivariate analysis. Data disaggregated by size excludes public contract hospitals due to
confidentiality requirements. Private hospital data disaggregated by size refers to both for-profit and not-for-profit hospitals. b Evans and Walker index of complexity
(appendix E). c Subject to outlier observations (median is 0.70). np Not published due to confidentiality requirements but included in totals where applicable. – Nil or
rounded to zero.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates.

188
Output

Public hospitals report a lower volume of average casemix-adjusted separations

than not-for-profit private hospitals. The extent of variation among public hospitals,
however, is much larger than that of the private hospitals in the sample. Public
contract hospitals report the highest average volume of activity compared to all
other hospitals included in the sample.

Emergency department services are concentrated in the public hospital sector. There
are no emergency departments in any of the not-for-profit private hospitals in the
sample. A similar pattern of activity is observed for outpatient services. A high
volume of outpatient service activity, on par with public hospitals, is reported for
public contract hospitals. Public contract hospitals, not surprisingly, show similar
characteristics to very large and large public hospitals.

Partial productivity measures

Rates of separations per non-medical staff member and per bed are higher among
private hospitals than among public hospitals. This differential is consistent across
all hospital sizes and is applicable to not-for-profit private hospitals. The same
trends are observed for patient days per non-medical staff member and per bed. For
these partial productivity measures, the public contract hospitals in the sample
generally report rates that are higher than public hospitals yet lower than private
hospitals.

Similar occupancy rates are reported for public and private hospitals, except in the
small and very small size category, where private hospitals report a higher rate than
public hospitals.1 This differential has the effect of reducing the average rate for all
public hospitals relative to all private hospitals. Not-for-profit hospitals report
relatively high average occupancy rates. Public contract hospitals report a similar
rate to private hospitals.

Public hospitals report a higher average length of stay (ALOS) than private
hospitals, except in the small and very small hospital category, where private
hospitals have a higher ALOS. This differential has the effect of increasing the
average value for all private hospitals relative to all public hospitals. Negligible
difference is observed in ALOS between not-for-profit and public hospitals, while
public contract hospitals report a similar ALOS to public hospitals.

1 The occupancy rates count same-day separations as one day’s length of stay.

MULTIVARIATE 189
ANALYSIS
Services

The data confirm that public hospital activity is concentrated in medical separations,
whereas private hospital activity is more evenly spread across medical and surgical
or other separations. The average share of surgical or other separations for public
contract hospitals is smaller than for private hospitals but higher than for public
hospitals.

A hospital’s share of same-day separations is greater for public hospitals in the

larger hospital sizes, but greater for private hospitals among the smaller hospital
sizes. Public contract hospitals report a similar rate to the public hospital average.

Consistent with the findings reported in chapter 5, the average cost weights indicate
that private hospitals undertake, on average, relatively more complex cases than
public hospitals. This differential is greatest in the small and very small size
category. The Commission found there to be little material difference between the
average public hospital cost weights for public and private hospitals combined
(appendix D).

A larger share of private hospitals in the sample, on average, are classified as

teaching hospitals, compared to public hospitals. This is apparent across all hospital
sizes except in the large category. These data, however, do not capture the extent of
teaching activity undertaken by the hospital, and therefore does not distinguish
between major teaching hospitals and hospitals offering relatively smaller teaching
functions.

Finally, based on the observed distribution of hospitals that belong to a network, it

may be expected that any potential efficiency effects related to network membership
will be detected among the larger size hospitals.

Patient characteristics

The Charlson index (Charlson et al. 1987) is a measure of the comorbidity of

patients. It is an odds-ratio of the risk of mortality within one year. Thus a Charlson
score of 6 indicates a 6:1 chance of the patient dying within one year. The Charlson
index for this study was prepared using administrative data based on the
ICD-10-AM codes (Quan et al. 2005; Sundarajan et al. 2004,).

Very large public and private hospitals treat patients of similar comorbidity levels
(based on the average Charlson score). Medium and large public hospitals treat
relatively more comorbid patients than private hospitals. Not-for-profit private
hospitals treat relatively more comorbid patients than do for-profit private patients.

190 PUBLIC AND PRIVATE

HOSPITALS
The most comorbid patients (Charlson score of 5 or more) collectively constitute a
larger share of patients in private hospitals than in public hospitals, except in the
medium hospital size category. This difference between public and private hospital
samples seems to be driven by the patient mix in not-for-profit private hospitals.

The Socio-Economic Index for Areas – Relative Socio-economic Advantage and

Disadvantage (SEIFA) is a measure of the relative advantage and disadvantage of
people in different geographic areas (ABS 2998a). For this study, the SEIFA 2001
index was calculated for each patient based on their postcode of the usual place of
residence.

SEIFA data indicate that patients from the most disadvantaged socio-economic
areas constitute a larger share of patients in public hospitals than in private
hospitals. This differential is particularly apparent in the small and very small size
category. With respect to patients’ socio-economics status, public contract hospitals
treat a similar patient profile to private hospitals. The majority of private hospital
patients are from major city or inner regional areas, whereas public hospital patients
are drawn from a broader range of areas.

Public hospitals treat a relatively larger proportion of patients aged less than
14 years, while private hospitals treat a relatively larger proportion of patients aged
50 to 69 years. Similar distributions are observed for all other age categories,
including patients aged 70 years and older. Public and private hospitals treat similar
proportions of patients on the basis of gender. Indigenous patients represent a larger
share of public hospital patients than private hospital patients. This difference is
consistent across all hospital sizes but particularly apparent among the smaller
hospitals.

8.3 Factors affecting hospital performance

The Commission reviewed the literature for a large number of Australian and
overseas studies that have benchmarked the performance of hospitals (a summary of
which is in appendix E). A number of submissions to this study also highlighted
factors that are thought to influence hospital performance. Not all factors identified
in the overseas literature and submissions, however, were included in this analysis.
Some variables were excluded because they were inconsistent with the specification
of the model — for example, occupancy rates were not included in the production
function because the variable is defined in terms of an output (patient days) divided
by an input (beds). In other instances, suggested variables were not included
because of a lack of suitable data.

MULTIVARIATE 191
ANALYSIS
In some instances, proxy variables were used instead. For example, since it is not
possible to identify each and every policy, regulation and legislation in every state
and territory, binary variables for each state and territory were used to control for
these and any other jurisdiction-specific effects. For example, the New South Wales
binary variable took on a value of ‘1’ if a hospital was located in that state, and ‘0’
if not. A variable was not defined for Queensland, because it was used as the
reference category and doing so would introduce collinearity in the model.

Hospital outputs and quality of care

Ideally, a hospital’s performance should be measured in terms of patient outcomes.

Patients seek hospital services in order to improve their physical and emotional
wellbeing relative to what would otherwise be the case. However, it is not
practicable to directly measure the changes to patient health outcomes. Instead, the
approach used here, is to use proxies for two dimensions of health outcomes —
hospital outputs and quality of care.

Hospitals are complex entities that provide a wide range of services. Furthermore,
hospitals vary significantly in terms of their functions and services, such as
respective shares of surgical and medical procedures; the delivery of outpatient and
emergency department care; and the provision of teaching services and clinical
research programs. As such, there is a strong argument that hospitals should be
modelled as multi-input multi-output firms (Butler 1995).

Hospital outputs

Admitted patient outputs used in the analysis were measured in terms of casemix-
adjusted separations grouped into four categories based on the Australian system of
major diagnostic categories (MDCs):
• acute separations — MDCs 1 to 9, 11 to 13, 16 to 18, 21 and 22
• pregnancy and neonate separations — MDCs 14 and 15
• mental and alcohol separations — MDCs 19 and 20
• other separations — MDC 23 (factors influencing health status and other
contacts with health services).

MDC 10 (endocrine, nutritional and metabolic diseases and disorders) is used to

normalise the output variables (for an explanation on the role of normalising the
output variables, see appendix E). This MDC was used as the dependent variable in

192 PUBLIC AND PRIVATE

HOSPITALS
the regression analysis. All separations were casemix adjusted using public hospital
cost weights (AR-DRG version 5.1).

The categories of non-admitted outputs used in the initial analysis were:

• accident and emergency services — number of accident and emergency
department presentations or visits
• allied health and dental services — number of occasions of service for allied
health, dental and other outpatient services
• mental and alcohol services — number of mental, alcohol and psychiatric
outpatient services
• dialysis and endoscopy — number of occasions of service for dialysis and
endoscopy
• diagnostic services — number of pathology and radiology services
• outreach services — number of community services, district nursing and other
outreach services.

There is no national casemix classification for outpatient services, and so these

services were not adjusted for inter-hospital differences in the types of cases treated.

Quality of hospital care

Several variables were available to the Commission to measure hospital quality,

including in-hospital mortality, infection rates and adverse events. As noted in
chapters 6 and 7, there are deficiencies in existing datasets on hospital-acquired
infection rates and adverse events. Some of the problems include possible
under-reporting and the difficulty in attributing the role of hospitals in contributing
to the cause of those events.

The Commission chose to use in-hospital mortality rates as the sole measure of
quality. However, this raises some issues. The rate of in-hospital mortality can vary
for reasons outside the control of hospitals. Some hospitals might specialise in
treating the most ill and at-risk patients. Other hospitals offer specialist palliative
care facilities.

To account for these external influences, the Commission risk adjusted the mortality
rates. The resulting risk-adjusted mortality ratios (RAMRs) (which are defined as
the ratio of the observed mortality rate divided by the predicted mortality rate) were
used in the stochastic frontier analysis. A positive coefficient for the RAMR
variable in the production function would mean that hospitals which have lower
mortality rates than predicted would have higher best practice frontiers. RAMRs
MULTIVARIATE 193
ANALYSIS
differ from the more well-known hospital standardised mortality ratios insofar that
the RAMRs do not account for differences in the services hospitals provide. A
summary of the variables used in the risk-adjustment process and the resulting
RAMRs is provided in appendix E.

Hospital inputs

Following common practice in this area of analysis, inputs into the production of
hospital services included:
• nursing staff — number of full-time equivalent nursing staff
• diagnostic staff — number of full-time equivalent pathology and radiology staff
• other staff — number of full-time equivalent domestic, administration and other
staff
• medical and surgical supplies — expenditure on medical and surgical supplies
• pharmaceutical supplies — expenditure on pharmaceutical supplies used in the
delivery of hospital services
• other inputs — expenditure on other hospital (non-labour) inputs, such as
administration and clerical, housekeeping, and repairs and maintenance not
counted earlier
• beds — number of beds of the hospital as a proxy for hospital capital. This is
given by the number of beds licensed in a private hospital, and the number of
beds recorded in the National Public Hospital Establishment Collection for
public hospitals.

The total number of beds is not an ideal measure of the usage of capital in a hospital
as it does not accurately reflect the differences in capital stock between hospitals.
Ideally, capital measures should be disaggregated such as by the number of ICU
beds, non-acute beds, palliative care beds, the number of same-day chairs, and the
number of operating theatres. Instead, differences in the capital of hospitals were
captured in the analysis by using variables that reflected differences in the roles and
functions of hospitals (see below).

Characteristics of patients

The characteristics of patients can affect hospital performance in two important

ways. First, hospitals that serve catchments of relatively older or sicker patients will
have higher activity levels and therefore will be observed to be providing more
services to their community. In this instance, it is not necessary to account for the
194 PUBLIC AND PRIVATE
HOSPITALS
additional services demanded by patient populations as this activity will be reflected
in the measures of casemix-adjusted separations.

Second, patients with more comorbidities consume more hospital resources per
episode of care. Failure to account for differences in patient populations can lead to
biased estimates of efficiency, because hospitals that serve relatively healthier
patient populations will be observed to be using fewer resources. This problem is
partly addressed with the casemix adjustment of hospital separations, where a
number of individual AR-DRGs are defined specifically to account for differences
in patient comorbidity. For example, the AR-DRG category ‘B03 Extra cranial
vascular procedures’ is divided into sub-groups: ‘B03A with catastrophic or severe
complications’ and ‘B03B without catastrophic or severe complications’.

There is a case to include other variables that account for differences in hospitals’
patient mix. Data for emergency department and outpatient services are not adjusted
for casemix. Even if the detailed AR-DRG-level data adequately accounted for
variation in patient morbidities, it is possible that the process of aggregating each of
the AR-DRGs may diminish some of the statistical variation in morbidity necessary
for robust estimation.

The patient-risk characteristics included in the model were:

• patient comorbidity — both the average Charlson score (Charlson et al. 1987)
across all admitted patients for a hospital, and the proportion of patients with
Charlson scores of 1, 2, 3, 4, 5 and 6 or more (where ‘1’ is the least comorbid
and ‘6 or more’ the most comorbid)
• socioeconomic status of patient — proportions of patients with SEIFA scores of
1, 2, 3, 4 and 5 (where ‘1’ is the most disadvantaged and ‘5’ is the most
advantaged)
• gender — proportion of females in the admitted patient population
• Indigenous status — proportion of the admitted patient population that
self-identified as Indigenous
• age profile — proportion of the population that was less than one-year old,
between 1 and 4 years old, between 5 and 14 years old, between 50 and 59 years
old, between 60 and 69 years old, and 70 years old and over
• remoteness of usual place of residence — proportion of admitted patients whose
usual place of residence was in major cities, inner regional, outer regional,
remote and very remote areas, as defined by the ABS Australian Standard
Geographical Classification – Remoteness Areas system (ASGC-RA).

MULTIVARIATE 195
ANALYSIS
A number of these factors are expected to be closely interrelated. For example, it is
expected that the most comorbid patients will, on average, also be the oldest.
Similarly, there is likely to be a strong association between a patient’s
socioeconomic status and the remoteness of their usual place of residence.

Role and function of hospitals

Another approach to addressing the differences between hospitals is in terms of

their roles and functions. Hospitals that take on teaching roles incur additional
overheads compared to those that do not teach. Public hospitals are increasingly
being organised into networks in which member hospitals adopt specialist functions.
For example, the larger hospitals in a local network might have level-III
intensive-care unit beds in order maximise the utilisation of those beds.

A number of variables were used to explore differences in roles and functions.

Teaching status is a binary variable that describes whether a hospital offers medical
and/or nursing teaching services (‘1’ indicates that it offers teaching, ‘0’ that it does
not). The variable does not reflect the extent of teaching provided.

Level-III intensive-care, palliative and residential care units are three binary
variables that indicate whether a hospital operates each of those units (‘1’ indicates
that it operates a particular unit, ‘0’ that it does not)

Proportion of patients treated with surgical and other procedures is a variable that
describes the extent to which a hospital specialises in surgical and other DRG cases,
or conversely, the degree to which hospitals undertake medical DRG cases. It is
often suggested that a difference between public and private hospitals is the ability
of private hospitals to maximise their productivity by specialising in elective
surgery procedures, which permits them to operate with higher levels of
productivity. On the other hand, public hospitals are unable to refuse medical
admissions, and since medical DRG cases have a greater likelihood of being
unplanned, medical DRGs become inherently more difficult for public hospitals to
manage. Ignoring the differences between surgical and medical cases has the
potential to distort the interpretations of efficiency measures.

Proportion of patients who are not treated as public patients is a proxy measure for
the different levels of resources used by hospitals to treat public and non-public
patients. It includes patients who are funded by private health insurance,
Department of Veterans’ Affairs, third-party motor vehicle accident, workers’
compensation patients, and self-funding. Public hospitals are funded with capped
budgets, at least when treating public patients. In contrast, the funding of non-public
patients is uncapped. It is possible that differences between capped and uncapped
196 PUBLIC AND PRIVATE
HOSPITALS
funding provides hospitals with the capacity to provide different service levels to
public and non-public patients.

Evans and Walker information indices (Evans and Walker 1972) are used as
measures of the relative complexity of work undertaken by hospitals. Two such
measures are considered here. The first is a measure of the complexity of hospital
work. The second is a measure of the complexity of hospital work while
recognising differences in hospital size. While larger hospitals generally treat more
complex cases than smaller hospitals, due to their size, they are also expected to
treat more complex cases. A detailed explanation and derivation of the indices is
given in appendix E.

Care must be exercised when choosing which combination of these variables to

include in the analysis, as they might be correlated and lead to distorted results. For
example, the presence of level-III intensive-care units may be correlated with the
Evans and Walker indices of hospital complexity. Similarly, it is possible that the
hospital roles and functions variables may be correlated with the patient
characteristic variables, if hospitals adopt certain functions because of their patient
mix.

A number of other variables were considered for inclusion, but were subsequently
dropped, because it was expected that they measure very similar effects and would
be too highly correlated with other variables. For example, it would be expected that
average length of stay and the proportion of same-day separations are highly
correlated, and would implicitly be reflected in the casemix-adjusted separations
measure of hospital outputs.

8.4 Factors contributing to best-practice benchmarks

The Commission analysed a number of models that have been used to examine the
performance of hospitals. Ten models are reported in appendix E, two of which are
reported in table 8.4. Each model differs in terms of the types of factors that were
tested for their relative importance, and in terms of the how the variables are
represented. Not all variables that were tested are reported in table 8.4.

The first model in table 8.4, the Cobb-Douglas model, includes variables on inputs
and outputs, factors describing both roles and functions of hospitals, as well as
patient characteristics. In addition to indicating the extent to which an independent
variable influences the dependent variable, the estimated coefficients in this model
also indicate how the frontier (or best-practice benchmark) is positioned for each
hospital. A positively-signed coefficient indicates that the variable has the effect of

MULTIVARIATE 197
ANALYSIS
shifting up a hospital’s best-practice benchmark. This can be interpreted to mean
that a hospital with the characteristics associated with the variable has a higher
hurdle to meet when being compared with other hospitals. Conversely, a
negatively-signed coefficient indicates the variable has the effect of lowering the
hospital’s best practice frontier, potentially making it easier for that hospital to meet
its best-practice benchmark.

The second model (the restricted translog model) is similar to the first but has one
important subtle difference. It includes ‘squared’ terms for the inputs and outputs to
more accurately reflect the presence of scale economies within the hospital sector.
By accounting for scale economies, the translog model better describes the data, and
it is this model that is used to generate the final technical efficiency scores in
table 8.5.

The interpretation of the input and output coefficients are more subtle, however. For
every pair of input and output coefficients, the first coefficient describes the
tendency of the best-practice frontier to be pushed up or pulled down, in much the
same manner as the Cobb-Douglas model. The second (that is, squared) variable,
describes the rate at which the frontier is pushed up or pulled down. For example, if
an input variable had its first coefficient signed positive and its squared term signed
negative, it would be possible to conclude that: hospitals with more of that input
would have their frontier increased (positive first coefficient), but that the rate at
which the benchmark was raised would diminish with further increases to the input
(negative second coefficient).

None of the coefficients indicate the impact of the variables on a hospital’s overall
technical efficiency. Technical efficiency is determined by the position of the
frontier after all adjustments are taken into account, as well as taking into account
the effect of random error and omitted variables.

The findings presented here are based on the coefficients which have the highest
level of significance. Broadly, the coefficients of outputs and inputs in the
production model correspond to prior expectations.

Outputs and inputs

The negative coefficient sign for the output variables in the Cobb-Douglas model
indicate that hospitals that provide a higher volume of services (both admitted
patient services and non-admitted occasions of service) have their best-practice
frontiers shifted downwards, when keeping all other factors unchanged. This means
that hospital benchmarks reward those hospitals that produce more with their given
resources. A number of the output variables become less statistically significant in

198 PUBLIC AND PRIVATE

HOSPITALS
the restricted translog model but their interpretation remains valid. For a number of
outputs a negative signed squared term indicates that as the output is increased, the
rate at which the frontier is shifted downwards is progressively diminished.

Table 8.4 Coefficient results of stochastic frontier analysis, 2006-07

Cobb-Douglas Restricted
modela translog modelb

Primary model
Inpatient services
Log of acute separations -0.506 *** -0.211
Log of acute separations — squared -0.022
Log of pregnancy & newborn separations -0.060 *** -0.052 ***
Log of pregnancy & newborn separations — squared -0.004
Log of mental & alcohol separations -0.106 *** -0.151 ***
Log of mental & alcohol separations — squared -0.019 ***
Log of other separations -0.151 *** -0.103 ***
Log of other separations — squared -0.016 ***
Non-admitted services
Log of emergency department visits -0.021 -0.069
Log of emergency department visits — squared 0.006
Log of allied & dental health services -0.050 *** 0.105 ***
Log of allied & dental health services –— squared -0.024 ***
Log of mental & alcohol services -0.011 0.029
Log of mental & alcohol services — squared -0.003
Log of outreach & district nursing services 0.004 0.010
Log of outreach & district nursing services — squared 0.000
Log of diagnostic services -0.041 *** -0.027
Log of diagnostic services — squared 0.000
Log of dialysis & endoscopy services 0.031 0.014
Log of dialysis & endoscopy services — squared -0.018
Inputs
Log of nursing staff 0.241 *** 0.678 ***
Log of nursing staff — squared -0.061 ***
Log of diagnostic staff 0.030 0.036
Log of diagnostic staff — squared -0.003
Log of other staff -0.161 *** -0.152
Log of other staff — squared 0.000
Log of beds 0.462 *** 0.075
Log of beds — squared 0.068 ***
Log drugs 0.068 *** -0.005
Log drugs — squared 0.011 ***
Log of medical & surgical supplies 0.015 0.246 ***
Log of medical & surgical supplies — squared -0.022 ***
Log of other inputs -0.012 -0.380 ***
Log of other inputs — squared 0.028 ***
(Continued next page)

MULTIVARIATE 199
ANALYSIS
Table 8.4 (continued)

Cobb-Douglas Restricted
model a translog model b

Patient comorbidities
Percent of patients with Charlson 6 + -6.518 ** -7.362 **
Percent of patients with Charlson 5 -1.520 ** -1.121
Percent of patients with Charlson 4 -2.641 *** -2.079 **
Average Charlson comorbidity score 0.394 *** 0.250 **
Patient SEIFA
Percent of patients from SEIFA 3 -0.307 *** -0.216 **
Percent of patients from SEIFA 2 -0.332 *** -0.322 ***
Percent of patients from SEIFA 1 -0.261 *** -0.238 ***
Role and functions
Teaching hospital 0.116 * 0.196 ***
Evans & Walker Index 1 -2.098 *** -1.777 ***
Evans & Walker Index 2 4.011 *** 3.246 ***
Percent of patients that are not public -1.160 *** -0.993 ***
Percent of separations that are surgical or other DRGs 1.131 *** 0.862 ***
State or territoryc
NSW -0.090 -0.098
Victoria -0.277 *** -0.249 ***
South Australia -0.230 *** -0.134
Western Australia -0.069 0.009
Tasmania 1.176 *** 1.001 ***
Northern Territory -0.217 -0.342 *
ACT -0.237 -0.253
Constant 3.644 *** 3.318 ***
Secondary model
Log σ v
2

Constant -2.543 *** -2.495 ***

Log σ 2
u
Constant -1.918 *** -2.560 ***
Model criteria
No. of observations 508 508
Log likelihood -311.3 -242.2
Wald χ2 7 345.2 9 830.1
Probability > χ 2
0.0000 0.0000
σv 0.280 0.287
σu 0.383 0.278
σ2 0.226 0.160
λ 1.367 0.968
a Model 5 (appendix E). b Model 10 (appendix E). c Queensland is not listed because it was used as the
reference category. *** Significant at the 1 per cent level. ** Significant at the 5 per cent level. * Significant at
the 10 per cent level.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates

200 PUBLIC AND PRIVATE

HOSPITALS
The positive coefficients on the input variables of the Cobb-Douglas model indicate
that hospitals that use relatively more of those inputs will have their best-practice
frontier moved up, while keeping other factors unchanged. For nursing staff, the
restricted translog indicates that increases to nurse staffing raises the best practice
benchmark at a diminishing rate. For hospital beds, the coefficient indicates that the
best-practice frontier is increased at an increasing rate.

The negative coefficients (Cobb-Douglas and translog models) of two inputs (other
labour services and other inputs) appear to reflect the effect of a number of
extraneous factors. This seems to be the result of the way the variables have been
defined to include all other (residual) inputs to production.

Patient characteristics

Both the broad categories of patient characteristics — patient comorbidity

(Charlson) and SEIFA — indices were statistically very significant. The groups of
variables suggest that in hospitals that treat relatively more comorbid or
economically disadvantaged patients, the best-practice benchmarks are lowered (in
both the Cobb-Douglas and translog models), particularly at the higher levels of
comorbidity and lowest levels of socioeconomic advantage.

Variables describing the patient’s age and the ASGC-RA of their usual place of
residence were not included in either Cobb-Douglas or translog models, and so were
not reported in table 8.4. Patient age was found to be related with comorbidity, and
the geographic remoteness of residence was thought to be too closely related with
the patient’s SEIFA index.

Role and function

Hospitals with a relatively higher share of non-public patients were found to have a
lower best-practice frontier. This may indicate that non-public patients, compared to
public patients, are more resource intensive for hospitals. This may reflect the way
in which hospitals differentiate their services to public and non-public patients by,
for example, offering additional services such as clinical interventions and private
ward accommodation to non-public patients.

A hospital’s share of surgical and other procedures was found to raise the hospital’s
best-practice frontier. This is equivalent to finding that hospitals that specialise in
medical DRG cases need to have the best-practice frontier adjusted downwards.
This finding strengthens the view that surgical procedures are less resource
intensive and simpler to manage than medical cases.

MULTIVARIATE 201
ANALYSIS
The two Evans and Walker indices should be interpreted jointly. The first index
indicates that the more complex the caseload that hospitals have, the further inwards
the best-practice frontier is positioned. The second index indicates, however, that
this effect is offset by hospital size. While larger hospitals are expected to take on
more complex cases, they do not appear to be taking on as many complex cases as
their size would suggest, so their best-practice benchmarks are adjusted
accordingly.

A hospital’s teaching status was found to raise a hospital’s best-practice frontier. If

trainee doctors and nurses were less productive than their fully trained counterparts,
one would expect the coefficient to be negatively signed. Instead it seems that the
binary nature of the variable is identifying extraneous influences on hospital
performance.

Other variables

Variables indicating the states and territories in which the hospitals were located
were included to control for differences in state and territory policies, regulations
and legislation, which cannot be reflected elsewhere in the model. The significance
of the coefficients for Victoria and Tasmania is likely to reflect, in part, the fact that
some of their hospital data was provided in aggregated form. Even though attempts
were made to rescale the data appropriately, it is possible that inconsistencies still
exist that are least are being controlled for by these variables (appendix E). The
significance of the Victorian coefficient may also pick up the effect of network
membership.

The RAMR variable, as a proxy for the quality of health care, was found to be
insignificant in the model and is not reported here. It was found to be highly
correlated with the various Charlson indices, and was consequently dropped from
the final analysis. The insignificance of the RAMR variable suggests that the quality
of hospital care, as modelled here, does not have a sufficiently strong effect on
hospital output. The significance of the Charlson and SEIFA variables suggests that
patient characteristics affect hospitals output directly. The Commission intends to
examine alternative methods to incorporate a quality control factor in future
analyses.

FINDING 8.1

A multivariate analysis of Australian hospital-level data established the

best-practice benchmarks for each hospital in the sample. The benchmarks were
influenced by a number of factors. The best-practice benchmarks were lower for
hospitals that treat:

202 PUBLIC AND PRIVATE

HOSPITALS
• highly morbid patients
• patients from lower socioeconomic communities
• relatively more medical cases, as these cases are more difficult to manage
• more complex cases, although this is less so for the largest hospitals.

8.5 Hospital efficiencies

The previous statistical analysis focused on identifying the factors that influence the
best-practice frontier of each hospital. The next step is to estimate how far hospitals
are from their respective frontiers.

A hospital’s technical efficiency is measured as an index with a possible value

ranging from zero to less than one. If a hospital has a technical efficiency index of
0.80, it is producing 20 per cent less output than it potentially could achieve if it met
its hypothetical benchmark. No hospital has a ranking of 1.00, as each hospital’s
benchmark is uniquely determined using all of the observations in the data set, not
just a few neighbouring peers. Put simply, even the best performing hospitals can
still learn from others on how to improve their performance.

The indices of technical efficiency from the restricted translog model (table 8.4) are
presented in table 8.5. The efficiency scores already account for differences in
hospital size. That is, the technical efficiency scores do not reflect the effects of
scale economies. Public, for-profit private and not-for-profit private hospitals within
this sample were operating with mean efficiencies between 0.75 and 0.80 of the best
practice efficiency in 2006-07, although this appears to be pulled downwards by the
small and very small hospitals.

There are perceptible differences between various groupings of public and private
hospitals, and with those hospitals that are not managed by governments as public
hospitals. However, care must be exercised when comparing across groups as there
is a large variation in the efficiencies among the members of these groups.

The mean efficiencies of very large and large private hospitals are estimated to be
slightly higher than similarly-sized public hospitals. In contrast, the mean technical
efficiencies of very small to medium sized public hospitals tend to be higher than
for similarly-sized private hospitals. However, caution needs to be exercised when
interpreting the results for very small to medium hospitals as there is a wide range
between the 5th and 95th percentile technical efficiency scores. For example, the
range for very small and small private hospitals is 0.203 to 0.919, suggesting that
there are a variety of activities among these hospitals that are yet to be fully
accounted in the model.
MULTIVARIATE 203
ANALYSIS
Table 8.5 Summary of hospital technical efficiency scores, 2006-07a
Public Private hospitals Public All
hospitals contract hospitals
All For Not-for- hospitals
profit profit

All hospitals
Mean rate 0.797 0.750 0.751 0.747 0.800 0.786
Median rate 0.816 0.822 0.818 0.838 0.805 0.816
5th percentile rate 0.643 0.313 0.313 0.203 0.583 0.553
95th percentile rate 0.901 0.916 0.917 0.906 0.908 0.906
No. of observations no. 368 122 94 28 18 508
Very large hospitals
Mean rate 0.813 0.819 0.834 0.795 np 0.814
Median rate 0.820 0.851 0.863 0.846 np 0.827
5th percentile rate 0.708 0.655 0.659 0.639 np 0.683
95th percentile rate 0.905 0.893 0.917 0.877 np 0.905
No. of observations no. 68 24 15 9 np 98
Large hospitals
Mean rate 0.810 0.813 0.810 0.819 np 0.809
Median rate 0.812 0.829 0.828 0.830 np 0.828
5th percentile rate 0.648 0.752 0.457 0.757 np 0.644
95th percentile rate 0.917 0.878 0.918 0.868 np 0.908
No. of observations no. 37 22 15 7 np 70
Medium hospitals
Mean rate 0.803 0.780 0.791 0.707 np 0.741
Median rate 0.815 0.841 0.820 0.876 np 0.780
5th percentile rate 0.622 0.427 0.470 0.029 np 0.491
95th percentile rate 0.907 0.931 0.931 0.906 np 0.915
No. of observations no. 45 38 33 5 np 83
Small and very small hospitals
Mean rate 0.788 0.641 0.640 0.642 np 0.766
Median rate 0.816 0.700 0.715 0.644 np 0.806
5th percentile rate 0.575 0.203 0.208 0.203 np 0.415
95th percentile rate 0.897 0.919 0.916 0.919 np 0.899
No. of observations no. 218 38 31 7 np 257
a Based on restricted translog model 10 (appendix E). np Not published due to confidentiality.

Source: Unpublished ABS and AIHW data; Productivity Commission estimates.

These findings are broadly consistent with findings of other research undertaken by
the Commission as part of its modelling of the potential benefits of the National
Reform Agenda in 2006. The Commission had reviewed published research on
hospital performance, from which it concluded that the gap between existing and
best-practice productivity might be in the order of 20–25 per cent for the Australian
(public and private) hospital sector as a whole (PC 2006a). This conclusion was

204 PUBLIC AND PRIVATE

HOSPITALS
based on a combination of the limited Australian research, particularly for the
private sector, and overseas studies.

What are some of the contributing factors?

Why is it then that some hospitals appear to be better performing than others?
Partial productivity, at least for admitted-patient care, is higher among private
hospitals than for public hospitals. The productivity difference is noticeable for each
of the partial measures of productivity — casemix-adjusted separations and patients
days, per non-medical staff and per bed. For example, the number of
casemix-adjusted separations per non-medical staff was 63.2 for not-for-profit
hospitals and 22.9 for public hospitals (table 8.2). Similar patterns are evident for all
sizes of public and private hospitals. This would suggest that private hospitals are
more technically efficient than public hospitals.

Public hospitals, however, provide more non-admitted patient care than do private
hospitals in this sample. Though the productivity numbers are not reported in table
8.2, productivity for these services will obviously be higher among public hospitals.

Public hospitals also undertake relatively more medical DRG separations

(75.7 per cent) than do not-for-profit hospitals (53.5 per cent) as a proportion of
their total activity. As estimated in table 8.4, medical DRG separations are more
difficult to manage than surgical and other DRG separations, so public hospitals are
regarded to have their best-practice benchmarks lowered in this regard. Similarly,
public hospitals treat relatively more patients from the more disadvantaged
socioeconomic regions of Australia than do private hospitals. For example,
40.3 per cent of all public hospital patients were from the most disadvantaged
regions (SEIFA 1).

Similarly, private for-profit hospitals treated the least morbid patients. The average
Charlson score for private for-profit hospitals was 0.42, compared to 0.55 for public
hospitals, 0.72 for not-for-profit hospitals and 0.70 for public contract hospitals.
These patterns are also evident in the highest Charlson comorbidity scores. Over
9 per cent of all the patients of not-for-profit hospitals had comorbidity scores of 5
or higher. These statistics contribute to lowering the best-practice frontiers of
not-for-profit and public contract hospitals (and to a lesser extent public hospitals)
further than for private for-profit hospitals.

MULTIVARIATE 205
ANALYSIS
The performance of smaller private hospitals

The efficiency scores for medium, small and very small private hospitals exhibited a
considerable degree of variation (table 8.5). This suggests a degree of unexplained
differences (heterogeneity) that has not been recognised in the model. Comparing
these efficiency scores with some of the data in table 8.3 provides insights on such
hospitals.

Small and very small hospitals appear to comprise two distinct groups — one that
specialises in high-volume same-day procedures and another that specialises in
long-stay cases. For example, the proportion of same-day cases in very small and
small private hospitals is 65 per cent, compared to 52 per cent for public hospitals.
Yet very small and small hospitals also have an ALOS of 6.5 days. These lengths of
stay are likely to represent non- and sub-acute services provided by otherwise acute
hospitals. The lengths of stay for these hospitals count against their reported
productivity, and at the same time, so does the lack of complexity (as represented
by the Evans and Walker indices). There is potential to more accurately model the
performance of these hospitals.

FINDING 8.2

After controlling for differences in services provided and types of patients treated,
the efficiency of public and private hospitals is, on average, similar. It was
estimated that the output of individual hospitals in both sectors is, on average,
around 20 per cent below best practice among the sampled hospitals. Among large
and very large hospitals, the scope to improve technical efficiency is slightly greater
for public hospitals. At the other end of the scale, the scope to improve efficiency is
higher for small and very small private hospitals, although these results may be
partly due to a number of factors that could not be accounted for in the analysis.

8.6 Further analysis

Given the delays in obtaining data faced by the Commission, the multivariate
analysis presented in this report estimates hospital production functions and
technical efficiency based on a single year of data (2006-07). Given the large
number of hospital observations in this data set, the results are expected to be
robust.

Nevertheless, the Commission intends over coming months to replicate this analysis
using a larger data set that includes data from the earlier years of 2003-04 to
2005-06. Future analysis will also focus on examining the performance of hospitals
for different peer groups (for example, to compare the performance of very large
206 PUBLIC AND PRIVATE
HOSPITALS
hospitals). The Commission will also extend this analysis to examine the
determinants of hospital costs.

The Commission intends to publish the results from this further analysis in
March 2010.

MULTIVARIATE 207
ANALYSIS
9 Informed financial consent

Key points
• Informed financial consent (IFC) occurs when patients undergoing treatment as a
private patient receive relevant cost information about their treatment prior to the
treatment taking place. IFC is important as it allows patients to make informed
decisions.
• The latest available data show that currently around 85 per cent of hospital medical
services provided to privately-insured patients do not have an associated gap
payment, suggesting that most patients do not have a problem with IFC.
Nevertheless, it is undesirable for any patients to incur gap payments without IFC.
• Survey data collected by Ipsos Australia are the only information source that is
available to consider the provision of IFC and out-of-pocket expenses. These data
are dated (2007 is the latest) and subject to potential sample-selection and self-
reporting bias and should be interpreted with caution. These data suggest that a
greater percentage of in-hospital services involve a gap than is actually the case.
• The 2007 Ipsos data for pre-planned admissions show that the IFC rate was:
– lower in private hospitals (around 80 per cent) than public hospitals (about
90 per cent)
– higher in Tasmania, South Australia, and inner regional areas across Australia
(around 90 per cent in each case)
– among specialists, lowest for paediatricians (around 60 per cent).
• The data also show that, at a national level, the IFC rate was relatively stable
between 2004 and 2007, and for a range of disaggregations.
• At a national level, the average gap for people who did not provide IFC was around
$800 and relatively stable between 2004 and 2007. However, few conclusions can
be made when disaggregating further, due to small sample sizes.
• The medical profession has sought to promote best practice with respect to IFC in
recent years, including through educational campaigns. In addition, individual
specialists are using internet-based packages to inform patients about their likely
out-of-pocket expenses.

The terms of reference ask the Commission to examine aspects of informed

financial consent (IFC) for privately-insured patients. IFC occurs when patients
receive relevant cost information, including notification of likely out-of-pocket

INFORMED 209
FINANCIAL CONSENT
expenses (gaps), desirably prior to agreeing to treatment. It is preferable for this
information to be provided in writing.

The provision of relevant cost information prior to treatment is desirable as it allows

patients to make more informed decisions about whether to continue to receive
treatment from the particular practitioner(s) and to understand the full cost
implications. However, it is not feasible for written notification to be provided in all
circumstances, such as in many emergency cases.

Excesses and copayments are not considered out-of-pocket expenses because they
are due to the patient’s choice of insurable cover. An excess is the amount that a
patient agrees to pay towards the cost of hospital treatment in exchange for a lower
insurance premium. A copayment is a set amount that insured patients pay for each
day they are in hospital in exchange for a lower insurance premium (PHIO 2009).

For the year to September 2009, there was no gap payment for 85 per cent of
hospital medical services provided to privately-insured patients (not including
hospital accommodation services) (PHIAC 2009a). For the remaining 15 per cent of
services that did involve a gap, 4 percentage points involved known-gap
agreements. In known-gap agreements there is a legal requirement for IFC to be
provided. Thus, it could be argued that no more than 11 per cent of services should
lack IFC. For the year to September 2009, this represented approximately 730 000
medical services (PHIAC 2009a).1 Of this 11 per cent, the Australian Society of
Anaesthetists (sub. 9) noted that 1–3 percentage points involve emergency cases, for
which it is not always realistic to expect IFC. Nevertheless, it is undesirable for any
non-emergency patients to incur out-of-pocket expenses without IFC, and desirable
for as many emergency patients to receive IFC as is practicable.

The Private Health Insurance Ombudsman (sub. 26) noted that there has been a
gradual decline in complaints received regarding IFC over recent years. The Private
Health Insurance Ombudsman received 76 complaints about medical gap issues in
2007-08 (of which almost all concerned IFC), which was 39 fewer than the previous
year, suggesting that the provision of IFC has been improving. However, there still
seems to be a number of patients that do not receive relevant cost information prior
to treatment.

Aside from analysing the provision of IFC and out-of-pocket expenses, this chapter
discusses some potential impediments to the provision of IFC, available IFC data

1 In 2008-09, there were 3 052 375 acute hospital episodes and 6 151 724 medical services
involving privately-insured patients, therefore privately-insured patients are on average treated by
approximately two medical service providers per episode of hospitalisation (PHIAC 2009a,
2009b).
210 PUBLIC AND PRIVATE
HOSPITALS
sources and possible data developments, and best-practice examples of IFC
provision.

9.1 Potential impediments to the provision of informed

financial consent
Key impediments to IFC can include short lead times prior to treatment, treatment
involving multiple service providers, and the complexity of some health insurance
policies.

Short lead times and multiple providers of services

It is generally more difficult for specialists or other service providers to obtain IFC
from patients for whom there is a short amount of time between the decision to
undertake surgery and the actual admission for surgery (short lead time). This is
because there is less available time to provide financial information to patients
enabling them to provide IFC in advance of the procedure.

This issue is not as pertinent for those specialists and service providers who discuss
in person the available treatment options with a patient, and thus have an
opportunity to also provide financial information. However, IFC can be more of an
issue for ‘downstream’ specialists and service providers who have less contact with
patients prior to surgery (for example, anaesthetists and surgeon’s assistants).
Internal polling conducted by the Australian Society of Anaesthetists supports this
view. The Australian Society of Anaesthetists (sub. 9) noted that its polling shows
that, while 93 per cent of anaesthetists believe they can obtain IFC within five days
of treatment, only 25 per cent believe they can obtain IFC when there are only two
days between the allocation of the task and the treatment date.

In this framework, emergency admissions can be considered to have no lead time

and this makes it very hard for specialists and service providers to obtain IFC.
Indeed, as the Australian Health Service Alliance (AHSA) (sub. 1) noted, it may not
be appropriate for specialists to obtain IFC if it delays the treatment of an
emergency patient.

Health insurance policies

The level of insurance benefit payable can be difficult for patients to ascertain and
this can hinder the ability to obtain IFC (Australian Medical Association, sub.
DR55). A patient’s out-of-pocket expenses are derived from the payable health
INFORMED 211
FINANCIAL CONSENT
insurer benefit, in conjunction with treatment charges and the Medicare benefit.
Some health insurance policies provide a higher benefit if certain requirements are
met, but only the minimum allowable benefit (25 per cent of the Medicare Benefits
Schedule fee) if these requirements are not met. These requirements can include the
relevant doctor being registered with the fund and accounts being ‘correctly’
processed. Uncertainty about the level of payable benefit could also make it difficult
for the doctor to provide the patient with an accurate estimate of out-of-pocket
expenses. Despite these potential issues with estimating likely private health
insurance benefits, many specialists do provide indicative private health insurance
and Medicare returns to show likely out-of-pocket expenses.

The Australian Society of Anaesthetists (sub. 9) noted that a lack of available

benefits schedules can hinder the proper provision of IFC. Health insurers usually
update the schedules of their more common ‘gap-cover’ type insurance policies
every November. However, the Australian Society of Anaesthetists suggested that
the updated schedules are not always available from the time at which they are
activated, and in some cases may even take up to a month to be published. If this
were the case, it would make it difficult for practitioners to calculate the expected
level of benefit that a patient would receive around this time of the year. The Private
Health Insurance Ombudsman stated that they have not received any complaints
from medical practitioners or consumers regarding the timely updating of schedules
(Private Health Insurance Ombudsman, pers. comm., 28 August 2009).

9.2 Informed financial consent data sources and their

suitability
The Australian Government Department of Health and Ageing has commissioned
three surveys on IFC in Australia. These were conducted by Ipsos Australia in
September 2004, September 2006 and May 2007. The surveys asked patients who
had recently made a private health insurance claim for an episode of hospitalisation
about the costs they incurred and the information they received about those costs
prior to treatment (Ipsos Australia 2005, 2007a, 2008). This is the best available
dataset to consider IFC and out-of-pocket expenses in the way stipulated by the
terms of reference. However, the surveys have some major limitations that will be
discussed below.

Ipsos has also estimated rates of IFC and out-of-pocket expenses for privately
insured patients in its biennial report Ipsos Healthcare and Insurance Australia
Report (Ipsos Australia 2007b) and in a member experience survey for Medibank
Private (2009). However, the latter survey was restricted to privately insured
patients who were treated in private hospitals.
212 PUBLIC AND PRIVATE
HOSPITALS
The Private Health Insurance Administration Council (PHIAC) collects industry
data on the out-of-pocket expenses incurred by privately-insured patients. The data
include information on ‘known-gap’ agreements, where:
… the medical practitioner agrees to accept a payment by the insurer in part satisfaction
of the amount owed and the patient has provided informed financial consent so that the
gap or out-of-pocket expense to be paid by the insured person is known in advance.
(PHIAC 2009a, p. 2)

The provision of IFC is a legal requirement of known-gap agreements and therefore

quantification of this segment can provide some information on the provision of
IFC. However, it cannot be considered an ideal measure of IFC as there is no
verification of whether the provision of IFC actually occurred. Furthermore, this
data source can at most be considered a lower bound on the provision of IFC, as it
does not include any estimates of IFC where a known-gap agreement is not in place.

Ipsos survey methodology

The samples for the 2004, 2006 and 2007 Ipsos surveys were drawn from
individuals who had made a recent insurance claim for hospital treatment as a
private patient, and whose claim had been settled prior to 1 September 2004,
15 September 2006 and 1 May 2007 respectively. Each survey involved the
distribution of 10 000 questionnaires, with the 2004, 2006 and 2007 surveys
receiving a response rate of 41, 46 and 42 per cent respectively. Public hospital
episodes and episodes covered by small health insurers were oversampled to
improve the accuracy of estimates relating to these groups. The sample population
was subsequently weighted so that it was representative of the population.

The Ipsos surveys have a number of limitations that need to be considered when
interpreting the results. A potential limitation of the surveys is that individuals may
have been more likely to respond if they did not receive sufficient information to
provide IFC or incurred significant out-of-pocket expenses. If such sample-selection
bias existed, it would result in rates of IFC being underestimated and out-of-pocket
expenses being overestimated.

A further limitation is that the surveys are based on patient perceptions and
recollections, and are not an audit of actual occurrences. While in hospital, patients
may be treated by a number of different practitioners, may receive a large amount of
healthcare information and may be worried about a range of other issues aside from
treatment costs. All these factors may lead to some patients incorrectly recalling the
gap they incurred and whether they provided IFC. Study participant Mark Sinclair
(sub. 8) noted that he has had cases as an anaesthetist where he obtained written
consent from patients regarding costs, yet was later told that the patient did not
INFORMED 213
FINANCIAL CONSENT
realise that they would have out-of-pocket expenses. The Private Health Insurance
Ombudsman (sub. 26) also noted that in the investigation of some IFC complaints
they have found that IFC was provided by the patient, even though the patient did
not later recall this being the case. However, the Private Health Insurance
Ombudsman noted that they believe patients recollections to be a ‘good guide’,
though not always accurate (Private Health Insurance Ombudsman, sub. 26, p. 5).

The possibility of sample-selection and/or self-reporting bias was tested by

comparing Ipsos results with data collected by PHIAC. This PHIAC data,
measuring medical gaps, is based on data submitted by health insurers for every
hospital admission by a privately-insured patient. The results for 2007 suggest that
the Ipsos surveys significantly overestimated the number of services where a gap
was paid and the average gap payment (table 9.1).

It is important to distinguish between IFC for individual services and IFC for a
patient’s whole episode of hospitalisation. IFC rates for individual services measure
the provision of IFC by each service provider or specialist. In contrast, IFC rates for
a patient’s whole episode of hospitalisation measure whether IFC is obtained by all
individual specialists or service providers by whom a patient is treated in an episode
of hospitalisation. As private patients are almost always treated by more than one
specialist or service provider in an episode of hospitalisation, the IFC rate for the
whole episode of hospitalisation should always be less than the IFC rate for
individual services. PHIAC does not publish data on the proportion of patients with
no gap for the whole episode of hospitalisation. However, it would be expected that
the proportion of patients in the population with no gap in 2007 for the whole
episode of hospitalisation was also greater than the figure suggested by the Ipsos
survey data.

Table 9.1 Comparison of Ipsos and PHIAC data, 2007

Units Ipsos PHIACa
Share of services with a gap paymentb % 28 18
Share of patients with a gap payment (whole episode)c % 42 na
Average gap across all servicesb $ 131 23
Average gap across all services where a gap occurredb $ 465 126
a PHIAC data is for 2006-07, as it is the closest to the sample period of the 2007 Ipsos survey. b This includes
all in-hospital services except hospital accommodation. c This excludes test-related specialities such as
pathology, radiology, ultrasound and x-ray. na not available.
Source: Ipsos survey data (unpublished); PHIAC (2007b).

A further limitation of the Ipsos surveys is that the most recent survey was
conducted in 2007, and thus the effects of any measures undertaken since 2007 to
increase the levels of IFC or to reduce out-of-pocket expenses cannot be assessed.
214 PUBLIC AND PRIVATE
HOSPITALS
Due to a lack of other suitable data sources, Ipsos survey data will be the primary
data source used to calculate statistics relating to IFC and out-of-pocket expenses in
this chapter. However, all results should be considered in light of the
aforementioned data limitations.

9.3 Rates of informed financial consent

Reporting rates of informed financial consent

Rather than reporting the proportion of all patients who received sufficient
information to provide IFC, this section will define the IFC rate to be the proportion
of patients who either had no gap or who received information and provided IFC on
any gap for their episode of hospitalisation. This measure of IFC recognises that
while patients value receiving information to provide IFC, they also value not
incurring any out-of-pocket expenses. This is the measure that is used by Ipsos
Australia (2005, 2007a, 2008) in its three reports on IFC.

As previously noted, it is important to distinguish between IFC for individual

services and IFC for episodes of hospitalisation. In this chapter the reported IFC
rates are based on episodes of hospitalisation that include all of the relevant aspects
of the episode (for example, hospital costs, medical costs, surgeon’s assistants and
prostheses). If IFC is not provided for one (or more) component of the service, then
IFC is not provided for the entire episode of hospitalisation. However, IFC rates for
individual services can be useful for analysing which specialists or service providers
are least likely to obtain IFC. Therefore, the IFC rates used for analysing the
provision of IFC for service providers and specialists are for individual services.

Emergency admissions

The Ipsos survey asked respondents whether their admission to hospital was
pre-planned or an emergency, with 22 per cent of respondents in 2007 indicating
that their admission was the latter.2

As previously mentioned, it would be expected that the provision of IFC would be

less prevalent for emergency admissions. However, Ipsos data suggests that in
2007, IFC was more likely to have occurred in emergency admissions than in

2 Ipsos Australia provided no further guidance as to the definition of pre-planned and emergency
admissions besides noting that maternity/obstetrics admissions were to be considered
pre-planned.
INFORMED 215
FINANCIAL CONSENT
pre-planned admissions (table 9.2). This difference is due to a greater percentage of
emergency admissions not incurring any gap payment. When only considering
privately-insured patients who incurred a gap, the proportion who provided IFC is
nonetheless similar in pre-planned and emergency admissions. While this result is
unexpected, any financial information that is provided to a patient in an emergency
situation is likely to be different to information provided in a pre-planned context,
as the patient may not be in a reasonable position to evaluate their treatment
options. For this reason, only pre-planned admissions are considered in the analysis
of IFC and out-of-pocket expenses, unless otherwise stated. Specialists who deal
with an emergency situation for a patient who had a pre-planned admission cannot
be removed from the sample.

Table 9.2 Informed financial consent rates for pre-planned and

emergency admissions, 2007a
Pre-planned Emergency b

% %
Informed financial consent 80 88
Informed financial consent (no gap)b 54 72
Informed financial consent (gap) 26 16
No informed financial consent 20 12
Total 100 100
a Test-related services are excluded in the calculation of total IFC rates. b Ipsos Australia provided no further
guidance as to the definition of pre-planned and emergency admissions besides noting that
maternity/obstetrics admissions were to be considered pre-planned. c This includes patients who were unsure
whether they incurred a gap.
Source: Ipsos survey data (unpublished).

Rates of informed financial consent

Ipsos survey data show that, in 2007, the rate of IFC was approximately 80 per cent
(table 9.3). Between 2004 and 2007, there was a small increase in the IFC rate, from
78 to 80 per cent.

216 PUBLIC AND PRIVATE

HOSPITALS
Table 9.3 Informed financial consent rates, 2004–2007a
2004 2006 2007

% % %
Informed financial consent 78 81 80
Informed financial consent (no gap)b 52 55 55
Informed financial consent (gap) 25 26 26
No informed financial consent 22 19 20
No informed financial consent for one servicec 16 14 15
No informed financial consent for multiple servicesd 6 5 5
Total 100 100 100
a Only patients who considered their admission to be pre-planned are included. Maternity/obstetrics
admissions are considered pre-planned. Test-related services are excluded in the calculation of total IFC
rates. b This includes patients who were unsure whether they incurred a gap. c Patients did not provide IFC to
one service provider. d Patients did not provide IFC to multiple service providers.
Source: Ipsos survey data (unpublished).

In 2007, of the 20 per cent of admissions where IFC was not provided on all gaps,
15 percentage points involved just one service provider failing to obtain IFC, while
5 percentage points involved more than one service provider failing to obtain IFC.

Informed financial consent rates by type of hospital provider

In 2007, the IFC rate in public hospitals was approximately 88 per cent compared to
80 per cent in private hospitals (table 9.4). This notable difference seems to be
primarily attributable to fewer privately-insured patients incurring gaps in public
hospitals. Between 2004 and 2007, the ratio of IFC rates between private and public
hospitals remained relatively stable.

The proportion of hospital admissions for which IFC was not provided on hospital
accommodation costs was 2 per cent or less for both public and private hospitals,
suggesting that a lack of IFC for accommodation costs is not an issue for either
sector.

INFORMED 217
FINANCIAL CONSENT
Table 9.4 Informed financial consent rates by hospital providera
Private hospitals Public hospitals

2004 2006 2007 2004 2006 2007

% % % % % %
Informed financial consentb 77 80 80 85 90 88
IFC (no gap)c 51 52 53 67 76 76
IFC (gap) 26 28 27 18 14 12
No informed financial consent 23 20 20 15 10 12
No IFC (medical)d 21 19 19 14 9 11
No IFC (accommodation)d 2 2 2 1 0 1
a Only patients who considered their admission to be pre-planned are included. Maternity/obstetrics
admissions are considered pre-planned. Test-related services are excluded in the calculation of total IFC
rates. b Due to rounding, in 2007 the IFC rate for private hospitals appears not to differ from the overall IFC
rate (table 9.3). c This includes patients who were unsure whether they incurred a gap. d The medical IFC
rate and accommodation IFC rate may not sum to the total IFC rate, as some patients may not have provided
IFC for both. They may also not equate due to rounding error.
Source: Ipsos survey data (unpublished).

Informed financial consent rates by jurisdiction

There is notable variation in the IFC rates between jurisdictions (table 9.5). In 2007,
Tasmania and South Australia had the highest rates of IFC among jurisdictions. In
contrast, after excluding the ACT and Northern Territory (due to insufficient sample
sizes), New South Wales had the lowest IFC rate. Between 2004 and 2007, the IFC
rates did not change significantly for any jurisdictions (figure 9.1).

218 PUBLIC AND PRIVATE

HOSPITALS
Table 9.5 Informed financial consent rates by jurisdiction, 2007a
NSW Vic Qld SA WA Tas

% % % % % %
IFC 77 80 82 88 80 91
IFC (no gap)b 53 53 58 63 48 72
IFC (gap) 24 27 25 25 32 18
No IFC 23 20 18 12 20 9
No IFC (medical)c 22 18 16 11 17 8
No IFC (accommodation)c 2 3 2 1 5 1
No. of observations 1037 889 540 269 287 103
a Jurisdiction refers to the patient’s jurisdiction of residence. According to the AIHW (2009a) approximately
98 per cent of separations occur in hospitals that are in the patient’s jurisdiction of residence. Only patients
who considered their admission to be pre-planned are included. Maternity/obstetrics admissions are
considered pre-planned. Test-related services are excluded in the calculation of total IFC rates. The sample
sizes for the ACT and Northern Territory were very small (52 and 8 observations respectively) and thus are
not published. b This includes patients who were unsure whether they incurred a gap. c The medical IFC rate
and accommodation IFC rate may not sum to the total IFC rate, as some patients may not have provided IFC
in both. They may also not equate due to rounding error.
Source: Ipsos survey data (unpublished).

Figure 9.1 Informed financial consent rates by jurisdictiona

100

80
IFC rate

60

40

20

0
NSW Vic Qld SA WA Tas

2004 2006 2007

a Only patients who considered their admission to be pre-planned are included. Maternity/obstetrics
admissions are considered pre-planned. Test-related services are excluded in the calculation of total IFC
rates.
Source: Ipsos survey data (unpublished).

Informed financial consent rates by location

In 2007, inner regional hospitals had the highest IFC rates (86 per cent), compared
to outer regional hospitals (76 per cent) and major city hospitals (80 per cent)
INFORMED 219
FINANCIAL CONSENT
(table 9.6). For private hospitals, the IFC rate was notably higher for hospitals
located in inner regional areas (85 per cent) compared to outer regional and major
city areas (71 and 79 per cent respectively). For public hospitals, the IFC rate was
also highest for hospitals located in inner regional areas (95 per cent), however this
result was not significantly different from the IFC rate for public hospitals located
in outer regional areas.

One possible explanation for these results may be that specialists in inner regional
areas have relatively strong community relationships, making it easier to provide
financial information to patients. In comparison, patients may be less likely to
personally know their specialists in major cities and outer regional areas (possibly
because specialists do not reside in the area), making the provision of IFC more
difficult.

Between 2004 and 2007, the IFC rate for hospitals located in major cities increased
slightly. This was largely due to an improvement in the IFC rate for private
hospitals in major cities.

While the improvement in the IFC rate for public hospitals in inner regional areas
was significant it was offset by a slight decrease for private hospitals. The changes
between 2004 and 2007 for the inner and outer regional areas were not significant.

Table 9.6 Informed financial consent rates by location and type of

providera
Major cities Inner regional Outer regional
2004 2006 2007 2004 2006 2007 2004 2006 2007

% % % % % % % % %
Private hospitals 75 79 79 87 88 85 72 77 71
Public hospitals 85 89 85 83 94 95 93 93 89
All hospitals 76 79 80 87 89 86 80 82 76
a Location based on ABS (2005) Australian Standard Geographical Classification. Data for remote and very
remote classifications are not published due to insufficient sample sizes. Only patients who considered their
admission to be pre-planned are included. Maternity/obstetrics admissions were considered pre-planned.
Test-related services are excluded in the calculation of total IFC rates.
Source: Ipsos survey data (unpublished).

Informed financial consent rates by medical specialist

Disaggregating IFC rates by specialist and service providers confirms that a lack of
IFC is more commonly associated with medical practitioner charges than with
hospital accommodation charges (table 9.7). In 2007, the IFC rate was highest for
hospital accommodation (98 per cent) and orthopaedic surgeons (95 per cent). In
220 PUBLIC AND PRIVATE
HOSPITALS
contrast, the IFC rate was lowest for paediatricians (61 per cent) and test-related
services (77 per cent). The IFC rate remained relatively stable across all specialties
between 2004 and 2007.

Table 9.7 Informed financial consent rates by medical practitioner or

service provider, 2007a
Medical specialist/service Sample size Patients that Patients using IFC rate
provider used provider provider that
had a gap

% % %
Paediatrician 226 8 45 61
Tests/pathology/radiology/
977 31 24 77
ultrasound/x-ray etc.
Anaesthetist 2 478 78 30 86
Specialist’s or surgeon’s
832 24 26 88
assistant
General surgeon 530 17 25 92
Oncologist 271 8 15 93
Obstetrician/gynaecologist 440 14 32 94
Cardiologist 200 6 9 94
Orthopaedic surgeon 418 13 45 95
Hospital (accommodation) 3 194 100 5 98
a These results should be interpreted with care as they do not necessarily provide an accurate indication of
which specialties have the lowest IFC rates. In particular, the sample sizes in the Ipsos survey for a number of
medical specialists or service providers were too small to make robust conclusions about rates of IFC. Only
patients who considered their admission to be pre-planned are included. Maternity/obstetrics admissions are
considered pre-planned.
Source: Ipsos survey data (unpublished).

When comparing rates of IFC for different service providers, it is important to

recognise that it can be more difficult for certain types of specialists to provide
sufficient information to enable their patients to give IFC. As mentioned previously,
specialists who have less direct contact with patients prior to surgery may find it
relatively more difficult to provide IFC. This could explain the lower rates of IFC
for anaesthetists, surgeon’s assistants and test-related specialists.

Ipsos survey data show that the IFC rates for different specialties are not notably
higher for admissions where the patient saw the admitting doctor at least five days
prior to the procedure taking place (table 9.8), but rather are broadly similar to the
results for the full sample.

INFORMED 221
FINANCIAL CONSENT
Table 9.8 Informed financial consent rates for patients with a lead time of
at least five days, 2007a
Medical specialist/service Sample Patients Patients using IFC rate Difference
provider size that used provider that from full
provider had a gap sampleb

% % % %
Paediatrician 200 10 48 59 -2
Tests/pathology/radiology/
713 33 26 76 -1
ultrasound/x-ray etc
Anaesthetist 1 753 80 32 85 -1
Specialist’s or surgeon’s
622 28 26 88 0
assistant
General surgeon 362 17 25 94 2
Oncologist 185 9 16 92 -1
Obstetrician/gynaecologist 358 17 34 94 0
Cardiologist 138 6 9 95 1
Orthopaedic surgeon 317 15 45 95 0
Hospital (accommodation) 2 189 100 5 97 -1
a Only patients who saw the admitting doctor at least five days prior to surgery are included. These results
should be interpreted with care as they do not necessarily provide an accurate indication of which specialties
have the lowest IFC rates. In particular, the sample sizes in the Ipsos survey for a number of medical
specialists or service providers were too small to make robust conclusions about rates of IFC. Only patients
who considered their admission to be pre-planned are included. Maternity/obstetrics admissions were
considered pre-planned. b Difference from full sample IFC rates (table 9.7).
Source: Ipsos survey data (unpublished).

9.4 Out-of-pocket expenses

The terms of reference ask the Commission to report the average out-of-pocket
expenses for patients who do not receive sufficient information to provide IFC. The
Commission has also been asked to report the range (the minimum to maximum) of
these costs, and to categorise the data by type of provider (public/private hospital
and medical specialty).

As was the case for IFC rates, out-of-pocket expenses are reported for the whole
episode of hospitalisation. The exception is that out-of-pocket expenses for
individual specialties are reported for individual services. In this context, the
average gap with no IFC is defined as the average gap incurred by privately insured
patients who do not receive sufficient information to provide IFC for at least one
gap.

In particular, as previously noted, results should be considered in light of the

limitations of the Ipsos surveys. Estimates regarding the average gap with no IFC
should be interpreted with care. This is because disaggregated groups other than

222 PUBLIC AND PRIVATE

HOSPITALS
private hospitals, New South Wales and Victoria, major city hospitals, hospital
accommodation and anaesthetist services have very small sample sizes.

The Ipsos survey data show that in 2007, the overall average gap with no IFC was
$847 (table 9.9). This is marginally greater than the average gap incurred by all
privately-insured patients that had a gap, but the difference is not significant. In
2007, the out-of-pocket expenses incurred with no IFC ranged from $5 to $19 827.

Table 9.9 Average gapa

Units 2004 2006 2007
Proportion of patients incurring a gap % 48 45 45
Average gap for patients incurring a gap $ 755 684 817
Average gap for patients who did not provide IFC $ 816 756 847
Minimum $ 2 4 5
Maximum $ 8 547 7 881 19 827
a Only patients who considered their admission to be pre-planned are included. Maternity/obstetrics
admissions are considered pre-planned. Test-related services are excluded in the calculation of average gaps.
Source: Ipsos survey data (unpublished).

Of the gaps where no IFC was provided, approximately 55 per cent were less than
$500, approximately 75 per cent were less than $1000, and approximately
90 per cent were less than $2000 (figure 9.2, figure 9.3). Only 2 per cent of gaps
with no IFC were greater than $4000. A number of these gaps over $4000 were due
to large gaps from specialist’s or surgeon’s assistants, while for very large gaps one
was due to a large gap from a cardiologist and one was due to a large gap for
hospital accommodation.

INFORMED 223
FINANCIAL CONSENT
Figure 9.2 Sample cumulative distribution function of gaps with no IFC,
2007a

100

80
Percentile

60

40

20

0
0 5000 10000 15000 20000
Gap ($)

a Only patients who considered their admission to be pre-planned are included. Maternity/obstetrics
admissions were considered pre-planned. Test-related services are excluded in the calculation of average
gaps.
Source: Ipsos survey data (unpublished).

Figure 9.3 Distribution of gaps with no IFC, 2007a

40
Per cent of observations

35
30
25
20
15
10
5
0
500 1000 1500 2000 2500 3000 3500 4000
Gap ($)

a Only patients who considered their admission to be pre-planned are included. Maternity/obstetrics
admissions were considered pre-planned. Test-related services are excluded in the calculation of average
gaps. Gaps over $4000 are not shown above and represent approximately 2 per cent of observations.
Source: Ipsos survey data (unpublished).

224 PUBLIC AND PRIVATE

HOSPITALS
Out-of-pocket expenses by type of hospital provider

The average gap with no IFC was larger for patients in private hospitals ($858)
compared to patients in public hospitals ($637) in 2007, however this difference
was not statistically significant (table 9.10). In private hospitals, the average gap
with no IFC was larger than the average gap for all patients who paid a gap and this
difference was statistically significant in 2004 and 2006.

In 2007, the out-of-pocket expenses for patients with no IFC ranged from $6 to
$19 827 for patients in private hospitals, compared to a range of $4 to $2030 for
patients in public hospitals.

Between 2004 and 2007, the proportion of patients incurring a gap in private
hospitals remained relatively stable. However, in public hospitals it fell
significantly.

In 2007, the average gap for all patients with a gap was very similar in public
hospitals ($788) and private hospitals ($818).

Table 9.10 Average gap by hospital provider, 2004–2007a

Private hospitals Public hospitals

Units 2004 2006 2007 2004 2006 2007

Proportion of patients
% 49 48 47 33 24 24
experiencing a gap
Average gap for patients
$ 746 689 818 915 575 788
that had a gap
Average gap for patients
$ 820 768 858 751 536 637
who did not receive IFC
Minimum $ 2 4 6 15 15 4
Maximum $ 8 547 7 881 19 827 4 700 3 400 2 030
a Only patients who considered their admission to be pre-planned are included. Maternity/obstetrics
admissions are considered pre-planned. Test-related services are excluded in the calculation of average gaps.
Source: Ipsos survey data (unpublished).

INFORMED 225
FINANCIAL CONSENT
Out-of-pocket expenses by jurisdiction

In 2007, the average gap with no IFC was greatest for patients in New South Wales
and Queensland, and smallest for patients in South Australia and Western Australia
(table 9.11). In each jurisdiction, the average gap with no IFC was not significantly
different from the average gap incurred by all patients experiencing a gap.

Table 9.11 Average gap by jurisdiction, 2007a

Units NSW Vic Qld SA WA Tas
Proportion of patients
% 47 47 42 37 52 28
experiencing a gap
Average gap for patients
$ 1 049 682 924 548 512 772
that had a gap
Average gap for patients
$ 1 052 710 1 040 421 433 930
who did not receive IFCb
Minimum $ 6 5 8 25 12 20
Maximum $ 7 050 19 827 10 500 3 100 2 572 5 487
a Jurisdiction refers to the patient’s jurisdiction of residence. According to the AIHW (2009a) approximately
98 per cent of separations occur in hospitals that are in the patient’s jurisdiction of residence. Only patients
who considered their admission to be pre-planned are included. Maternity/obstetrics admissions are
considered pre-planned. Test-related services are excluded in the calculation of average gaps. The sample
sizes for the ACT and the Northern Territory were very small (52 and 8 observations respectively) and thus
are not published. b These figures are based on very small sample sizes for all jurisdictions aside from New
South Wales and Victoria and should be interpreted with care.
Source: Ipsos survey data (unpublished).

Out-of-pocket expenses by location

The average gap with no IFC in 2007 was greatest for patients in hospitals located
in major cities, and smallest for patients in hospitals located in outer regional areas
(table 9.12). These differences could be because hospitals in major cities undertake
more complex procedures, while regional hospitals tend to provide less complex
procedures. The difference between the average gap with no IFC and the average
gap for all patients who paid a gap was only significantly different for major city
hospitals in 2004 and 2006.

226 PUBLIC AND PRIVATE

HOSPITALS
Table 9.12 Average gap by region, 2004–2007a
Major cities Inner regional Outer regional

Units 2004 2006 2007 2004 2006 2007 2004 2006 2007
Proportion of
patients % 51 48 48 34 33 33 38 36 47
experiencing a gap
Average gap for
patients that had a $ 804 716 867 345 461 520 686 497 536
gap
Average gap for
patients who did not $ 864 799 901 364 452 581 769 471 440
receive IFC.
Minimum $ 2 4 6 4 6 8 40 35 5
Maximum $ 8 547 7 881 19 827 2 082 2 220 2 790 2 400 1 550 1 669
a Location based on ABS (2005) Australian Standard Geographical Classification. Data for remote and very
remote classifications are not published due to insufficient sample sizes. Only patients who considered their
admission to be pre-planned are included. Maternity/obstetrics admissions are considered pre-planned. Test-
related services are excluded in the calculation of average gaps.
Source: Ipsos survey data (unpublished).

Out-of-pocket expenses by medical specialist

In 2007, the average gap with no IFC was greatest among patients treated by
obstetricians/gynaecologists ($753) and orthopaedic surgeons ($720), and smallest
for paediatricians ($197) (table 9.13). However, it is important to remember that
almost all these figures are based on small sample sizes and should be interpreted
with care.

One possible reason that patients treated by some service providers incur greater
out-of-pocket expenses may be that these service providers perform more complex
procedures or use more expensive equipment.

The Ipsos data also suggest that between 2004 and 2007, the average gap where no
IFC was provided did not change significantly for any speciality.

INFORMED 227
FINANCIAL CONSENT
Table 9.13 Average gap by medical practitioner or service provider, 2007a
Medical specialist/service Average gap Average gap Minimum gap Maximum gap
provider amount for incurred by incurred by incurred by
patients who patients who patients who patients who did
had a gap did not provide did not provide not provide IFC
IFC IFC

$ $ $ $
Obstetrician/gynaecologist 828 753 39 4 000
Orthopaedic surgeon 841 720 30 2 750
Oncologist 920 677 20 5 600
Cardiologist 633 600 20 19 727
Specialist’s or surgeon’s
536 461 7 6 000
assistant
General surgeon 518 444 12 4 200
Hospital (accommodation) 353 410 1 10 500
Anaesthetist 320 308 4 2 610
Tests/pathology/radiology/
287 292 5 4 656
ultrasound/x-ray etc.
Paediatrician 207 197 16 900
a Gap figures for most specialties are based on small sample sizes and should be interpreted with care. Only
patients who considered their admission to be pre-planned are included. Maternity/obstetrics admissions are
considered pre-planned.
Source: Ipsos survey data (unpublished).

FINDING 9.1

According to the Private Health Insurance Administration Council, around

90 per cent of hospital services for privately-insured patients do not have out-of-
pocket expenses that require informed financial consent. Complaints data collected
by the Private Health Insurance Ombudsman suggest that the rate of informed
financial consent has been increasing in recent years.

FINDING 9.2

The incidence and average size of out-of-pocket expenses for privately-insured

patients appear to be overstated in available survey data collected by Ipsos, due to
sample-selection and self-reporting bias. Subject to this qualification, the data
suggest that privately-insured patients have a higher rate of informed financial
consent and lower out-of-pocket expenses in public hospitals. Few conclusions can
be made about out-of-pocket expenses due to small sample sizes.

228 PUBLIC AND PRIVATE

HOSPITALS
9.5 Future data improvements
The terms of reference ask the Commission to consider developments that would
improve the feasibility of future comparisons. As previously mentioned, the Ipsos
survey data have potential limitations of self-selection and reporting bias, which
could limit their usefulness in providing information regarding IFC and out-of-
pocket expenses. Therefore, future IFC surveys should focus on overcoming these
limitations. A robust dataset on IFC would have the following features:
• To address self-selection bias, collection methods would ensure that the sample
was representative of all patients who had utilised their private health insurance
for hospital visits.
• To address reporting bias, data would not be wholly reliant on patient
recollections.

One option would be to include survey questions relating to IFC in the ABS
National Health Survey (NHS), as this would remove any self-selection problem
and also include self-insured individuals. However, it would still rely on patient
recollections and may not consist of enough people with a recent hospital episode.
The NHS (ABS 2009b) reports that 53 per cent of the population hold private health
insurance. Therefore, if weighting is not considered, the sample of 22 000 people in
the NHS would include approximately 12 000 people with private health insurance.

According to PHIAC data the NHS is likely to contain approximately 3800

privately-insured individuals who had a recent episode of hospitalisation, including
approximately 500 who were admitted to a public hospital as a private patient.3
However, to properly analyse the difference in the provision of IFC for public and
private hospitals, approximately 1000 public hospital observations would be
required. Thus it is unlikely that the NHS would provide a sufficient sample for an
analysis of IFC to be conducted.

Another option would be to require privately-insured patients to indicate on their

health insurance claim form whether they gave IFC. This information could then be
provided to PHIAC and published with information it already releases on
out-of-pocket expenses. This option would remove any self-selection bias as all

3 In 2008-09, there were 3 052 375 acute hospital episodes involving privately insured patients,
including 435 193 acute public hospital episodes. In 2008-09, there were 9 676 645 individuals
privately-insured for hospital treatment (PHIAC 2009b). This calculation does not take into
account that some individuals are admitted to hospital more than once in a year. For a sample
size similar to the Ipsos surveys, approximately one-third of people with private health insurance
would need to have had a recent episode of hospitalisation, which is somewhat unlikely.
Furthermore, to properly analyse the difference in the provision of IFC for public and private
hospitals, approximately 1000 public hospital observations would be required.
INFORMED 229
FINANCIAL CONSENT
privately-insured patients would complete health insurance claim forms. It would
also remove some self-reporting problems, as the length of time between the
hospital episode and filling out a claim form would be less than that involved in
responding to the Ipsos survey.

Following the release of the Discussion Draft for this study, the Consumers Health
Forum (sub. DR59) endorsed this proposal. However, the Australian Medical
Association (sub. DR55) and the Australian Government Department of Health and
Ageing (sub. DR69) noted that there can be significant time lags between the giving
of IFC for the procedure and the submission of an insurance claim form, and thus
patients may still forget whether they provided IFC. The Australian Medical
Association further noted that in a significant proportion of complaints surrounding
a lack of IFC, doctors have documentary evidence of IFC having been provided.

In order to address the problem of patients not accurately recalling their

experiences, it might be possible for medical specialists and service providers to
include as part of the billing and insurance-claim process an indication of whether
documented evidence of IFC is held for the relevant item. This information could in
turn be used by PHIAC to monitor rates of IFC.

FINDING 9.3

A more robust future data source on informed financial consent (IFC) could be
created by requiring privately-insured patients to indicate on their health insurance
claim form whether they provided IFC prior to the procedure. Alternatively,
medical specialists and service providers could be required to include as part of the
billing and insurance-claim process an indication of whether documented evidence
of IFC is held for the relevant item. This information could be collected and
reported by the Private Health Insurance Administration Council.

9.6 Best-practice examples of IFC

The terms of reference ask the Commission to identify best-practice examples of
where IFC is provided for every procedure. Emphasis is to be put on best-practice
examples that occur in specialties where a lack of IFC is most common. The
Commission requested examples in the Issues Paper and Discussion Draft but only
a limited number of examples were provided.

Table 9.7 suggests that patients are most likely to pay a gap without IFC when using
the services of a paediatrician, an anaesthetist, a specialist’s or surgeon’s assistant
or when undergoing diagnostic tests such as pathology, radiology, ultrasound and
x-ray. However, as previously mentioned, these results should be interpreted with
230 PUBLIC AND PRIVATE
HOSPITALS
care as they do not necessarily provide an accurate indication of which specialties
have the lowest IFC rates.

In recent years, a number of medical specialist groups have undertaken education

campaigns regarding IFC, demonstrating their desire to improve the rates of IFC
provision. These included the following campaigns (Australian Medical
Association, sub. 28):
• The Australian Medical Association’s campaign called ‘Let’s talk about fees’,
which provided doctors with materials to read and share with patients.
• The Australian Diagnostic Imaging Association’s campaign, which involved the
creation of a website where members publish fee information.
• The Australian Society of Anaesthetists’ IFC campaign, which included among
other initiatives:
– publishing a newsletter on IFC
– mailing out an ‘IFC campaign kit’ to anaesthetists that contained resources on
how to improve IFC in anaesthetic practice
– holding IFC educational meetings for anaesthetists in capital cities and major
regional centres.

These education campaigns can be seen to complement the best-practice examples

outlined below.

Meditrust

Meditrust is an organisation that helps anaesthetists to provide patients with

information prior to their procedure. This information is delivered through a website
or toll-free phone number and includes the expected out-of-pocket
expenses (box 9.1).

It may be more difficult for anaesthetists than other specialists to obtain IFC from
patients. This is because an anaesthetist often needs to make separate contact with
the patient aside from the patient’s contact with their surgeon, which can be
especially difficult when there is a short lead time prior to the procedure.
Meditrust’s system facilitates this anaesthetist/patient contact and may remove the
necessity for a meeting before the day of surgery.

INFORMED 231
FINANCIAL CONSENT
 Box 9.1 Meditrust and IFC
The Meditrust system provides information to patients in the following way:
• The anaesthetist provides the surgeons with whom they regularly work a note that
lists the different procedures specific to that surgeon.
• Any patient to be treated by the anaesthetist receives a copy of the note from the
surgeon indicating the procedure they will undergo.
• The patient enters a surgeon-specific password into the Meditrust website.
• Patients select their procedure and private health insurer from lists.
• Patients are presented with information relevant to the anaesthetic for that specific
procedure. This includes information relating to:
– administration of the anaesthetic itself
– an estimated total fee, the rebate and expected out-of-pocket expenses.
• Patients are provided with the anaesthetist’s phone number to contact if they have
any questions.
• The anaesthetist receives an email notifying them that the patient has accessed the
information and has provided IFC.
Mark Sinclair (an anaesthetist who uses Meditrust) noted in his submission that:
… patients who are not ‘computer literate’ and/or do not have Internet access are given a
toll-free phone number to ring. A Meditrust staff member asks for the names of the relevant
surgeon and anaesthetist, and the information is mailed to the patient as a hard copy, free of
charge. (sub. 8, p. 2)

Source: Dr. Mark Sinclair (sub. 8, p. 2)

Sportsmed SA

Sportsmed SA is a South Australian healthcare provider specialising in sports

medicine, orthopaedics, podiatry and physiotherapy. Sportsmed SA’s hospital
admissions only relate to orthopaedic cases.

Aside from emergency admissions that occur on weekends, Sportsmed SA claims

that they obtain IFC for all orthopaedic surgeon and hospital accommodation fees.
When surgery is recommended for a patient, they are given a ‘comprehensive
financial quotation for the operation, hospital fees and incidental charges at the time
of booking’ (Sportsmed SA 2009, p. 11).

Sportsmed SA’s hospital is collectively owned by ten orthopaedic surgeons, who

have their private practices on site (Sportsmed SA 2009). This collective ownership,

232 PUBLIC AND PRIVATE

HOSPITALS
coupled with a well functioning administrative process for providing financial
information, enables Sportsmed SA to obtain IFC in all cases.

South Australian public hospitals

In South Australia, Rights of Private Practice agreements require that salaried

doctors in public hospitals only bill the full Medicare Benefits Schedule fee (SA
Department of Health, sub. 4). Therefore, privately-insured patients in public
hospitals should not incur any out-of-pocket expenses in relation to treatment
performed by salaried doctors. Visiting medical officers are able to sign up to such
agreements, but not all do so . These doctors are able to charge as they wish, leaving
open the possibility that IFC is not provided on all out-of-pocket expenses in SA
public hospitals.

FINDING 9.4

The medical profession has sought to promote best practice for informed financial
consent in recent years. This has included educational campaigns for practitioners
and internet-based packages to inform consumers of their likely expenses.

INFORMED 233
FINANCIAL CONSENT
10 Indexation of the Medicare Levy
Surcharge income thresholds

Key points
• The Medicare Levy Surcharge (MLS) was introduced by the Australian Government
in 1997 as part of a suite of measures designed to arrest the decline in private
health insurance (PHI) membership, and was first applied for the 1997-98 financial
year. The MLS only applies to taxpayers that do not have private patient hospital
cover for themselves and all family members.
• The MLS was initially applied at a rate of 1 per cent of taxable income for singles
who earned $50 000 or more a year, and to families with one dependent child who
earned $100 000 or more a year (with the threshold increasing by $1500 for each
dependent child after the first).
• The MLS income thresholds remained unchanged until 2008-09, when the
Australian Government lifted the singles threshold to $70 000 a year and the
families threshold to $140 000 a year. These amendments also introduced annual
indexation of the thresholds.
• The Australian Government lifted the thresholds and introduced indexation to
‘refocus the MLS on those with higher income’.
• The terms of reference for this study ask the Commission to advise the Australian
Government on the most appropriate indexation factor for the MLS thresholds. The
Commission has examined four possible indexation factors: average weekly
ordinary time earnings (AWOTE), average weekly total earnings, the consumer
price index and the wage price index.
• To assess the merits of these options the Commission estimated and compared the
proportion of single and family taxpayers that would have been subject to the MLS,
had the thresholds been indexed, between 1999-2000 and 2007-08, by each of the
four indexes.
• The Commission found that the proportion of taxpayers subject to the MLS would
have increased under all indexation options relative to the proportion of taxpayers
subject to the MLS in 1999-2000. However, the proportion of taxpayers subject to
the MLS would have increased least if AWOTE was used to index the MLS
thresholds.
• The Commission also investigated using an indexation measure based on high
income earners, such as the ninth decile of taxable income. This was not practical.
Therefore, the Commission suggests that AWOTE is the most appropriate
indexation factor, because it is most likely to maintain the Australian Government’s
goal of keeping the MLS targeted at high income earners.

INDEXATION OF THE 235

MLS THRESHOLDS
The terms of reference for this study request advice on the most appropriate
indexation factor for the Medicare Levy Surcharge (MLS) income thresholds. The
MLS is levied on Australian taxpayers who earn above a specified income threshold
and do not have private patient hospital cover for themselves and all family
members. This chapter considers four potential indexation factors for the MLS
income thresholds.

10.1 Background to the Medicare Levy Surcharge

The MLS was introduced in 1997, and first applied for the 1997-98 financial year,
as part of a package of measures designed to stem the decline in private health
insurance (PHI) membership in the Australian community, and to maintain the
private hospital system as a ‘vital complement to the long term viability of
Medicare and the public hospital system’ (Wooldridge 1997). Other measures
included a 30 per cent rebate on PHI premiums (introduced in January 1999) and
the Lifetime Health Cover community rating scheme (introduced in July 2000).1

The MLS first applied to taxable income in the 1997-98 financial year.2 It was
applied to singles who earned $50 000 or more a year, and to families3 with one
dependent child who earned $100 000 or more a year, if they did not have
appropriate private patient hospital cover.4 For families with more than one

1 In 2005, the rebate was increased to 35 per cent for people aged 65–69, and to 40 per cent for
those aged over 70. The 2009-10 Commonwealth Budget proposed to means test eligibility for
the rebate, however legislation giving effect to this proposal is yet to be passed through
parliament. The policy, if implemented, would involve the rebate being progressively reduced for
higher income ranges, and totally withdrawn for the highest income range. The Lifetime Health
Cover rating scheme provides for progressive increases in premiums payable by those taking up
PHI after the age of 30.
2 Taxpayers are defined as people who have a net tax liability greater than zero. Taxable income
equals assessable income minus deductions. Income for MLS purposes equals taxable income
plus reportable fringe benefits plus the net amount on which family trust distribution tax has been
paid, minus any post-June 1983 component of an Employment Termination Payment where the
maximum tax rate is zero. The 2008-09 Commonwealth Budget included measures to expand the
MLS income definition to include salary-sacrificed superannuation contributions and net losses
from financial investments.
3 For MLS purposes a person is considered to be a member of a family if they contribute to the
maintenance of a dependant, including a spouse, even if the spouse has their own income. Any
parent (including a sole parent) who contributes to the maintenance of a dependent child or
children is considered to be a member of a family.
4 An appropriate insurance policy for MLS purposes is one that does not have an excess greater
than $500 for singles or greater than $1000 for families.
236 PUBLIC AND PRIVATE
HOSPITALS
dependent child, the threshold increased by $1500 for each dependent child after the
first. The MLS applied at a rate of 1 per cent of all income. The income thresholds
set in 1997 remained unadjusted until October 2008.

In 2008, a number of changes were made to the MLS. These included provision for
annual increases in the thresholds by means of indexation, and an increase in the
income threshold levels, which had remained unchanged since the MLS was
introduced for the 1997-98 financial year (box 10.1).

Box 10.1 Recent changes to the Medicare Levy Surcharge

Amendments to the Medicare Levy Surcharge (MLS) were passed by Parliament in
October 2008, and came into effect for the 2008-09 financial year. The income
thresholds at which the MLS became payable in the 2008-09 financial year increased
from:
• $50 000 to $70 000 for singles
• $100 000 to $140 000 for couples and families with one dependent child.
The rise in the annual income threshold for families with more than one dependent
child remained unchanged at $1500 for each dependent child after the first.
The recent amendments also introduced annual indexation. Average weekly ordinary
time earnings (AWOTE) was chosen as the indexation factor. The singles threshold is
now indexed annually by multiplying the 2008-09 surcharge threshold by the indexation
factor and rounding the result down to the nearest multiple of $1000 (provided that the
indexation factor is greater than one). The MLS threshold for couples and families with
one dependent child will continue to be set at twice the singles threshold.
The indexation factor is the index number for the quarter ending 31 December in the
previous financial year, divided by the index number for the quarter ending
31 December 2006, calculated to three decimal places. The index number used for the
December quarter is the estimate of full-time adult AWOTE for the middle month of the
quarter (November) that is first published by the ABS.
In the 2009-10 budget, the Australian Government proposed changes that would
increase the MLS for high income earners without PHI to:
• 1.25 per cent, for single people earning more than $90 000 and for families (with
less than two dependent children) earning more than $180 000
• 1.5 per cent, for single people earning more $120 000 and for families (with less
than two dependent children) earning more than $240 000.
Source: Tax Laws Amendment (Medicare Levy Surcharge Thresholds) Bill (no. 2) 2008; Treasury (2009a).

There has been extensive discussion regarding the appropriate approach to

indexation. While the October 2008 amendments to the MLS provided for

INDEXATION OF THE 237

MLS THRESHOLDS
indexation of the thresholds on the basis of changes to average weekly ordinary
time earnings, the terms of reference for this study request the Commission to
advise on the most appropriate indexation factor.

10.2 Why index the MLS thresholds?

When introduced in 1997, the MLS was focused on encouraging ‘high income
earners who can afford to take out private health insurance to do so’, and the
threshold levels at which the MLS was applied were set accordingly
(Wooldridge 1996, p. 8576). When the Australian Government adjusted the MLS
income thresholds in the 2008-09 Budget, it indicated that this was done in order to
‘refocus the MLS on those with higher income’ (Treasury 2008b, p. 33).

Without indexation of the income thresholds, an increasing proportion of taxpayers

will be subject to the MLS over time, due to a combination of rising real incomes
and wage inflation. For example, in 1997-98 around 8 per cent of single taxpayers
exceeded the singles threshold. However with no indexation of the MLS thresholds,
this proportion increased to approximately 33 per cent of single taxpayers by
2007-08. Therefore, without indexation, more taxpayers become liable for the MLS
and the MLS is less effective over time at solely targeting high income earners.

Study participants who commented on the MLS thresholds were generally

supportive of indexation, although they expressed a variety of views about the
appropriate indexation factor.

10.3 Possible indexation factors

Indexation involves adjusting a dollar amount over time in line with changes in an
index, with the index often based on a measure of changes in prices or wages. This
is intended to provide an estimate of equivalent dollar amounts over time, adjusting
for movements in measured prices or wages.

There are a number of potential price and wage measures that could be used to
index the MLS income thresholds. The four measures are:
• average weekly ordinary time earnings (AWOTE)
• average weekly total earnings (AWTE)
• consumer price index (CPI)
• wage price index (WPI) (table 10.1).

238 PUBLIC AND PRIVATE

HOSPITALS
Table 10.1 Possible indexation factors for the Medicare Levy Surcharge
Indexation factor Description
Average weekly ordinary • Measures growth in average weekly pre-tax earnings from standard
time earnings (AWOTE) hours of full-time work for adult wage and salary earners.
• Includes award, workplace and enterprise bargaining payments,
penalty payments, shift allowances, commissions and retainers,
bonuses, incentives, profit sharing payments, workers compensation
and salary payments to directors.
• Excludes amounts that are salary sacrificed, non-cash components
of salary packages, overtime payments, retrospective pay, pay in
advance, leave loadings, severance, termination and redundancy
payments and other sources of income such as capital gains.
• Does not cover a number of workers, including self-employed
persons and owners of unincorporated businesses.
• Index varies not only with changes in wage levels but also according
to changes in average hours worked and composition of the
workforce.
Average weekly total • Includes both AWOTE and overtime earnings.
earnings
Consumer price index • Measures change over time in the price of a specified ‘basket’ of
goods and services, which comprise a high proportion of household
expenditures.
Wage price index • Measures change over time in wages paid for a fixed amount of
labour.
• Controls for changes in income levels resulting from increases in
hours worked, or changes in the composition of the workforce.
• Does not cover other sources of income such as capital gains.
Source: ABS (Average Weekly Earnings, Australia, Cat. no. 6302.0; Consumer Price Index, Australia,
Cat. no. 6401.0; Labour Price Index, Australia, Cat. no. 6345.0).

The key differences between these measures are that:

• AWOTE and AWTE both measure pre-tax weekly earnings but AWOTE
includes standard hours of full-time work and AWTE includes overtime.
• WPI measures the change over time in wages paid for a fixed amount of labour
and controls for changes in hours worked and the composition of the workforce.
• CPI does not measure earnings. It measures changes in a group of
commonly-purchased household items.

These are all published approximately two months after the end of the quarter
making each a practical option for indexing the MLS thresholds.

Study participants expressed a range of views on the most appropriate indexation

factor for the MLS. The Australian Nursing Federation (sub. 17; DR57) supported
AWOTE or AWTE being used, noting that these measures most accurately reflect
changes in earnings. The Australian Government Department of Health and Ageing
(DOHA) (sub. 32; DR69) also supported AWOTE being used as an indexation

INDEXATION OF THE 239

MLS THRESHOLDS
factor for the MLS thresholds, noting that it is most relevant because it measures the
‘normal’ earnings of workers (since it excludes overtime).

Access Economics (sub. DR60) and the Australian Medical Association (sub. 28)
preferred the WPI. The Australian Medical Association observed that:
[O]ther measures of earnings, such as survey-based AWE [Average Weekly Earnings]
and AWOTE, are affected significantly by changes in the composition of employment.
As such, the trends in AWOTE bear little relation to the experience of the typical
householder. (sub. 28, p. 7)

Following the release of the Discussion Draft for this study, the Australian Medical
Association (sub. DR55) also suggested the national accounts measure of average
earnings — average non-farm compensation per employee — that is published by
the ABS (2009a). This measure includes compensation in the form of in-kind
benefits and employer contributions to superannuation and workers compensation,
in conjunction with wages and salaries paid in cash. However, given that the MLS
is only calculated on the basis of taxable income, this measure does not seem
suitable.

The Royal Australasian College of Surgeons (sub. 30, p. 4) noted that ‘the public
understanding of CPI makes it the appropriate indexation methodology’. The
Australian Health Insurance Association also argued in favour of the CPI because it
is used to index other thresholds and payment levels:
The use of the CPI would ensure a consistent policy approach to the adjustment of
Australian Government health and welfare thresholds and payments, as the CPI is also
used to adjust:
• the Medicare Levy Low Income threshold;
• the Medicare Safety Net;
• the PBS Safety Net;
• the Baby Bonus; and
• Family Tax Benefits A and B. (sub. 18, p. 10)

However, not all government thresholds and payments are indexed by the CPI. As
noted by DOHA:
AWOTE is the indexation measure used for a number of other income thresholds (e.g.,
the concessional superannuation contributions cap and the low-rate threshold for
superannuation lump sum payments). (sub. 32, p. 26)

240 PUBLIC AND PRIVATE

HOSPITALS
In addition, it is important that the MLS indexation factor tracks taxable income
over time, given the goal of keeping the MLS targeted at high income earners. As
noted by Catholic Health Australia (sub. 20, p. 2):
It is important that whatever index is chosen, it provides a reasonable representation of
movements in income levels.

10.4 Assessment of potential indexation factors

To assess potential indexation factors, the Commission examined what the MLS
income thresholds would have been if indexed between 1997-98 and 2007-08 by
one of the four measures outlined above. Historically, the CPI has increased at a
slower rate than wages and so it would have led to the lowest thresholds between
1997-98 and 2007-08. If the singles threshold had been indexed by AWOTE or
AWTE from 1997-98 to 2007-08, it would have been around 17 per cent higher in
2007-08 than if it had been indexed by the CPI (figure 10.1).

Figure 10.1 MLS income thresholds if there had been indexation, singlesa

AWOTE AWTE CPI WPI

80

70
$'000/year

60

50

40
97-98 98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08

a Data for WPI are only available from August 1997. Therefore, an indexation factor for 1998-99 could not be
calculated. The hypothetical WPI thresholds were indexed instead by AWOTE for this year because it is also a
wage measure.
Source: ABS (Average Weekly Earnings, Australia, Cat. no. 6302.0; Consumer Price Index, Australia,
Cat. no. 6401.0; Labour Price Index, Australia, Cat. no. 6345.0), Productivity Commission estimates.

INDEXATION OF THE 241

MLS THRESHOLDS
Using these hypothetical thresholds and data on the distribution of income, the
Commission calculated the proportion of single taxpayers and family taxpayers who
would have been subject to the MLS had indexation occurred.

If the MLS singles threshold had been indexed, the threshold would, in 2007-08,
have been $77 000 (with AWOTE indexation), $76 000 (AWTE), $65 000 (CPI)
and $69 000 (WPI). Similarly, if the MLS families threshold had been indexed, the
threshold for families would, in 2007-08, have been $154 000 (with AWOTE),
$152 000 (AWTE), $130 000 (CPI) and $138 000 (WPI).

To ensure the MLS remains focused on high income earners, the indexation factor
used needs to be commensurate with the changes in their income. The Commission
used data on the distribution of income to estimate the proportion of taxpayers who
would have had incomes above the MLS income thresholds under each of the four
indexation options from 1999-2000 to 2007-08, (box 10.2).

Box 10.2 Income distribution data used in MLS analysis

The Commission obtained data from the Australian Government Department of the
Treasury on the distribution of income used from 1999-2000 to 2005-06, based on
confidentialised sample unit record data extracted by the Australian Taxation Office
(ATO). These samples have typically been extracted on an annual basis shortly after
the publication of Taxation Statistics.
In the period after the sample data extraction, some tax returns continue to be lodged
by ‘late lodgers’. Therefore, the aggregate number of taxpayers is likely to be greater
than estimated using the relevant sample data.
For the years 2006-07 and 2007-08, the analysis is based on data recently provided to
the Treasury by the ATO. The data for 2007-08 may not be complete. However, due to
the Tax Bonus, the number of returns processed for 2007-08 in a 12-month period
exceeds what would be expected in a normal year.
Consistent data for the years 1997-98 and 1998-99 were not available.

If the MLS singles threshold had been indexed between 1999-2000 and 2007-08,
the proportion of single taxpayers subject to the MLS would have been most stable
had AWOTE been used (figure 10.2). Therefore, indexing the MLS by AWOTE
would have gone closest to achieving the Australian Government’s objective of
keeping the MLS focused on high income earners among singles.

Nevertheless, even with AWOTE indexation, the proportion of single taxpayers that
would have been subject to the MLS would have increased from 11.4 to
15.3 per cent between 1999-2000 and 2007-08.

242 PUBLIC AND PRIVATE

HOSPITALS
Figure 10.2 Proportion of single taxpayers subject to the MLS under
alternative indexation options

AWOTE AWTE CPI WPI No indexation

35
Per cent of single taxpayers

30

25

20

15

10

0
99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08
Financial year

Source: Australian Government Department of the Treasury (unpublished); ABS (Average Weekly Earnings,
Australia, Cat. no. 6302.0; Consumer Price Index, Australia, Cat. no. 6401.0; Labour Price Index, Australia,
Cat. no. 6345.0); Productivity Commission estimates.

In contrast, the proportion of single taxpayers subject to the MLS would have
increased significantly if the CPI (from 11.4 to 20.5 per cent) or WPI (from 11.4 to
18.4 per cent) were used. Therefore, the CPI and WPI are clearly unsuitable if the
intention of the Government is for the MLS to target high income earners among
singles.

If AWTE had been used, a slightly higher proportion of single taxpayers would
have been subject to the MLS in some years than if AWOTE had been used.

Given that no indexation actually took place between 1999-2000 and 2007-08,
33 per cent of single taxpayers were potentially subject to the MLS in 2007-08.

As discussed above, when the MLS was introduced, the threshold for families was
set at twice the singles threshold. If the MLS families threshold had been indexed,
between 1999-2000 and 2007-08, the proportion of families subject to the MLS
would have been most stable had AWOTE been used (figure 10.3). Therefore,
indexing the MLS income thresholds by AWOTE would also have best met the
Australian Government’s objective of keeping the MLS focused on high income
earners of families.
INDEXATION OF THE 243
MLS THRESHOLDS
Figure 10.3 Proportion of taxpayers who were members of a family subject
to the MLS under alternative indexation optionsa

AWOTE AWTE CPI WPI No indexation

35

30
Per cent of family taxpayers

25

20

15

10

0
99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08
Financial year

a For purposes of simplicity, the $1500 threshold increase for a second and each additional child is not taken
into account in this analysis. Therefore, these estimates slightly overstate the proportion of family taxpayers
who would have been subject to the MLS.
Source: Australian Government Department of the Treasury (unpublished); ABS Source: ABS (Average
Weekly Earnings, Australia, Cat. no. 6302.0; Consumer Price Index, Australia, Cat. no. 6401.0; Labour Price
Index, Australia, Cat. no. 6345.0); Productivity Commission estimates.

Nevertheless, with AWOTE, the proportion of family taxpayers that would have
been subject to the MLS would still have increased, from 9.4 per cent in 1999-2000
to 14.1 per cent in 2007-08.

The proportion of family taxpayers subject to the MLS would have increased by
much more had the CPI (from 9.4 to 19.5 per cent) or WPI (from 9.4 to
17.2 per cent) been used as the indexation factor. Hence the CPI and WPI are
unsuitable if the intention of the Government is for the MLS to target high income
families.

If AWTE had been used, a slightly higher proportion of family taxpayers would
have been subject to the MLS in some years than if AWOTE had been used.

The analysis in this chapter shows that if the MLS income thresholds had been
indexed, the proportion of single and family taxpayers subject to the MLS would

244 PUBLIC AND PRIVATE

HOSPITALS
each have increased even if AWOTE had been used. The other indexation factors
would have had led to more people being within the scope of the MLS.

The Government’s objective is for the MLS to target high income earners. But the
four indexation options examined above would not have held the proportion of
taxpayers subject to the MLS constant.

The Commission investigated using an indexation factor that specifically measured

income changes for high income earners, rather than average earnings. Since
8 per cent of single taxpayers were subject to the MLS when it was introduced, the
Commission considered the feasibility of indexing the thresholds based on
movement in the ninth decile (ninetieth percentile) of incomes for singles.5 The
Commission could not find data published annually by income decile. Thus, at this
point in time this option is not feasible. However, this option could be considered in
the future if data were available in a robust and timely manner.

In light of the above, the Commission considers that AWOTE is the most
appropriate indexation factor for the MLS thresholds, and is more likely to meet the
Australian Government’s goal of the MLS being targeted at high income earners
than if other indexation factors were used.

FINDING 10.1

Average weekly ordinary time earnings is the most appropriate indexation factor
for the Medicare Levy Surcharge income thresholds.

5 An alternative option would be to index the thresholds based on movements in the eighty-fifth
percentile of incomes for singles, as around 85 per cent of single taxpayers were below the MLS
income threshold for singles after it was increased in 2008-09.
INDEXATION OF THE 245
MLS THRESHOLDS
A Public consultation

Table A.1 Submissions received

Participant Submission numbera
ACT Health DR52
Access Economics DR60
Australasian College for Emergency Medicine 14
Australian and New Zealand College of Anaesthetists 11
Australian Commission on Safety and Quality in Health Care 24
Australian Council on Healthcare Standards 13#
Australian Government Department of Health and Ageing DR69, 32
Australian Health Insurance Association DR58, 18
Australian Health Service Alliance DR53, 1
Australian Healthcare and Hospitals Association 33
Australian Medical Association DR55, 28
Australian Nursing Federation DR57, 17
Australian Private Hospitals Association DR71, 25#
Australian Society of Anaesthetists 9
Australian Unity 31
Becton Dickinson 29
Bio21 Australia Limited 35
Catholic Health Australia DR62, 20
Centre for Health Communication, UTS 3
Centre for Health Economics, Monash University 7
Centre for Health Economic Research and Evaluation, UTS DR68
Children’s Hospitals Australasia 21#
Cochrane Consumer Network DR47
Consumers Health Forum in Australia DR59
Croakey 2#
Dr. John Deeble AO DR56
Department of Health, Government of Western Australia DR72
Doctors Reform Society of Australia DR50
Epworth Healthcare DR70
Flinders University – Centre for Clinical Change 10#
Gerry Carton Consulting Pty Ltd 12
(Continued next page)

PUBLIC 247
CONSULTATION
Table A.1 (Continued)

Participant Submission numbera

Grattan Institute DR66
Hanlon, Mark DR46
Harper, Richard 6
Health Services Association of New South Wales DR54
Health Services Union DR63
Healthcare Associated Infection Unit, Communicable Disease Control 38
Directorate, Department of Health WA
Healthscope Limited DR67, 42*
Health Services Association of NSW Branch DR54
Medical Technology Association of Australia DR48
Melbourne Institute of Applied Economic & Social Research, 16
University of Melbourne
National Coalition of Public Pathology DR49
NSW Department of Health 40, 41
NSW Health Surgical Services Taskforce DR43
Pharmaceutical Society of Australia 19
Private Cancer Physicians of Australian and Haematology and 36
Oncology Clinics of Australia
Private Health Insurance Intermediaries Association 5
Private Health Insurance Ombudsman 26
Queensland Health 27
Queensland Nurses’ Union DR51
Repatriation Commission 39
Rhonda Kerr and Associates, Health Facility Planning DR44, 34
Royal Australasian College of Surgeons 30#
Royal College of Pathologists of Australasia 23
SA Department of Health DR45, 4
Sinclair, Mark 8#
Tasmanian Department of Health and Human Services DR61, 37#
UnitingCare Health 16
a A hash (#) indicates the submission includes attachments. An asterix (*) indicates the submission is ‘In
Confidence’.

248 PUBLIC AND PRIVATE

HOSPITALS
Table A.2 Visits
Participant (grouped by visit location)

Canberra
Australian Bureau of Statistics
Australian Centre for Economic Research on Health
Australian Health Insurance Association
Australian Healthcare and Hospitals Association
Australian Institute of Health and Welfare
Australian Medical Association
Australian Private Hospitals Association
Catholic Health Australia
Department of Health and Ageing (Australian Government)
Department of Veterans’ Affairs (Australian Government)
Women’s and Children’s Hospitals Australasia
Melbourne
Australian Health Service Alliance
Australian Society of Anaesthetists
Business Council of Australia
Centre for Health Economics, Monash University
Harper, Richard
Health Insurance Restricted Membership Association of Australia
Healthscope Limited
Medibank Private
Victorian Department of Human Services
Visasys
Sydney
Australian Commission on Safety and Quality in Health Care
NSW Department of Health
Ramsay Health Care

PUBLIC 249
CONSULTATION
Table A.3 Participants in initial roundtable
Canberra 30 June 2009
ACT Department of Health
Australian Bureau of Statistics
Australian Centre for Economic Research on Health
Australian Commission on Safety and Quality in Health Care
Australian Council on Healthcare Standards
Australian Health Insurance Association
Australian Health Service Alliance
Australian Healthcare and Hospitals Association
Australian Institute of Health and Welfare
Australian Medical Association
Australian Nursing Federation
Australian Private Hospitals Association
Catholic Health Australia
Centre for Health Economics Research and Evaluation
Consumers’ Health Forum
Department of Health and Ageing (Australian Government)
Department of Veterans’ Affairs (Australian Government)
Healthscope Limited
Queensland Health
Ramsay Health Care
SA Department of Health
Tasmanian Department of Health and Human Services
Victorian Department of Human Services
WA Department of Health

250 PUBLIC AND PRIVATE

HOSPITALS
Table A.4 Participants in Discussion Draft roundtable
Canberra 22 October 2009
ACT Department of Health
Australian Bureau of Statistics
Australian Council on Healthcare Standards
Australian Commission on Safety and Quality in Health Care
Australian Healthcare and Hospitals Association
Australian Health Insurance Association
Australian Health Service Alliance
Australian Institute of Health and Welfare
Australian Medical Association
Australian Nursing Federation
Australian Private Hospitals Association
Catholic Health Australia
Centre for Health Economics Research and Evaluation
Consumers’ Health Forum
Department of Health (Government of Western Australia)
Department of Health and Ageing (Australian Government)
Department of Health and Families (NT Government)
Healthscope Limited
NSW Department of Health
Queensland Health Department
Ramsay Health Care
SA Department of Health
Tasmania Department of Health and Human Services
Victorian Department of Human Services

PUBLIC 251
CONSULTATION
Table A.5 Teleconference participants
17 September 2009 (Hospital and medical costs)
ACT Department of Health
Catholic Negotiating Alliance
Department of Health and Ageing (Australian Government)
Healthscope Limited
Mater Health Services
Mercy Health and Aged Care
NSW Department of Health
NT Department of Health and Families
Queensland Health
SA Department of Health
St Andrew's Hospital
Tasmanian Department of Health and Human Services
UnitingCare Health
Victorian Department of Health
WA Department of Health

23 November 2009 (Multivariate analysis)

Australian Health Insurance Association
Australian Health Service Alliance
Australian Institute of Health and Welfare
Australian Private Hospitals Association
Catholic Health Australia
Department of Health and Ageing (Australian Government)
Healthscope Limited
NSW Department of Health
Queensland Health
SA Department of Health
Victorian Department of Health
WA Department of Health

252 PUBLIC AND PRIVATE

HOSPITALS
B National Healthcare Agreement
performance indicators

The National Healthcare Agreement (NHA) is one of six national agreements

incorporated in the current Intergovernmental Agreement on Federal Financial
Relations (box B.1) (COAG 2008c). The NHA provides governments with a
structure for the funding and delivery of services across the health sector. It defines
the respective roles and responsibilities of the Australian and state and territory
Governments, and sets out mutually agreed objectives and outcomes for the sector.
Monitoring and reporting of government performance against agreed outcomes and
benchmarks will be conducted using a set of performance indicators designed for
that purpose.

B.1 The National Healthcare Agreement

The NHA had its origins in the 20 December 2007 meeting of the Council of
Australian Governments (COAG). COAG agreed to commence a program of
substantive reform in order to increase productivity, address emerging inflationary
pressures and improve the quality of services delivered to the Australian
community. Health and ageing was one of seven areas identified for reform.

The National Health and Hospitals Reform Commission (NHHRC) was established
in February 2008 to support reform in the area of health and ageing. Terms of
reference provided to the NHHRC included provision of advice on a framework for
the next Australian Health Care Agreements (AHCAs), and development of a
long-term health reform plan (COAG 2007; NHHRC 2008, 2009).

While previous AHCAs were bilateral agreements between the Australian

Government and each state and territory, the current NHA is a single agreement
between the Australian and all state and territory governments. It took effect
1 July 2009 and will be reviewed every four to five years, commencing midway
through the first four to five year period (COAG 2008d).

NHA PERFORMANCE 253

INDICATORS
 Box B.1 National Agreement Reporting
In November 2008, the Council of Australian Governments (COAG) endorsed a new
Intergovernmental Agreement on Federal Financial Relations (IGA) (2008b). The IGA
provides ‘an overarching framework for the Commonwealth’s financial relations with
the States and Territories’. In addition, the IGA sets out ‘roles and responsibilities of
each level of government and an improved focus on accountability for better outcomes
and better service delivery’.
The six National Agreements incorporated in the IGA are the:
• National Healthcare Agreement
• National Education Agreement
• National Agreement for Skills and Workforce Development
• National Affordable Housing Agreement
• National Disability Agreement
• National Indigenous Reform Agreement.
Each National Agreement contains objectives, outcomes, outputs and performance
indicators for the sector, as well as performance benchmarks, policy directions and
priority reform areas. National Agreements also set out the respective roles and
responsibilities of the Australian and state and territory governments in the delivery of
services. The performance of all governments in achieving mutually agreed outcomes
and benchmarks will be monitored and assessed by the COAG Reform Council and
reported publicly on an annual basis.
National Partnerships (NPs) are another form of agreement that fund specific projects
and facilitate and/or reward states and territories that deliver on nationally significant
reforms. They are bilateral agreements between the Australian and individual state and
territory governments. NPs that relate to the health sector include the National
Partnership Agreement on Hospital and Health Workforce Reform, the National
Partnership Agreement on Preventive Health and the National Partnership Agreement
on Closing the Gap in Indigenous Health Outcomes.
Source: COAG (2008c).

The NHA has the overarching objective ‘to improve health outcomes for all
Australians and the sustainability of the Australian health system’ (COAG 2008d).
Developed in the context of growing challenges to the sustainable provision of
healthcare, it recognises the need for reform of the health sector as a whole in order
to achieve this objective. Challenges include access to services, the growing burden
of chronic disease, population ageing and escalating costs associated with new
health technologies (COAG 2007; NHHRC 2008, 2009).

254 PUBLIC AND PRIVATE

HOSPITALS
Unlike previous agreements, which focused exclusively on public hospitals, the
NHA extends across preventative, primary, sub-acute, acute and aged care, and is
intended to incorporate private sector services where relevant (COAG 2008d;
DOHA 2009f). It directly addresses issues of inequitable access to healthcare for
Indigenous Australians, residents of rural and remote areas and the
socioeconomically disadvantaged. The NHA is also designed to address concerns
about the long-term sustainability of the health system.

In another departure from previous agreements, the NHA addresses issues of

governance. It clarifies roles and responsibilities of the Australian and state and
territory governments in delivering health services. It sets out mutually agreed
objectives and intended outcomes across the continuum of care, and specifies policy
directions and reform areas that governments have undertaken to prioritise. The
comparative performance of governments in achieving objectives and outcomes will
be monitored and assessed against agreed progress and output indicators.

The NHA is organised around agreed long-term objectives in seven areas, one of
which is ‘hospital and related care’ (table B.1). Intended outcomes and associated
performance indicators (progress measures and outputs) are set out for each of the
objectives (table B.2). This structure recognises that, while hospitals are integral to
a comprehensive healthcare system, they do not operate in isolation from other parts
of the health sector (NHHRC 2008, 2009). Hospital performance is affected not
only by internal activities, but also by the performance of, and interaction between,
acute, sub-acute and primary healthcare services.

Table B.1 Objectives of the National Healthcare Agreement

Area Long-term objectives

Prevention Australians are born and remain healthy.

Primary and Community Australians receive appropriate high quality and affordable primary
Health and community health services.
Hospital and Related Care Australians receive high quality hospital and hospital-related care that
is appropriate and timely.
Aged Care Older Australians receive appropriate high quality and affordable
health and aged care services.
Patient Experience Australians have positive health and aged care experiences which
take account of individual circumstances and care needs.
Social Inclusion and Australia’s health system promotes social inclusion and reduces
Indigenous Health disadvantage, especially for Indigenous Australians.
Sustainability Australians have a sustainable health system.
Source: COAG (2008d).

NHA PERFORMANCE 255

INDICATORS
Table B.2 National Healthcare Agreement Reporting Structure
Outcome Progress measure Output

Prevention
Children are born and remain Proportion of babies born with Immunisation rates for
healthy. low birth weight. vaccines in the national
Australians have access to the Incidence/prevalence of schedule.
support, care and education they important preventable Cancer screening rates
need to make healthy choices. diseases. (breast, cervical, bowel).
Australians manage the key risk Risk factor prevalence. Proportion of children with
factors that contribute to ill fourth year developmental
health. health check.
Primary and community health
The primary healthcare needs of Access to general practitioners, Number of primary care
all Australians are met effectively dental and other primary services per 1000 population
through timely and quality care in healthcare professionals. (by location).
the community. Proportion of diabetics with Number of mental health
People with complex care needs HbA1c below 7 per cent. services.
can access comprehensive, Life expectancy (including the Proportion of people with
integrated and coordinated gap between Indigenous and selected chronic disease
services. non-Indigenous). whose care is planned
Infant/young child mortality rate (asthma, diabetes, mental
(including the gap between health).
Indigenous and Number of women with at least
non-Indigenous). one antenatal visit in the first
Potentially avoidable deaths. trimester of pregnancy.
Treated prevalence rates for
mental illness.
Selected potentially
preventable hospitalisations.
Selected potentially avoidable
general practitioner type
presentations to emergency
departments.
Hospital and related care
Australians receive high quality Waiting times for services. Rates of services provided by
hospital and hospital-related Selected adverse events. public and private hospitals.
care that is appropriate and Unplanned/unexpected
timely. readmissions.
Survival of people diagnosed
with cancer.

(Continued on next page)

256 PUBLIC AND PRIVATE

HOSPITALS
Table B.2 (continued)

Outcome Progress measure Output

Aged care
Older Australians receive high Residential and community Number of older people
quality, affordable health and aged care services receiving aged care services by
aged care services that are per 1000 population aged type (in the community and
appropriate to their needs and 70+ years. residential settings).
enable choice and seamless, Selected adverse events in Number of aged care
timely transitions within and residential care. assessments conducted.
across sectors. Number of younger people with
disabilities using residential,
Community Aged Care Package
and Extended Aged Care at
Home services.
Number of people 65+ receiving
sub-acute and rehabilitation
services.
Number of hospital patient days
by those eligible and waiting for
residential aged care.
Patient experience
All Australians experience best Nationally comparative
practice care suited to their information that indicates
needs and circumstances levels of patient satisfaction
informed by high quality health around key aspects of care
information. they received.
Patients experience seamless
and safe care when
transferring between settings.
Social inclusion and Indigenous health
Indigenous Australians and Age standardised mortality. Indigenous Australians in the
those living in rural and remote Access to services by type of health workforce.
areas or on low incomes service compared to need.
achieve health outcomes Teenage birth rate.
comparable to the broader
Hospitalisation for injury and
population.
poisoning.
Children's hearing loss.
Sustainability
Australians have a sustainable Net growth in health workforce Number of accredited or filled
health system that can respond (doctors, nurses, midwives, clinical training positions.
and adapt to future needs. dental practitioners,
pharmacists).
Allocation of health and
aged-care expenditure.
Cost per casemix-adjusted
separation for both acute and
non-acute care episodes.
Source: COAG (2008d).

NHA PERFORMANCE 257

INDICATORS
The NHA sets out agreed performance benchmarks against three aspects of
‘hospital and related care’ (COAG 2008d) that, along with performance indicators,
will be considered in assessment of the comparative performance of governments
against the NHA. These aspects are:
• administration — within five years implement a nationally-consistent approach
to activity-based funding for public hospital services, which also reflects the
community service obligations for small and regional hospital services
• emergency departments — 80 per cent of emergency department presentations
are seen within clinically recommended triage times as recommended by the
Australasian College for Emergency Medicine by 2012-13
• quality and safety — the rate of Staphylococcus aureus (including Methicillin-
resistant Staphylococcus aureus (MRSA)) bacteraemia is no more than
2 per 10 000 occupied bed days for acute care public hospitals by 2011-12 in
each state and territory.

B.2 Monitoring and reporting

The COAG Reform Council (CRC) will monitor and assess government
performance in relation to the agreed objectives, outcomes, outputs and
performance benchmarks. Performance will be reported publicly on an annual basis,
commencing with the 2008-09 financial year. Data will be provided to the CRC by
the Steering Committee for the Review of Government Service Provision
(COAG 2008c).

Hospital and related care

Under the NHA, and like the AHCAs, public hospital funding is the joint
responsibility of Australian and state and territory governments. States and
territories are responsible for providing health and emergency services through the
public hospital system. These services are to be accessible to all eligible Australians
free of charge, within clinically appropriate periods, on the basis of clinical need.
States and territories also have responsibility for ensuring that those who elect to be
treated as private patients in public hospitals do so on the basis of informed
financial consent (COAG 2008d).

Governments have agreed to particular policy directions and priority areas for
reform in order to achieve the agreed outcomes and objectives (COAG 2008d). In
relation to ‘hospital and related care’, the long-term objective is for ‘Australians [to]

258 PUBLIC AND PRIVATE

HOSPITALS
receive appropriate high quality and affordable hospital and hospital-related care’.
Related policy directions and priority areas for reform are provided in box B.2.

Box B.2 Policy directions and priority reform areas

The following policy directions and priority reform areas have been agreed by the
Australian, state and territory governments. They include those specified against the
‘hospital and related care’ objective, as well as those specified under other objectives
but related to hospital performance.

Hospital and related care

Agreed policy directions include:
• reduce elective surgery and emergency department waiting times
• increase technical efficiency of public hospital services
• improve safety and quality of care, and patient access to performance information
• more effective assessment and support of patients before admission to, and on
discharge from, acute-care settings.
Agreed priority areas for reform include:
• develop nationally consistent activity-based funding for public hospital services
• implement improvements in hospital quality and safety
• increase the proportion of elective surgery patients treated within clinically
recommended waiting times
• improve access to rehabilitation, post-acute and transition care services
• improve assessment of relative performance of public and private hospitals
• improve quality of data on non-admitted patient services
• improve levels of informed financial consent for private patients in public and private
hospitals.

Areas other than hospital and related care

Agreed policy directions for each target area are:
• aged care — provide continuity of care across hospitals, community and aged care
• sustainability — reward allocative efficiency across preventative, primary, acute,
sub-acute, rehabilitation and aged care services.
(Continued on next page)

NHA PERFORMANCE 259

INDICATORS
 Box B.2 (continued)

Agreed priority areas for reform for each target area are:
• aged care — provide older patients in hospitals with timely access to appropriate
sub-acute care, including rehabilitation
• sustainability — move to a proper long-term share of Commonwealth funding for the
public hospital system.
Source: COAG (2008d).

Performance indicators

Performance indicators to be reported under the NHA largely reflect the agreed
policy directions and priority reform areas. ‘Hospital and related care’ performance
indicators (progress measures and outputs) presented in table B.3 include items
from a proposed NHA indicator set released in 2008 (AIHW 2008a). Further work
to develop these indicators has been undertaken, but is yet to be publicly released.
NHA indicators for other areas that relate to hospital performance are listed in
table B.4.

Table B.3 Hospital and related care performance indicators

Progress measure Output

Waiting times for: Rates of services provided by public and private

• elective surgery hospitals per 1000 weighted population by
patient type.
• emergency department services.
Selected adverse events in acute and sub-acute
care settings, including:
• adverse drug events
• Staphylococcus aureus (including MRSA)
bacteraemia
• pressure ulcers
• falls resulting in patient harm
• intentional self-harm.
Unplanned or unexpected readmissions within
28 days of selected surgical admissions.
Survival of people diagnosed with cancer
(5 year relative rate).
Source: COAG (2008d); AIHW (2008a).

260 PUBLIC AND PRIVATE

HOSPITALS
Table B.4 Other NHA indicators related to hospital performance
Area Progress measure Output

Primary and community health Selected potentially

preventable hospitalisations.
Aged care Number hospital patient days by
those eligible and waiting for
residential aged care.
Patient experience Nationally comparable
information that indicates levels
of patient satisfaction around
key aspects of care they
received.
Social inclusion and Indigenous Access to services by type of
health service compared to need.
Sustainability Cost per casemix-adjusted
separation for both acute and
non-acute care episodes.
Source: COAG (2008d).

NHA PERFORMANCE 261

INDICATORS
C Other health performance monitoring
frameworks

The National Healthcare Agreement (NHA) and associated performance indictors

are described in appendix B. Two other national health performance monitoring
frameworks developed prior to the NHA performance indicators are the National
Health Performance Framework (NHPF) and the Report on Government Services
health performance monitoring framework. These frameworks focus on the health
system as a whole, or large components of it, and both include performance
monitoring of public hospitals.

C.1 National Health Performance Framework

The NHPF was developed to report the performance of the Australian health system
at a national level. The NHPF was developed by the National Health Performance
Committee (NHPC) at the request of the Australian Health Ministers’ Conference
and was published in 2001 (NHPC 2001).1 In August 2001, Australian Health
Ministers agreed to this overarching performance framework for use in reporting
across all areas of the health system.

The NHPF has a broader focus than the national reporting previously undertaken by
the NHPC, which had focused on performance of acute hospital inpatient services.
The NHPF focuses on overall health systems performance, which includes not only
acute inpatient services, but also services such as community health, general
practice and public health. The NHPF also differs from previous Australian
frameworks as it focuses not only on system performance, but also on health status

1 The NHPC’s mission was to foster the use of benchmarking based on national performance
measures and indicators to improve the quality of care of health services. The group was a
standing committee of the National Health Information Management Principal Committee, which
in turn advised the Australian Health Ministers’ Advisory Council on matters including
information requirements and technology planning. The NHPC comprised representatives from
the Australian, State and Territory Governments and a number of other organisations, including
the Australian Health Insurance Association, the Australian Private Hospitals Association and the
Australian Institute of Health and Welfare. Some NHPC functions have now been assumed by
the National Health Information Standards and Statistics Committee.
OTHER MONITORING 263
FRAMEWORKS
and health determinants. It also includes areas such as capability and sustainability
that had not been widely reported in the past.
The NHPF was seen as a structure to guide the understanding and evaluation of
health service performance in Australia. The framework consists of three tiers
(table C.1):
• health status and outcomes
• determinants of health
• health system performance.

Table C.1 The National Health Performance Framework

Health status and outcomes
How healthy are Australians? Is it the same for everyone? Where is the most opportunity for
improvement?
Health Conditions Human Function Life Expectancy and Deaths
Wellbeing
Prevalence of disease, Alterations to body, Broad measures of Age and/or condition
disorder, injury or structure or function physical, mental and specific mortality rates.
trauma or other (impairment), activities social wellbeing of
health-related states. (activity limitation) and individuals and other
participation derived indicators such
(restrictions in as disability-adjusted
participation). life expectancy.
Determinants of health
Are the factors determining health changing for the better? Is it the same for everyone? Where and
for whom are they changing?
Environmental Socioeconomic Community Health Behaviours Person-related
Factors Factors Capacity Factors
Physical, chemical Socioeconomic Characteristics of Attitudes, beliefs, Genetic-related
and biological factors such as communities and knowledge and susceptibility to
factors such as air, education, families such as behaviours, e.g. disease and other
water, food and soil employment, population patterns of eating, factors such as
quality resulting from per capita density, age physical activity, blood pressure,
chemical pollution expenditure on distribution, health excess alcohol cholesterol levels
and waste disposal. health, and literacy, housing, consumption and and body weight.
average weekly community smoking.
earnings. support services
and transport.
(Continued next page)

264 PUBLIC AND PRIVATE

HOSPITALS
Table C.1 (continued)

Health system performance

How well is the health system performing in delivering quality health actions to improve the health
of all Australians? Is it the same for everyone?
Effective Appropriate Efficient

Care, intervention or action Care/intervention/action Achieving desired results with

achieves desired outcome. provided is relevant to the most cost effective use of
client’s needs and based on resources.
established standards.

Responsive Accessible Safe

Service provides respect for Ability of people to obtain The avoidance or reduction to
persons and is client health care at the right place acceptable limits of actual or
orientated. It includes respect and right time irrespective of potential harm from health care
for dignity, confidentiality, income, physical location and management or the environment
participation in choices, cultural background. in which health care is delivered.
promptness, quality of
amenities, access to social
support networks, and choice
of provider.
Continuous Capable Sustainable
Ability to provide uninterrupted, An individual’s or service’s System’s or organisation’s
coordinated care or service capacity to provide a health capacity to provide infrastructure
across programs, practitioners, service based on skills and such as workforce, facilities and
organisations and levels over knowledge. equipment, and be innovative
time. and respond to emerging needs
(research, monitoring).
Source: NHPC (2001).

Questions are posed for each tier and dimension and it was anticipated by the
NHPC that performance indicators would be chosen or developed to provide
answers about the performance of the system (NHPC 2001). Equity is considered to
be integral to each of the three tiers and is represented in each by the question ‘is it
the same for everyone?’ Quality is also an integral part of the framework, and the
dimensions considered in determining the quality of the health system are very
similar to those measuring health system performance.

Indicators within the NHPF

The NHPC was also tasked with identifying and developing indicators to be
reported against the NHPF. The selection criteria used by the NHPC to select the
indicators are shown in box C.1. An indicator could provide information in several
dimensions across the framework.

OTHER MONITORING 265

FRAMEWORKS
 Box C.1 Selection criteria used by the NHPC for health performance
indicators
Generic indicators when used at a program level to whole-of-system level should have
all or some of the following qualities. They should:
1. Be worth measuring.
The indicators represent an important and salient aspect of the public’s health or the
performance of the health system.
2. Be measurable for diverse populations.
The indicators are valid and reliable for the general population and diverse
populations (that is, Aboriginal and Torres Strait Islander populations, sex,
rural/urban, socioeconomic etc.)
3. Be understood by people who need to act.
People who need to act on their own behalf or that of others should be able to
readily comprehend the indicators and what can be done to improve health.
4. Galvanise action.
The indicators are of such a nature that action can be taken at the national, state,
local or community level by individuals, organised groups and public and private
agencies.
5. Be relevant to policy and practice.
Actions that can lead to improvement are anticipated and feasible — they are
plausible actions that can alter the course of an indicator when widely applied.
6. Reflect results of actions when measured over time.
If action is taken, tangible results will be seen indicating improvements in various
aspects of the nation’s health.
7. Be feasible to collect and report.
The information required for the indicator can be obtained at reasonable cost in
relation to its value and can be collected, analysed and reported on in an
appropriate time frame.
8. Comply with national processes of data definitions.
Source: NHPC (2002).

The NHPC reported indicator data against the NHPF in its National Report on
Health Sector Performance Indicators for 2001 and 2003 (NHPC 2002 and 2004).
The 2003 Report contained 44 indicators, with eight reported against health status
and outcomes, 11 against determinants of health and 25 reported against health
system performance (table C.2). The NHPF has since been reported as part of the
Australia’s Health report published by the Australian Institute of Health and
Welfare, most recently in 2008 (AIHW 2008c).

266 PUBLIC AND PRIVATE

HOSPITALS
Table C.2 Indicators reported in the National Report on Health Sector
Performance, 2003
Health status and outcomes
Health Conditions Human Function Life Expectancy and Deaths
Wellbeing
Incidence of cancer. Severe or profound Life expectancy. Potentially
Incidence of heart core activity limitation. Psychological avoidable deaths.
attacks. distress. Infant mortality.
Mortality for
national Health
Priority Area
diseases and
conditions.
Determinants of health
Environmental Socioeconomic Community Health Behaviours Person-related
Factors Factors Capacity Factors
Children Income Informal care. Adult smoking. Low birthweight
exposed to inequality. Risky alcohol babies.
tobacco consumption. High blood
smoke in the Fruit and pressure.
home. vegetable intake.
Availability of Physical inactivity.
fluoridated
Overweight and
water.
obesity.
Health system performance
Effective Appropriate Efficient
Unsafe sharing of needles. Appropriate use of Hospitals costs.
Teenage purchase of antibiotics. Length of stay in hospital.
cigarettes. Management of diabetes.
Cervical screening. Delivery by caesarean
Breast cancer screening. section.
Childhood immunisation. Hysterectomy rate.
Influenza vaccination.
Potentially preventable
hospitalisations.
Survival following acute
coronary heart disease event.
Cancer survival.
(Continued next page)

OTHER MONITORING 267

FRAMEWORKS
Table C.2 (continued)

Responsive Accessible Safe

Waiting times in emergency Bulk billing for non-referred Electronic prescribing and
departments. (GP) attendances. clinical data in general
Availability of GP services. practice.
Access to elective surgery. Adverse events treated in
hospitals.
Continuous Capable Sustainable
Enhance primary care Accreditation in general Health workforce.
services. practice.
Health assessments by
GPs.
Source: NHPC (2004).

Although the primary purpose of the NHPF was performance measurement at the
national level, the framework was intended to support performance measurement at
all levels of the health system. A number of groups involved in health performance
indicator development have adopted this framework for use within specific project
areas and in publications. For example, the health performance indicator
frameworks contained within the Report on Government Services have been aligned
as much as possible with the NHPF (SCRGSP 2009). In addition, the Aboriginal
and Torres Strait Islander Health Performance Framework is based on the NHPF
(AHMAC 2006). A set of key performance indicators for Australian public mental
health services was also developed using the NHPF (NMHWG 2005).

C.2 Review of Government Service Provision

The Review of Government Service Provision (the Review) has developed
performance monitoring frameworks that have been applied across a number of
areas of government service provision, including health services. Health services
examined in the Review include public hospitals, primary and community health,
breast cancer detection and management and specialist mental health management.
Data are reported against these frameworks on an annual basis in the Report on
Government Services (the Report) (box C.2).

Health services are included in the Report as they are an important component of
government service provision. Over 40 per cent of expenditure within the scope of
reporting of the Report on Government Services 2009 was accounted for by health
services (SCRGSP 2009).

268 PUBLIC AND PRIVATE

HOSPITALS
 Box C.2 Aims of the Review of Government Service Provision
Heads of government (now the Council of Australian Governments or COAG)
established the Review of Government Service Provision (the Review) to provide
information on the effectiveness and efficiency of government services in Australia. A
Steering Committee, comprising senior representatives from the central agencies of all
governments, manages the Review with the assistance of a Secretariat provided by the
Productivity Commission.
The Review was established in 1993 to:

• provide ongoing comparisons of the performance of government services

• report on service provision reforms that governments have implemented or that are
under consideration.
The Review has produced 14 editions of the annual Report on Government Services
since it was established, with the most recent being published in January 2009.
Source: SCRGSP (2009).

General framework

The Report’s general performance framework is set out in figure C.1. The
framework depicts the Review’s focus on outcomes, consistent with demand by
governments for outcome-oriented performance information. This outcome
information is supplemented by information on outputs. Output indicators are
grouped under ‘equity’, ‘effectiveness’ and ‘efficiency’ headings (SCRGSP 2009).

Outcome indicators provide information on the impact of a service on the status of

an individual or a group, and on the success of the service area in achieving its
objectives. Outputs are the actual services delivered. While the aim of the Review is
to focus on outcomes, they are often difficult to measure. The Report therefore
includes measures of outputs, with an understanding that there is a correlation
between those outputs and desired outcomes, and that the measures of outputs are
proxies for measures of outcomes.

OTHER MONITORING 269

FRAMEWORKS
Figure C.1 Report on Government Services general framework

Equity of
Equity of access
Equity Access outcome
indicators
indicators

Objectives Access
Access
indicators

Program
Appropriateness
Effectiveness Appropriateness effectiveness
indicators
indicators
PERFORMANCE
Quality
Quality
indicators

Technical Cost
Inputs per
Efficiency efficiency effectiveness
output unit
indicators indicators

Outputs Outcomes

Source: SCRGSP (2009).

A comprehensive view of performance reporting is taken by the Review, and its

frameworks incorporate indicators across all relevant dimensions of performance,
namely effectiveness, efficiency and equity. There are inherent tradeoffs in
allocating resources and dangers in analysing only some aspects of a service. For
example, a unit of service may have a high cost but be more effective than a
lower-cost service, and therefore be more cost effective. It is also important that
services are provided equitably.

Equity of access indicators relate to the gap in service delivery outputs and
outcomes between special-needs groups and the general population. Effectiveness
indicators measure how well the outputs of a service achieve the stated objectives of
that service. Effectiveness comprises appropriateness indicators, which measure
how well services meet client needs, and quality indicators, which reflect the extent
to which a service is suited to its purpose and conforms to specifications.
Effectiveness also includes access indicators whereby all Australians are expected
to have adequate access to services. This notion of access differs from that of equity
of access, which is concerned with access by special-needs groups. Efficiency
indicators measure how well services use their resources (inputs) to produce outputs
for the purpose of achieving desired outcomes.

270 PUBLIC AND PRIVATE

HOSPITALS
Health performance framework

The performance framework for health services in the Report on Government

Services reflects both the general Review framework and the NHPF. In the Report
on Government Services 2004, the Review of Government Service Provision sought
to align the health framework with the NHPF as far as possible. Complete
alignment was not possible, given the different terms of reference of the two
committees. The health framework differs from the general Review framework in
two respects. First, it includes four subdimensions of quality — safety,
responsiveness, capability and continuity — and, second, it includes an extra
dimension of effectiveness — sustainability (figure C.2). These additions are
intended to address the following key performance dimensions of the health system
in the NHPF that were not explicitly covered in the general Review framework:
• safety — the avoidance, or reduction to acceptable levels, of actual or potential
harm from health care services, management or environments, and the
prevention or minimisation of adverse events associated with health care
delivery
• responsiveness — the provision of services that are client-oriented and respectful
of clients’ dignity, autonomy, confidentiality, amenity, choices, and social and
cultural needs
• capability — the capacity of an organisation, program or individual to provide
health care services based on appropriate skills and knowledge
• continuity — the provision of uninterrupted, timely, coordinated healthcare,
interventions and actions across programs, practitioners and organisations
• sustainability — the capacity to provide infrastructure (such as workforce,
facilities and equipment), be innovative and respond to emerging needs
(NHPC 2001).

OTHER MONITORING 271

FRAMEWORKS
Figure C.2 Performance indicator framework for health services

Equity of access Equity of

Equity Access
indicators outcome
indicators

Access Access Indicators

Appropriateness
Appropriateness
indicators

Safety Safety indicators

Effectiveness
Responsiveness
Objectives Responsiveness
indicators
Quality
Program
Capability Capability indicators effectiveness
indicators
PERFORMANCE

Continuity Continuity indicators

Sustainability
Sustainability
indicators

Cost
Technical efficiency
Efficiency Inputs per output unit effectiveness
indicators
indicators

Outputs Outcomes

Source: SCRGSP (2009).

Specific performance indicator frameworks

The Review of Government Service Provision has used the health performance
framework to develop:
• detailed performance indicator frameworks for public hospitals and primary and
community health services

272 PUBLIC AND PRIVATE

HOSPITALS
• specific frameworks to examine the appropriate mix of services and
service-delivery mechanisms for two health management issues: breast cancer
and mental health.

Figures C.3 and C.4 depict the public hospitals performance indicator framework
and the maternity services indicator framework. Maternity services are included as
part of public hospital reporting in the Report on Government Services, as they are
an important component of services provided within public hospitals. Maternity
services accounted for 9.2 per cent of total acute separations in public hospitals and
around 11.0 per cent of the total cost of all acute separations in public hospitals in
2006-07 (SCRGSP 2009).

The frameworks depict the dimensions of both the Review of Government Service
Provision and the NHPF. The frameworks are populated with the performance
indicators. The choice of indicators has been strongly influenced by the priorities of
Australian, state and territory governments. For example, reducing elective surgery
waiting times has long been a priority of governments around Australia and waiting
times for elective surgery are included in the framework. As there has been a degree
of alignment between the Review framework and the NHPF, and both frameworks
reflect governments’ priorities, a number of indicators are common to both the
Review framework and the NHPF. In addition, the approach taken by the Review is
to use indicators that are already in use in Australia or internationally. Adopting
these indicators can lower the costs of, and reduce delays in, reporting performance.

The framework identifies those indicators that are not complete or directly
comparable. This signifies the Review’s approach of using acceptable, albeit less
than perfect indicators with appropriate caveats, rather than reporting no data at all
for an indicator. Data are generally presented for those jurisdictions that can
currently report, rather than waiting until data are available for all jurisdictions. The
framework also identifies those indicators that are yet to be developed or where data
are not available. This shows that even though reporting for these indicators is not
currently possible, these areas are still a priority of governments.

OTHER MONITORING 273

FRAMEWORKS
Figure C.3 Performance indicators for public hospitals

Source: SCRGSP (2009).

274 PUBLIC AND PRIVATE

HOSPITALS
Figure C.4 Performance indicators for maternity services

Equity Access
Caesareans for
selected primiparae
Access
Inductions for
selected primiparae
Appropriateness
Vaginal delivery
following previous Baby’s Apgar
primary caesarean score at 5
minutes

Fetal death rate

Objectives Perineal status
Quality Safety
after vaginal birth
Neonatal
Effectiveness death rate
Responsiveness

Perinatal
PERFORMANCE
Capability death rate

Continuity Gestation
standardised
Recurrent cost per perinatal
Sustainability mortality ratio
maternity separation

Total cost per

Efficiency
maternity separation
Mother’s
post-natal average
length of stay
Key to indicators
Outputs Outcomes
Data for these indicators are comparable, subject to
Text
caveats to each chart or table
Text Data for these indicators not complete, or not direc ly
comparable
Text These indicators yet to be developed or data not
collected for this Report

Source: SCRGSP (2009).

OTHER MONITORING 275

FRAMEWORKS
D Constructing estimates of hospital
and medical costs

Constructing cost estimates for this study has been a major challenge because:
• existing data collections are limited by inconsistent collection methods and
missing information
• differences between hospitals in the types of patients treated and services
provided make like-for-like comparisons difficult.

This appendix details how the Commission has sought to address the data
limitations, and take account of the diversity and complexity of hospitals, by
drawing on various data sources and, where necessary, incorporating adjustments to
make the data more comparable. However, the Commission readily acknowledges
that significant data shortcomings have limited its ability to construct
fully-comparable costs. The Commission therefore stresses that the cost estimates in
this report should be treated as experimental.

D.1 National Hospital Cost Data Collection

The National Hospital Cost Data Collection (NHCDC) is a voluntary collection of
public and private hospital cost and activity information that is collected each
financial year. The purpose of the NHCDC is to ‘produce national cost weights for
Australian Refined Diagnosis-Related Groups (AR-DRGs) and other statistics
relevant for hospital service costing and planning’ (Australian Government
Department of Health and Ageing (DOHA) 2008c, p.7). The first round of the
NHCDC was collected in 1996-97, and the most recent round (Round 12) was
collected in 2007-08.1

The collection and reporting process for the NHCDC has several steps, as outlined
in box D.1. These steps ultimately result in the production of two reports — the
Cost Report and the Peer Group Report.

1 Rounds 8, 9 and 10 (2003-04 to 2005-06) of the NHCDC only include data for public hospitals
(DOHA 2009a).
CONSTRUCTING COST 277
ESTIMATES
 Box D.1 NHCDC collection and reporting process
The collection and reporting process of the National Hospital Cost Data Collection
(NHCDC) is as follows:

Stage 1: Preparation
The preparation for data collection is the process that is followed at the start of each
new round. Collection is initially undertaken by public hospitals within each state and
territory, and by private hospitals and private hospital groups.
Guidelines for the collection of data are stipulated within the Hospital Reference
Manual which is released around August or September of the year prior to the
collection period.
Participant training is generally conducted by the Australian Government Department
of Health and Ageing (DOHA), and/or the relevant state or territory coordinator,
between January and March of the collection year.

Stage 2: Collection
The data collection component of the costing process is undertaken in collaboration
with the state and territory coordinators of public hospitals, private hospitals and private
hospital groups. They undertake initial quality assurance checks before data are
submitted to DOHA for further verification.

Stage 3: National receipt and processing

After the receipt of data, DOHA checks, processes and constructs the final files
required to produce a national database. DOHA then produces estimates for the total
hospital population based on the sample collection.

Stage 4: Analysis and reporting

Analysis and reporting of the data is finalised by the NHCDC team within DOHA. A
number of reports, including the Cost Report and the Peer Group Report (for public
hospitals), are then produced.
Source: DOHA (2008c).

In this report, the Commission has used unpublished NHCDC data from Round 12
(2007-08) to generate its cost estimates. The 2007-08 NHCDC was contributed to
by 241 public hospitals and 109 private hospitals (tables D.1, D.2 and D.3). This
covered 89 per cent of public acute separations and 72 per cent of private acute
separations (DOHA 2009a).2

2 Separations data supplied by state and territory coordinators for the NHCDC is the source of a
population estimate of 4 508 000 public-hospital separations. The AIHW (2009a) estimated that
the total number of public hospital acute separations in 2007-08 was 4 462 000, implying a
coverage of just over 90 per cent. Private hospital acute separations data used to calculate
278 PUBLIC AND PRIVATE
HOSPITALS
Table D.1 NHCDC sample by jurisdiction and region, 2007-08a, b
Tas, NT
NSW Vic Qld SA WA & ACT b Australia
Public hospitals
Major City
No. of hospitals 41 27 14 8 11 2 103
No. of separations 960 597 882 312 412 617 244 162 298 823 76 462 2 874 973
Inner Regional
No. of hospitals 30 17 10 10 2 4 73
No. of separations 284 773 192 251 164 273 21 797 15 671 79 147 757 912
Outer Regional
No. of hospitals 11 5 6 17 5 8 52
No. of separations 45 734 25 022 122 029 42 299 36 840 61 665 333 589
Remote
No. of hospitals – – 1 3 2 3 9
No. of separations – – 1 118 6 163 10 426 38 594 56 301
Very Remote
No. of hospitals – – – 1 – 3 4
No. of separations – – – 1 640 – 8 172 9 812
Total
No. of hospitals 82 49 31 39 20 20 241
No. of separations 1 291 104 1 099 585 700 037 316 061 361 760 264 040 4 032 587
Private hospitals
Major City
No. of hospitals 22 24 12 8 7 3 76
No. of separations 337 391 388 412 268 514 119 880 178 887 35 280 1 328 364
Inner Regional
No. of hospitals 9 5 13 – 1 2 30
No. of separations 51 169 34 265 104 423 – 18 269 32 669 240 795
Outer Regional
No. of hospitals – – 2 – – 1 3
No. of separations – – 32 927 – – 5 592 38 519
Remote
No. of hospitals – – – – – – –
No. of separations – – – – – – –
Very Remote
No. of hospitals – – – – – – –
No. of separations – – – – – – –
Total
No. of hospitals 31 29 27 8 8 6 109
No. of separations 388 560 422 677 405 864 119 880 197 156 73 541 1 607 678
a Regions are based on ABS Australian Standard Geographical Classification, Cat. no. 1216.0 b Separations
are not casemix adjusted. c Data for Tasmania, the Northern Territory and the ACT are aggregated to protect
the confidentiality of the small number of hospitals in each of these jurisdictions. – Nil.
Source: DOHA (unpublished data).

NHCDC coverage is extracted from the Private Hospital Data Bureau collection
(DOHA 2009a).
CONSTRUCTING COST 279
ESTIMATES
Table D.2 NHCDC sample by jurisdiction and hospital size, 2007-08a
Tas, NT
NSW Vic Qld SA WA & ACT b Australia
Public hospitals
Very large
No. of hospitals 21 18 12 5 5 4 65
No. of separations 804 241 781 645 493 750 216 403 236 970 180 549 2 713 558
Large
No. of hospitals 18 11 10 1 3 3 46
No. of separations 261 183 212 144 178 845 10 904 49 246 65 119 777 441
Medium
No. of hospitals 15 6 2 4 6 – 33
No. of separations 130 778 56 962 13 108 35 352 59 553 – 295 753
Small
No. of hospitals 22 5 2 10 3 1 43
No. of separations 82 647 37 253 7 740 34 965 12 886 6 035 181 526
Very small
No. of hospitals 6 9 5 19 3 12 54
No. of separations 12 255 11 581 6 594 18 437 3 105 12 337 64 309
Total
No. of hospitals 82 49 31 39 20 20 241
No. of separations 1 291 104 1 099 585 700 037 316 061 361 760 264 040 4 032 587
Private hospitals
Very large
No. of hospitals 8 6 8 3 4 – 29
No. of separations 213 022 184 939 233 570 73 199 159 193 – 863 923
Large
No. of hospitals 5 7 5 2 1 3 23
No. of separations 62 148 110 889 78 008 25 723 18 269 47 774 342 811
Medium
No. of hospitals 9 11 6 1 1 2 30
No. of separations 71 495 99 731 57 841 11 804 12 090 20 175 273 136
Small
No. of hospitals 8 4 7 2 2 1 24
No. of separations 39 750 24 044 35 853 9 154 7 604 5 592 121 997
Very small
No. of hospitals 1 1 1 – – – 3
No. of separations 2 145 3 074 592 – – – 5 811
Total
No. of hospitals 31 29 27 8 8 6 109
No. of separations 388 560 422 677 405 864 119 880 197 156 73 541 1 607 678
a Hospital size defined by annual casemix-adjusted separations as follows: very large (more than 20 001),
large (10 001 to 20 001), medium (5001 to 10 000), small (2001 to 5000), and very small (up to 2000).
Casemix adjustment for the purpose of allocating hospitals to a size group was undertaken by DOHA using
separate cost weights for public and private hospitals. The number of separations in the table above are not
casemix adjusted and so may not correspond to hospital category size, which is based on casemix-adjusted
separations. b Data for Tasmania, the Northern Territory and the ACT are aggregated to protect the
confidentiality of the small number of hospitals in each of these jurisdictions. – Nil.
Source: DOHA (unpublished data).

280 PUBLIC AND PRIVATE

HOSPITALS
Table D.3 NHCDC sample by region and hospital size, 2007-08a
Inner Outer Very
Major city regional regional Remote Remote Australia
Public hospitals
Very large
No. of hospitals 52 10 3 – – 65
No. of separations 2 285 222 294 971 133 365 – – 2 713 558
Large
No. of hospitals 27 16 2 1 – 46
No. of separations 435 041 269 478 40 880 32 042 – 777 441
Medium
No. of hospitals 10 13 10 – – 33
No. of separations 96 749 106 318 92 686 – – 295 753
Small
No. of hospitals 8 19 12 4 – 43
No. of separations 48 991 70 554 41 141 20 840 – 181 526
Very small
No. of hospitals 6 15 25 4 4 54
No. of separations 8 970 16 591 25 517 3 419 9 812 64 309
Total
No. of hospitals 103 73 52 9 4 241
No. of separations 2 874 973 757 912 333 589 56 301 9 812 4 032 587
Private hospitals
Very large
No. of hospitals 29 – – – – 29
No. of separations 863 923 – – – – 863 923
Large
No. of hospitals 14 8 1 – – 23
No. of separations 207 621 113 309 21 881 – – 342 811
Medium
No. of hospitals 20 9 1 – – 30
No. of separations 188 105 73 985 11 046 – – 273 136
Small
No. of hospitals 13 10 1 – – 24
No. of separations 68 715 47 690 5 592 – – 121 997
Very small
No. of hospitals – 3 – – – 3
No. of separations – 5 811 – – – 5 811
Total
No. of hospitals 76 30 3 – – 109
No. of separations 1 328 364 240 795 38 519 – – 1 607 678
a Regions are classified according to the Australian Standard Geographical Classification (detailed in
ABS 2005). Hospital size defined by annual casemix-adjusted separations as follows: very large (more than
20 001), large (10 001 to 20 001), medium (5001 to 10 000), small (2001 to 5000), and very small (up to
2000). Casemix adjustment for the purpose of allocating hospitals to a size group was undertaken by DOHA.
b The number of separations in the table are not casemix-adjusted and may not correspond to hospital
category size, which is based on casemix-adjusted separations. – Nil.
Source: DOHA (unpublished data).

CONSTRUCTING COST 281

ESTIMATES
The NHCDC data provided to the Commission differ from those used in the
Round 12 Cost Report (DOHA 2009a). In particular, the unpublished NHCDC data
provided to the Commission by DOHA are unweighted, and so may not necessarily
be representative of the broader hospital sector. In addition, some separations were
excluded from the cost analysis due to small sample sizes. DRGs were excluded if
there were fewer than 30 separations at a national level, or fewer than five
separations in a particular jurisdiction, or if separations occurred in fewer than three
hospitals.

The Commission also excluded a number of DRGs from the cost analysis, on the
advice of study participants. DRGs relating to mental diseases and disorders or drug
and alcohol use (those beginning with either a ‘U’ or a ‘V’) were not included
because of ‘the combination of lack of robust classification systems and very
different models of paying for care in different jurisdictions’ (Australian Health
Service Alliance, sub. 1, p. 3). The Commission also took account of advice from
the Australian Health Service Alliance (sub. 1) that the rehabilitation DRGs be
excluded, due to the potential for heterogeneity. Similarly, the ‘error DRGs’ (those
with the AR-DRG prefix ‘9’) were not included due to their (unknown)
heterogeneity.

Data not included in the NHCDC

While the NHCDC is the most useful source of hospital cost information, there are a
number of cost areas for which it does not provide information.

Most notably, the NHCDC does not include costs for a large proportion of medical
and diagnostics expenditure in private hospitals because these items are often billed
directly to patients. Information regarding medical costs for private patients in
public hospitals is also not included for this reason. The Commission has to some
extent been able to adjust the data for this lack of information by including medical
and imaging costs from the Hospital Casemix Protocol (HCP) dataset. This is
discussed further in section D.2.

Information about capital costs is also deficient in the NHCDC. There are no data
on the user cost of capital, and costing practices unique to Victoria mean that there
are no depreciation costs for public hospitals in that jurisdiction. This is discussed
further in section D.6.

Costs associated with blood products are also not included in the NHCDC.
Similarly, teaching costs are not specifically identified, implying that either these
costs are not included, or they are incorporated into other categories.

282 PUBLIC AND PRIVATE

HOSPITALS
Inconsistencies within the NHCDC

There are a number of differences both within and between the private and public
sectors that are likely to impact on the cost estimates (Tasmanian Department of
Health and Human Services, sub. 37). Differences in reported costs result from
factors such as different reporting practices and obligations, and admission
practices.

Publicly available documentation on different reporting practices is limited. For

example, there is considerable ambiguity regarding how administrative overhead
costs are treated, and the extent to which they are included in the NHCDC. It is
similarly unclear how teaching, training and research costs are treated.3

Another key difference in the reporting of costs between public and private sector
hospitals in the NHCDC is the predominance of ‘cost modelling’ to produce cost
estimates in the private sector, in comparison to ‘patient costing’ for the majority of
public hospitals.

Cost modelling involves allocating aggregate costs to individual separations through

the use of national utilisation averages, or service weights. Service weights are
derived from individual studies designed specifically for the resource area to which
they relate, and reflect the cost of intermediate resources used in each respective
DRG. More than 90 per cent of private hospitals providing data to the NHCDC
report cost-modelled data (DOHA 2008c).

One potential issue with cost modelling is that it can lead to a ‘systemic under-
costing of high-cost activity and over-costing of low-cost activity’, due to the
potential for averaging of costs within hospitals and within DRGs (Tasmanian
Department of Health and Human Services, sub. 37, pp. 7–8).

In contrast, patient costing involves attributing costs directly to patients as they

occur, often through the use of automated clinical information systems. DOHA
(sub. 32) noted that approximately 75 per cent of public-sector cost data are patient-
costed. It is worth noting that some patient-costed sites do use service weights to
allocate costs for resources in various disciplines.

As well as differences in the way data are collected and reported, differing
admission practices and access to hospitals lead to variation in the average costs
reported across jurisdictions. For example, in public hospitals in New South Wales,

3 The Commission understands that 132 of the 241 hospitals that submitted data to Round 12 of
the NHCDC separately identify teaching, training and research costs and exclude this from the
analysis. Western Australian hospitals are among those that do not separately identify these
costs (Department of Health, Government of Western Australia, sub. DR72).
CONSTRUCTING COST 283
ESTIMATES
South Australia and the ACT, there has been a shift over recent years from
admitting chemotherapy patients to treating them as non-admitted patients
(AIHW 2009a). Furthermore, states and territories may differ in the extent to which
certain types of services are provided in non-hospital settings, such as community
health centres.

NHCDC costs relating to public hospitals in different jurisdictions are to some

extent based on different costing standards due to different reporting requirements.
For example, Victorian public hospital costs are compliant with the Clinical Costing
Standards Association of Australia and are subsequently mapped to the NHCDC
cost structure (DOHA 2009a). This results in differences in the areas of nursing,
medical, on-costs and ward supplies cost buckets, as outlined in table D.4.
Similarly, the NSW Government has noted that its intensive-care unit and
emergency-department funding models may increase costs in those areas, while
decreasing costs attributed to diagnostics and imaging (DOHA 2009a).

284 PUBLIC AND PRIVATE

HOSPITALS
Table D.4 Differences between Victoria and other jurisdictions for NHCDC
cost buckets
NHCDC cost bucket Description of Victorian data
Ward medical Includes surgical and non-surgical medical costs.
Ward nursing In addition to nursing salaries and wages, includes the direct
costs of running a ward (such as consumables and transport) and
overhead costs (such as power and light, catering and cleaning).
Includes 'non-ward' costs for admitted patients including Hospital
in the Home, transit lounge, maternity and post-domiciliary
nursing care.
Pathology Similar across jurisdictions, including departmental salaries and
wages, consumables and overhead costs.
Imaging Similar across jurisdictions, including departmental salaries and
wages, consumables and overhead costs.
Allied health Includes similar costs to other jurisdictions (predominantly
departmental salaries and wages for allied health staff).
Pharmacy Similar across jurisdictions, including departmental salaries and
wages, drugs, consumables and overhead costs.
Critical care Similar across jurisdictions, including departmental medical and
nursing salaries and wages, consumables and overhead costs for
intensive-care units, neonatal intensive-care units, special-care
nurseries, coronary-care units and high-dependency units.
Operating room Similar across jurisdictions, including departmental salaries and
wages, consumables (including anaesthetic drugs) and overhead
costs.
Emergency departments Similar across jurisdictions, including departmental salaries and
wages, consumables and overhead costs.
Ward supplies and other Medical supplies are rolled into the clinical area where the
overheads expense is incurred, such as ward nursing or a clinical unit.
Supplies are not restricted to direct departments.
Specialist procedure suites Some specialist suites costs, such as catheterisation laboratory
or bone marrow procedure rooms may be reported in the ward
medical bucket.
Prostheses Reported as prosthesis if identifiable by the health service. May
be allocated to operating room bucket if not distinguished from
operating room costs.
On-costs Included directly in the department costs such as a ward or
imaging department.
Hotel services Generally distributed to other buckets directly based on activity
type, as indirect costs in Victorian cost data.
Depreciation Capital costs are not included in Victorian cost data.
Source: Victorian Department of Health (unpublished).

D.2 Hospital Casemix Protocol

Hospital Casemix Protocol (HCP) data are collected as part of the regulation of
private health insurance, and are used as the source of private medical, imaging and
diagnostic costs for this study. The HCP has clinical, demographic, benefit and
charge data for privately-insured admitted-patient episodes nationally from
CONSTRUCTING COST 285
ESTIMATES
1996-97. The HCP for 2007-08 covered 592 public hospitals and 299 private
hospitals (table D.5).

It is important to note that the HCP differs from the NHCDC in that it contains
amounts charged to patients and benefits paid by insurers, rather than hospital
expenditure (costs).

The collection of HCP data is a two-step process involving the provision of patient
information from hospitals to health insurers and then from health insurers to
DOHA. In the first step, hospitals are required to provide information to health
insurers within six weeks of the insured person being discharged from hospital. In
the second step, health insurers are required to provide data to DOHA within twelve
weeks of the insured person being discharged from hospital.

Table D.5 Hospital Casemix Protocol descriptive statistics, 2007-08

Units Public hospitals Private hospitals
Separations No. 299 122 1 874 341
Hospitals No. 592 299
Medical item charges $m 190 2 340
Source: DOHA (unpublished data).

Limitations of the HCP collection

The HCP is considered to be representative of all separations for which private

health insurance is claimed. This is in contrast to the NHCDC data, which are from
a voluntary sample of hospitals. HCP data relating only to those hospitals which had
submitted NHCDC data was requested by the Commission, however, the data
supplied to the Commission by DOHA were for the full HCP collection. This means
that the cost estimates presented in this study involve combining average costs of
DRGs from different populations.

Patients who did not make a private health insurance claim, including Department
of Veterans’ Affairs patients, are excluded from the HCP data. In 2007-08, these
patients accounted for around 90 per cent of separations in public hospitals (most of
whom are public patients) and 20 per cent of separations in private hospitals
(AIHW 2009a). This means that any private costs associated with the provision of
medical and diagnostic services for these patients are not included in the cost
estimates.

A major deficiency of the HCP is that public hospitals often fail to allocate
separations to individual DRGs for their private patients. In 2007-08, around
80 per cent of separations for private patients in public hospitals were classified as
286 PUBLIC AND PRIVATE
HOSPITALS
‘ungroupable’ in the HCP. In contrast, the HCP data relating to private hospitals is
of a much higher quality, with only around 1 per cent of separations classified as
ungroupable in 2007-08.

Medical and diagnostics costs would be understated in the Commission’s

experimental cost estimates, particularly for public hospitals, if the ungroupable
HCP separations were not assigned to individual DRGs. The Commission
apportioned the ungroupable HCP medical and diagnostic costs across DRGs using
the methods outlined in box D.2.

Box D.2 Allocation of ungroupable HCP medical and diagnostics costs

across DRGs
A large proportion of public hospital separations in the Hospital Casemix Protocol
dataset are not assigned to specific DRGs but rather are classified as ‘ungroupable’.
This means that, without adjustments, the medical and diagnostics costs associated
with these separations would not be included in cost estimates. ‘Ungroupable’ medical
and diagnostics costs have been incorporated into cost estimates by allocating them
across all DRGs in each sector, and for each jurisdiction.
Total ungroupable costs were first scaled to reflect that the NHCDC is only a sample of
public hospital separations. Ungroupable costs were then allocated across DRGs on
the basis of weighted separations. That is, each DRG was allocated the proportion of
ungroupable costs corresponding to the relevant number of weighted separations. The
number of private separations in public hospitals for each DRG was obtained from the
National Hospital Morbidity Database. Because of the low quality of the public-hospital
HCP data, average private medical charges from private hospital patients were then
used to weight these separations.

The effect of incorporating these costs varied according to the number of

ungroupable separations in the public-hospital HCP data, relative to the total
number of separations in the NHCDC data, and the total amount of charges
associated with ungroupable DRGs. At a national level, inclusion of the
ungroupable HCP costs increased the Commission’s estimate of medical and
diagnostics costs in public hospitals by $34 per casemix-adjusted separation
(table D.6). Costs increased by $12 per casemix-adjusted separation in private
hospitals.

CONSTRUCTING COST 287

ESTIMATES
Table D.6 Ungroupable separations for private patients by sector,
2007-08a
Tas,
NT &
Units NSW Vic Qld SA WA ACTb Aust.

Public hospitals
Ungroupable HCP separations ‘000 141 49 13 9 8 8 232
Total HCP separations ‘000 144 86 14 25 9 11 299
Per cent ungroupable % 97.9 56.3 92.3 37.2 96.6 74.9 77.6
Ungroupable HCP medical
charges $m 102 27 6 5 5 3 150
Total HCP medical charges $m 104 51 6 11 5 6 190
Per cent ungroupable % 98.5 53.6 93.4 42.0 97.4 61.4 78.8
Private hospitals
Ungroupable HCP separations ‘000 6 9 7 3 6 1 32
Total HCP separations ‘000 474 473 397 157 197 74 1 874
Per cent ungroupable % 1.3 1.9 1.8 1.7 2.8 1.4 1.7
Ungroupable HCP medical
charges $m 6 6 8 3 5 1 28
Total HCP medical charges $m 651 582 539 194 243 91 2 340
Per cent ungroupable % 1.0 1.0 1.4 1.3 1.9 1.1 1.2
a Ungroupable separations are those assigned the AR-DRG code 960Z. b Data for Tasmania, the Northern
Territory and the ACT are aggregated to protect the confidentiality of the small number of hospitals in each of
these jurisdictions.
Source: DOHA (unpublished data); Productivity Commission estimates.

D.3 Cost components

A measure of total cost was created by combining the different costs associated with
an episode of care. As discussed in section D.1, cost components within the
NHCDC vary in terms of consistency and comparability, both across and within
sectors. Costs are accordingly presented in five broad groupings, as set out in
table D.7, in order to aid comparability.

288 PUBLIC AND PRIVATE

HOSPITALS
Table D.7 Components of total cost estimatesa
Component Cost bucket Description

General Ward nursing Nursing salaries and wages in general ward areas. Ward
hospital nursing costs may also be found in other buckets that have a
medical salary and wages component, such as critical care,
operating rooms, specialist procedures suites, emergency
departments, imaging, pathology, allied health and pharmacy.
Non-clinical This cost bucket includes all other costs of service provision for
salaries each inpatient separation during the collection period. These
costs are primarily other salaries and wages such as
patient-care assistants.
Allied health Costs of clinical services which are delivered by allied health
professionals who have direct patient contact in areas such as
audiology, physiotherapy, podiatry and dietetics.
Critical care Covers costs incurred in both intensive care and coronary care
units.
Operating room Costs attributed to the area of a hospital where significant
surgical procedures are carried out under surgical conditions,
under the supervision of qualified medical practitioners.
Specialist Costs incurred in areas where diagnostic and therapeutic
procedure suites procedures are performed under the direction of suitably
qualified medical practitioners.
Ward supplies & Costs for goods and services, medical and surgical supplies,
other overheads ward overheads and clinical department overheads.
On-costs Includes cost items such as superannuation, termination
payments, workers compensation and long service leave.
Hotel services Includes food service, linen and grocery supplies.
Pharmacy Pharmacy The cost of providing a pharmacy. This includes the purchase,
production, distribution, supply and storage of drugs and
clinical pharmacy services. Pharmacy costs reported in critical
care, operating rooms, specialist procedures suites, emergency
departments, pathology, and imaging are not included in this
bucket.
Emergency Emergency Area of the hospital where patients who present in an
departments unscheduled manner can be triaged, assessed and treated.
These costs relate to emergency patients who are
subsequently admitted.
Prostheses Prostheses Prostheses appearing on hospital accounts and costs incurred
by the hospital. Prostheses acquired by patients or their
doctors directly (rather than by the hospital) will not show up on
hospital accounts and are not reported.
Capital Depreciation The cost of depreciation for items that are durable, that can
support production for an appreciable period of time and are
purchased outright or donated. Depreciation costs are sourced
from the NHCDC, with the exception of public hospitals in
Victoria and Queensland which were derived from data
published in SCRGSP (2009).
User cost of Estimates of the opportunity cost of funds tied up in the capital
capital used to deliver services. Derived from data published by the
ABS (2008e) and SCRGSP (2009) (see section D.6).

(Continued next page)

CONSTRUCTING COST 289

ESTIMATES
Table D.7 (continued)

Component Cost bucket Description

Medical & Ward medical Salaries and wages of all medical officers (incl. sessional
diagnostics payments). Medical costs may also be found in other buckets
that have a medical salary and wages component, such as
critical care, operating rooms, specialist procedures suites,
emergency departments, imaging, pathology, allied health and
pharmacy.
Imaging Costs of diagnostic and therapeutic imaging. Excludes imaging
costs reported in critical care, operating rooms, emergency
departments, specialist procedures suites, pharmacy, and
pathology.
Pathology Costs of diagnostic clinical laboratory testing for the diagnosis
and treatment of patients. Excludes pathology costs reported in
critical care, operating rooms, emergency departments,
specialist procedures suites, pharmacy, and imaging.
Medical charges Total charge for medical and diagnostic items as presented in
medical records associated with the episode of care. This
component includes medical charges that are billed directly to
the patient, and are sourced from HCP data.
a Cost buckets are cost categories incurred by the hospital and are drawn from the NHCDC, with the
exception of the medical charges category, which is drawn from the HCP.
Source: DOHA (2008b, 2008c).

The first component — labelled ‘general hospital’ — comprises general cost items
that are often under the control of a hospital.

Emergency departments and pharmacy costs are not included with ‘general
hospital’ items because of significant differences between public and private
sectors. Emergency departments are predominantly in the public sector, and
typically involve significant fixed costs.

Pharmacy costs for private hospitals are likely to be significantly understated in the
NHCDC as they are subsidised by the Australian Government under the
Pharmaceutical Benefits Scheme (PBS) (Dr. John Deeble, sub. DR56; NSW
Department of Health, sub. 41; DOHA, sub. 32).

The Australian Institute of Health and Welfare (AIHW 2009d) recently estimated
the expected private hospital cost of pharmaceuticals. In 2005-06, pharmaceutical
costs accounted for around 3.7 per cent of private hospital expenditure. If private
hospitals had faced the same pharmaceutical costs as public hospitals, taking into
account differences in casemix, pharmaceuticals would have accounted for 6.4 per
cent of private hospital expenditure. This suggests that private hospitals have either
substantially lower pharmaceutical costs, or up to 40 per cent of the pharmaceutical
costs for patients in private hospitals are met by external arrangements, such as the
PBS (AIHW 2009d).

290 PUBLIC AND PRIVATE

HOSPITALS
Pharmacy costs for public hospitals are also likely to be understated. For
jurisdictions other than Victoria, the NHCDC cost bucket for pharmacy does not
include the cost of pharmaceuticals used in critical care, operating rooms,
emergency departments, pathology, imaging, and specialist procedures suites. The
extent to which pharmacy costs are included in other cost buckets varies between
jurisdictions, due to differences in reporting practices. For those jurisdictions that
submit sufficiently detailed data (New South Wales, Queensland, South Australia,
Tasmania and the Northern Territory), it is estimated that around 76 per cent of all
pharmaceutical costs are captured in the pharmacy cost bucket, with 12 per cent
reported under operating rooms and 8 per cent under critical care (table D.8).

Prostheses are presented separately due to the different ways in which the costs are
realised in both sectors (Australian Health Services Alliance, sub. 1; Catholic
Health Australia, sub. 20). Prostheses in the public sector are typically purchased
from relatively restricted lists at comparatively low costs, due to the presence of
bulk purchasing arrangements.

In the private sector, most prostheses are purchased by the hospital and supplied to
the patient by the hospital, although the choice of prosthesis is made by the treating
doctors. Benefits for prostheses are payable to hospitals by private health insurers
on the basis of amounts determined by the Minister for Health and Ageing, as
presented in the Prostheses List (Catholic Health Australia, sub. DR62). Where
there is a gap between the benefit paid by the fund and the prosthesis charge, this is
typically paid by the patient to the hospital, and so is included in the NHCDC.
Study participants indicated that private sector arrangements generally involve the
use of a wider range of products, often at a noticeably greater cost (section D.8).

The experimental nature of the capital cost estimates necessitates that they be
presented separately. The estimation of capital costs, as required by the terms of
reference for the study, has been particularly challenging because of significant data
constraints. Details about these estimates are presented in section D.6.

The medical and diagnostics component contains medical, imaging and pathology
costs from the NHCDC, and medical charges from the HCP. As the HCP medical
charge contains both medical and diagnostic costs, it is appropriate to group them
all together in the interests of comparability.

The NHCDC cost bucket for ward medical excludes medical salaries and wages
reported in imaging, pathology, critical care, operating rooms, emergency
departments, specialist procedures suites, allied health, and pharmacy. This means
that public-patient medical costs will be understated in the Commission’s estimates
to the extent that the NHCDC includes medical costs in the general hospital cost

CONSTRUCTING COST 291

ESTIMATES
buckets (including critical care, operating rooms and specialist procedures suites),
emergency departments and pharmacy costs.

Data for New South Wales, Queensland, South Australia, Tasmania and the
Northern Territory suggest that around two-thirds of NHCDC medical costs are
captured in the ward medical, pathology and imaging cost buckets, with the
remaining medical costs being recorded in operating rooms, critical care and
emergency departments (table D.8). National medical and diagnostics costs per
casemix-adjusted separation would rise from $798 to $1065 if one-third of medical
costs were recorded in operating rooms, critical care and emergency departments,
and these were reallocated to the medical and diagnostics component (table D.9).
Under this scenario, medical and diagnostics costs per casemix-adjusted separation
for patients in public hospitals would still be $281 less than medical and diagnostic
costs experienced by patients in private hospitals — a difference of around 21 per
cent.

Table D.8 Distribution of NHCDC pharmacy and medical costs for

selected patient-costed public hospitals, by cost bucket,
2007-08a
NHCDC cost bucket Pharmacy costs Medical salaries and wages
Per cent Per cent
Ward medical – 61.2
Ward nursing – –
Non-clinical salaries 0.1 –
Pathology 0.2 0.9
Imaging 0.5 2.6
Allied health – 0.1
Pharmacyb 76.4 0.1
Critical care 8.2 8.2
Operating rooms 12.1 19.4
Emergency departments 2.0 7.4
Supplies and ward overheads – –
Specialised procedure suites 0.5 0.2
Prostheses 0.1 –
Staff on-costs – –
Hotel – –
Depreciation – –
Total cost 100.0 100.0
a Includes patient-costed data from public hospitals in NSW , NT, QLD, NT, SA and TAS. Disaggregation of
medical and pharmacy costs for other jurisdictions was not able to be obtained as they do not provide data at
a sufficient level of disaggregation, or do not submit patient-costed data b Victorian pharmacy costs are only
included in the pharmacy bucket and not in other buckets. – Nil or rounded to zero.
Source: DOHA (unpublished).

292 PUBLIC AND PRIVATE

HOSPITALS
Table D.9 Public hospital medical salaries and wages included in other
NHCDC cost buckets, 2007-08
Costs per casemix-adjusted separation NSW, QLD, SA, Tasmania, NT Australiab
and ACTa

Diagnosticsc 268 270

Ward medical 408 490
HCP medical and diagnostic charges 55 37
Medical salaries and wages included in General
Hospitald 176 211
Medical salaries and wages included in
Emergencyd 47 56
Total medical and diagnostics costs 953 1065
a Percentage of medical costs included in general hospital and emergency for QLD, NSW, TAS, NT and SA
(as presented in table D.8) is assumed to apply to the ACT. b Estimates are based on the assumption that the
allocation of medical salaries and wages to other NHCDC cost buckets, as presented in table D.8, is
consistent across all jurisdictions. c Diagnostics costs are NHCDC cost buckets for pathology and imaging. d
Amounts calculated using proportions from table D.8. Components may not sum to total due to rounding.
Source: Productivity Commission estimates.

D.4 Cost indicators

Two commonly-used measures of hospital costs were estimated for this study:
• cost per casemix-adjusted separation — the average cost of treating a range of
different diagnoses, after taking into account differences in the complexity of
required treatments (casemix adjustment)
• cost per separation — the average cost of treating a group of patients with
clinically-similar diagnoses.

Clinically-similar diagnoses were defined according to the widely-accepted system

of AR-DRGs (box D.3). This classification system provides a clinically-meaningful
way of relating types of patients treated to required resources (DOHA 2004).
Individual DRGs represent a class of patients with similar clinical conditions who
require similar hospital services (AIHW 2009a; DOHA, sub. 32).

Some participants were concerned that individual DRGs are not sufficiently
homogeneous to enable like-for-like comparisons (for example, Queensland Health,
sub. 27; Tasmanian Department of Health and Human Services, sub. 37; Women’s
and Children’s Hospitals Australasia, sub. 21). It is inevitable that any patient
classification system will have some heterogeneity within individual categories, as
no single patient is identical to another, and so the question is whether such
heterogeneity is significant and likely to prejudice any cost comparison.

CONSTRUCTING COST 293

ESTIMATES
 Box D.3 Classifying episodes of care — Diagnosis-Related Groups
The Diagnosis-Related Group (DRG) system is a taxonomy of hospital outputs that is
used to document, apportion and control costs in hospitals. The purpose of the DRGs
is to relate the mix of patients treated, or casemix, to the resource demands and
associated costs experienced by a hospital. Separations are categorised on the basis
of three main principles:
• clinical meaning — diagnoses within each DRG are to be clinically similar
• resource homogeneity — treatment of diagnoses within each DRG should utilise a
similar type and amount of resources
• exclusivity — diagnoses should only correspond to a single DRG.
The DRG system currently used in Australia reflects local clinical practice and is
referred to as Australian Refined Diagnosis-Related Groups (AR-DRGs), with the latest
version (6.0) released in 2008. The NHCDC and HCP data presented in this study use
AR-DRG version 5.1.
A DRG system groups episodes of patient care into categories differentiated by factors
such as main diagnosis, clinical procedures, gender, age, and the presence of
additional diagnoses or complications. At the highest level, episodes of care are
classified into a Major Diagnostic Category (MDC). Diagnoses in each MDC
correspond to a single body system or cause of disease, broadly reflecting the
specialty providing care. All possible principal diagnoses in the AR-DRG classification
system fall into one of 23 mutually exclusive MDCs, and into one of 665 AR-DRGs.
Source: Bridges, Haas and Mazevska (1999); DOHA (2004); Erlandsen (2008).

The Commission notes that factors such as patient age, severity of conditions, and
the presence of comorbidities, are included in the AR-DRG system, and so are, to
some extent, controlled for. The AR-DRG system has been refined over a period of
more than a decade with input from national, state and territory health departments
so that only patients with similar clinical conditions and resource requirements are
grouped into the same DRG (DOHA 2004).

The AR-DRG system only applies to admitted patients, and so it was not possible to
compare costs for other hospital services. Admitted-patient services accounted for
71 per cent of the costs incurred by overnight acute hospitals in 2007-08
(AIHW 2009a).4

4 Victoria admits patients for treatments that other jurisdictions may administer as non-admitted
(outpatient) services, such as chemotherapy and dialysis, and so Victoria may account for a
disproportionate share of national costs for admitted-patient services.
294 PUBLIC AND PRIVATE
HOSPITALS
The grouping of similar outputs by DRG, and casemix adjustment when comparing
costs for more than one DRG, is an important step in making cost comparisons
more meaningful. The details of casemix adjustment are outlined in box D.4.

Box D.4 Cost per casemix-adjusted separation

Casemix adjustment involves weighting the separations for each DRG by its relative
complexity, and is often used to improve comparisons between different hospitals.
The complexity of an episode of care in the context of hospital costing refers to the
expected resources that are to be used in treatment. The complexity of a DRG is
measured by its relative cost weight — the average cost of that DRG across all
relevant hospitals divided by the average cost of all DRGs.
The cost per casemix-adjusted separation is given as:
Total exp enditure
∑i
Separation s i × Re lative cos t weight i

where there are i number of DRGs.

The denominator in this expression is the number of casemix-adjusted separations and
is used to adjust the number of separations for their relative complexity in calculating a
per unit cost. Casemix adjustment can be performed at different levels of aggregation,
such as by jurisdiction, region or size.
Source: AIHW (2009a).

D.5 Tax exemptions

Public and not-for-profit private hospitals are partially exempt from paying
fringe-benefits tax (FBT) and are not required to pay payroll tax. Private not-for-
profit hospitals are also entitled to income tax exemptions (including capital gains
tax), and goods and services tax, stamp duty and land tax concessions. As
‘deductible gift recipients’ they are entitled to receive income tax deductible gifts
and tax deductible contributions (KPMG 2009). These concessions can assist public
and not-for-profit private hospitals in recruiting and retaining staff
(Treasury 2008a). The terms of reference for this study require the Commission to
take account of FBT exemptions when comparing costs.

The FBT and payroll-tax concessions mean that the cost of offering a given level of
post-tax remuneration is likely to be greater for a for-profit hospital, than for a
public or not-for-profit private hospital. That is, the concessions confer a cost

CONSTRUCTING COST 295

ESTIMATES
advantage on public and not-for-profit private hospitals by effectively subsidising
their labour costs.

As the cost of labour faced by public and not-for-profit hospitals is reduced by the
tax concessions, this is likely to distort resource allocation. Reducing the price of
labour relative to capital and other inputs for public and not-for-profit private
hospitals provides an incentive for them to be more labour intensive than for-profit
private hospitals that are not afforded these concessions. No adjustments for this
distortion have been made in this study.

The Commission has, however, sought to ensure that costs are compared on a
like-for-like basis by removing the additional tax burden that for-profit hospitals,
compared to public and not-for-profit private hospitals, incur due to not having
access to the FBT and payroll-tax exemptions.

Fringe-benefits tax exemption

The fringe-benefits tax (FBT) exists to ensure that remuneration from employers is
treated consistently, regardless of the form in which the income is received. It is
paid by employers at the top marginal tax rate plus the Medicare levy
(46.5 per cent).

The FBT exemption for public and not-for-profit hospitals can provide them with a
cost advantage that aids in recruiting and retaining staff (Treasury 2008a).
Individuals working for these hospitals are able to increase their post-tax
remuneration by taking some of their pay package as fringe benefits. The exemption
is capped at $17 000 per employee (ATO 2007). The cap prevents overuse,
constrains the impact of the concession on competitive neutrality, and limits the
foregone tax revenue to the Australian Government to $7905 per employee
(46.5 per cent of $17 000).

However, there are a number of items that are excluded from the $17 000 cap on the
FBT exemption for public and not-for-profit hospitals. These include meal
entertainment (such as a doctor’s expenses on a restaurant meal at a social
occasion), entertainment-facility leasing expenses and car parking. There is little
information on the use of these uncapped FBT exemptions, and so the Commission
has not been able to specifically adjust for them in its cost estimates.

The proportionate increase in post-tax remuneration that can be achieved by using

the capped FBT exemption will depend on a worker’s pre-tax salary. Based on the
average salaries of different occupations, the capped FBT exemption has the
potential to increase post-tax remuneration by a greater percentage for other

296 PUBLIC AND PRIVATE

HOSPITALS
personal care staff compared to salaried medical officers (figure D.1). However,
participants noted that the NSW Government has a policy of taking 50 per cent of
the tax savings that public hospital employees in NSW would otherwise enjoy by
taking part of their package as fringe benefits (for example, Mark Hanlon,
sub. DR46).

Figure D.1 Maximum effect of the capped fringe-benefits tax exemption on

post-tax remuneration, by occupationa

12.0

10.0

8.0
Per cent

6.0

4.0

2.0

0
Salaried Nurses Diagnostic and Administrative Domestic and Other personal
medical officers allied health and clerical staff other staff care staff

a Reduced tax as a percentage of average salaries of full-time equivalent staff in public acute and psychiatric
hospitals. It is assumed that employees minimise their tax liabilities and realise the exemption up to the
$17 000 cap.
Source: Productivity Commission estimates.

FBT payments are included in the NHCDC on-cost bucket (DOHA 2008c).
However, it is not possible to separately identify FBT in the NHCDC data. The
Commission therefore had to estimate the impact of the FBT exemption indirectly.

The capped FBT concession for public and not-for-profit private hospitals is
estimated to have cost the Australian Government $270 million in foregone revenue
in 2007-08 (Treasury 2008b). This was equivalent to around 1.4 per cent of the total
wage bill of public and private not-for-profit hospitals in 2007-08 (AIHW 2009a,
ABS 2008e).

As for-profit and not-for-profit private hospitals are not distinguished in the

NHCDC, estimating and removing the ‘excess’ FBT incurred by for-profit hospitals
required a number of adjustments. First, the amount by which the for-profit private
hospital wage bill was to be reduced was estimated. This amount was then
expressed as a proportion of the total wage bill of private hospitals (for-profit and
CONSTRUCTING COST 297
ESTIMATES
not-for-profit hospitals combined), as these hospitals are not identified separately in
either the NHCDC or the HCP datasets. The cost buckets that relate specifically to
wages and salaries were then reduced by this percentage.

In estimating the percentage by which the private for-profit wage bill needs to be
reduced, it was assumed that use of the capped FBT exemption is the same across
both public and private not-for-profit hospitals. It was also assumed that if for-profit
private hospitals had access to the capped FBT exemption, they would utilise it in
the same way as public and not-for-profit hospitals.

The estimated total cost of the capped FBT exemption ($270 million in 2007-08)
was first apportioned between public and private not-for-profit hospitals, according
to the relative size of their total wage bills.5 On this basis, around $246 million, or
about 90 per cent, of the tax benefit from the capped FBT exemption, was estimated
to have gone to public hospitals, and around $24 million — around 1.4 per cent of
the total wage bill of private not-for-profit hospitals (ABS 2008e) — went to private
not-for-profit hospitals (table D.10).

Table D.10 Distribution of benefits from the capped FBT exemption by

sector
Hospital type Total wage Proportion of foregone
expenditurea FBT revenue

$m %
Private
For-profit 1 700 724 ..
Not-for-profit 1 701 072 9
Public 16 410 900 91
a Total wage expenditure figures are from 2006-07, as private wage expenditure figures for 2007-08 are not
currently available. .. Not applicable.
Source: ABS Labour Price Index, Australia, Cat. no. 6345.0; AIHW (2008b); Productivity Commission
estimates.

If private for-profit hospitals had utilised the capped FBT exemption to the same
extent as other hospitals, private for-profit hospitals would have received a tax
benefit in the order of $24 million.6 This amounts to around 0.7 per cent of the total
wage bill for all private hospitals, and is the factor by which private hospital labour
costs were reduced to take into account the differences in access to the capped FBT

5 Private hospital wage data were not available for 2007-08, so wage relativities for 2006-07 were
used to apportion the 2007-08 FBT cost across public and private not-for-profit hospitals.
6 This is around 1.4 per cent of the wage bill for private for-profit hospitals. This calculation
assumes that the employment behaviour of private for-profit hospitals would not have changed
with access to the capped FBT exemption.
298 PUBLIC AND PRIVATE
HOSPITALS
exemption. In particular, NHCDC cost buckets for ward medical, ward nursing and
non-clinical salaries were reduced by 0.7 per cent for private hospitals. This
averaging approach takes full account of the capped FBT disadvantage faced by the
private for-profit sector by apportioning it across the entire private hospital sector.

Payroll taxes

All states and territories levy a payroll tax on employers that have total wage and
salary payments exceeding specified tax-free thresholds. As noted above, public and
not-for-profit private hospitals are exempt from payroll tax.

The states and territories individually administer payroll taxes, so there are different
tax rates and thresholds across jurisdictions. Payroll-tax rates range from
4.7 per cent in Queensland to 6.8 per cent in the ACT. The wage-bill threshold at
which payroll taxes become payable range from a wage bill of $550 000 in Victoria
to $1.5 million in the ACT.

The Commission was advised that payroll taxes are supposed to be excluded from
the NHCDC data, and so the Commission did not adjust the data to reflect the
different payroll-tax regimes applying to for-profit private hospitals relative to
public and for-profit hospitals. It should, nevertheless, be noted that the impact of
payroll-tax exemptions on labour costs is not trivial.

The payroll tax concessions represent a significant cost that is often not explicitly
taken into account. As shown in figure D.2, the impact of payroll-tax exemptions on
public hospitals is also likely to vary markedly between jurisdictions, depending on
the tax rate applied. Across all jurisdictions, the concession is worth around
$970 million for public hospitals alone. This represents around 5.4 per cent of the
total wage and salary bill for public hospitals in Australia (AIHW 2009a).

CONSTRUCTING COST 299

ESTIMATES
Figure D.2 Estimated benefit to public hospitals of payroll-tax exemptions,
2007-08a

400

300
$ million

200

100

0
NSW Vic Qld SA WA Tas ACT NT

a Based on the number of full-time equivalent employees and total wage bill for each jurisdiction.

Source: ACT Revenue Office (2009a, 2009b); AIHW (2009a); State Revenue Office (Vic) (2009a, 2009b);
Office of State Revenue (Qld) (2009a, 2009b, 2009c); Office of State Revenue (NSW) (2007, 2008), Revenue
SA (2008, 2009), Office of State Revenue (WA) (2007); Department of Treasury and Finance (Tas) (2008a,
2008b); Territory Revenue Office (NT) (2008a, 2008b); Productivity Commission estimates.

D.6 Capital costs

The terms of reference require the Commission to take into account capital costs
when comparing costs for clinically-similar procedures performed by public and
private hospitals. Capital costs comprise two components:
• depreciation — the reduction in the value of an asset due to usage or
obsolescence
• the user cost of capital (UCC) — the opportunity cost of funds tied up in the
capital used to deliver services. That is, the return that could be generated if the
funds tied up in the capital used to provide hospital services were employed in
their next best use.

In most cases, depreciation is recorded at a DRG level in the NHCDC, and so

identifying this cost has been relatively straightforward. In contrast, comparing the
UCC on a like-for-like basis is difficult because it is not included in the NHCDC,
and other data sources are limited by inconsistent collection methods and missing
information. This reflects differences in the rationale for, and relative importance
attached to, public reporting of the amount of capital used in the public,
not-for-profit, and for-profit sectors.

300 PUBLIC AND PRIVATE

HOSPITALS
Where capital costs are not available in the NHCDC, the Commission has largely
drawn on the methodology and data that jurisdictions have for some years
contributed to for national reporting of public hospital costs under the auspices of
the Steering Committee for the Review of Government Service Provision
(SCRGSP). The SCRGSP (2009) — comprising representatives from the
Australian, state and territory governments — reported that, in 2006-07, capital
costs accounted for around 10 per cent of the average cost per casemix-adjusted
separation for inpatient services at major public acute hospitals.

The SCRGSP methodology for estimating the UCC of public hospitals involves
calculating the return foregone on the next best investment, estimated at a rate of
8 per cent of the value of assets (box D.5). To ensure like-for-like comparisons in
this study, the Commission has used the same approach when calculating the UCC
for private hospitals.

Box D.5 SCRGSP methodology for calculating public hospital capital

costs
The SCRGSP methodology for calculating capital costs for public hospitals is as
follows:
• Asset values for land, buildings and equipment, and depreciation data for buildings
and equipment are provided by state and territory governments
• The user cost of capital (UCC) for each asset class in each state is calculated by
multiplying the value of the jurisdiction’s assets by a UCC rate (8 per cent).
• The resulting capital cost (depreciation and UCC) for each asset class is then
divided by the number of casemix-adjusted separations and multiplied by an
‘admitted-patient cost proportion’ to obtain a capital cost per casemix-adjusted
separation for admitted patients. Asset values and depreciation data for Victoria and
Western Australia are only for admitted patients and thus the admitted-patient cost
proportion is one for these jurisdictions.
• The next step is to calculate a total capital cost (excluding land) per
casemix-adjusted separation. This is done by adding the capital cost per
casemix-adjusted separation for buildings and equipment, and subtracting interest
payments per separation. Land is excluded, as differences in property values
obscure the differences between how well hospitals are managed. Interest
payments represent a UCC, and so are subtracted to avoid double counting.
Source: SCRGSP (2009).

The estimation of the UCC is considerably more difficult for private hospitals
compared to public hospitals, as the asset values of private hospitals are not publicly
available and need to be estimated. The absence of this information presents a

CONSTRUCTING COST 301

ESTIMATES
considerable obstacle in estimating comparable and robust capital costs, particularly
in the private sector. As detailed below, the asset values used to estimate the UCC
for private hospitals were derived from investment and depreciation data collected
by the ABS (2008e). However, this data may not cover all capital investment in
private hospitals. As parties other than the private hospital operator invest in
hospital capital, it is likely that the value of leased assets are not fully covered in
asset estimates.

NSW Department of Health (sub. 41; sub. DR64) and Dr. John Deeble (sub. DR56)
favoured a different approach in which profits were used to measure the UCC for
private hospitals. Using profits to measure the UCC of private hospitals is likely to
be misleading because many private hospitals are run on a not-for-profit basis. As
noted by Catholic Health Australia, the large number of hospitals it represents are
motivated by benefits other than just profits:
Catholic hospitals also have a mission focus which is often reflected in providing a
wider range of treatments, such as palliative care, than might be the case than if the
hospital was purely focused on profit maximisation. It also means that some Catholic
hospitals are located in geographic regions which might not necessarily be attractive to
for-profit operators. (sub. 20, p. 2)

A further problem with the approach suggested by NSW Department of Health and
Dr. Deeble is that it confuses profits recorded for accounting purposes with the
economic concept of the UCC. Accounting profits measure the difference between
revenue and the amounts paid for inputs, rather than their opportunity costs. Two
companies could use identical amounts of capital — and hence have the same UCC
— but record very different profits for accounting purposes because of differences
in their use of debt and rented capital items.

Nevertheless, despite using a different methodology, NSW Department of Health

and appear to have reached a similar conclusion to that found by the Commission.
In particular, NSW Department of Health (sub. 41) estimated that the average
amount of capital used per bed in public hospitals is much higher than in private
hospitals ($388 000 versus $244 000 per bed). The Commission’s experimental
estimates also show that public hospitals have a higher capital cost per casemix-
adjusted separation than private hospitals (chapter 5).

Capital costs for public hospitals

Average depreciation costs are included in the NHCDC by DRG for all jurisdictions
except Victoria. However, the data for Queensland exclude building depreciation,
which accounts for the majority of depreciation in other jurisdictions.

302 PUBLIC AND PRIVATE

HOSPITALS
The depreciation values reported by the SCRGSP (2009) for Victorian and
Queensland public hospitals were used to approximate average depreciation by
DRG for these jurisdictions. To reflect that the NHCDC is a sample of hospitals, the
reported depreciation values for Victorian and Queensland hospitals were multiplied
by the percentage of public separations in the NHCDC for each jurisdiction. These
total depreciation amounts were then inflated to 2007-08 levels using a state,
territory and local government gross fixed capital formation index published by the
AIHW (2009c). Depreciation was apportioned across DRGs using a weighted
average of the other jurisdictions’ depreciation profiles, using weights were based
on the jurisdiction’s share of total separations.

It is unclear whether leasing and interest costs are included in the NHCDC cost
buckets for depreciation and/or ward supplies and other overheads. The NHCDC
Hospital Reference Manual states that costs associated with major leases are to be
grouped with corporate overhead costs and included in the ward supplies and other
overheads cost bucket (DOHA 2008c).7 The treatment of leasing and
interest-related costs is also likely to differ between sectors and jurisdictions.

The UCC for each jurisdiction was based on the 2006-07 admitted-patient UCC for
buildings and equipment (minus interest payments) published by the SCRGSP
(2009). To reflect the fact that the NHCDC only represents a sample of all hospital
episodes, the UCC figure for each jurisdiction was multiplied by the percentage of
that jurisdiction’s public-hospital separations that were included in the NHCDC.
These UCC figures were then inflated to 2007-08 values using a state, territory and
local government gross fixed capital formation index published by the
AIHW (2009c). To obtain an average UCC by DRG for each jurisdiction, the
estimated total UCC was allocated according to the proportion of a jurisdiction’s
public hospital depreciation attributed to each DRG.

The reported public hospital asset values on which the UCC is derived suggest that
Australian public hospitals had assets worth approximately $21.9 billion in
2007-08. NSW Department of Health (sub. 41) noted that this figure is consistent
with work carried out by Dr. Deeble for the governments of Victoria, Queensland
and South Australia over the last ten years. Nevertheless, NSW Department of
Health (sub. 41, p. 3) observed that ‘nobody knows exactly how much capital is
currently used by the public hospitals’. This is partly due to inconsistent accounting
practices regarding depreciation and the valuation of assets among governments.
This might explain why the public hospital assets that Victoria reports to the
SCRGSP seem to be an underestimate when compared to those of New South
Wales and Queensland (figure D.3).

7 These corporate costs are allocated across different DRGs on the basis of bed days.

CONSTRUCTING COST 303

ESTIMATES
The use of public-private partnership arrangements, and the contracting out of
public-hospital services to private operators, may lead to an understatement of
assets used to provide public-hospital services in some jurisdictions.

Figure D.3 Public hospital asset values (excluding land), 2007-08a

6
$ billion

0
NSW Vic Qld WA SA Tas, NT,
ACT

a Asset values provided to SCRGSP by Victoria and Western Australia only apply to admitted patients. These
asset values have been adjusted to apply to both admitted and non-admitted patients using the admitted-
patient cost proportion. All asset values have been inflated to 2007-08 levels using a state, territory and local
government gross fixed capital formation index published by the AIHW (2009c).
Source: SCRGSP (2009).

Data constraints prevented the calculation of the UCC estimates by region or

hospital size. For these disaggregations, the Australia-wide UCC for public
hospitals was apportioned across groupings by the number of casemix-adjusted
separations.

When disaggregating by hospital size or region, the estimates of average

depreciation for public hospitals by DRG do not include Victorian depreciation data
or Queensland building depreciation data — since neither are reported in the
NHCDC — and thus are understated.

Capital costs for private hospitals

Depreciation values for acute overnight private hospitals by DRG are included in
the NHCDC.

Asset values are currently not reported for acute overnight private hospitals, making
the calculation of the UCC difficult. The Commission estimated asset values for
private hospitals from investment and depreciation data collected by the

304 PUBLIC AND PRIVATE

HOSPITALS
ABS (2008e). The estimation method involved a perpetual inventory model similar
to that used by Webster et al. (1998) (box D.6). Using this method the Commission
estimated an asset value for acute overnight private hospitals, based on 2003-04 to
2006-07 data on capital expenditure and depreciation from the ABS (2008e).8 The
resulting regression estimates indicated that the average rate of depreciation for
private hospitals was approximately 6 per cent per annum.

Setting the value of the investment time horizon variable, H, between 15 and
20 years seems to be appropriate, because the resulting annual amounts of
investment before 2003-04 are similar to the values for private hospitals between
2003-04 and 2006-07. This would result in an estimated total value of assets for
acute overnight private hospitals of between $3.5 billion and $4.0 billion in
2006-07. After inflating the value of assets to 2007-08 levels using the private gross
fixed capital formation index (AIHW 2009c), this would imply a range of between
$3.6 billion and $4.1 billion for 2007-08.

An investment time horizon of 17 years was used to estimate private hospital asset
values, resulting in an estimate of approximately $3.9 billion for 2007-08. A UCC
estimate for all Australian acute overnight private hospitals was then calculated by
multiplying the estimated asset value by the UCC rate. The UCC rate used is
8 per cent, which is the rate used by the SCRGSP (2009).

As the ABS data are from a census of all acute overnight private hospitals, the UCC
estimate was reduced using separation data so that it was in proportion to the
sample in NHCDC data. No inpatient admitted-patient cost proportion was available
for private hospitals, and so it was assumed that the admitted-patient cost proportion
for acute overnight private hospitals was 100 per cent. The Commission
acknowledges that this assumption is likely to overstate the UCC for admitted-
patient services in private hospitals.

The national estimate of the UCC was apportioned to private hospitals in each
jurisdiction by the proportion of total gross capital expenditure (minus land) in each
jurisdiction between 2002-03 and 2006-07 (ABS 2008e). Finally, the UCC values
were apportioned to individual DRGs according to the proportion of a jurisdiction’s
total private hospital depreciation attributable to each DRG.

The capital costs relating to diagnostic services are not included in the
aforementioned calculation of private hospital capital costs as private hospitals

8 Gross capital expenditure was used as a proxy for net capital expenditure (gross capital
expenditure less the trade-in values of replaced items and receipts for sales of replaced items)
because the latter was not available. Gross capital expenditure on land was excluded in the
estimation of asset values.
CONSTRUCTING COST 305
ESTIMATES
generally do not own the diagnostic equipment used in their hospital. However, the
capital costs for this equipment are included in medical and diagnostics costs as the
HCP charge data used implicitly incorporates a fee to cover capital costs.

As was the case for public hospitals, the UCC estimates by region or hospital size
were calculated by apportioning the Australia-wide UCC for private hospitals
across groupings by the number of casemix-adjusted separations.

Box D.6 Estimating asset values

The following example described by Webster et al. (1998), assumes capital prices are
fixed, straight line depreciation, and constant annual amounts of net investment prior to
the base period. Furthermore, it is assumed that:
I n is capital expenditure in year n
Dn is depreciation in year n
K n is the capital stock in year n

d is the rate of depreciation

K 0 is the base period capital stock
D0 is the (constant) annual amount of depreciation on the base period capital stock.

It is also assumed that capital investment occurs in the middle of the year and thus the
resulting capital depreciates only for half of that year. Therefore, the following
relationships hold:
K 1 = K 0 − D0 + I1 × (1 − d 2)
D1 = D0 + I1 × d 2
n −1
K n = K n −1 + I n × (1 − d 2) − D0 − d ∑I
i =1
i

n −1
Dn = D0 + I n × d 2 + d ∑I
i =1
i

(continued)

306 PUBLIC AND PRIVATE

HOSPITALS
 Box D.6 (continued)
Because the values for I t and Dt (for t = 1,...,n) are known, the equations relating to
depreciation all take the form:
y t = D0 + x t × d
where y t is Dt and x t is I t .

It is therefore possible to estimate D0 and d by regression.

Next, it is assumed that the investments that contributed to the base period capital
stock occurred in equal annual amounts over some (unknown) time horizon. For a
given time horizon, H, the amount of annual investment can then be calculated and the
relations above can be used to calculate the capital stock for subsequent years.
Source: Webster et al. (1998).

Benchmarking against asset data for major hospital groups

To assess whether the estimated $3.5–4.0 billion range for the total value of assets
for acute overnight private hospitals for 2006-07 was plausible, published data for
two major private hospital operators — Ramsay Health Care and Healthscope —
were examined.

Healthscope (2007) reported that it had property, plant and equipment (excluding
land) worth $560 million at 30 June 2007. It was estimated that Ramsay Health
Care had property, plant and equipment (excluding land) worth $990 million at the
same point in time.9

Both organisations have only a few free-standing day facilities and these do not
account for a significant share of total assets (Ramsay Health Care, pers. comm.
23 September 2009; Healthscope, pers. comm. 24 September 2009). Therefore, the
value of acute overnight private hospitals (excluding land) owned by both Ramsay
Health Care and Healthscope is considered to be approximately $1.55 billion in
2006-07.

9 Ramsay Health Care (2007) reported that it had property, plant and equipment worth
$1.16 billion in 2006-07, but does not separately publish the value of its land. Approximately
15 per cent of the total assets of both Healthscope (2007) and Australian public hospitals
(SCRGSP 2009) are reported to be attributable to land. If the same proportion applied to
Ramsay Health Care, then it would have property, plant and equipment worth $990 million.
This does not include Ramsay’s UK hospital operations, as they were purchased in
November 2007, but it does include the three hospitals it owns in Indonesia.
CONSTRUCTING COST 307
ESTIMATES
It is estimated that Ramsay Health Care and Healthscope accounted for around
48 per cent of acute overnight private hospital separations in 2006-07.10 If it is
assumed that the cost of capital per separation was similar across all private hospital
providers, then this would imply that the value of all acute overnight private
hospital assets was approximately $3.23 billion.

However, this is likely to be an underestimate of the actual value of acute overnight

private hospitals because the value of property, plant and equipment reported in the
annual reports of both Healthscope and Ramsay Health Care do not actually
represent the market value of these assets, but are more reflective of the cost
incurred when the assets were purchased, less depreciation (Ramsay Health Care,
pers. comm. 23 September 2009; Healthscope, pers. comm. 24 September 2009).

While the market value of assets is not published for either company, it is possible
to infer an upper bound using the company’s enterprise value (box D.7). The
enterprise value of acute overnight private hospitals (excluding land) is estimated to
be about $9 billion.11 If this were an estimate of the total assets it would mean that
the value of each business (goodwill) is equal to zero, which is not plausible.
Indeed, the majority of the difference between the upper bound and the estimated
value of assets (excluding land) is likely to be attributable to the value of the
business.12

10 According to Ramsay Healthcare (2009), it currently admits over 750 000 patients per annum
in Australia. Between 2005-06 and 2007-08, admissions in Ramsay Health Care hospitals rose
by 4.35 per cent per annum (Ramsay Health Care 2007, 2008). Assuming a similar growth rate
in 2008-09, it is estimated that the number of separations in Ramsay Health Care hospitals was
approximately 690 000 in 2006-07. Approximately 450 000 separations were recorded in
Healthscope hospitals in 2006-07 (Healthscope, pers. comm. 1 October 2009). Australian acute
overnight private hospital separations reported by the AIHW (2009) were 2 371 000 in
2006-07, implying an estimated market share for Ramsay Health Care and Healthscope of
around 48 per cent of all private hospital separations.
11 At 30 June 2007, Ramsay Health Care had a market capitalisation of approximately
$1.94 billion and net debt worth approximately $730 million. It therefore had an enterprise
value of approximately $2.67 billion. At 30 June 2007, Healthscope had a market capitalisation
of approximately $1.24 billion and net debt worth approximately $550 million. It therefore had
an enterprise value of approximately $1.79 billion. Assuming a market share for Ramsay Health
Care and Healthscope of approximately 48 per cent, the upper bound of the enterprise value of
all overnight acute private hospitals in Australia (excluding land) is approximately $9.3 billion.
12 When Ramsay Health Care bought Affinity Holdings in 2005, Affinity had an enterprise value
of approximately $1.4 billion. This included property, plant and equipment of $820 million at
market value, implying the market value of the business was approximately $580 million, or 41
per cent of the enterprise value. The proportion of a firm’s enterprise value that is attributable to
the market value of the business will differ between companies. However, if this same
percentage was applied to the estimated enterprise value for all acute overnight private
hospitals, it would imply that the total asset value of acute overnight private hospitals in
Australia was approximately $5.5 billion.
308 PUBLIC AND PRIVATE
HOSPITALS
 Box D.7 Enterprise value
The enterprise value of a company is an indicator of how the market values the
company. It can be represented as follows:
EV = MC + D
where EV is enterprise value, MC is market capitalisation (share price × no. of
ordinary shares) and D is the net debt (short term debt + long term debt – cash – cash
equivalents).
That is, EV is equal to the company’s market capitalisation — its share price
multiplied by the number of shares — plus its net debt (debt minus cash and cash
equivalents).
The market value of a company (as estimated by enterprise value) can broadly be
considered to consist of the market value of its fixed assets and the market value of the
business (including goodwill).
The difference between the value of property, plant and equipment and the enterprise
value will be largely attributable to the sum of the:
• market value of the business (including goodwill)
• difference between reported and market values of property, plant and equipment.
Source: McClure (2004).

Another reason why the Commission’s estimate of $3.5–4.0 billion for the value of
acute overnight private hospitals might be an underestimate is the use of operating
leases. While the value of hospitals that are operated under finance leases are
included in property, plant and equipment, the value of hospitals that are operated
under operating leases are not. Both Ramsay Health Care and Healthscope operate a
small number of hospitals under operating leases (Ramsay Health Care, pers.
comm. 23 September 2009, Healthscope, pers. comm. 24 September 2009).
However, some of these hospitals are operated as public hospitals. Furthermore, the
Commission understands that most of the hospitals with operating leases are
relatively small. Therefore, the Commission is of the view that not including
operating leases is unlikely to result in an underestimation of overnight acute
private hospital asset values of more than a few hundred million dollars.

Dr. Deeble (sub. DR56) calculated the implied depreciation rates for public and
private hospitals based on estimates prepared for this study’s Discussion Draft and
concluded that the depreciation rates were significantly different. However, these
rates were not based on the same depreciated asset values of public and private
hospitals used in the calculation of costs for this study’s Discussion Draft.
Specifically, Dr. Deeble reverse-engineered the total UCC for both public and
private hospitals using the 8 per cent UCC rate, but did not appear to adjust for
CONSTRUCTING COST 309
ESTIMATES
interest payments that are removed in the calculation of the UCC (box D.5). The
depreciation rates implicit in the capital cost calculations detailed in this report are
4.52 per cent and 8.18 per cent for public and private hospitals respectively.

In conclusion, given the published data of Healthscope and Ramsay Health Care
and the issues of operating leases and market valuations, the estimate of between
$3.5–4.0 billion for the value of acute overnight private hospitals may be an
underestimate of the actual asset value. As noted previously, the estimated value of
public-hospital assets may also be underestimated due to under-reporting of capital
used in public-private partnership arrangements, and the contracting out of public-
hospital services to private operators. The approaches used to estimate capital costs
and apportion them across DRGs are summarised in table D.11. The estimated
capital costs by jurisdiction are reported in table D.3.

Table D.11 Summary of sources and methods used to estimate capital

costs
Public hospitals Private hospitals

Cost of capital
Depreciation • NHCDC (DOHA, • NHCDC (DOHA,
unpublished) for all states unpublished)
except Victoria
• Victorian depreciation values
sourced from
SCRGSP (2009)
User cost of capital • SCRGSP (2009) • Commission estimates of
private hospital asset values,
based on ABS (2005, 2006,
2007, 2008).
Apportioning across DRGs
Depreciation • NHCDC for all states except • NHCDC
Victoria
• Victorian depreciation values
allocated across DRGs on
the basis of a weighted
average of the other
jurisdictions based on
separations.
User cost of capital • UCC values apportioned • UCC values apportioned
according to the proportion of according to the proportion of
total depreciation associated total depreciation associated
with each DRG. with each DRG.

310 PUBLIC AND PRIVATE

HOSPITALS
Table D.12 Estimated capital costs per casemix-adjusted separation,
2007-08a
Dollars

Tas, NT
NSW Vic Qld SA WA & ACTb Australia
Public hospitals
User cost of capital
per separation 290 212 390 252 237 301 279
Depreciation per
separation 148 147 170 129 123 146 147
Total cost of capital
per separation 439 359 560 381 359 447 426
Private hospitals
User cost of capital
per separation 97 94 104 69 129 126 100
Depreciation per
separation 113 145 118 89 152 219 130
Total cost of capital
per separation 210 240 223 158 281 345 230
a Australian totals may not add due to rounding. b Data for Tasmania, the Northern Territory and the ACT are
aggregated to protect the confidentiality of the small number of hospitals in each of these jurisdictions.
Source: Productivity Commission estimates.

Sensitivity analysis of public and private sector asset values

Because of the considerable uncertainty around the capital costs presented in this
report (especially those based on private hospital and public hospital asset values),
the Commission undertook a sensitivity analysis to analyse how different asset
values would alter the UCC per casemix-adjusted separation. As previously
discussed, the Commission considers that its estimates of private hospital asset
values could be underestimated, while there are also some questions regarding the
estimates of public hospital asset values. It was therefore thought useful to examine
the implications of varying asset values.

NSW Department of Health (sub. 41) estimated that the value of acute overnight
private hospitals was approximately $6 billion in 2007-08, compared to the
Commission’s estimate of around $3 billion. The sensitivity analysis was therefore
done for a range of $3–6 billion for the value of acute overnight private hospitals.
For public hospitals, a range of $18–24 billion was considered sufficient, given the
possible data inconsistencies. The UCC per casemix-adjusted separation was
calculated for both public and private hospitals for different asset values within
these ranges.

CONSTRUCTING COST 311

ESTIMATES
The Commission found that, if public hospitals assets were equal to $24 billion and
private hospital assets were equal to $6 billion, then there would be a difference of
almost $135 between the UCC per casemix-adjusted separation for public and
private hospitals (figures D.4 and D.5).

Figure D.4 Sensitivity analysis for private hospital user cost of capital

250
UCC per casemix-adjusted

200
separation

150

100

50

0
2 3 4 5 6 7
Value of assets ($billion)

Source: Productivity Commission estimates.

Figure D.5 Sensitivity analysis for public hospital user cost of capital

350
UCC per casemix-adjusted

300

250
separation

200

150

100

50

0
17 18 19 20 21 22 23 24 25
Value of assets ($billion)

Source: Productivity Commission estimates.

In contrast, if private hospital asset values were equal to $6 billion, and public
hospital asset values were equal to $18 billion, then the difference in UCC per
casemix-adjusted separation would still be approximately $55.

312 PUBLIC AND PRIVATE

HOSPITALS
In conclusion, the sensitivity analysis showed that, for a range of different asset
values, the capital cost per casemix-adjusted separation in public hospitals is
consistently higher than in private hospitals.

D.7 Relative complexity

The large volume of renal dialysis patients in public hospitals has a major impact on
the relative complexity of the public hospital casemix. Renal dialysis admissions
account for around 19 per cent of all public hospital separations, and have a
relatively low cost weight (table D.13). In contrast, around 3.7 per cent of private
hospital admissions are for renal dialysis. Calculating average cost weights without
renal dialysis separations increases the average public hospital cost weight from
0.96 to 1.01 and decreases the private hospital cost weight from 1.09 to 0.98
(table D.14).

Table D.13 Renal dialysis and chemotherapy separations as a percentage

of all separations by sector, 2007-08a
Tas, NT
NSW Vic Qld SA WA & ACTb Australia
Public hospitals
Renal Dialysis (L61Z) 19.5 18.2 17.9 15.7 18.1 29.2 19.0
Chemotherapy (R63Z) 0.2 5.8 3.1 0.0 5.6 1.8 2.8
Private hospitals
Renal Dialysis (L61Z) 3.7 0.0 8.7 0.0 0.0 0.0 3.7
Chemotherapy(R63Z) 5.7 9.6 3.3 10.7 10.7 8.0 7.1
All hospitals
Renal Dialysis (L61Z) 15.8 13.1 14.4 11.3 11.8 23.0 14.6
Chemotherapy (R63Z) 1.5 6.9 3.2 3.0 7.4 3.1 4.1
a Renal dialysis and chemotherapy separations are expressed as a percentage of all separations in this cost
analysis. A number of DRGs were not included in the cost analysis, as outlined in section D.1. b Data for
Tasmania, the Northern Territory and the ACT are aggregated to protect the confidentiality of the small
number of hospitals in each of these jurisdictions.
Source: DOHA (unpublished).

Another DRG with a high volume and low cost is chemotherapy (R63Z). It
accounts for around 2.8 per cent of public hospital separations and 7.1 per cent of
private hospital separations. Removing this chemotherapy DRG from calculations,
in addition to the prior removal of all renal dialysis cases, causes the relative
complexity of treatment across the two hospital sectors to converge to a relative cost
weight of 1.00 for both public and private hospitals.

CONSTRUCTING COST 313

ESTIMATES
Table D.14 Impact of renal dialysis and chemotherapy separations on
average cost weights by sector, 2007-08a
Tas, NT
NSW Vic QLD SA WA & ACTb Australia

Public hospitals
All DRGs with > 30 seps 1.01 0.92 0.96 1.01 0.94 0.85 0.96
All DRGs with > 30 seps,
without L61Zc 1.07 0.95 1.00 1.02 0.97 1.00 1.01
All DRGs with > 30 seps,
without L61Z and R63Zd 1.03 0.97 0.99 0.98 0.99 0.98 1.00
Private hospitals
All DRGs with > 30 seps 1.13 1.08 1.09 1.07 1.04 1.00 1.09
All DRGs with > 30 seps,
without L61Zc 1.02 0.94 1.02 0.93 0.90 0.88 0.98
All DRGs with > 30 seps,
without L61Z and R63Zd 1.03 0.98 1.01 0.98 0.95 0.90 1.00
a Average cost weight is the ratio of the average cost of all separations in a jurisdiction, relative to all
separations. b Data for Tasmania, the Northern Territory and the ACT are aggregated to protect the
confidentiality of the small number of hospitals in each of these jurisdictions. c L61Z refers to separations
involving renal dialysis. d R63Z refers to separations involving chemotherapy.
Source: Productivity Commission estimates.

The DRG system includes adjacent categories (adjacent DRGs, or ADRGs)

indicating the presence of comorbidities or complications which can increase the
expense associated with treatment.13 Less complex DRGs can be thought of as
‘bounded’ in their complexity — separations involving greater resource
consumption should be categorised as belonging to an DRG corresponding with a
higher level of resource consumption. DRGs that end with the suffix ‘A’ are
‘unbounded’ in their potential resource consumption in that they involve severe or
catastrophic complications or comorbidities, and by definition they correspond to
the highest consumption of resources within the ADRG (DOHA 2004). DRGs
ending with the suffix ‘Z’ are similarly unbounded in that they are not split by
resource requirements. In contrast, DRGs ending with a ‘B’, ‘C’ or ‘D’ are bounded
in that there is a higher category of resource usage.

If there was a noticeable difference between sectors in the ‘complexity’ within more
complex DRGs, it would be expected that removing the unbounded DRGs from the
cost analysis may bring the comparative estimates of cost per casemix-adjusted
separation closer together.

13 For example, in AR-DRG version 5.1, the ADRG (F62) relating to heart failure includes two
‘splits’ indicating different levels of resource consumption — one involving heart failure with
catastrophic complications or comorbidities (F62A), and one without (F62B).
314 PUBLIC AND PRIVATE
HOSPITALS
Table D.15 suggests that there may be some difference in average cost per casemix-
adjusted separation for different levels of resource requirements. For the most
complex DRGs (suffix ‘A’), the difference between public and private average costs
is around 8 per cent of the average public cost. For other DRG levels, the difference
is generally less, with the exception of ‘D’ DRGs, of which there are only four
included in this analysis.

Table D.15 Cost per casemix-adjusted separation for adjacent DRGs,

Australia, 2007-08a
Number of Public hospitals Private hospitals Difference
DRGs

DRGs with ‘A’ suffix 194 4 346 3 971 375

DRGs with ‘B’ suffix 197 4 259 4 267 -8
DRGs with ‘C’ suffix 50 4 241 4 301 - 60
DRGs with ‘D’ suffix 4 4 345 3 855 490
DRGs with ‘Z’ suffix 147 4 330 4 149 180
a DRGs with less than 30 separations in both public and private hospitals are excluded. Costs are casemix-
adjusted using combined DRG-level costs weights for both public and private hospitals.
Source: Productivity Commission estimates.

Table D.16 shows that, while there may be some difference in complexity, it does
not impact significantly on overall relative costs of public and private hospitals.
Removing those DRGs that are unbounded in their complexity does not
significantly impact on the cost difference between sectors.

Table D.16 Cost per casemix-adjusted separation for adjacent DRGs,

Australia, 2007-08a
Public hospitals Private hospitals Difference

All DRGs 4 302 4 172 130

DRGs with ‘A’, ‘B’, ‘C’, or ‘D’ suffix 4 291 4 285 106
DRGs with ‘B’, ‘C’, or ‘D’ suffix 4 255 4 274 -19
a DRGs with less than 30 separations in both public and private hospitals are excluded. Costs are casemix-
adjusted using combined DRG-level costs weights for both public and private hospitals.
Source: Productivity Commission estimates.

D.8 Prostheses costs

There are significant differences between the public and private prostheses costs.
This is particularly apparent at a DRG level. Of the 20 DRGs with the greatest
average cost per separation, 19 have a public hospital prosthesis cost that is less
than 90 per cent of the private cost, and seven of the 20 have a public hospital

CONSTRUCTING COST 315

ESTIMATES
prosthesis cost that is less than 50 per cent of the private cost (table D.17). Across
the twenty DRGs presented, the public prosthesis cost is around 55 per cent that of
the private prosthesis cost.

It is important to recognise that on the basis of the cost data presented, no firm
conclusions can be drawn as to whether the source of the difference is differential
pricing or the use of different prostheses across sectors. This is due to the lack of
available price comparisons across sectors for identical items. BUPA Australia
(2004) have previously presented evidence suggesting that suppliers of prostheses
charge different unit prices across the two sectors, stating that the cost faced by the
public sector is 55 per cent of that paid by BUPA Australia themselves for the same
item.

A number of cardiac procedures also display vastly different prosthesis costs across
sectors, although this may be both a product of different pricing and use of different
products across sectors. For example, the average public sector prosthesis cost
associated with percutaneous coronary intervention without acute myocardial
infarction involving the use of stents (DRG F15Z) is estimated to be around one
quarter of the prosthesis cost in the private sector. However, use of drug-eluting
stents — which may cost three to four times as much as bare-metal stents — is
higher in the private sector than in the public sector, and is a likely driver of the
sectoral differences in prosthesis costs for this procedure (Harper 2007; McLean
and Clark 2008).

However, a wider choice of more expensive devices is not necessarily the sole
cause of higher prostheses prices in the private sector. The two DRGs with the
costliest prostheses in the private sector (F01A and F01B) involve the implantation
or replacement of an automated implantable cardioverter-defibrillator (AICD)
(table D.17). As specified by the Prostheses List, benefits that are payable by
private health funds for these devices on the list range from $36 400 up to $52 000
(DOHA 2009d). Given that the average prostheses cost for these DRGs is between
$12 100 and $13 900 in the public sector, there appears to be a difference of over
$22 000 between the average prosthesis cost in the public sector and the least costly
device available in the private sector.14

14 The Commission understands that prosthesis costs for these DRGs are not necessarily restricted
to the AICD, but also involve a number of other costly components. Public costs mentioned
above include these components, whereas the private cost refers only to the AICD.
316 PUBLIC AND PRIVATE
HOSPITALS
Table D.17 Prosthesis costs for selected DRGs, 2007-08a
Public sector Private sector

DRG Descriptionb Separations Average Separations Average

cost cost

No. $ No. $
F01A Implantation or replacement of AICD, 1 079 13 849 652 55 490
total system w cs cc
F01B Implantation or replacement of AICD, 885 12 154 957 49 753
total system w/o cs cc
I06Z Spinal fusion w deformity 314 16 936 257 28 546
D01Z Cochlear implant 370 21 043 276 21 918
F02Z AICD component 177 7 880 79 18 638
implantation/replacement
I01Z Bilateral or multiple major joint 576 9 533 1 544 16 848
procedures of lower extremity
I09A Spinal fusion w cs cc 813 10 294 981 16 742
F12Z Cardiac pacemaker implantation 4 959 3 225 4 231 13 368
I03A Hip revision w cs cc 484 7 760 537 12 990
I09B Spinal fusion w/o cs cc 1 516 6 761 4 577 12 939
I11Z Limb lengthening procedures 124 3 589 56 10 971
I03C Hip replacement w/o cs cc 7 091 5 605 10 128 10 838
F17Z Cardiac pacemaker replacement 1 819 3 286 1 682 10 670
I03B Hip replacement w cs cc or hip revision 5 440 4 498 3 591 9 599
w/o cs cc
I05Z Other major joint replacement and limb 1 145 4 964 1 731 8 790
reattachment procedures
I04Z Knee replacement and reattachment 10 907 6 010 17 464 8 443
F03Z Cardiac valve proc w CPB pump w 371 5 780 579 6 706
invasive cardiac inves
F04A Cardiac valve proc w CPB pump w/o 1 672 4 965 1 212 6 578
invasive cardiac inves w cs cc
F04B Cardiac valve proc w CPB pump w/o 814 4 511 874 5 485
invasive cardiac inves w/o cs cc
a Table includes 20 DRGs with the highest prosthesis costs per separation. Public and private sectors share
the same top 20 DRGs. b w: with. w/o: without. cc: complications and comorbidities. cs: catastrophic or severe.
proc: procedure. AICD: automated implantable cardioverter-defibrillator CPB: cardiopulmonary bypass. inves:
investigation.
Source: Productivity Commission estimates.

D.9 Costs for patients funded by the Department of

Veterans’ Affairs
The Department of Veterans’ Affairs (DVA) is responsible for providing health care
to veterans and their dependants on behalf of the Repatriation Commission
(box D.8). In 2006-07, DVA-funded patients represented around 2.8 per cent of all
separations in public hospitals and 7.1 per cent of separations in private hospitals
CONSTRUCTING COST 317
ESTIMATES
(AIHW 2008b). As a client of both the public and private hospital sectors across
Australia, DVA’s experience could provide useful insights into the relative
performance of the two sectors.

Box D.8 Health care arrangements for veterans and their dependants
The Repatriation Commission is responsible under the Veterans’ Entitlements Act 1986
(Cwlth) for the provision of health services to eligible veterans and their dependants.
This responsibility is administered on behalf of the Repatriation Commission by the
Department of Veterans’ Affairs (DVA) and covers a range of available health care,
including general practitioner and allied-health treatment, in-home care and support,
and hospital care in both public and private hospitals.
In providing these services, over $4 billion was spent in the last year, with $1.7 billion
being spent on hospital services. In funding veteran health care, DVA covers the full
cost of treatment — there are no ‘gap’ payments made by veterans.
Currently there are around 272 000 veterans that are eligible for health services
provided by the Repatriation Commission. The Repatriation Commission notes that
there is a high risk of complications developing over the course of hospitalisation of
veterans given their age profile — 91 per cent of eligible veterans are over the age of
55, and 67 per cent over the age of 75. This risk is a potentially significant burden in
terms of cost to DVA.
Source: Repatriation Commission (sub. 39).

The Commission obtained data from DVA on the costs it has incurred in procuring
hospital services for veterans and their dependants. DVA identified the top 20
DRGs in terms of total cost between 2003-04 and 2006-07.15

A number of study participants cautioned that DVA patients are not necessarily
representative, with the procedures they undergo — and the difficulties associated
with them — likely to differ from those of the broader population (for example,
ACT Health, sub. DR52). This may be the case where the DVA patient cohort is
comprised exclusively of veterans. However, given that more than 50 per cent of
DVA’s patients are dependents — typically spouses of veterans, and often without
war-related illnesses — it is reasonable to expect that there are commonalities with
the general population. Procedures common to DVA patients could also be common
to those not eligible for DVA-provided health care but of similar demographic
profile. Further, DVA patients are often treated in the same hospitals and by the
same clinicians as other private patients. As such, the DVA data may provide a

15 Excluding mental health and rehabilitation DRGs and services involving sub-acute and
non-acute care.
318 PUBLIC AND PRIVATE
HOSPITALS
broad indication of the robustness of the Commission’s general findings based on
the NHCDC and HCP.

One way of assessing how similar DVA and other patients are is to compare their
average length of stay (ALOS) for a given DRG. ALOS is admittedly a crude
measure of patient heterogeneity, as it can be affected by a range of factors,
including comorbidities, age-related factors, clinical practice, and
purchasing/funding models. Nevertheless, a higher ALOS might be expected for
DVA patients because they tend to be older than the general population with a
higher incidence of comorbidities. Among the 20 DRGs for which the Commission
obtained DVA data, almost all had a higher ALOS for DVA patients than for the
NHCDC sample used in the Commission’s cost analysis (table D.18). Excluding
same-day procedures (lens procedures and renal dialysis), ALOS was on average
16 per cent higher for DVA patients.

Another concern expressed by study participants was that the DVA data are for
payments based on prices negotiated between DVA and the providers of hospital
services, rather than the cost of providing those services. The extent to which there
is a mark up over costs could vary across jurisdictions for public hospitals and
between different operators of private hospitals.

The Commission understands that there are notable differences between

jurisdictions in how contracts are structured between DVA and public hospitals. For
example, in some jurisdictions the cost of prostheses is included in a ‘bundled
charge’. Other jurisdictions charge for prostheses separately via the hospitals, in an
arrangement similar to that between DVA and private hospitals. The cost of
prostheses is included in the analysis below to ensure comparability between
hospitals.

It is also important to note that DVA contracts with private hospitals do not cover
payments to medical specialists, non-salaried allied health, diagnostic, radiology,
and pathology services. These payments are settled separately by DVA with the
specialists, and recorded in the data as a separate medical payment.

CONSTRUCTING COST 319

ESTIMATES
Table D.18 Comparison of average length of stay for DVA and NHCDC
patients, selected DRGs, 2006-07a
DVA populationb NHCDC sample
Public Private Public Private
DRG Descriptionc hospitals hospitals hospitals hospitals
I04Z Knee Replacement and Reattachment 8.6 8.3 7.1 7.4
A06Z Tracheostomy or Ventilation >95 hours 26.9 33.7 29.4 31.9
F12Z Cardiac Pacemaker Implantation 6.2 4.9 4.6 4.1
F15Z Percutaneous Coronary Intervention W/O AMI W 4.0 3.3 2.3 2.2
Stent Implantation
E65A Chronic Obstructive Airways Disease W 8.5 11.8 7.4 11.1
Catastrophic or Severe CC
I03C Hip Replacement W/O Catastrophic or Severe CC 8.4 8.4 6.8 7.2
I03B Hip Replacement W Cat or Sev CC or Hip 13.8 12.9 12.1 10.7
Revision W/O Cat or Sev CC
F62B Heart Failure and Shock W/O Catastrophic CC 6.0 8.5 4.8 7.5
I08A Other Hip and Femur Procedures W Catastrophic 14.0 18.1 14.7 16.5
or Severe CC
E62A Respiratory Infections/Inflammations W 10.5 14.4 10.0 12.8
Catastrophic CC
F42B Circulatory Disorders W/O AMI W Invasive 2.5 1.9 1.9 1.5
Cardiac Inves Proc W/O C
B63Z Dementia and Other Chronic Disturbances of 17.1 14.4 12.3 15.0
Cerebral Function
E62B Respiratory Infections/Inflammations W Severe or 6.6 9.4 5.8 8.1
Moderate CC
F62A Heart Failure and Shock W Catastrophic CC 11.6 15.5 10.5 14.5
E65B Chronic Obstructive Airways Disease W/O 4.7 8.2 4.5 7.2
Catastrophic or Severe CC
B70A Stroke W Catastrophic CC 16.5 21.3 16.5 19.2
G02A Major Small and Large Bowel Procedures W 17.4 18.4 17.1 16.7
Catastrophic CC
F08B Major Reconstruct Vascular Procedures W/O CPB 8.3 8.3 7.3 7.4
Pump W/O Catastrophe
a DRGS are ranked by total cost across sectors by the Department of Veterans’ Affairs for the four-year period
2003-04 to 2006-07. Renal dialysis (L61Z) and sameday lens procedures (C16B) are excluded from this table
as they have an average length of stay (ALOS) of one day by definition. ALOS for the DVA population is for
2006-07. The NHCDC sample data is for 2007-08, and includes both public and private patients. b ALOS for
the DVA population is the total number of occupied bed days divided by the number of separations for each
selected DRG. c w: with. w/o: without. cc: complications and comorbidities. cs: catastrophic or severe. AMI:
acute myocardial infarction. CPB: cardiopulmonary bypass. inves: investigation.
Source: Department of Health and Ageing (unpublished data); Department of Veterans’ Affairs
(unpublished data).

DVA patients in public hospitals are admitted as private patients and so are entitled
to choose their doctor. As a result, medical costs for DVA patients in public
hospitals are a combination of items billed by hospitals (services provided by
salaried medical officers) and items billed separately by private medical

320 PUBLIC AND PRIVATE

HOSPITALS
specialists.16 Prostheses are generally paid for separately in public hospitals by
DVA. To ensure comparability between public and private hospitals, medical items
billed by both hospitals and specialists are included in the analysis below.17

Among the 20 DRGs for which DVA provided data to the Commission, 70 per cent
(14 DRGs) had a lower cost per separation in public hospitals in 2006-07
(table D.19). However, the difference in cost between the public and private sectors
was relatively small on average across the 20 DRGs (cost per separation in public
hospitals about 4 per cent lower than private hospitals).

Nevertheless, many of the DRGs had a cost difference that was relatively large.
Around two-thirds of DRGs had a cost per separation in public hospitals that was
more than 10 per cent lower or higher than in private hospitals. At the extremes:
• cost per separation in public hospitals for percutaneous coronary intervention
without acute myocardial infarction, with stent implantation (F15Z), was 42 per
cent ($8449) lower than in private hospitals
• cost per separation in public hospitals for dementia and other chronic
disturbances of cerebral function (B63Z) was 50 per cent greater ($3943) than in
private hospitals.

Cardiac procedures involving large prostheses costs — in particular, stenting and

cardiac pacemaker implantation (F12Z and F15Z) — had a cost per separation that
was more than 10 per cent lower in public hospitals, compared to private hospitals.
This is broadly consistent with the Commission’s DRG-level cost estimates. Public
hospitals also had a lower cost for treating heart failure and shock with and without
catastrophic complications or comorbidities (F62A and F62B).

16 Costs for non-salaried medical officers are standard across both sectors, according to a fee set
by DVA.
17 For DVA patients in public hospitals, medical and prostheses costs were identified by DVA as
costs incurred between the date of admission and date of separation. This may overstate the
costs associated with a hospital episode of care, if the patient incurred health costs outside a
hospital on the admission or separation date. The impact of this is considered to be negligible.
DVA further advised that, particularly for public hospitals, there are a range of cost components
that are not readily attributable to DRGs and so may be excluded from the cost estimates (DVA,
pers. comm. 20 November 2009.
CONSTRUCTING COST 321
ESTIMATES
Table D.19 Separations and episode costs for DVA patients, selected
DRGs, 2006-07a
Separations Cost per separation
DRG Descriptionb Public Private Publicc Privated
no. no. $ $
I04Z Knee replacement and reattachment 101 2 147 21 375 21 518
A06Z Tracheostomy or ventilation >95 hours 235 197 80 069 82 370
F12Z Cardiac pacemaker implantation 244 1 368 18 476 21 292
C16B Lens procedures, sameday 831 7 881 3 436 3 387
F15Z Percutaneous coronary intervention w/o 97 1 335 11 512 19 961
AMI w stent implantation
E65A Chronic obstructive airways disease w cs 2 180 1 608 6 734 8 008
cc
I03C Hip replacement w/o cs cc 231 935 19 428 22 446
I03B Hip replacement w cs cc or hip revision 405 634 23 229 24 680
w/o cs cc
F62B Heart failure and shock w/o catastrophic 2 470 2 065 4 726 6 047
cc
I08A Other hip and femur procedures w cs cc 661 354 19 008 19 065
E62A Respiratory infections/inflammations w 1 278 706 9 436 10 032
catastrophic cc
F42B Circulatory disorders w/o AMI w invasive 259 2 898 5 049 5 693
cardiac inves proc w/o cc
L61Z Admit for renal dialysis 22 437 12 744 516 399
B63Z Dementia and other chronic disturbances 1 050 583 11 264 7 526
of cerebral function
E62B Respiratory infections/inflammations w 1 494 1 187 5 425 6 612
severe or moderate cc
F62A Heart failure and shock w catastrophic cc 906 624 9 662 10 921
E65B Chronic obstructive airways disease w/o 1 771 1 489 3 892 5 509
cs cc
B70A Stroke w catastrophic cc 763 272 14 694 12 960
G02A Major small and large bowel procedures w 226 356 27 608 23 665
catastrophic cc
F08B Major reconstructive vascular procedures 107 551 20 312 18 614
w/o CPB pump w/o catastrophic cc
a Top 20 DRGs ordered in terms of total cost incurred by DVA over the four-year period 2003-04 to 2006-07.
Activity in standalone day procedure centres was excluded. DVA advised that, particularly for public hospitals,
there are a range of cost components that are not readily attributable to DRGs and so may be excluded from
the cost estimates. b w: with. w/o: without. cc: complications and comorbidities. cs: catastrophic or severe. AMI:
acute myocardial infarction. CPB: cardiopulmonary bypass. inves: investigation. c Public costs include data
supplied by DVA as hospital, medical and prostheses costs. Costs of public hospital episodes are indicative
because they include South Australian costing rates that have yet to be finalised. d Private costs include data
supplied by DVA as hospital medical, prostheses, theatre, accommodation, bundled and other costs. Medical
costs include diagnostics costs and allied health costs. Pharmacy and Intensive Care Unit costs are not
included.
Source: Department of Veterans’ Affairs (unpublished data); Productivity Commission estimates.

322 PUBLIC AND PRIVATE

HOSPITALS
The DVA cost data provide a useful point of comparison with the Commission’s
DRG-level cost estimates, although such a comparison needs to be viewed in light
of the abovementioned qualifications. To enable such comparisons, the
Commission’s 2007-08 estimates were deflated to 2006-07 values using the total
Health Price Index (AIHW 2009c). It was found that:
• The cost per separation for DVA patients in public hospitals is within 90 to
110 per cent of the Commission’s estimate for six out of the 20 DRGs. The cost
for DVA patients was more than 10 per cent below the Commission’s estimate
for seven DRGs, and more than 10 per cent above for the remaining seven
DRGs.
• The cost per separation for DVA patients in private hospitals appears to be more
comparable to the Commission’s estimates. The cost for DVA patients in private
hospitals was within 90 to 110 per cent of the Commission’s estimate for ten out
of the 20 DRGs. The cost for DVA patients was more than 10 per cent below the
Commission’s estimate for three DRGs, and more than 10 per cent above for the
remaining seven DRGs.

CONSTRUCTING COST 323

ESTIMATES
E Multivariate analysis in detail

The purpose of this appendix is to detail the data and statistical techniques the
Commission has used its multivariate analysis of the performance of public and
private hospitals. A summary of previous selected studies is presented in
section E.1. A description of the methods applied is given in section E.2. Data
sources and the Commission’s approach to assembling the dataset are outlined in
section E.3. The variables used in the analysis are described in section E.4. Results
of the analysis and post-estimation statistics are presented in section E.5. The
Commission’s proposed future analysis is discussed in section E.6.

E.1 Previous studies

There are a large number of multivariate studies of hospital performance that have
been undertaken worldwide. Despite this large number, only a few have used
Australian data. O’Neill et al. (2008), for example, in a detailed study of 79 data
envelopment analysis (DEA) studies did not include any Australian studies in their
review. A similar pattern can be gleaned from literature reviews by Butler (1995),
Peacock et al. (2001), Hollingsworth (2008) and Hollingsworth and Peacock (2008).

This is not to say that there have not been any Australian studies. The Commission
reviewed thirteen of the more commonly cited Australian studies published since
the mid-1990s. These include Butler (1995), SCRCSSP (1997), Webster, Kennedy
and Johnson (1998), Yong and Harris (1999), Wang and Mahmood (2000a, 2000b),
Paul (2002), Queensland Department of Health (2004), Mangano (2006), Jensen,
Webster and Witt (2007), Gabbitas and Jeffs (2008), and Chua, Palangkaraya and
Yong (2008, 2009).

A summary of the methods and data used in the overseas and Australian studies is
given in table E.1. The table is organised according to the type of function (cost or
production) and modelling techniques used (DEA, stochastic frontier analysis
(SFA), stochastic distance function (SDF) or other). Studies that employed more
than one modelling technique (such as Webster, Kennedy and Johnson 1998) are
therefore reported more than once.

MULTIVARIATE 325
ANALYSIS IN DETAIL
Table E.1 Selected literature review
Author(s) and No. of hospitals Dependent Independent variables External factors Quality or
year published and year(s) variable patient safety

Cost function – Stochastic frontier analysis

Herr (2008) 1594 German Total No. of cases, no. of weighted No subsidies Occupancy
public, non-profit (adjusted) cases, unit prices for doctors, dummy, East rate, nurse-bed
private, and for- costs. nurses, other staff, no. of beds, dummy, female ratio, average
profit private surgery ratio, total adjusted ratio, 75+ ratio. length of stay
hospitals, costs per bed, total adjusted (ALOS),
2001-2003. costs per weighted case. mortality rate.
Yaisarwang 131 US Vets Total Medical, nursing and other Intensive care unit In-hospital
and Burgess Affairs hospitals, (adjusted) salaries, no. of operating beds, intensity index, mortality rate,
(2006) 2000. costs. outpatient services, inpatient urban, teaching and readmission
services, access indicators psychiatric hospital rate, length of
(occupancy rate, waiting days, status. stay for
market penetration). readmissions,
average days
to readmit.
Jacobs (2001) 232 National Cost Index Emergency room (ER) visits, Transfers to and None.
Health Service (actual cost casemix weight, index of from a hospital,
hospitals, divided by unexpected ER visits, occasions patients under 15,
1995-96. expected of outpatient services. patients over 60,
cost). female patients,
teaching, market
forces factor.
Wang and 113 NSW public Total Inpatient casemix index, Dor and Farley ALOS of acute
Mahmood hospitals (in two variable occasions of service, ER visits, index, inpatient separations.
(2000a) peer groups – cost. input price of medical staff, casemix index.
large and small) average non-medical costs,
1997-98. average available beds,
percentage sameday
separations.
Yong and 35 large Victorian Total Weighted-inlier equivalent Metropolitan hospital, Occupancy
Harris (1999) acute public operating separations (WIES), occasions of teaching status. rate,
hospitals for expend., service, emergency services, staff per WIES.
1994-95. admitted average medical wage, nursing
patient cost. wage, other staff wage, hotelling
wage, medical support staff
wage, size (number of beds).
Rosko and 195 Pennsylvania Total costs. Inpatient separations, outpatient Severity of illness None.
Chilingerian acute care visits, wage rate, average price index, teaching
(1999) hospitals, 1989. of capital, casemix index. variables,
Herfindahl index.
Linna (1998) Finnish hospitals Net Inpatient admissions, accident Research and Readmission
from 1988 to operating and emergency visits, hourly development rate.
1994. cost. wage index, index on local variable, teaching
government expenditure, time dummy.
dummy.
Webster, 280 Australian Total Bed unit costs, materials unit None. None.
Kennedy, private hospitals operating costs, staff unit costs, revenue
Johnson (1998) in 1994-95. expenditure (output), occupied bed days,
. squared and cross terms.

(Continued next page)

326 PUBLIC AND PRIVATE

HOSPITALS
Table E.1 (continued)

Author(s) and No. of hospitals Dependent Independent variables External factors Quality or
year published and year(s) variable patient safety
Zuckerman, 1600 US hospitals Total Medicare admissions, Medicare Percent male Transfers from
Hadley and in 1984 and 1985. operating post admission days, patients, percent another
Iezzoni (1994) cost. non-Medicare admissions and older patients, hospital,
non-Medicare post-admission scores for disease mortality rates
days, outpatient visits, average status, plus a large of certain
salary per FTE (full-time number of factors patients.
equivalent), average capital descr bing
cost per bed. characteristics of
hospitals.
Vitiliano and 443 US nursing Total costs. Patient days, admissions and Voluntary, public, None.
Toren (1994) homes for 1987 transfers, per cent low care corporate,
and 1990. patients, wages of medical aids, proprietorship,
registered nurse wages, partnership.
property expenses (per square
feet).
Cost function – Ordinary least squares
Dor and Farley 500 US acute Total Inpatient discharges, casemix Severity of illness None.
(1996) non-federal variable index, outpatient services, index, source of
general hospitals. (operating) surgery share, ER visits, average hospital funding.
cost. salary, average capital price.

Butler (1995) 121 Queensland Average ALOS, occupancy rate, case None. None.
public hospitals cost per flow rate, no. of beds.
and 35 private casemix-
hospitals. adjusted
separation.
Scott and 76 Scottish acute Total No. of acute discharges, no. of None. None.
Parkin (1995) hospitals for variable other discharges, acute length of
1992-93. cost. stay (LOS), other LOS, outpatient
and ER visits, beds.
Granneman, 867 US hospitals Total No. of acute inpatient, sub- Revenue sources, None.
Brown and in 1982. annual cost. acute, and intensive care days location dummies,
Pauly (1986) and discharges, and accident per capita income of
and emergency visits, region, teaching
outpatient and other visits, status and presence
wage rates for four categories. of particular
facilities.
Single output production function – Stochastic frontier analysis
Herr (2008) 1594 German No. of No. of doctors, no. of nurses, No subsidies Occupancy
public, non-profit cases, no. no. of other staff, no. of beds, dummy, East rate, nurse-bed
private, and for- of weighted total adjusted costs per bed, dummy, female ratio, ALOS,
profit private cases. total adjusted costs per ratio, 75+ ratio, morality rate.
hospitals, 2001– weighted case. surgery ratio.
2003.

(Continued next page)

MULTIVARIATE 327
ANALYSIS IN DETAIL
Table E.1 (continued)

Author(s) and No. of hospitals Dependent Independent variables External factors Quality or
year published and year(s) variable patient safety
Mangano 116 Victorian Total WIE No. of FTE nurses, no. of FTE Teaching and None.
(2006) public hospitals, separations, medical support staff, no. of metropolitan
1992-93 to total inpatients admin and clerical staff and no. location status.
1995-96. treated. of FTE hotelling staff, average
no. of available beds.
Brown (2003) 20 per cent Inpatient No. of FTE employees, no. of Share of None.
sample of separations. beds, capital expenses, admissions enrolled
hospitals in 17 casemix index. in health
US states, 1992 management
to 1996. organisations, share
enrolled in preferred
provider
organisations,
teaching dummy,
public & for-profit
status.
Webster, 300 private Revenue, No. of FTE staff, no. of beds, Hi tech dummies. None.
Kennedy, hospitals for composite of cost of materials, (plus squared
Johnson (1998) 1994-95. occupied bed and cross terms).
days.
Multi-output production function – Data envelopment analysis
Chua, 123 Victorian Total WIES No. of FTE doctors, no. of FTE Second-stage Tobit Risk-adjusted
Palangkaraya public hospitals registered and other nurses, no. regression testing unplanned
and Yong between of FTE admin, domestic and for the effects of readmissions
(2009) 2003-04 and other staff, no. of beds, hospital (output).
2004-05. expenditures on drug, medical competition.
and surgical supplies.
Vitikainen, 40 Finnish Casemix- Hospital operating costs. None. None.
Street and public acute adjusted
Linna (2009) hospitals in inpatient
2005. admissions
(episodes,
days and
cases),
outpatient
visits and ER
visits
Nayar and 53 non-federal Casemix- No. of total staff, no. of beds, Teaching FTEs (as Percent of
Ozcan (2008) hospitals in adjust. costs (excluding payroll and an output). patients
Virginia in 2003. separations, costs), total assets. receiving:
outpatient antibiotics;
visits oxygenation;
(including and aged 65+
accident and given
emergency). pneumoccal
vaccination.
Mangano 100 Victorian WIES, total No. of FTE non-medical staff, None. None.
(2006) public hospitals, inpatients average no. of available beds.
1992-93 to treated.
1995-96.

(Continued next page)

328 PUBLIC AND PRIVATE

HOSPITALS
Table E.1 (continued)

Author(s) and No. of hospitals Dependent variable Independent variables External Quality or
year published and year(s) factors patient safety
Harrison and Between 471 Admissions, outpatient No. of FTE staff, no. of beds, None. None.
Sexton (2006) and 480 private, visits. operating expenses, no. of
public, not-for- services.
profit for 1998
and for 2001.
Queensland Queensland Weighted separations, No. of FTE staff, non-labour None. None.
Department of public hospitals outpatient occasions of costs and gross asset values
Health (2004) for 2000-01 to service, other admitted
2002-03. care .
Biørn et al Unspecified no. Casemix-adjusted No. of FTE physicians, no. of Dummies None.
(2003) of Norwegian separations, fee-weighted other FTE staff, medical costs, for funding
hospitals outpatient visits . total expenses. source and
between 1992 university
and 2000. affiliation
and
location.
Hofmarcher, 93 Austrian Patient days, no. of No. of medical staff, no. of None. None.
Paterson, and hospitals discharges, LDF points. para-medical staff, no. of
Riedel (2002) between 1994 admin. staff, no. of beds, no.
and 1996. of wards, Index of casemix
complexity.
Al Shammari 15 Jordanian Patient days, minor No. of physicians, no. of None. None.
(1999) hospitals, 1991– operations, major health personnel, no. of bed
1993. operations. days.
Wang and 113 NSW public Inpatient casemix index, No. of doctors, no. of nurses, None. ALOS of
Mahmood hospitals for inpatient admissions, no. of non-medical staff, no. of acute
(2000b) 1997. outpatient visits, ER visits. beds, other expenses. separations.
Webster, 301 private Inpatient days, surg. No. of FTE medical staff, None. None.
Kennedy, hospitals for days, non-patient contract value of visiting
Johnson (1998) 1994-95. services, nursing home medical officers, no. of FTE
days, surg. proc., nurses, no. of FTE other staff,
inpatient separations, ER no. of beds, cost of materials.
visits, comp. output.
Burgess and 2420 US Acute inpatient days, No. of registered nurses, no. None. None.
Wilson (1998) hospitals with casemix-adjusted of practice nurses, no. of other
100+ beds, discharges, long-term clinical staff, no. of non-clinical
1985 to 1988. care days, no. of staff, no. of acute beds, no. of
outpatient visits, long-term beds, casemix
ambulatory surgeries, index.
inpatient surgeries.
O’Neill (1998) 40 Philadelphia Casemix-adjust. inpatient No. of FTE staff, no. of beds, Capital None.
and Pittsburgh medical separations, operational expenditure intensity
hospitals (27 casemix-adjust. inpatient (excluding payroll and capital). index for
urban and 13 surgical separations, specialist
teaching) with casemix-adjust. units.
300+ beds in outpatients, no. of trained
1992. residents.
SCRCSSP 109 Victorian Three categories of WIES No. of FTE non-medical staff, None. Unplanned
(1997) public hospitals outputs. no. of FTE medical staff, all readmission
for 1994-95. FTE staff, non-salary costs, rates.
medical salaries, total salaries.

(Continued next page)

MULTIVARIATE 329
ANALYSIS IN DETAIL
Table E.1 (continued)

Author(s) and No. of hospitals Dependent variable Independent variables External Quality or
year published and year(s) factors patient safety
Ferrier and 360 US rural No. of acute days, No. of FTE staff, no. of None. Occupancy
Valdmanis hospitals for subacute days, no. of beds, size, regional rate.
(1996) 1989. intensive days, no. of location, ownership.
surgeries, discharges,
outpatients
Bedard and 58 New York No. of inpatient No. of FTE staff, no. of None. None.
Wen (1990) and West separations, no. of beds; cost of labour,
Pennsylvania surgical operations, no. of non-payroll expenditure.
hospitals 1974 outpatient visits.
to 1979.
Morey and 105 North No. of patient days for Cost of nursing services, No. of None.
Dittman (1996) Carolina persons aged under 14, cost of ancillary services intensive-care
hospitals in patient days for persons (for example, radiology), beds, acute
1978. aged 14 to 65, patient cost of administration and beds and
days for persons aged general services. other beds,
over 65. percent each
of intensive-
care patient
days,
intensive or
acute-care
patient days,
capital value
of hospital.
Borden (1988) 52 New Jersey No. of cases treated for No. of total FTE staff, no. None. None.
hospitals 1979 high most common of FTE nurses, no. of
to 1984. diagnosis-related groups beds, other non-payroll
(DRGs), all other DRG expenses.
separations combined.
Multi-output production function with some outputs defined as undesirable – Data envelopment analysis
Clement et al. 667 hospitals No. of births, outpatient No. of FTE registered None. Risk-adjusted
(2008) from 10 US surgeries, ER visits, nurses, no. of FTE acute
states for 2000. outpatient visits, practice nurses, no. of myocardial
casemix-adjusted other FTE staff, no. of infraction,
admissions. beds, and capital. congestive
heart failure,
stroke,
gastrointestinal
haemorrhage,
pneumonia.
Multi-output production function – Stochastic distance function
Ferrari (2006) 52 Scottish Inpatients index, No. of medical staff, no. of None. None.
public hospitals outpatients et al. services nursing staff, no. of other
for 1991-92 to index. staff, no. of beds, capital.
1996-97.

(Continued next page)

330 PUBLIC AND PRIVATE

HOSPITALS
Table E.1 (continued)

Author(s) and No. of hospitals Dependent variable Independent External factors Quality or
year published and year(s) variables patient safety
Siciliani (2006) 17 Italian No. of discharges, No. of physicians None. None.
hospitals surgical discharges, and nurses, no. of
between 1996 medical discharges. other personnel, no.
and 1999. of beds.
Paul (2002) 223 NSW public No. acute inpatient seps, No. of FTE staff, no. Research, rurality, Standardised
hospitals in non- and sub-acute bed- of beds, capital, cost index of education mortality ratio.
1995-96. days, OOS, Inpatient of materials, no. of and occupation,
seps separated into services, no. of teaching.
public and private, and diagnoses.
were unweighted.
Löthgren (2000) 26 Swedish No. of operations, no. of Cost expenditure, None. None
county hospitals physician visits, no. of no. of beds.
1989–1994. inpatient admissions.
Gerdtham, 26 Swedish No. of operations, no. of Cost expenditure, Reimbursement None
Löthgren, county hospitals physician visits, no. of no. of beds. mechanism,
Tambour and 1989–1995. inpatient admissions. university hospital
Rehnberg status, patient age.
(1999)
Grosskopf, 108 Not-for- No. of acute patient days, No. of physicians, None None
Margaritis and profit and public no. of intensive care no. of FTE non-
Valdmanis hospitals in inpatient days, no. of medical staff, net
(1995) California and inpatient and outpatient plant assets.
New York in surgeries, no. of ER
1982. visits.
Malmquist productivity change (including when some outputs are undesirable)
Weng et al. 65 Iowa Average speeds of: No. of staff members, None. None.
(2009) hospitals treatment per case, no. of available beds.
between 2001 swing bed service, no.
and 2005. of admitted patients, no.
of swing bed patients.
Arocena and 20 Costa Rican No. of casemix-adjusted No. of FTE None. No. of
Garcia-Prado public hospitals discharges, no. of physicians, no. of FTE casemix-
(2007) between 1997– casemix-adjust. nurses, no. of beds, adjusted
2001. outpatient services. expenditure on goods hospital
and services. readmissions.
Chen (2006) 40 Taiwanese No. of seps, no. of No. of doctors, no. of Second stage ALOS and
public and surgeries, no. of nurses, no. of beds, regression of public occupancy rate
private intensive cares, no. cost of other medical status, severity of in a
hospitals. outpatient visits. supplies, no. of illness, Herfindahl second-stage
doctors and nurses index. regression
per department.
Sola and Prior 8 private and 12 No. of acute days, no. No. of FTE health None No. of
(2001); Prior public hospitals of long stay days, staff, no. of FTE other infections.
(2006) for 1990–1993. intensive days, no. of staff, no. of beds, cost
visits. of materials.
Maniadakis and 75 Scottish No. of ER patients, no. No. of doctors, no. of None None
Thanassoulis hospitals for of inpatients, no. of day nurses, no. of other
(2000) 1991-92 to cases, no. of staff, no. of beds,
1995-96. outpatients. cubic metre floor
space.

(Continued next page)

MULTIVARIATE 331
ANALYSIS IN DETAIL
Table E.1 (continued)
Author(s) and No. of hospitals Dependent variable Independent variables External factors Quality or
year published and year(s) patient safety
Webster, 280 private No. of occupied bed No. of FTE staff, no. None None
Kennedy, hospitals for days. of beds, cost of
Johnson (1998) 1991-92 to materials.
1994-95.
Linna (1998) Finnish No. of inpatient Hourly wage index, RandD variable, Readmission
hospitals from admissions, no. of index on local teaching dummy. rate
1988 to 1994. AandE visits. government
expenditure, time.
Färe, 19 OECD No. of bed days, no. of No. of physicians, no. None. Life
Grosskopf, countries from discharges. of beds; No. of expectancy
Lindgren and 1974 to 1989. physicians per for women
Poullier (1997) person, beds per over 40,
person. reciprocal of
infantry
mortality rate.
Burgess and 1545 profit, No. of inpatient days, No. of registered and None. None.
Wilson (1995) non-profit, Vets no. of casemix practice nurses, no. of
Aff., and Local separations, no. of long other clinical staff, no.
Govt hospitals stay days, no. of of non-clinical staff
for 1985–1988. outpatients, no. of ER No. of acute and
surgeries, no. of long-term beds, value
inpatient surgeries. of capital, casemix
severity.
Färe, 39 Michigan No. of acute care No. of doctors, no. of None. None.
Grosskopf and hospitals with patients, no. of ICU FTE non-doctor staff,
Valdmanis 200+ beds in patients, no. of emerg. no. of admissions, no.
(1989) 1982. patients, and no. of of beds.
surgeries.

No. of hospitals and Dependent variables Independent variables

year(s)

Patient-level modelling
Chua, 130 Victorian public Aggregate index of No. of episodes of care, proportion with: heart
Palangkaraya hospital admitted standardised hospital disease, admissions via emerg. department,
and Yong patients with heart mortality rate old, with high Charlson score, and with private
(2008) disease, 2000-01 to health insurance. Dummies for hospital
2004-05. location and status
Jensen, 130 Victorian public Readmission for AMI within 6 Charlson comorbidity index, gender, country
Webster and hospitals admitted months, or death within 30 of birth, Indigenous status, marriage status,
Witt (2007) patients with heart days of admission, mortality SEIFA index, hospital status (private, public
disease, 1996 to 2005. within 30 days of an teaching, public non-teaching).
unplanned 6-month
readmission.
Dormont and 36 French public Average cost per stay, for Gender, age profile, length of stay, hospital
Milcent (2004) hospitals 1994–1997. acute myocardial infarction admission, home admission, methods of
treatment.

332 PUBLIC AND PRIVATE

HOSPITALS
Some lessons from Australian and overseas studies

An examination of the Australian studies provides the following indicative

conclusions:
· private hospitals are less costly than public hospitals (when medical costs are
excluded)
· private hospitals give rise to better health outcomes than public hospitals
· for-profit private hospitals are more technically efficient than not-for-profit
private hospitals
· metropolitan public acute hospitals are more technically and cost efficient than
smaller rural hospitals.

A review of the overseas literature, however, generates some different impressions

with respect to the comparison between public and private hospitals:
· public hospitals are generally more technically efficient than not-for-profit
hospitals, which in turn are more efficient than for-profit hospitals (for example,
Hollingsworth 2008)
· teaching hospitals are generally less efficient than non-teaching hospitals,
possibly due to their more complex workloads (for example,
Hollingsworth 2008)
· larger hospitals tend to be more efficient than smaller hospitals, possibly due to
greater opportunities for scale economies (for example, Prior 2006; Vitikainen
et al. 2009)
· urban hospitals tend to be more efficient than non-urban hospitals (for example,
Färe, Grosskopf and Valdmanis 1989).

These, sometimes contradictory, impressions should not be generalised for public

and private hospitals in Australia, and possibly overseas because of the:
· limited scope of the studies
· inadequate representation of hospital services
· inadequate representation of health outcomes, quality and patient safety
· method by which factors outside the control of hospitals are controlled
· country-specific dimensions that affect the way in which public and private
hospitals are managed and the services they provide.

Even though the Commission is unable to draw firm conclusions about the studies’
findings, lessons can be drawn about the methods employed in each of these studies.

MULTIVARIATE 333
ANALYSIS IN DETAIL
Scope of studies

To date, no known Australian study has examined the comparative performance of

public and private hospitals nationally. Of the studies reviewed by the Commission,
most Australian studies examined the performance of public hospitals of one
jurisdiction (commonly New South Wales or Victoria) (Chua, Palangkaraya and
Yong 2008, 2009; Jensen, Webster and Witt 2007; Mangano 2006; Paul 2002;
SCRCSSP 1997; Wang and Mahmood 2000a, 2000b; Yong and Harris 1999). Only
three studies in the Commission’s literature review examined the performance of
both public and private hospitals, and these were limited to one jurisdiction (Butler
1995; Chua, Palangkaraya and Yong 2008, 2009). Only one study was conducted on
a national scale, but was limited to private hospitals (Webster, Kennedy and
Johnson 1998).

Inadequate representation of hospital services

A hospital’s performance should, ideally, be judged in terms of the cost of

providing incremental improvements to its patients’ health outcomes (Melbourne
Institute of Applied Economic and Social Research, sub. 16). However, this is a
problem for hospital-level studies because health-outcome measures cannot be
readily constructed.1 Instead, hospital performance is typically modelled by
separately accounting for the intermediate outputs of hospitals (such as inpatient
services, emergency department visits, and outpatient services) and the measurable
aspects of quality and patient safety.

While the majority of Australian studies have sought to adjust for the casemix
differences of inpatient services, not all have included emergency department visits
and outpatient services as intermediate outputs (for example, Chua, Palangkaraya
and Yong 2009; Mangano 2006; SCRCSSP 1997; Webster, Kennedy and Johnson
1998). This is particularly important when comparing public and private hospitals,
given that relatively more public hospitals operate emergency departments than
private hospitals.

Health outcomes, quality and patient safety

While some studies have directly measured patient health outcomes (for example,
Chua, Palangkaraya and Yong 2008; Jensen, Webster and Witt 2007), the majority
of Australian studies either ignored or only gave a cursory treatment to patient

1 This tends not to be an issue for patient-level studies (which make use of the incidence of
mortality) and country-level studies (which make use of life expectancies and disability-adjusted
life expectancies).
334 PUBLIC AND PRIVATE
HOSPITALS
health outcomes, quality and patient safety. The same can be said for most of the
overseas studies.

There appear to be two broad approaches to measuring quality and patient safety:
· Indirect (or proxy) variables are used to describe the level of patient care in a
hospital. These include the average length of stay, the occupancy rate, and the
ratio of clinical workforce per bed or patient (for example, Chen 2006; Ferrier
and Valdmanis 1996; Herr 2008).
· Direct variables of quality and patient safety. The most commonly used
measures are readmission rates and mortality rates (for example, Linna 1998;
Nayar and Ozcan 2008; Yaisarwang and Burgess 2006).

Factors outside the control of hospitals

Finally, most Australian studies did not adequately account for factors outside the
control of hospitals (for example, Queensland Department of Health 2004; Webster,
Kennedy and Johnson 1998). Again, the same can be said for many overseas studies
(for example, Färe, Grosskopf and Valdmanis 1995; Maniadakis and
Thanassoulis 2000).

Where external factors have been taken into account, they have tended to include:
· patient characteristics, such as:
– patient comorbidities (for example, Zuckerman, Hadley and Iezzoni 1994)
– gender and age profile of patients (for example, Zuckerman, Hadley and
Iezzoni 1994)
– patient socioeconomic characteristics (for example, Jensen, Webster and Witt
2007; Paul 2002)
· financial incentives of hospitals, such as:
– source of patient revenues — the extent to which a hospital is funded using a
prospective payment system or operates under capped budgets (for example,
Brown 2003; Dor and Farley 1996)
– market power of the hospital (for example, Chua, Palangkaraya and
Yong 2009; Rosko and Chilingerian 1999)
· geographic characteristics, such as:
– hospital location (for example, Granneman, Brown and Pauly 1986;
Herr 2008)

MULTIVARIATE 335
ANALYSIS IN DETAIL
· hospital roles, functions and specialisation, such as:
– whether it is a teaching or university hospital, the extent of research and
development (for example, Linna 1998; Yong and Harris 1999)
– the presence of specialist facilities or technologies (for example,
O’Neill 1998; Yaisarwang and Burgess 2006)
– the extent to which the hospital participates in inter-hospital transfers (for
example, Jacobs 2001).

There is a risk that hospital efficiency estimates would be biased if any of these
‘external’ factors are ignored. Worthington (2004), for example, argued that
ignoring patient characteristics could result in estimates of hospital efficiency
representing differences in patient characteristics rather than the hospital’s
performance.

Ownership or financial incentives?

One striking difference between the Australian and overseas studies is the
comparative efficiency of public and private hospitals. While it is conceivable that
private hospitals are more (cost) efficient in Australia and less technically efficient
overseas, it is possible that these findings reflect other confounding factors
(Hollingsworth 2008). One such factor is the way in which public and private
hospitals are funded.

There are three mechanisms by which publicly- and privately-owned hospitals are
funded:
· prospective payment systems (PPS) — in which hospitals are paid a fixed price
for each unit of output they provide
· per diem funding — where hospitals are paid for each patient in accordance with
the number of days spent in hospital
· global budget caps — where hospital budgets are capped.

Publicly-owned hospitals have traditionally been funded under capped global

budgets and privately-owned hospitals have been funded by private insurers on a
per diem basis. PPS funding (or casemix funding as it is known in Australia) is
increasingly being adopted to fund both public and private hospitals in Australia
and overseas.

336 PUBLIC AND PRIVATE

HOSPITALS
When hospitals are compared in terms of their funding mechanisms, a tentative
conclusion is that PPS funding is at least as efficient as funding under capped
budgets, and that both are more efficient than per diem funding. For example:
· US hospitals funded under the Medicare PPS were observed to have lower costs
than those that did not (Rosko and Chilingerian 1999; Zuckerman et al. 1994)
· Norwegian hospitals that were funded by PPS were found to be more efficient
than those that were funded by global budgets (Biørn et al. 2003)
· even though public hospitals in Germany were found to be more cost and
technically efficient than private hospitals (Herr 2008), the author noted that this
might be because public hospitals were funded under global budget caps and
private hospitals were paid on a per diem basis
· there is some evidence that US hospitals that receive prospective payment
funding are more technically efficient than those that are funded on a per diem
basis (Bedard and Wen 1990; Morey and Dittman 1996), though Borden (1988)
came to the opposite conclusion
· the introduction of PPS funding arrangements in Taiwan has led to
improvements in productivity and quality, and improvements were strongest
among public hospitals. PPS was observed to lead to excessive medical services
among private hospitals (Chen 2006)
· Löthgren (2000) and Gerdtham et al. (1999) each found that Swedish hospitals
funded with capped budgets were more efficient than those that were funded on
an output basis, but the authors acknowledged that they did not distinguish
between PPS and per diem funding arrangements.

A related confounding factor is that the generosity of the payer may also make a
difference to the reported efficiency. For example, Dor and Farley (1994) found that
US Medicare and private health insurance (PHI) pay relatively more than Medicaid
(and residual purchasers) and as a consequence, experienced higher hospital costs.
A third confounding factor is the role played by health management organisations,
which Brown (2003) found to make private hospitals more efficient than
not-for-profit private hospitals.

A key lesson for this study is to distinguish between ownership and funding models,
to the extent that such data are available.

MULTIVARIATE 337
ANALYSIS IN DETAIL
E.2 Commission’s approach to modelling hospital
performance
Hospitals are complex in the services they provide. There is also considerable
diversity between them in terms of the services they provide and their patients.
Hospitals can be compared in terms of technical and cost efficiency.

The Commission’s analysis in this report focuses on understanding the factors that
drive technical efficiency in the hospital sector. To achieve this, the first stage of
analysis is based on a pooled dataset of all hospitals in the sample for a single year
(2006-07). The pooled sample allows for variations in efficiency to be detected on
the basis of hospital size, indicating the extent to which scale economies exist
across the hospital sector. The pooled sample also allows for the number of
observations in the dataset to be preserved, which improves the accuracy of the
estimated model.

The Commission intends to undertake further analysis over coming months of

hospital performance in terms of both hospital outputs and costs, using a longer
dataset from 2003-04 to 2006-07. It is intended that the results from this analysis
will be published in March 2010.

The following discussion provides an overview of the techniques the Commission

has used for this first stage of its analysis.

Production function

In the first stage of analysis, hospital performance is modelled on the basis of an

output-oriented production function, where a hospital’s output (volume and type of
services provided) is assumed to depend on its use of inputs (resources such as staff
and capital). In the context of an output-oriented production model, hospital
performance is measured in terms of the hospital’s capacity to maximise its output
for a given set of inputs. This is known as technical efficiency (Coelli et al. 2005).

The efficiency of an individual hospital can be assessed by comparing its actual

output to the optimal level of output that could be achieved if the hospital adopted
best-practice production techniques. Using the available data in the sample, a
production ‘frontier’ is constructed which represents the optimal level of output
achievable. In this method of benchmarking hospital performance, in general,
production functions are widely used because they do not rely on any assumptions
about the behaviour of hospitals in relation to inputs and output prices.

338 PUBLIC AND PRIVATE

HOSPITALS
The production model is founded on the following form:

y i = f ( xi ) (1)

where yi is the output and xi is the vector of inputs for hospital i. Following
Kumbhakar and Lovell (2000), at this stage of analysis, the production model is
expressed as a deterministic function. Random variation will be introduced at a later
point.

When applied in a benchmarking framework, the optimal level of output that could
be achieved by a best-practice hospital is represented by:

y* = f ( x) (2)

where y* is the output of the best-practice hospital, and x is the vector of inputs that
generates the optimal level of output.

From these equations, the technical efficiency (TE) of a given hospital can be
computed. The efficiency score for a given hospital reflects the extent to which its
output falls below the optimal level of output achievable. Specifically, the scope of
technical efficiency (TE) of hospital i is measured by the ratio of its actual output
(yi) to the optimal output achievable (y*), as defined by:

yi
TE i = (3)
y*

The value of TEi will be between zero and one, where a value closer to one
indicates that the hospital is closer to full technical efficiency.

Estimating the frontier

The assessment of hospital performance involves estimating the ‘frontier’ that

benchmarks the optimal level of performance, and then computing the extent to
which each hospital falls below this frontier. One of the most commonly applied
methods to undertake these steps is stochastic frontier analysis (SFA). This
econometric technique was originally developed by Aigner, Lovell and
Schmidt (1977) and Meeusen and van den Broeck (1977) to study the efficiency and
productivity of firms. A good introductory summary of SFA can be found in
Coelli et al. (2005) and a more advanced treatment in Kumbhakar and
Lovell (2000).

MULTIVARIATE 339
ANALYSIS IN DETAIL
In SFA, the extent to which each hospital falls short of the benchmarked frontier
(that is, the extent of its inefficiency) is captured by the error term of the regression.
A key feature of SFA is that the error term is divided into two components:
· random error due to measurement errors, the omission of variables which cannot
be measured, and other random factors that affect output
· an error term that captures the extent to which the individual hospital falls short
of maximising its output for a given set of inputs (that is, its technical
inefficiency).

When introducing the two error components into the production function, the
stochastic frontier regression is modelled as:

y i = f ( x i ) + (v i − u i ) (4)

where yi is output, xi is a vector of inputs, vi is the random error, and ui is the

measure of technical inefficiency), for hospital i. It is assumed that both vi and ui are
independently and identically distributed; that vi follows a normal distribution with
a zero mean and constant variance; and that ui is a non-negative value and follows a
non-normal distribution that can be pre-defined as half-normal, truncated
half-normal, exponential or gamma.

The error component ui is interpreted to capture the technical inefficiency of each

hospital. Although the choice of the distribution for ui will affect the calculated
efficiency scores, there is evidence to suggest that it has a relatively lesser effect on
the ordinal rankings of the scores within a sample (Kumbhakar and Lovell 2000).
Conventionally, the technical efficiency score of each hospital is expressed in
logarithmic form such that the measured effects can be interpreted as proportional
changes, as follows:

Indexi = exp(−u i ) (5)

Figure E.1 illustrates the estimation of the production model using SFA regression.
The estimated function plots the relationship between input and outputs, shaped to
reflect diminishing returns to scale. Firstly, the model is estimated to pass through
the mean of the data (in this example, observation points A, B, C and D). This
generates the deterministic component of the production function, MM′.

Next, the production function MM′ is adjusted for each hospital by the component
of the random error that cannot be attributed to technical inefficiency (vi). This
establishes each hospital’s stochastic frontier. In this example, the production
function MM′ is adjusted by the amounts va and vb for hospitals A and B
respectively, establishing their respective stochastic frontier points A′ and B′. If vi is

340 PUBLIC AND PRIVATE

HOSPITALS
positive, the stochastic frontier will shift above the deterministic production
function (as for hospital B). If this random error is negative, the stochastic frontier
will shift below it (as for hospital A).

Figure E.1 Illustration of SFA production model

Output
B′
ub
vb

B
M′
D

va
A′
C
ua

M Input

Source: Adapted from Coelli et al. (2005).

Having established a stochastic frontier for each hospital that accounts for
hospital-specific random error, the difference between each hospital’s actual output
and its frontier can be attributed to its technical inefficiency (as represented by the
error component ui). In this example, the technical inefficiency of hospitals A and B
is represented by ua and ub respectively.

SFA offers some recognised advantages of over alternative estimation techniques.

Compared to standard ordinary least squares (OLS) regression, SFA formally
allows a role for random error in the estimation of efficiency measurements. OLS
estimation would construct a production frontier through the mean of the data (as in
the initial step of SFA) but would not adjust for hospital-specific random error when
computing the distance between the frontier and actual output (as in the next step of
SFA). As a further point of difference, SFA allows for a component of the errors to
be skewed (that is, non-symmetrically-distributed) whereas OLS imposes the
assumption that the whole error term is symmetrically distributed.

Another common technique applied in efficiency measurement is data envelopment

analysis (DEA), which uses piece-wise linear programming to estimate the
production function. A key difference between DEA and SFA is that SFA generates
parameters on the basis of a functional form, whereas DEA generates estimates

MULTIVARIATE 341
ANALYSIS IN DETAIL
based on the values of the observations rather than an assumed functional form
(Coelli et al. 2005).

The non-parametric approach of DEA may be considered an advantage because it

means that fewer restrictions are imposed on the model, and there are less risks
associated with misspecified functional forms (Nguyen and Coelli 2009). However,
non-parametric estimation presents several drawbacks. First, the significance of the
relationship between inputs and outputs cannot be statistically tested (PC 1999a;
Siciliani 2006). Without testing their significance, explanatory variables may be
inappropriately included in the frontier model. Second, non-parametric estimation is
more sensitive to the presence of outliers, which may distort the construction of the
production frontier and overstate the computed efficiency scores (Siciliani 2006).
Third, non-parametric estimation does not formally allow for technical inefficiency
to be distinguished from all other hospital-specific random error (Nguyen and Coelli
2009). For these reasons, the Commission has chosen to undertake SFA in favour
over DEA.

Accounting for quality of care

While estimating the volume of output delivered as a function of inputs, the

production model also needs to account for hospital resources that are allocated to
the quality of care that a hospital delivers, and include appropriate measures of
quality in the regression.

Before their inclusion, quality indicators need to be adjusted to control for

differences in the risk characteristics of the patients admitted to different hospitals.
In this context, risk refers to the extent to which patients’ characteristics affect the
likelihood of a successful treatment outcome, independent of the actions of the
hospital. It may be expected that a hospital which admits relatively ‘low-risk’
patients will require fewer resources per separation, meaning that it can deliver a
relatively larger volume of output for a given level of input. This will gives rise to
higher efficiency scores, all other factors equal, compared to a hospital which
admits relatively high-risk patients.

To adjust for patient characteristics, the Commission has used hospital-level

variables that reflect the composition of each hospital’s patient mix. For example,
patients’ gender is captured by a measurement of the proportion of a hospital’s
patients who are female.

Many of the available hospital-level quality indicators are measured as rates (for
example, mortality rates and readmission rates). This means that the estimated

342 PUBLIC AND PRIVATE

HOSPITALS
regressors of the model must be specified to fall between pre-determined upper and
lower bounds, as estimated by a Tobit model:

qi * = f ( z i ) + ε i
qi = qi * if q L < q i* < qU
(6)
= qL if q i* ≤ q L
= qU if q i* ≥ qU

where q* is the latent variable of the quality indicator, qi is the observed value of the
quality indicator, zi are the patient characteristics assumed to influence q* , qL and qU
are the lower and upper bounds of the quality indicator, and εi is the error term, for
hospital i. As with other censored regression models, parameters are estimated using
maximum likelihood methods.

The estimated results of the Tobit regression are used to compute the standardised
value of the quality indicator. This is computed by dividing the observed values by
the estimated values. This is commonly applied to mortality rates, where a
standardised value less than one indicates that a hospital is performing better than
expected (the actual mortality rate is lower than predicted), while a value greater
than one indicates an unfavourable performance (the actual mortality rate is higher
than predicted) (Ben-Tovim et al. 2009). The standardised values of the quality
indicator are included as regressors in the output equation.

Other factors influencing efficiency

The production function estimates a hospital’s level of output as a direct function of

its inputs. However, it is acknowledged that there are additional factors — known as
covariates — that influence a hospital’s production process and, therefore, its
reported efficiency score. The appropriate method to incorporate such factors into
the model depends on whether the factors are considered to be within the control of
the hospital or not.

Factors which are considered to be outside of the hospital’s control contribute to

setting the position of the frontier. In this case, the covariates can be included in the
production model, regressed directly against output. Factors which are considered
to be within the hospital’s control contribute to variations in efficiency below the
benchmarking frontier. In this case, the covariates can be modelled as a function of
the random errors of the output model.

MULTIVARIATE 343
ANALYSIS IN DETAIL
The two steps of this regression are defined as:
M
ln y i = β 0 + ∑ β mi ln x mi + (vi − u i ) (7)
m =1

J
μiu = δ 0 + ∑ δ ji ln z ji + ξi (8)
j =1

where yi, xi, vi and ui are as previously defined, μ iu is the conditional mean of ui, zi is
the vector of additional factors, and ξi is the error term. Factors which are within the
hospital’s control are included in xi, whereas factors which are outside of the
hospital’s control are included in zi.

Model specification

Given that hospitals produce a range of outputs (rather than a single output), a
stochastic frontier specification which allows for multiple outputs is used. Known
as an (output) stochastic distance function, it is defined as:

DOi ( xi , yi ) = min{TE : yi / TE ∈ P( xi )} (9)

where yi is the vector of outputs, xi is the vector of inputs, and TE is the minimum
amount by which output can be reduced and still remain producible with the given
set of inputs (Kumbhakar and Lovell 2000).

When applied to the production model, several functional forms are applicable.
One of the most basic and widely-applied functional forms is the Cobb-Douglas
model, which regresses the terms in first-order form only. The functional form can
be expanded with the inclusion of second-order quadratic and cross-terms that allow
for interaction effects among the variables, as is applicable for a multi-output,
multi-input production model (Paul 2002). The following equation specifies a
production model in an expanded multi-input, multi-output form, known as a
transcendental logarithmic (translog) distance function:
M K
ln DOi ( xi , yi ) = α 0 + ∑ α m ln ymi + ∑ β k ln xki
m =1 k =1
M M K K
+ 0.5∑ ∑ α mn ln ymi ln yni + 0.5∑∑ β kl ln xki ln xli (10)
m =1 n =1 k =1 l =1
K M
+ ∑∑ δ km ln xki ln ymi
k =1 m =1

344 PUBLIC AND PRIVATE

HOSPITALS
where DOi is the distance to the frontier (taking a value between 0 and 1), yk
represents output, xm represents input, M is the number of outputs, and K is the
number of inputs. As is common practice, all terms are specified in natural
logarithms, so that the measures represent proportional values rather than absolute
levels. The first line of equation (10), comprising first-order variables only,
represents the standard Cobb-Douglas form. The inclusion of the higher-order
squared terms in the second and third lines represents the complete translog
function.

The Cobb-Douglas model is widely applied as it is more parsimonious and

computationally simpler to estimate than the higher order, more flexible functional
forms. Compared to the more flexible functional forms, the limited number of
parameters in the Cobb-Douglas model means there is less risk of multicollinearity
and less loss in degrees of freedom. Furthermore, the coefficients of the
Cobb-Douglas model are relatively more straightforward to interpret as elasticity
values.

However, the simplicity of the Cobb-Douglas model restricts its estimation power.
For example, the introduction of the squared terms can be used to detect scale
economies, while the further inclusion of cross-terms in the translog model can
detect elasticity of substitution between inputs, production coefficients between
inputs and outputs, and marginal rates of transformation between outputs (Nguyen
and Coelli 2009; Siciliani 2006). All this means is that the Cobb-Douglas model is a
relatively inflexible form and is not likely to completely fit the curvature of the
production function.

In this analysis, the Commission estimated both the Cobb-Douglas and a restricted
version of the translog model and then compared measures of their goodness-of-fit
and predictive performance. Higher-order functional forms are expected to provide
a more accurate fit of the observed data. These models, therefore, are expected to
generate higher efficiency scores because they contain less unexplained variation
that would otherwise be attributed to random error or inefficiency. Nguyen and
Coelli (2009) presented a meta-analysis of hospital efficiency studies which
substantiated this observation. When selecting the model to apply, it is also
recognised that higher-order functional forms are likely to incur more
computational difficulties, due to the large number of multiplicative parameters
contained in the model.

MULTIVARIATE 345
ANALYSIS IN DETAIL
For the models to comply with standard economic regularity properties, and for an
empirical equation to be estimated, homogeneity constraints need to be imposed
(Coelli et al. 2005; O’Donnell and Coelli 2005). The constraint of homogeneity of
degree one in outputs is defined as:
M M M K M

∑ α m + ∑∑ α mn ln y n + ∑∑ δ km ln xk = 1
m =1 m =1 n =1 k =1 m =1 (11)

This constraint is satisfied if:

M M K M

∑α
m=1
M = 1, ∑α
m =1
mn = 1 for all n, and ∑∑ δ
k =1 m=1
km = 0 for all k . (12)

According to Lovell et al. (1994), the homogeneity condition is equivalently

satisfied by normalising equation (10) by one of the outputs (yL), as follows:

⎛ D (x , y ) ⎞ M −1
⎛y ⎞ K
ln⎜⎜ i i i ⎟⎟ = α 0 + ∑ α m ln ⎜⎜ mi ⎟⎟ + ∑ β k ln xki
⎝ y Li ⎠ m =1 ⎝ y Li ⎠ k =1
M −1 M −1
⎛y ⎞ ⎛y ⎞ K K
+ 0.5 ∑ ∑ α mn ln ⎜⎜ mi ⎟⎟ ln ⎜⎜ ni ⎟⎟ + 0.5∑∑ β kl ln xki ln xli (13)
m =1 n =1 ⎝ y Li ⎠ ⎝ y Li ⎠ k =1 l =1

K M −1
⎛y ⎞
+ ∑ ∑ δ km ln xki ln ⎜⎜ mi ⎟⎟
k =1 m =1 ⎝ y Li ⎠

This can be condensed to:

⎛D ⎞ y
ln ⎜⎜ i ⎟⎟ = TL ( xi , i , α , β , δ ) (14)
⎝ y Li ⎠ y Li

where TL(.) refers to the translog function.

The expression can be re-arranged and specified as a stochastic distance function

with the inclusion of the technical efficiency component and random error term, as
follows:

yi
ln Di − ln y L = TL ( xi , ,α , β , δ ) (15)
y Li

yi
− ln y L = TL ( xi , , α , β , δ ) − ln Di (16)
y Li

346 PUBLIC AND PRIVATE

HOSPITALS
 yi
− ln y M = TL ( xi , , α , β , δ ) + (v i − u i ) (17)
y Mi

where –lnDi = (vi – ui) and vi and ui are as previously defined.

E.3 Data sources

Data for public and private hospitals, detailed at both patient and establishment
levels, had to be sourced from several different data collections and then merged to
create the final data set. Details of the data sources and the process of accessing and
assembling the dataset are outlined below.

Public hospital data

Establishment-level data for public hospitals were drawn from the National Public
Hospital Establishments Database (NPHED), which is held by the Australian
Institute of Health and Welfare (AIHW).

Patient-level data for public hospitals were drawn from the National Hospital
Morbidity Database (NHMD), which is also held by the AIHW.

Private hospital data

Establishment-level data for private hospitals were drawn from the Private Health
Establishments Collection (PHEC), which is held by the Australian Bureau of
Statistics (ABS). The collection is drawn from a census of private hospitals (acute
and psychiatric) and free-standing day facilities (ABS 2008f).

Patient-level data for private hospitals were drawn from the National Hospital
Morbidity Database (NHMD), which is held by the AIHW. Although the PHEC
held by ABS contains patient data, the Commission does not regard these data to be
useful for this study because they are not casemix-adjusted and do not include the
details required on patient morbidity.

Accessing hospital data

To access data for the purpose of this analysis, the Commission obtained the
consent of the state and territory health departments for the AIHW to release public
hospital patient and establishment data to the ABS. The Commission also obtained

MULTIVARIATE 347
ANALYSIS IN DETAIL
the consent of 130 privately-owned hospitals for the state and territory health
departments to provide additional information that would allow the private hospital
patient data held by the AIHW to be matched with the establishment-level data held
by the ABS. After excluding free-standing day facilities and non- and sub-acute
facilities, there were 122 private acute hospitals in the sample.

The ABS undertook the analysis with the assistance of the Commission. This
arrangement was to facilitate access to the private hospital information held by the
ABS, and to safeguard the data drawn from both ABS and AIHW sources.

Assembling the data

The first step in assembling the dataset was to match the patient-level morbidity
data needed with each hospital. The morbidity data were then aggregated to the
establishment-level data. Hospital-level patient variables were created which
represented the shares of patients with given patient-level characteristics.

In the case of private hospitals, the patient-level data contained in the NHMD (held
by the AIHW) had to be matched with the corresponding establishment-level data
contained in the PHEC (held by the ABS).

Furthermore, several adjustments to the dataset needed to be made to handle

reporting inconsistencies.
· A number of Victorian hospitals are incorporated into regional networks. As a
result, much of the establishment-level data for these hospitals are available at
the network level and needed to be rescaled to match establishment-level data.
Rescaling was achieved by disaggregating the networked data on the basis of the
number of hospitals contained in the network, and weighting the values on the
basis of each hospital’s number of casemix-adjusted separations. To capture
potential efficiency effects of belonging to a network and indicate networked
hospitals, a dummy variable denoting network membership was created.
· A single observation was provided for Tasmanian public hospitals. The names
and the number of beds are known for each Tasmanian hospital, but not the
number of casemix-adjusted separations. The establishment- and patient-level
data of the single Tasmanian observation were disaggregated on the basis of the
number of acute and non-acute beds. On the basis of the hospital’s name and
address, the Australian Standard Geographic Classification – Remoteness Area
(ASGC-RA) classification of each hospital was computed. The limitation of this
approach is that it blurs the distinction between the functions of Tasmanian acute
and non-acute hospitals.

348 PUBLIC AND PRIVATE

HOSPITALS
· A single observation was provided for Tasmanian private hospitals. However,
since the number of private hospitals in Tasmania was not known with certainty,
the observation could not be disaggregated, although the scale of these hospitals
is expected to be very small.

Representativeness of the sample

Ideally, the data contained in the sample for analysis should be representative of all
Australian hospitals. In this study, however, data on private hospitals was only
made available on a voluntary basis and therefore do not necessarily represent the
full range of private hospitals in Australia.

In particular, a larger proportion of for-profit private hospitals made their data

available to this study than not-for-profit private hospitals. For-profit hospitals
accounted for 57 per cent of Australia’s 289 private acute hospitals (AIHW 2009a).
In contrast, 85 per cent of the private hospital sample comprises for-profit hospitals.

This presents two concerns. First, it means that the not-for-profit hospitals are
relatively under-represented compared to for-profit hospitals. Second, it means that
the dataset is potentially subject to sample-selection bias, as the private hospitals
included in the study are not a random selection. If the factors which affect hospital
efficiency also affect the likelihood that a hospital agreed to participate in the study,
the efficiency estimates may be biased.

The Commission considered potential methods to overcome this sampling issue,

including the Heckman correction procedure (Heckman 1976). However, given that
there is no common statistical technique to address this issue in this field of
analysis, and given the time constraints of this study, the Commission’s analysis
proceeded without such sampling correction. It is acknowledged, therefore, that the
findings only apply to the hospitals included in the study, and the Commission
cautions readers from drawing conclusions for all hospitals in Australia.

In its further analysis, the Commission intends to examine the degree to which the
sample of hospitals included in the analysis adequately represents the population of
hospitals Australia-wide, and further investigate methods to address potential
sampling bias.

Final dataset

The AIHW provided a range of hospital-level data from the NHMD that correspond
to 703 public hospital observations in its NPHED and 130 private hospital

MULTIVARIATE 349
ANALYSIS IN DETAIL
observations that agreed to participate in this study. After removing acute, sub-acute
non-acute, psychiatric hospitals and free-standing day facilities, there were 508
acute hospital observations in the sample. Of these, 368 were public hospitals and
another 18 that are ordinarily classified as public hospitals by the AIHW, but which
are typically managed by non-government entities to provide public hospital
services for state and territory governments. These are referred to as ‘public
contract’ hospitals. There were also 122 private acute hospital observations in the
sample (table E.2).

Table E.2 Hospital sample by size, region and sector, 2006-07a

Major cities Outside major cities All
Public Private Public Public Private Public hospitals
contract contract
Very large 53 np np 15 np – 98
Large 21 16 np 16 6 np 70
Medium 14 26 – 31 12 – 83
Small & very small 8 np – 210 np np 257
All hospitals 96 93 15 272 29 3 508
a Hospital location is defined by the Australia Standard Geographical Classification (ABS 2001). Hospital size
is defined by number of casemix-adjusted separations per year, where very large refers to 20 001 or more
casemix-adjusted separations; Large is defined as 10 001 to 20 000 casemix-adjusted separations per year;
medium is defined as 5001 to 10 000 casemix-adjusted separations per year; small is defined as 2001 to 5000
casemix-adjusted separations per year; and very Small is defined as 2000 or fewer casemix-adjusted
separations per year. Sample refers to all the acute hospitals included in the Commission’s multivariate
analysis. np Not published due to confidentiality. – Nil or rounded to zero.
Source: Productivity Commission estimates based on unpublished ABS and AIHW data.

E.4 Variables
This section describes the variables selected for use in the analysis and discusses
some associated sampling issues. Full details of the variables used in the analysis,
including their definitions and summary statistics, are presented at the end of the
section in table E.3.

Drawing on the literature review, variables used in the analysis are grouped as:
· outputs
· quality and patient safety
· inputs
· other factors that describe establishment characteristics, hospital roles and
functions, financial incentives and patient characteristics.

350 PUBLIC AND PRIVATE

HOSPITALS
Outputs

Ideally, a hospital’s performance should be measured in terms of patient outcomes.

Individuals seek hospital services in order to improve their physical and emotional
wellbeing relative to what would otherwise be the case. However, it is not
practicable to directly measure changes to patient health outcomes. Instead, the
approach used here is to measure health outcomes along two dimensions — hospital
outputs and quality of care.

Hospitals are complex entities that provide a wide range of services. This is a strong
argument that hospitals should be modelled as multi-input multi-output firms
(Butler 1995). Hospitals vary significantly in terms of the surgical and medical
procedures they provide. Many provide some sort of outpatient services, emergency
departments and a number provide teaching services while others maintain research
and development programs.

Inpatient services

The Melbourne Institute of Applied Economic and Social Research suggested that a
reasonable compromise would be to model inpatient activity at the major diagnostic
category (MDC) level:
… considering the need to keep model specification parsimonious in empirical
analysis, this approach probably represents a reasonable compromise. (sub. 16, p. 4)

However, a concern is that since there are 23 MDCs, this would represent too many
variables, particularly when more complex functional forms are considered. The
categories of inpatient outputs used in this study are:
· acute separations — casemix-adjusted separations for MDCs 1 to 9, 11 to 13, 16
to 18, 21 and 22)
· pregnancy and neonate separations — casemix-adjusted separations for MDCs
14 and 15
· mental and alcohol separations — casemix-adjusted separations for MDCs 19
and 20
· other separations — casemix-adjusted separations for MDC 23
· endocrine, nutritional and metabolic diseases and disorders — casemix-adjusted
separations for MDC 10. This was the dependent variable for the model.

Pregnancy and neonate MDCs were kept separate from the majority of acute care
separations, as pregnancy and neonates do not generally constitute a disease or

MULTIVARIATE 351
ANALYSIS IN DETAIL
illness. Similarly, mental and alcohol separations were also kept separate because of
concerns over the robustness of measuring cost weights for these categories.

Public hospital cost weights were used for both public and private hospitals. In the
estimation, each of the output categories (except for the last) were normalised by
the dependent variable (MDC 10). All variables were expressed in natural
logarithms, and where a natural number was reported as zero, its corresponding
natural logarithm was changed to zero.

Non-admitted occasions of service

There is no national casemix classification for outpatient services, so there is a

greater need to provide a detailed level of aggregation of these hospital activities
than it is for admitted patient care. The output categories are:
· accident and emergency services — the number of accident and emergency
department presentations or visits
· allied health and other services — the number of occasions of service for allied
health, dental and other outpatient services
· mental and alcohol services — the number of mental, alcohol and psychiatric
outpatient services
· dialysis and endoscopy — the number of occasions of service for dialysis and
endoscopy
· diagnostic services — the number of pathology and radiology services
· outreach services — the number of community services, district nursing and
other outreach services.

Each of these output categories were divided by the reference category. Each output
was expressed in terms of natural logarithms.

The Commission included a binary variable to indicate whether a hospital is a

teaching hospital (‘1’ if it is teaching hospital, ‘0’ otherwise). However, no
distinction was drawn between medical and nursing teaching functions, or the
intensity of the teaching effort. The variable represents all forms of teaching
functions — major and minor.

Given the procedure of normalising hospital outputs, the coefficients of the output
variables on the right-hand side would be expected to take on a positive value.
However, to make interpretation simpler, the dependent variable was multiplied by
minus one to ensure that the right-hand side output variables take on a negative
value. This assists in the interpretation of the coefficients — each of the output
352 PUBLIC AND PRIVATE
HOSPITALS
variables are expected to take a negative value (reflecting the marginal rate of
transformation between the reference category and outputs) and a positive value for
each of the inputs.

Quality and patient safety

A number of variables were available to the Commission to measure hospital

quality, including:
· in-hospital mortality
· infection rates
· adverse events.

As noted in chapter 7, there are limits to both adverse events and hospital infections
data due to under-reporting and the difficulty in attributing the role of hospital in
contributing the cause of those events. As a result, these were not considered in this
analysis of hospital performance, though they will be reconsidered in further work.
Robust data on re-admissions were not available to the Commission.

Drawing on the practice of previous studies, in-hospital mortality rates were used as
a measure of the quality of hospital services. Based on a review of literature into the
standardisation of hospital mortality ratios (Ben-Tovim et al. 2009), the following
variables were included:
· average comorbidity — the average Charlson Index of comorbidity
· distribution of comorbidity — the proportion of hospital separations that were
associated with each of the seven indices of comorbidity (0, 1, 2, 3, 4, 5 and 6 or
more) (Charlson et al. 1987)2
· age — the proportions of patients who are in youngest and oldest age groups
· gender — the proportion of patients who are female
· socioeconomic status — the proportion of patients who reside in areas of the
highest quintiles of socioeconomic disadvantage, as measured by the
Socio-economic index for Areas — Index of Relative Disadvantage and
Advantage (ABS 2008g)

2 The Commission explored the possibility of employing the Multipurpose Australian Comorbidity
Scoring System (Preen et al. 2006) but chose not to use this approach because the data available
to the Commission were neither linked between different hospitals or within the same hospital
over time.
MULTIVARIATE 353
ANALYSIS IN DETAIL
· remoteness of residence — the proportion of patients whose usual place of
residence was in inner regional, outer regional, remote and very remote
communities (as defined by the Australian standard Geographic Classification–
Remoteness Area)
· Indigenous — the proportion of patients that identified themselves as
Indigenous.

Unlike in other studies, no account was made for differences in the procedures
undertaken by hospitals, as these are formally considered in the analysis of hospital
production.

The Tobit regression generates the predicted mortality rates for each hospital. Using
the estimates, risk-adjusted mortality ratios (RAMR) are derived. Lower ratios
indicate lower relative mortalities after adjusting for patient differences. If a
hospital faces a trade off between improving the quality of care and producing
additional services, it is expected that the coefficient of the RAMR variable will be
positive with respect to the output variables.

Inputs

Following common practice in this area of analysis, inputs into the production of
hospital services include:
· nursing staff — number of nursing staff (measured in terms of full-time
equivalents)
· diagnostic staff — number of diagnostic (pathology and radiology) staff
(measured in terms of full-time equivalents)
· other staff — number of domestic, administration and other staff (measured in
terms of full-time equivalents)
· medical and surgical supplies — expenditure on medical and surgical supplies
used
· pharmaceutical supplies — expenditure on pharmaceuticals
· other inputs — expenditure on other hospital inputs, such as administration and
clerical, housekeeping, and repairs and maintenance
· beds — number of beds of the hospital as a proxy for hospital capital. This is
given by the number of beds licensed in a private hospital, and the number of
beds published by the AIHW for public hospitals.

354 PUBLIC AND PRIVATE

HOSPITALS
The total number of beds is not a satisfactory measure of the usage of capital in a
hospital. The number of beds does not adequately reflect the change in capital stock
over time or between hospitals. Ideally, capital measures should be disaggregated
into the main categories of hospital activity — such as the number of ICU beds,
non-acute beds, palliative care beds, the number of sameday chairs, the number of
operating theatres, and so on. Instead, differences in the capital of hospitals were
captured with variables that reflected differences in the roles and functions of
hospitals (discussed below).

Since the number of doctors working in private hospitals is not known, the number
of medical staff has been excluded from the analysis. All efficiency scores derived
from the analysis are to be interpreted as the efficiency of the hospital, and not
specifically of the hospital and the medical workforce.

Each of the coefficients of these variables, for a Cobb-Douglas specification, is

expected to take a positive sign.

Patient-risk characteristics

Although it is posited above that differences in the level of patient risk might be
represented in a measure of quality, it is feasible that patient-risk characteristics
might directly influence the level of hospital output. For example, more morbid
populations may compel hospitals to undertake additional services, to be more
productive with the resources that they have. The patient-risk characteristics
explored here include the same set described in the section quality and patient
safety.

Hospital roles and functions

A number of other variables were included in the analysis to account for the
differences between hospitals in terms of the services they provide, the resources
they use and the patients they treat.

Admissions from an emergency department — the number of accident and

emergency visits divided by the total number of inpatient separations is used as a
proxy for the extent to which emergency patients are admitted hospitals. A number
of commentators have said to the Commission that the presence of an accident and
emergency department can reduce the throughput of inpatient services, particularly
if there are insufficient beds available to accommodate the variability of demand. If
this were the case, then the coefficient on this variable would be negative.

MULTIVARIATE 355
ANALYSIS IN DETAIL
Same-day separations as a share of total separations — a number of study
participants have said to the Commission that private hospitals would appear to be
more technically efficient than public hospitals because the former undertake
relatively more same-day separations. If same-day separations constitute best
practice, and the variable were included in the main model (equation 7), the
coefficient on the variable would be positive. If, on the other hand, same-day
separations permit hospitals to reach best practice, the coefficient on the same-day
separations variable would be positive in the second model (equation 8).

Proportion of patients treated with surgical and other procedures is a variable that
describes the extent to which a hospital specialises in surgical and other DRG cases,
or conversely, the degree to which public hospitals undertake medical DRG cases.
Some participants to this study have argued that a difference between public and
private hospitals is the ability of private hospitals to maximise their productivity by
specialising in elective surgery procedures, which permits them to operate with
higher levels of productivity. On the other hand, public hospitals are unable to
refuse medical treatment, and since medical DRG cases have a greater likelihood of
being unplanned, medical DRGs become inherently more difficult for public
hospitals to manage. Ignoring the differences between surgical and medical cases
has the potential to distort the interpretations of efficiency measures.

As noted earlier, the lack of detailed capital data limits the ability of this type of
analysis to distinguish between hospitals on the basis of their inputs. Instead, a
number of surrogate variables were used to test the extent to which there were such
differences.

Hospital services can also differ in terms of the level of acuity in the services they
provide. For example, hospitals that maintain level III intensive care units have
different resourcing requirements to those than those that maintain residential aged
care units and palliative care units. These three influences are represented with three
binary variables (with ‘1’ indicating that these services or units are provided, ‘0’ if
they are not).

Proportion of patients who are not treated as public patients is a proxy measure for
the different levels of resources used by hospitals to treat public and non-public
patients. It includes patients who are funded by private health insurance,
Department of Veterans’ Affairs, third-party motor vehicle accident, workers’
compensation patients, and self-funding. Public hospitals are funded with capped
budgets, at least when treating public patients. In contrast, the funding of non-public
patients is uncapped. It is possible that differences between capped and uncapped
funding enables hospitals to provide different service levels to public and
non-public patients.

356 PUBLIC AND PRIVATE

HOSPITALS
Evans and Walker indices

The Evans and Walker information indices are measures of the relative complexity
of work undertaken by hospitals. They are based on work undertaken by Thiel
(1967) in the field of information theory. Evans and Walker (1972) postulated a
relationship between the complexity of work undertaken by a hospital and the
information the hospital learns from undertaking that work. By establishing a link
between complexity and information gain, the authors were able to adapt
information indexes as proxies for hospital complexity.

In general, the amount of information a hospital learns from an admission is

inversely related to the likelihood of that case occurring within the system and the
likelihood of that hospital treating that particular case. If an event is almost certain
to take place, such as a routine case from which the hospitals learns little, the
hospital attracts a relatively low index of information gain (Butler 1988). In
contrast, more complex (and presumably rarer cases) attract more information gain.

Evans and Walker offer two indices. They differ in terms of the assumptions about
the prior knowledge of probabilities. The first assumes there is no prior knowledge
of the distribution of cases among hospitals. This is a measure of the complexity of
a hospital’s caseload (Evans and Walker 1972). The index X i1 is given as:

X i1 = ∑ H 1j pij (18)
j

which is a weighted average of the standardised complexity indexes H 1j of each

AR-DRG, where the weights pij are the share of the ith’s hospital’s cases being
classified as the jth AR-DRG.

To derive H 1j , the index of complexity for the jth AR-DRG is used:

⎛ ⎞
⎜q ⎟
H 1j = ∑ qij ln ⎜ ij ⎟ (20)
1
i ⎜ ⎟
⎝ I ⎠

Equation (20) describes the information gain rising from the probability of the jth
AR-DRG being treated by the ith hospital. The smaller the qij, the larger will be its
natural logarithm. Pre-multiplying gives the probability of that information gain
occurring. If in the absence of any information of the actual distribution of cases,
the probability of a case going to any hospital is the same for all hospitals, and is
equal to the inverse of the number of hospitals 1/I.

MULTIVARIATE 357
ANALYSIS IN DETAIL
H 1j is standardised to ensure that the index has a mean of one:

H 1j
H =1
(21)
∑H q
j 1
j j
j

This second measure of a hospital’s relative complexity takes into account the
relative differences in hospital size. In this index, it is assumed that the prior
probability of a case occurring is equal to the actual proportion of all cases in the
system treated by the hospital. This means that the larger the hospital, the higher
will be the probability that it will treat a case entering the system (Butler 1995).
While larger hospitals may treat more complex cases than smaller hospitals, they
are also expected to treat more complex cases.

The second Evans and Walker index X i2 resembles the first, insofar that it is equal
to the weighted average of standardised complexity cases H 2j :

X i2 = ∑ H 2j pij (22)
j

However, the corresponding measure of information gain differs in that it is now

influenced by the probability pi that a case will go to the ith hospital is given by:

⎛ qij ⎞
H 2j = ∑ qij ln ⎜ ⎟ (23)
i ⎝ pi ⎠

As with the first index, equation (23) is standardised to ensure that the index has a
mean of one:

H 2j
H 2j = (24)
∑H
j
2
j qj

What is the Commission measuring?

In this study, the Commission has compared the performance of all acute hospitals
in one sample. That is, all hospitals — large and small, urban and rural — were
compared in a single multivariate analysis. The typical practice in benchmarking is
to identify relevant ‘peers’ against which hospital can be compared. For example,
large metropolitan teaching hospitals are compared against other large metropolitan
teaching hospitals, in order to learn about ways these hospitals might improve their
performance. This practice of stratifying the sample according to key hospital
characteristics, however, is not necessarily useful in an analytical context, because it
358 PUBLIC AND PRIVATE
HOSPITALS
cannot address an important research question: how significant are factors such as
location and size in determining a hospital’s performance? How can the impact of a
hospital’s size or location on efficiency be assessed if hospitals are only compared
with those of the same size or location?

The Commission’s analysis therefore is based on a pooled sample of all hospitals in

the study, as the econometric model is designed to account for differences in
hospitals which would typically be used to define ‘peer groups’. For example, the
inclusion of the explanatory variables measuring hospital size, location, teaching
status and emergency services is designed to control for the effects of these factors
on hospital output and efficiency. Using the stochastic frontier regression technique,
the model can identify variation in hospital output that is specifically due to the
inefficient use of inputs, and not due to differences in a hospital establishment’s
characteristics.

E.5 The results

Before attempting to estimate the technical efficiencies, the Commission undertook
to identify a suitable measure of the quality of hospital care. The approach used here
was to risk-adjust in-hospital mortality rates using a Tobit regression, and then to
include the estimated risk-adjusted mortality ratios (RAMRs) into the estimation of
hospital performance.

Risk-adjustment analysis

As noted earlier, in-hospital mortality is probably the only reliably measured

hospital health outcome. Other measures, such as adverse events and infections, are
generally not well reported. But, mortality rates do not always provide an indication
of the quality of care in a hospital — a number of other factors outside the control
of hospitals (such as the patient’s comorbidities) can contribute to patient mortality.

Three sets of Tobit regressions were analysed. Model 1 considered each of the
major categories of variables — patient comorbidities, socioeconomic status, place
of residence, gender, Indigenous status and age profile. Model 2 excludes gender
and the younger age profiles which appear to be insignificant as a group. It tests
specifically for the effect of place of residence. Model 3 is identical to model 2
apart from replacing the place of residence variables with socioeconomic status of
the patient (table E.4).

MULTIVARIATE 359
ANALYSIS IN DETAIL
Table E.3 Description and summary statistics of variables, 2006-07a
Variable Description Mean Std. dev.

Hospital outputs — Admitted patients

Number of separations Total number of separations 11 549.38 15 382.78
Acute separations Number of casemix-adjusted separations (defined by MDC) 9 526.72 14 786.51
Pregnancy and neonate separations Number of casemix-adjusted separations (defined by MDC) 1 175.80 2 557.90
Mental and alcohol separations Number of casemix-adjusted separations (defined by MDC) 536.67 1 021.36
MDC 10 separations Number of casemix-adjusted separations (defined by MDC) 267.68 427.37
Other separations Number of casemix-adjusted separations (defined by MDC) 526.80 908.15
Average cost weight Ratio 0.8953 0.370
Hospital outputs — Non-admitted patient services
Accident and emergency services Number of occasions of service 11 436.72 16 190.93
Allied health and dental services Number of occasions of service 26 842.13 63 441.71
Mental and alcohol services Number of occasions of service 536.67 1 021.36
Dialysis and endoscopy services Number of occasions of service 158.19 1 457.82
Community outreach and district nursing services Number of occasions of service 7 526.32 22 949.87
Pathology and radiology services Number of occasions of service
Hospital inputs
Nursing staff Number of full-time equivalents 211.96 339.17
Diagnostic staff Number of full-time equivalents 63.55 145.77
Other staff Number of full-time equivalents 141.88 243.78
Total beds Total number 118.03 151.22
Drug costs $’000s 306.47 13.09
Other hospital costs $’000s 1 934.16 11.88
Medical and surgical supplies cost $’000s 606.21 14.58
(Continued next page)

360 PUBLIC AND PRIVATE

HOSPITALS
Table E.3 (continued)
Variable Description Mean Std. dev.

Roles and functions

Teaching Dummy variable (1=yes; 0=otherwise) 0.3209 0.4673
Network Dummy variable (1=yes; 0=otherwise) 0.0846 0.2786
Share of patients that were not public patients Share of total separations 0.3915 0.3804
Surgical and other DRGs Share of total separations 0.3104 0.2387
Same-day separations Share of total separations 0.4585 0.1785
Accident and emergency rate Ratio to total separations 2.1304 2.8140
Transfers to aged care Share of total separations 0.0102 0.0164
Transfers to acute hospitals Share of total separations 0.0687 0.0622
Transfers to other hospitals Share of total separations
Evans and Walker Index 1 Rate 0.5557 0.5241
Evans and Walker Index 2 Rate 0.4904 0.4211
Palliative-care unit Dummy variable (1=yes; 0=otherwise) 0.1122 0.3159
High intensive care unit Dummy variable (1=yes; 0=otherwise) 0.1772 0.3822
Residential care unit Dummy variable (1=yes; 0=otherwise) 0.0039 0.0627
Patient characteristics
Female patients Share of total patients 0.5372 0.0800
Aged less than 1year Share of total patients 0.0191 0.0325
Aged 1-4 years Share of total patients 0.0252 0.0438
Aged 5-14 years Share of total patients 0.0346 0.0526
Aged 50-59 years Share of total patients 0.1334 0.0511
Aged 60-69 years Share of total patients 0.1451 0.0516
Aged 70+ years Share of total patients 0.2866 0.1395
From major city Share of total patients 0.3880 0.4279
From inner regional Share of total patients 0.3187 0.3591
From outer regional Share of total patients 0.2124 0.3250
From remote Share of total patients 0.0376 0.1427
From very remote Share of total patients 0.0433 0.1782
(Continued next page)

361
Table E.3 (continued)
Variable Description Mean Std. dev.
SEIFA 1 Share of total patients 0.3352 0.3464
SEIFA 2 Share of total patients 0.2326 0.2854
SEIFA 3 Share of total patients 0.1819 0.2272
SEIFA 4 Share of total patients 0.1354 0.1851
SEIFA 5 Share of total patients 0.1149 0.1974
Charlson score 1 Share of total patients 0.0819 0.0448
Charlson score 2 Share of total patients 0.0997 0.1014
Charlson score 3 Share of total patients 0.0150 0.0134
Charlson score 4 Share of total patients 0.0133 0.0262
Charlson score 5 Share of total patients 0.0317 0.0523
Charlson score 6 or higher Share of total patients 0.0050 0.0079
Average Charlson score Score 0.5396 0.4481
Quality indicator
Mortality rate Rate 0.0133 0.0310
a Statistics for the minimum and maximum observations were suppressed for confidentiality reasons.

Source: Unpublished ABS and AIHW data; Productivity Commission estimates.

362 PUBLIC AND PRIVATE

HOSPITALS
Between the three models, patient characteristics prove to have a significant
influence on hospital mortality rates. In particular, hospitals which have
proportionally more patients in older age-groups (70 years or older), with higher
Charlson scores (5 or over), and that identify with Indigenous status are expected to
report higher mortality rates. Hospitals’ patient profiles according to patient gender,
usual place of residence, and socio-economic status (the latter based on the SEIFA)
were not found to be significant in most cases (table E.4).

In terms of overall fit (log likelihood) and parsimony of variable choice (Akaike and
Bayesian Information Criteria tests), there is little to separate the three models. The
younger age profiles and gender were generally poor explanators, and so were
dropped from the analysis altogether. The choice between models 2 and 3 is almost
arbitrary. The residuals of the third model were used for the predicted mortality
rates in table E.5.

The predicted and RAMRs are reported in table E.5 for private, public and public
contract hospitals. A RAMR value less than one indicates that a hospital’s actual
mortality rate is less than predicted, given its patient profile, while a value greater
than one indicates the reverse. On average, the private hospitals in this study
reported lower RAMRs than public and public contract hospitals. It is of interest to
note that the RAMRs of public contract hospitals are slightly lower than public
hospitals, with whom they are likely to share a similar pattern of activity. The
RAMRs are further disaggregated in table E.6 according to hospital size.

Care needs to be taken when interpreting RAMRs in relation to hospital quality. For
example, the average RAMR for public hospitals (0.632) does not mean that
patients die at twice the rate than in private hospitals (0.305) (table E.5). The
purpose of the regression is to adjust hospital mortality rates for the profile of
patients they treat. The Tobit regression is only intended to provide an indication of
the extent to which patient-risk characteristics influence hospital mortality rates, and
are not designed to account for the different activities that hospitals undertake (that
is, their casemix). The estimated mortality ratios are then used as a control for
quality in the output regression. Variables to measure a hospital’s casemix are not
included in the mortality rates regression, as they are already included as a direct
component of the output stochastic frontier regression, and inclusion of these
factors in the mortality rates is likely to generate collinearity.

The reported RAMRs should not be compared to other reported mortality measures
(such as Hospital Standardised Mortality Rates, HSMRs).

MULTIVARIATE 363
ANALYSIS IN DETAIL
Table E.4 Results of Tobit regression of mortality rates, 2006-07
Variable Model 1 Model 2 Model 3
Comorbidity
Share of patients with Charlson 6 or more 2.817 *** 2.810 *** 2.829 ***
Share of patients with Charlson 5 0.142 *** 0.146 *** 0.156 ***
Share of patients with Charlson 4 -0.013 -0.011 0.007
Share of patients with Charlson 3 -0.295 *** -0.292 *** -0.296 ***
Share of patients with Charlson 2 -0.027 * -0.027 * -0.023
Average Charlson score -0.007 -0.001 -0.004
Age
Share of patients aged 70 or more 0.049 *** 0.050 *** 0.051 ***
Share of patients aged between 60 and 69 -0.064 ** -0.064 *** -0.066 ***
Share of patients aged between 50 and 59 -0.065 ** -0.061 *** -0.058 **
Share of patients aged between 5 and 14 -0.019
Share of patients aged between 1 and 4 0.026
Share of patients aged under 1 -0.017
Indigenous status 0.008 0.010 *** 0.013 **
Female 0.007
Patient’s usual place of residence
Proportion of patients from inner regional areas 0.005 0.004 *
Proportion of patients from outer regional areas 0.009 ** 0.008 ***
Proportion of patients from remote areas 0.009 0.008
Proportion of patients from very remote areas 0.006 0.005
SEIFA classification of patient’s residence
Proportion of SEIFA 4 0.002 0.001
Proportion of SEIFA 3 0.003 0.005
Proportion of SEIFA 2 0.000 0.005
Proportion of SEIFA 1 0.002 0.008 *
Constant -0.004 0.000 0.001
Sigma 0.018 *** 0.017 *** 0.018 ***
Model criteria
Log likelihood 1 244.94 1 244.33 1 241.81
Likelihood Ratio χ 2
591.10 589.89 584.84
Probability > χ 2
0.0000 0.0000 0.0000
Akaike Information Criterion -2 441.9 -2 456.6 -2 450.5
Bayesian Information Criterion -2 340.3 -2 388.9 -2 382.8
No. of observations 508 508 508
*** Significant at the 1 per cent critical level. ** Significant at the 5 per cent critical level. * Significant at the
10 per cent critical level.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates.

364 PUBLIC AND PRIVATE

HOSPITALS
Table E.5 Predicted mortality rates and risk-adjusted mortality ratios,
by sector, 2006-07
Public hospitals Public contract Private hospitals All hospitals
hospitals
Predicted mortality rates
Mean 0.022 0.030 0.022 0.022
Median 0.019 0.017 0.015 0.018
Standard deviation 0.010 0.040 0.041 0.023
Minimum 0.007 0.012 0.006 0.006
Maximum 0.083 0.186 0.434 0.434
Weighted averagea 0.023 0.032 0.019 0.023
RAMRsb
Mean 0.632 0.540 0.305 0.550
Median 0.593 0.420 0.189 0.517
Standard deviation 0.380 0.563 0.324 0.399
Minimum – 0.074 – –
Maximum 2.793 2.583 1.860 2.793
Weighted averagea 0.530 0.383 0.327 0.471
Number of observations 368 18 122 508
a Weighted average by casemix-adjusted separations. b RAMR – Risk-adjusted mortality ratio is equal to the
actual (observed) mortality rate divided by the predicted mortality rate. – Nil or rounded to zero.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates.

MULTIVARIATE 365
ANALYSIS IN DETAIL
Table E.6 Risk-adjusted mortality ratios, by sector and hospital size,
2006-07a
Very large Large Medium Very small All
and small
Public hospitals
Mean 0.506 0.472 0.532 0.718 0.632
Median 0.506 0.441 0.481 0.685 0.593
Standard deviation 0.195 0.269 0.325 0.431 0.380
Minimum 0.072 – – – –
Maximum 0.889 1.043 1.590 2.793 2.793
Number of observations 68 37 45 218 368
Public contract hospitals
Mean np np np np 0.540
Median np np np np 0.420
Standard deviation np np np np 0.563
Minimum np np np np 0.074
Maximum np np np np 2.583
Number of observations np np np np 18
Private hospitals
Mean 0.357 0.316 0.274 0.297 0.305
Median 0.340 0.236 0.185 0.064 0.189
Standard deviation 0.256 0.267 0.270 0.432 0.324
Minimum – – – – –
Maximum 0.908 0.908 0.908 1.860 1.860
Number of observations 24 22 38 38 122
All hospitals
Mean 0.457 0.432 0.414 0.662 0.550
Median 0.469 0.415 0.330 0.636 0.517
Standard deviation 0.221 0.277 0.390 0.465 0.399
Minimum – – – – –
Maximum 0.908 1.124 1.691 2.793 2.793
Number of observations np np np np 508
a RAMR – Risk-adjusted relative mortality ratio is equal to the actual (observed) mortality rate divided by the
predicted mortality rate. np Not published due to confidentiality concerns. – Nil or rounded to zero.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates

Stochastic frontier analysis

Two distinct types of production (distance) functions were modelled using the
2006-07 data — a Cobb-Douglas and a restricted translog function (as it was not
technically possible to solve the full version of the translog function). The results
for a number of versions of the Cobb-Douglas and a restricted translog are
presented in tables E.7 and E.8.

366 PUBLIC AND PRIVATE

HOSPITALS
Table E.7 Results of Cobb-Douglas stochastic frontier analysis,
2006-07
Model 1 Model 2 Model 3 Model 4 Model 5
Primary model
Inpatient services
Log of acute separations -0.546 *** -0.542 *** -0.488 *** -0.545 *** -0.506 ***
Log of preg. & neonate seps. -0.057 *** -0.014 -0.066 *** -0.059 *** -0.060 ***
Log of mental & alcohol seps. -0.102 *** -0.108 *** -0.121 *** -0.100 *** -0.106 ***
Log of other separations -0.146 *** -0.186 *** -0.146 *** -0.144 *** -0.151 ***
Non-admitted services
Log of emergency dept. visits -0.014 * 0.009 -0.022 -0.011 -0.021
Log of allied & dental services -0.044 *** -0.080 *** -0.048 ** -0.049 *** -0.050 ***
Log of mental & alcohol serv. -0.011 0.001 -0.006 -0.013 -0.011
Log of outreach & dist. nurs. 0.002 -0.005 0.003 0.004 0.004
Log of diagnostic services -0.038 ** -0.013 -0.041 *** -0.037 ** -0.041 ***
Log of dialysis & endoscopy 0.036 0.010 0.057 0.027 0.031
Quality
RAMR -0.035 0.057 0.022 -0.017
Inputs
Log of nursing staff 0.177 *** 0.268 *** 0.187 *** 0.188 *** 0.241 ***
Log of diag. staff 0.033 * 0.092 *** 0.024 0.032 0.030
Log of other staff -0.147 *** -0.006 -0.120 ** -0.140 ** -0.161 ***
Log of beds 0.436 *** 0.681 *** 0.443 *** 0.449 *** 0.462 ***
Log of drugs 0.064 ** -0.036 0.075 *** 0.061 ** 0.068 ***
Log of med.& surg. supplies 0.025 0.142 *** 0.017 0.020 0.015
Log of other inputs -0.016 -0.088 *** -0.018 -0.008 -0.012
Role and functions
Emergency to admission ratio 0.008
Teaching hospital 0.100 0.106 0.097 0.116 *
Level III ICU 0.004 0.013
Palliative care unit -0.026 0.007
Residential care unit -0.187 -0.171
Evans & Walker Index 1 -1.997 *** -2.030 *** -2.015 *** -2.098 ***
Evans & Walker Index 2 3.970 *** 4.029 *** 3.944 *** 4.011 ***
% of seps surgical or other 1.117 *** 0.958 *** 1.152 *** 1.131 ***
% non-public patients -1.089 *** -1.118 *** -1.123 *** -1.160 ***
Patient characteristics
% with Charlson 6 + -0.866 8.763 -1.912 -6.518 **
% with Charlson 5 0.710 2.677 0.825 -1.520 **
% with Charlson 4 -0.584 2.200 -0.712 -2.641 ***
% with Charlson 3 0.717 0.021 0.915
% with Charlson 2 1.025 2.691 ** 1.036
Average Charlson -0.062 -0.729 -0.072 0.394 ***
(Continued next page)

MULTIVARIATE 367
ANALYSIS IN DETAIL
Table E.7 (continued)

Model 1 Model 2 Model 3 Model 4 Model 5

Patient SEIFA
% from SEIFA 4 -0.045 0.429 ** -0.045
% from SEIFA 3 -0.245 * 0.235 -0.251 * -0.307 ***
% from SEIFA 2 -0.228 0.198 -0.233 -0.332 ***
% from SEIFA 1 -0.162 0.314 ** -0.173 -0.261 ***
Patient place of residencea
Major city 0.378 *** 0.137 0.248 **
Outer regional 0.498 *** 0.016 0.114
Remote 0.826 *** -0.066 0.052
Very remote 0.475 -0.078 -0.098
Hospital locationa
Major city -0.078 0.149 **
Outer regional -0.350 *** -0.002
Remote -0.720 *** -0.102
Very remote -0.576 * -0.163
State or territoryb
NSW -0.087 -0.054 -0.083 -0.072 -0.090
Victoria -0.254 *** -0.161 ** -0.224 *** -0.253 *** -0.277 ***
South Australia -0.244 *** 0.104 -0.224 -0.240 *** -0.230 ***
Western Australia -0.077 0.164 -0.037 ** -0.055 -0.069
Tasmania 0.976 *** 0.403 1.050 *** 0.977 *** 1.176 ***
Northern Territory -0.190 0.211 -0.105 -0.190 -0.217
ACT -0.123 -0.161 0.152 -0.129 -0.237
Constant 3.802 *** 2.683 *** 3.521 *** 3.747 *** 3.644 ***
Secondary model
Log σ v
2

Constant -2.664 *** -2.585 *** -2.452 *** -2.654 *** -2.543 ***
Log σ 2
u
Constant -1.915 *** -1.289 *** -1.953 *** -1.866 *** -1.918 ***
Model criteria
No. of observations 508 508 508 508 508
Log likelihood -297.4 -400.6 -319.3 -305.3 -311.3
Wald χ
2
7 969.7 5 215.4 6 997.2 7 663.4 7 345.2
Probability > χ 2 0.0000 0.0000 0.0000 0.0000 0.0000
Akaike Inference Criterion 704.8 885.3 718.6 704.6 700.5
Bayesian Inference Criterion 937.6 1062.9 887.8 903.4 865.5
σv 0.264 0.275 0.295 0.265 0.280
σu 0.384 0.525 0.376 0.393 0.383
σ2 0.217 0.351 0.228 0.225 0.226
λ 1.454 1.911 1.283 1.483 1.367
a Inner regional is the reference region. b Queensland is the reference jurisdiction. *** Significant at the
1 per cent level, ** Significant at the 5 per cent level, * Significant at the 10 per cent level.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates.

368 PUBLIC AND PRIVATE

HOSPITALS
Table E.8 Results of translog stochastic frontier analysis, 2006-07
Model 6 Model 7 Model 8 Model 9 Model 10
Primary model
Inpatient services
Log of acute separations -0.302 -0.265 -0.160 -0.171 -0.211
Log of acute seps – sq -0.015 -0.022 -0.030 -0.029 -0.022
Log of preg. & neonate seps -0.049 *** -0.035 *** -0.049 *** -0.050 *** -0.052 ***
Log of preg. & neon. – sq -0.005 -0.009 *** -0.005 -0.004 -0.004
Log of mental & alc. -0.148 *** -0.167 *** -0.156 *** -0.155 *** -0.151 ***
Log of mental & alc – sq -0.019 *** -0.023 *** -0.020 *** -0.020 *** -0.019 ***
Log of other seps -0.096 *** -0.095 *** -0.087 *** -0.090 *** -0.103 ***
Log of other seps – sq -0.017 *** -0.029 *** -0.017 *** -0.017 *** -0.016 ***
Non-admitted services
Log of ED visits -0.061 -0.099 ** -0.073 * -0.074 * -0.069
Log of ED visits – sq 0.007 0.016 ** 0.006 0.007 0.006
Log of allied & dental 0.128 *** 0.160 *** 0.115 *** 0.117 *** 0.105 ***
Log of allied & denta – sq -0.027 *** -0.035 *** -0.026 *** -0.026 *** -0.024 ***
Log of mental & alc 0.035 * 0.029 0.039 ** 0.042 ** 0.029
Log of mental & alc – sq -0.005 *** -0.001 -0.005 -0.005 ** -0.003
Log of outreach & dist. 0.007 0.005 -0.004 -0.001 0.010
Log of outreach – sq 0.001 0.001 0.002 0.002 0.000
Log of diagnostic -0.034 -0.044 -0.025 -0.029 -0.027
Log of diagnostic – sq 0.003 0.004 0.000 0.001 0.000
Log of dialysis & endoscopy 0.021 -0.013 0.071 0.025 0.014
Log of dial & endo. – sq -0.008 ** -0.018 0.006 -0.009 -0.018
Quality
RAMR -0.256 * -0.050 -0.200 -0.216
RAMR – sq 0.109 0.015 0.087 0.090
Inputs
Log of nursing staff 0.533 *** 0.708 *** 0.664 *** 0.657 *** 0.678 ***
Log of nursing staff - sq -0.051 ** -0.058 ** -0.066 *** -0.066 *** -0.061 ***
Log of diag. staff 0.032 0.087 *** 0.029 0.030 0.036
Log of diag. staff - sq -0.004 -0.009 -0.006 -0.006 -0.003
Log of other staff -0.129 0.028 -0.163 -0.155 -0.152
Log of other staff - sq -0.002 -0.014 0.007 0.007 0.000
Log of beds -0.007 0.030 *** 0.032 0.039 0.075
Log of beds - sq 0.075 *** 0.092 *** 0.071 *** 0.069 *** 0.068 ***
Log of drugs -0.010 -0.073 *** -0.005 -0.004 *** -0.005
Log of drugs - sq 0.011 *** 0.014 *** 0.011 *** 0.011 *** 0.011 ***
Log of med.& surg. supplies 0.290 *** 0.175 ** 0.246 *** 0.238 *** 0.246 ***
Log of med. & surg. - sq -0.027 *** -0.006 -0.023 *** -0.022 *** -0.022 ***
Log of other inputs -0.383 *** -0.423 *** -0.380 *** -0.375 -0.380 ***
Log of other inputs -sq 0.030 *** 0.024 *** 0.029 *** 0.029 *** 0.028 ***
Role and functions
Teaching hospital 0.168 *** 0.176 *** 0.185 *** 0.196 ***
Level III ICU 0.057 0.062
Palliative care unit -0.035 -0.025
(Continued next page)

MULTIVARIATE 369
ANALYSIS IN DETAIL
Table E.8 (continued)

Model 6 Model 7 Model 8 Model 9 Model 10

Residential care unit -0.403 -0.610 *
Evans & Walker Index 1 -2.047 *** -1.732 *** -1.574 *** -1.777 ***
Evans & Walker Index 2 3.658 *** 3.230 *** 3.086 *** 3.246 ***
% non-public patients -0.962 *** -0.998 *** -0.996 *** -0.993 ***
% of seps surg. or other 0.856 *** 0.767 *** 0.761 *** 0.862 ***
Patient characteristics
% with Charlson 6 + -0.374 13.307 ** -7.362 **
% with Charlson 5 1.822 7.850 *** -1.121
% with Charlson 4 0.692 6.384 *** -2.079 **
% with Charlson 3 1.235 4.308
% with Charlson 2 1.424 4.271 ***
Average Charlson -0.362 -1.769 *** 0.250 **
Patient SEIFA
% from SEIFA 4 -0.182 0.080
% from SEIFA 3 -0.216 0.097 -0.216 **
% from SEIFA 2 -0.272 ** 0.037 -0.322 ***
% from SEIFA 1 -0.178 0.167 -0.238 ***
Patient place of residencea
Major city 0.409 *** 0.122 0.190 *** 0.191 ***
Outer regional 0.423 *** -0.008 -0.032 -0.034
Remote 0.837 *** 0.030 -0.147 -0.148
Very remote 0.499 -0.084 -0.182 * -0.188 *
Hospital locationa
Major city -0.070
Outer regional -0.332 ***
Remote -0.757 ***
Very remote -0.637 **
State or territoryb
NSW -0.089 * -0.075 -0.088 -0.079 -0.098
Victoria -0.219 *** -0.137 * -0.203 *** -0.212 *** -0.249 ***
South Australia -0.165 ** 0.089 -0.141 -0.144 * -0.134
Western Australia 0.007 0.207 ** 0.034 0.026 0.009
Tasmania 0.737 ** -0.411 0.937 *** 0.938 *** 1.001 ***
Northern Territory -0.356 ** -0.031 -0.227 -0.229 -0.342 *
ACT -0.171 -0.167 0.054 0.056 -0.253
Constant 3.864 *** 3.578 *** 3.281 *** 3.323 *** 3.318 ***
Secondary model
Log σ v
2

Constant -2.744 *** -2.646 *** -2.500 *** -2.511 *** -2.495 ***
Log σ 2
u
Constant -2.430 *** -1.961 *** -2.553 *** -2.520 *** -2.560 ***
(Continued next page)

370 PUBLIC AND PRIVATE

HOSPITALS
Table E.8 (continued)

Model 6 Model 7 Model 8 Model 9 Model 10

Model criteria
No. of observations 508 508 508 508 508
Log likelihood -305.3 -292.6 -242.9 -244.7 242.2
Wald χ
2
7 663.4 8 637.8 9 650.9 9 584.7 9 830.1
Probability > χ 2 0.0000 0.0000 0.0000 0.0000 0.0000
Akaike Inference Criterion 704.6 705.9 601.8 599.4 596.1
Bayesian Inference Criterion 903.4 959.7 847.2 832.0 833.2
σv 0.254 0.266 0.287 0.285 0.287
σu 0.297 0.375 0.279 0.284 0.278
σ2 0.152 0.212 0.160 0.162 0.160
λ 1.171 1.408 0.973 0.996 0.968
a Inner regional is the reference region. b Queensland is the reference jurisdiction. sq Indicates a squared
term. *** Significant at the 1 per cent level, ** Significant at the 5 per cent level, * Significant at the 10 per cent
level.
Source: Unpublished ABS and AIHW data; Productivity Commission estimates

Models 1 and 6 include all the variables available for each of the functional forms.
Models 2 and 7, respectively, include outputs and inputs as well as the major
patient-risk characteristics (such as Charlson comorbidity scores, SEIFA indices).
They do not include those variables that describe the roles and functions of
hospitals. It is worth noting the high degree of collinearity between these variables
and the RAMR (which includes a number of these variables in its estimation).

Models 3 and 8 include the hospital outputs and inputs, the RAMR and all the
variables describing hospital roles and functions and hospital location. It is worth
observing that dummy variables indicating the presence of intensive care, palliative
care and residential aged care units were not significant. The coefficients for both
Evans and Walker indices confirm that the complexity of hospital services is a
determinant of the dependent variable. Models 4 and 9 are similar to models 3 and 8
but with selected hospital roles and function variables excluded.

In models 5 and 10, the RAMR is replaced by the patient-risk characteristics. Not
all of the Charlson and SEIFA variables were included, as collinearity was evident
within members of each set. Models 5 and 10 reflect the synthesis of models 3 and
4, and 8 and 9 respectively. The Akaike and Bayesian information criteria tests
indicate that models 5 and 10 are to be preferred, followed by models 4 and 9, for
the Cobb-Douglas and restricted translog functions respectively.

MULTIVARIATE 371
ANALYSIS IN DETAIL
In interpreting the coefficients (from models 5 and 10), the following observations
can be made:
· Hospitals that treat relatively more comorbid partients (Charlson index) and
patients from more disadvantage areas (SEIFA index) have lower frontiers (best-
practice benchmarks).
· Hospitals that treat relatively more non-public patients (that is, patients who
elect to be funded by private health insurance, the Department of Veterans’
Affairs, third-party motor vehicle accident schemes or are self-funded) tend to
have lower frontiers. This may reflect the additional resources employed by
hospitals to treat these patients.
· Hospitals that undertake relatively more surgical and other procedures (as
opposed to medical procedures) tend to have higher frontiers. This may be
because that medical procedures are inherently more difficult to manage,
possibly because of their relatively unplanned nature.
· The coefficients for Victoria and Tasmania remain relatively significant in all
specifications. This is likely to reflect the effects of having to disaggregate the
data for these jurisdictions from a single public hospital observation.

Other variables, such as average length of stay and the proportion of same-day
separations, were not considered in the analysis because shorter lengths of stay and
higher turnover of patients is reflected in the greater level of inpatient separations.

Efficiency results

Efficiency results are presented in tables E.9 to E.11 for models 4, 5, 9 and 10. After
taking into account the various factors that influence their performance, the average
efficiency of all hospitals was broadly similar. The mean technical efficiencies
across the major hospital categories (public, private, public contract) were between
0.75 to 0.80 (models 9 and 10) (table E.9). The median efficiencies across the same
categories ranged between 0.81 and 0.83 (model 9), and between 0.81 and 0.84
(model 10), suggesting a degree of skewness in efficiency scores (table E.9).

The use of the translog functional form is intended to ‘net out’ the effects of scale
economies, although using the mean efficiency scores, it is possible to discern
differences in the technical efficiencies of hospitals of different size. For example,
the mean technical efficiency score was about 0.766 for the smallest hospitals
(table E.11, model 10) and 0.814 for very large hospitals (table E.10, model 10).

The median is a better measure of central tendency than the mean, given the
skewness in the data. There is a perceptible difference between the major hospital

372 PUBLIC AND PRIVATE

HOSPITALS
categories. The median technical efficiency score for large and very large private
hospitals was slightly higher than for public hospitals, except for large hospitals in
model 9 (table E.10). For example, the median technical efficiency of large and
very large private hospitals was 0.829 and 0.851 respectively, compared to 0.812
and 0.820 for public hospitals of the same size (model 10, table E.10).

Table E.9 Technical efficiency scores, all hospitals, 2006-07a

Model 4 Model 5 Model 9 Model 10
All hospitals
Mean 0.729 0.732 0.783 0.786
Median 0.772 0.776 0.814 0.816
5th percentile 0.366 0.408 0.501 0.553
95th percentile 0.895 0.893 0.904 0.906
No. of observations 508 508 508 508
Public hospitals
Mean 0.743 0.746 0.794 0.797
Median 0.771 0.774 0.814 0.816
5th percentile 0.478 0.503 0.627 0.643
95th percentile 0.890 0.886 0.902 0.901
No. of observations 368 368 368 368
Private hospitals
Mean 0.677 0.680 0.746 0.750
Median 0.771 0.785 0.817 0.822
5th percentile 0.092 0.105 0.311 0.313
95th percentile 0.905 0.905 0.913 0.916
No. of observations 122 122 122 122
For-profit hospitals
Mean 0.663 0.667 0.749 0.751
Median 0.765 0.777 0.816 0.818
5th percentile 0.062 0.079 0.311 0.313
95th percentile 0.911 0.909 0.918 0.917
No. of observations 94 94 94 94
Not-for-profit hospitals
Mean 0.721 0.722 0.736 0.747
Median 0.797 0.796 0.828 0.838
5th percentile 0.110 0.136 0.175 0.203
95th percentile 0.880 0.888 0.898 0.906
No. of observations 28 28 28 28
Public contract hospitals
Mean 0.791 0.787 0.801 0.800
Median 0.826 0.814 0.818 0.805
5th percentile 0.511 0.523 0.580 0.583
95th percentile 0.911 0.906 0.907 0.908
No. of observations 18 18 18 18
a Results based on models 4 and 5 (Cobb-Douglas) and models 9 and 10 Logarithmic quadratic. The 5% and
95% percentile values are equivalent to the minimum and maximum scores after removing for the outliers in
the estimated distribution.

MULTIVARIATE 373
ANALYSIS IN DETAIL
Table E.10 Technical efficiency scores, large and very large hospitals,
2006-07a
Large hospitals Very large hospitals
Model 4 Model 5 Model 9 Model 10 Model 4 Model 5 Model 9 Model 10
All hospitals
Mean 0.776 0.776 0.807 0.809 0.761 0.763 0.814 0.814
Median 0.793 0.799 0.827 0.828 0.765 0.770 0.814 0.827
5th percentile 0.565 0.556 0.656 0.644 0.587 0.585 0.715 0.683
95th percentile 0.892 0.891 0.907 0.908 0.895 0.890 0.908 0.905
No. of obs. 70 70 70 70 98 98 98 98
Public hospitals
Mean 0.763 0.764 0.808 0.810 0.750 0.754 0.811 0.813
Median 0.785 0.773 0.826 0.812 0.756 0.762 0.810 0.820
5th percentile 0.567 0.581 0.668 0.648 0.557 0.585 0.729 0.708
95th percentile 0.886 0.891 0.917 0.917 0.895 0.893 0.908 0.905
No. of obs. 37 37 37 68 68 68 68
Private hospitals
Mean 0.788 0.789 0.810 0.813 0.785 0.784 0.823 0.819
Median 0.827 0.827 0.824 0.829 0.805 0.811 0.850 0.851
5th percentile 0.662 0.644 0.751 0.752 0.647 0.645 0.670 0.655
95th percentile 0.881 0.878 0.887 0.878 0.885 0.879 0.894 0.893
No. of obs. 22 22 22 22 24 24 24 24
For-profit hospitals
Mean 0.780 0.784 0.808 0.810 0.793 0.793 0.839 0.834
Median 0.823 0.823 0.824 0.828 0.827 0.821 0.858 0.863
5th percentile 0.565 0.558 0.465 0.457 0.587 0.586 0.689 0.659
95th percentile 0.892 0.889 0.920 0.918 0.893 0.891 0.918 0.917
No. of obs. 15 15 15 15 15 15 15 15
Not-for-profit hospitals
Mean 0.807 0.799 0.815 0.819 0.772 0.770 0.797 0.795
Median 0.831 0.832 0.830 0.830 0.801 0.795 0.825 0.846
5th percentile 0.739 0.699 0.751 0.757 0.647 0.651 0.643 0.639
95th percentile 0.851 0.850 0.858 0.868 0.852 0.847 0.876 0.877
No. of obs. 7 7 7 7 9 9 9 9
a Results based on models 4 and 5 (Cobb-Douglas) and models 9 and 10 Logarithmic quadratic. The 5% and
95% percentile values are equivalent to the minimum and maximum scores after removing for the outliers in
the estimated distribution.

374 PUBLIC AND PRIVATE

HOSPITALS
Table E.11 Technical efficiency scores, small and very small, and
medium hospitals, 2006-07a
Small and very small hospitals Medium hospitals
Model 4 Model 5 Model 9 Model 10 Model 4 Model 5 Model 9 Model 10
All hospitals
Mean 0.701 0.708 0.762 0.766 0.738 0.746 0.791 0.741
Median 0.771 0.766 0.808 0.806 0.765 0.777 0.819 0.780
5th percentile 0.173 0.202 0.378 0.415 0.506 0.499 0.557 0.491
95th percentile 0.890 0.879 0.902 0.899 0.913 0.915 0.915 0.915
No. of obs. 257 257 257 257 83 83 83 83
Public hospitals
Mean 0.737 0.739 0.786 0.788 0.748 0.754 0.797 0.803
Median 0.776 0.781 0.815 0.816 0.762 0.760 0.812 0.815
5th percentile 0.404 0.408 0.556 0.575 0.528 0.558 0.607 0.622
95th percentile 0.889 0.880 0.895 0.897 0.904 0.901 0.915 0.907
No. of obs. 218 218 218 218 45 45 45 45
Private hospitals
Mean 0.495 0.505 0.622 0.641 0.725 0.727 0.785 0.780
Median 0.601 0.605 0.694 0.700 0.798 0.803 0.838 0.841
5th percentile 0.046 0.063 0.175 0.203 0.103 0.118 0.448 0.427
95th percentile 0.907 0.909 0.916 0.919 0.925 0.926 0.928 0.931
No. of obs. 38 38 38 38 38 38 38 38
For-profit hospitals
Mean 0.471 0.480 0.628 0.640 0.732 0.734 0.797 0.791
Median 0.607 0.605 0.716 0.715 0.795 0.802 0.826 0.820
5th percentile 0.046 0.063 0.186 0.208 0.178 0.205 0.503 0.470
95th percentile 0.907 0.908 0.916 0.916 0.925 0.926 0.928 0.931
No. of obs. 31 31 31 31 33 33 33 33
Not-for-profit
hospitals
Mean 0.598 0.614 0.597 0.642 0.679 0.682 0.707 0.707
Median 0.596 0.606 0.634 0.644 0.878 0.883 0.874 0.876
5th percentile 0.110 0.136 0.175 0.203 0.028 0.031 0.029 0.029
95th percentile 0.875 0.875 0.913 0.919 0.905 0.908 0.898 0.906
No. of obs. 7 7 7 7 5 5 5 5
a Results based on models 4 and 5 (Cobb-Douglas) and models 9 and 10 Logarithmic quadratic. The 5% and
95% percentile values are equivalent to the minimum and maximum scores after removing for the outliers in
the estimated distribution.

These differences in the means and medians are relatively small, particularly when
it is recognised that there are significant variations within each group of hospitals.
For example, the range between the 5th and 95th percentile for very large private
hospitals is 0.655 and 0.893 (model 10, table E.10). This implies that the differences
in the means between very large public and private hospitals may be negligible.
That said, in terms of median scores, the relative rankings between public and
private hospitals remained the same, regardless of the functional form

MULTIVARIATE 375
ANALYSIS IN DETAIL
(Cobb-Douglas and restricted translog) and choice of variables, with the exception
of large hospitals in model 9 (table E.10).

In contrast, the median efficiencies of very small and small private hospitals were
lower than for public hospitals (for example, the efficiency scores of very small and
small private hospitals efficiency was 0.700 compared to 0.816 for public hospitals,
for model 10, table E.11). The greater dispersion of efficiency among small and
very small private hospitals, for example with efficiencies between 0.203 and 0.919
in model 10 (compared with public hospitals 0.575 to 0.897) suggests a degree of
variability that has not been adequately captured in the model.

Finally, some correlation statistics were calculated for three variables of interest on
the efficiency scores (table E.12). Occupancy rates were positively correlated with
efficiency scores for all hospitals, public and private, and to some extent, public
contract hospitals. Average length of stay (ALOS) is an important contributor to
private hospital efficiency — hospitals with higher ALOS were less efficient.
Finally complexity, as measured by cost weights, indicated that public hospitals
with the higher cost weights were more efficient, while the private hospitals with
the lower cost weights were more efficient.

Table E.12 Correlation coefficients between selected variables and

technical efficiency scores
Model 4 Model 5 Model 9 Model 10
Occupancy rate
All hospitals 0.1616* 0.3222* 0.2162* 0.3361*
Publics 0.2264* 0.2333* 0.2386* 0.2555*
Privates 0.1855* 0.1964* 0.2899* 0.3051*
Public contract 0.3517 0.3279 0.5356* 0.4973*
ALOS
All hospitals -0.2749* -0.3530* -0.2411* -0.3913*
Publics 0.0981 0.0752 0.0725 0.0920
Privates -0.4564* -0.4478* -0.3954* -0.3606*
Public contract 0.1809 0.1762 0.3313 0.3720
Cost weight
All hospitals -0.0800 -0.2390* -0.0186 -0.1890*
Publics 0.2151* 0.1799* 0.2236* 0.2344*
Privates -0.3242* -0.3254* -0.2005* -0.1847*
Public contract -0.0934 -0.1333 0.0186 0.0924
*Significant at the 5 per cent level.

Source: Unpublished ABS and AIHW data; Productivity Commission estimates

376 PUBLIC AND PRIVATE

HOSPITALS
E.6 Proposed future analysis
Given the data delays faced by the Commission, the multivariate analysis presented
in this report estimates hospital production functions and technical efficiency based
on a single year of data (2006-07). Given the large number of hospital observations
in this data set, the results are expected to be robust.

Nevertheless, the Commission intends over coming months to replicate this analysis
using a larger data set that includes data from the earlier years of 2003-04 to
2005-06. Future analysis will also focus on examining the performance of hospitals
for different peer groups (say, to compare the performance of very large hospitals).
The Commission will also extend this analysis to examine the determinants of
hospital costs.

The Commission intends to publish the results from this further analysis in
March 2010.

MULTIVARIATE 377
ANALYSIS IN DETAIL
378 PUBLIC AND PRIVATE
HOSPITALS
F State-level data on hospital-acquired
infections

Government monitoring of hospital-acquired infections is largely undertaken by

state governments, reflecting their role as providers of public hospitals and
regulators of private hospitals. Such monitoring is not done on a
nationally-consistent basis, but public and private hospitals are included in most
cases.

New South Wales is the only state with a dedicated infection surveillance program
under which the data reported to government are limited to public hospitals.1 The
Northern Territory Government also confines its infection monitoring to public
hospitals.2 The ACT Government collects data from just one private hospital and
two public hospitals, and so it would not be possible to maintain confidentiality in a
public-private comparison.3 Nevertheless, it is likely that hospitals that are not
required to report data to governments would still monitor their infection rates and
participate in voluntary cross-hospital reporting programs, such as the Clinical
Indicator Program (CIP) managed by the Australian Council on Healthcare
Standards.

The Commission did not request infections data from individual hospitals (or
groups of hospitals managed by the same entity) because it would be difficult to
maintain confidentiality, and the collection methods and definitions may not be

1 The NSW Government’s infection surveillance program is mandatory for public hospitals
(NSW Department of Health 2005). Reported data include surgical-site infections following
selected procedures, central-line associated bloodstream infections, Staphylococcus aureus
bacteraemia, and methicillin-resistant Staphylococcus aureus cases in intensive-care units
(NSW Department of Health 2008). Private hospitals are encouraged to use the same indicator
framework and should report data to their infection control and/or quality committee, and
medical advisory committee.
2 NT public hospitals submit infections data to the Clinical Indicator Program, which is managed
by the Australian Council on Healthcare Standards.
3 The ACT Government routinely collects data on bloodstream infections, and surgical-site
infections associated with selected procedures (joint arthroplasty, cardiac surgery and caesarean
sections) (Bull et al. 2008).
STATE-LEVEL 379
INFECTIONS DATA
comparable with other hospitals (or groups). Data was not requested from the CIP,
given the limitations with that information source (discussed in chapter 6).

F.1 Victoria
There are two potential sources of infections data in Victoria:
• Victorian Nosocomial Infection Surveillance System (VICNISS)
• Victorian Admitted Episodes Dataset (VAED).

Victorian Nosocomial Infection Surveillance System

VICNISS was established in 2002 and is funded by the Victorian Government to

monitor infections in public hospitals. The VICNISS Coordinating Centre collects
and analyses data from individual hospitals, and reports quarterly to participating
hospitals and the Victorian Department of Health.

All public hospitals report to VICNISS. Private hospitals recently expressed an

interest in also participating, and this has so far led to seven private hospitals
reporting data. The Commission understands that all of these are larger private
hospitals (more than 100 beds). It is expected that this development will ‘eventually
enable a comprehensive data collection of surgical procedures in Victoria and allow
comparisons between all hospitals, both public and private’ (Victorian Department
of Human Services 2008b, p. iv).

VICNISS surveillance methods differ according to hospital size. Hospitals with

100 or more beds (type-1 hospitals) are subject to three components based on the
US NNIS system. These components are the surveillance of surgical-site infections
(SSIs), intensive-care units (ICUs) and neonatal ICUs. VICNISS uses the
NHSN/NNIS risk index (described in box 6.1) to risk adjust SSI rates, although the
methodology is modified for operations where the use of a laparoscope influences
the risk of developing an SSI (for example, appendectomy and cholecystectomy)
(Victorian Department of Human Services 2008b). The most recent published data
for SSIs are shown in table F.1.

Surveillance for hospitals with fewer than 100 beds (type-2 hospitals) involves
monitoring processes that have been demonstrated to affect outcomes and, for
hospitals with high surgical throughput, reporting selected infection rates. While the
prevalence of MRSA is not reported for type-1 hospitals, it is one of the infection
rates that type-2 hospitals can report. The VICNISS Coordinating Centre stratifies
the type-2 hospital data into small hospitals (1–14 acute beds), medium hospitals
380 PUBLIC AND PRIVATE
HOSPITALS
(15–49 acute beds) and large hospitals (50–99 acute beds) and reports infections per
occupied bed day. Data for type-2 hospitals on rates of methicillin-resistant
Staphylococcus aureus (MRSA) and bloodstream infections (BSIs) are shown in
table F.2.

Table F.1 SSI rates for Victorian public hospitals by procedure and
risk category, 2007a
Risk categoryb
0 1 2 3
Coronary artery bypass – 1.0 (0.5–1.7) 1.6 (0.7–3.2) –
grafts, deep and organ
space
Colon surgery – 4.9 (2.5–8.7) 9.5 (6.3–13.6) 11.2 (6.3–18.1)
Caesarean section 1.5 (1.1–1.9) 1.3 (0.6–2.5) – –
Hip arthroplasty deep 0.9 (0.4–1.7) 1.9 (1.3–2.7) – –
and organ space
Knee arthroplasty deep 1.3 (0.6–2.4) 0.8 (0.3–1.6) – –
and organ space
a Hospitals with 100 or more acute beds (VICNISS type-1 hospitals). SSI rates are expressed in terms of
infections per 100 procedures. b Risk categories are based on the NHSN/NNIS risk index for SSIs (detailed in
box 6.1). Numbers in parentheses are 95 per cent confidence intervals. – Nil or rounded to zero.
Source: Victorian Department of Health (unpublished VICNISS data).

Table F.2 MRSA and BSI rates for Victorian public hospitals by
hospital size, 2004–2007a
Hospital size MRSAb BSIsb
1–14 acute beds 0.5 (0.2–0.8) 0.1 (0.0–0.3)
15–49 acute beds 0.4 (0.3–0.7) 0.3 (0.1–0.4)
50–99 acute beds 1.0 (0.8–1.4) 0.7 (0.5–0.9)
Total 0.7 (0.5–0.8) 0.4 (0.3–0.5)
a Hospitals with fewer than 100 acute beds (VICNISS type-2 hospitals). b Infection rates are expressed as
infections per 10 000 occupied bed days. Numbers in parentheses are 95 per cent confidence intervals.
Source: Victorian Department of Human Services (2008b).

Victorian Admitted Episodes Dataset

The VAED contains data on all episodes of care for admitted patients in public and
private hospitals in Victoria. Hospitals are required to provide these data to the
Victorian Department of Health.

The Victorian Government provides data from the VAED to a national database —
the National Hospital Morbidity Database (NHMD) managed by the Australian

STATE-LEVEL 381
INFECTIONS DATA
Institute of Health and Welfare (AIHW) — as part of its healthcare agreement with
the Australian Government. Other jurisdictions have similar arrangements with the
Australian Government, and national coding standards have been established to
ensure data are reported consistently. However, Victoria has supplementary coding
standards to gather extra information for its own purposes beyond what is required
at the national level. This includes a prefix on diagnosis codes that can, among other
things, be used to identify conditions that arose during an episode of care. This
prefix has been used for many years in Victoria, and will be utilised by the
Victorian Department of Health to derive the condition-onset flag recently
mandated for the NHMD.

The Commission obtained data from the Victorian Department of Health that uses
the VAED condition-onset prefix, in combination with codes for specific infection
organisms, to identify hospital-acquired cases of MRSA and vancomycin-resistant
enterococci (VRE).4 These data may slightly understate the number of infections
for technical reasons associated with the coding of diagnoses.5 To test this, the
Commission compared public-hospital MRSA data from the VAED with that
reported by VICNISS for type-2 hospitals. As expected, the MRSA infection rate
was slightly lower using VAED data (table F.3).

The VAED data show that private hospitals had lower rates of hospital-acquired
MRSA and VRE than public hospitals between 2005-06 and 2007-08 (figures F.1
and F.2). This pattern was also evident when the data were stratified by region and
whether the patient spent time in an ICU (tables F.4 and F.5). Between 2005-06 and
2007-08, the infection rates for both MRSA and VRE in both public and private
hospitals were greater in metropolitan hospitals than in rural hospitals. This may
reflect the fact that metropolitan hospitals are more likely to treat complex cases
with a greater risk of infection.

4 Cadwallader et al. (2001) also used data from hospital medical records to identify infections.
They found that this approach was comparable to an infection-surveillance program in
identifying SSIs following orthopaedic surgery in a WA teaching hospital in the late 1990s.
More recently, Jackson, Michel, Roberts, Jorm and Wakefield (2009) have developed and
validated a method for using data from hospital medical records that include a condition-onset
flag to identify and classify hospital-acquired diagnoses (including hospital-acquired infections).
5 Data were derived from the VAED by identifying cases that had a C-prefix diagnosis for
Staphylococcus aureus (ICD-10-AM code B95.6) or Streptococcus group D (B95.2), combined
with a code for methicillin-resistant agent (Z06.32) or vancomycin-resistant agent (Z06.41).
This might exclude some Staphylococcus aureus and Streptococcus group D infections that are
identified by a combined ‘infection site and organism code’ specifying both (a) that there is an
infection and (b) the organism is Staphylococcus aureus or Streptococcus group D.
382 PUBLIC AND PRIVATE
HOSPITALS
Table F.3 Comparison of VAED and VICNISS data for public-hospital
MRSA infections
VAEDa VICNISSb

2005–2008c 2004–2007d
No. of MRSA infections 101 82
Acute occupied bed days 1 736 866 1 226 952
Infection rate (per 10 000 acute occupied bed days) 0.58 0.67
a All public hospitals. MRSA infections were derived from the VAED by counting separations with diagnosis
codes indicating a ‘C-prefix’ infection for B95.6 (Staphylococcus aureus as the cause of diseases classified to
other chapters) and Z06.32 (methicillin-resistant agent) indicating the Staphylococcus aureus infection is
methicillin resistant. b Public hospitals with fewer than 100 beds (VICNISS type-2 hospitals). c 1 July 2005 to
30 June 2008. d 1 May 2004 to 31 December 2007.
Source: Victorian Department of Health (unpublished VAED data); Victorian Department of Human
Services (2008b).

Figure F.1 Hospital-acquired MRSA infections in Victoria by sector,

2005–2008a

0.7
Infection rate per 10 000
occupied bed days

0.6
0.5
0.4
0.3
0.2
0.1
0.0
Public Private
Hospital type

2005-06 2006-07 2007-08

a Excluding same-day separations. MRSA infections derived from the VAED only include separations that had
a diagnosis code indicating a ‘C-prefix’ infection for B95.6 (Staphylococcus aureus as the cause of diseases
classified to other chapters) and Z06.32 (methicillin-resistant agent) indicating the Staphylococcus aureus
infection is methicillin resistant. This excludes Staphylococcus aureus infections identified by a combined
‘infection site and organism code’ specifying both (a) that there is an infection, and (b) the organism is
Staphylococcus aureus. As a result, the number of MRSA infections may be underestimated.
Source: Victorian Department of Health (unpublished VAED data).

STATE-LEVEL 383
INFECTIONS DATA
Figure F.2 Hospital-acquired VRE infections in Victoria by sector,
2005–2008a

0.20
Infection rate per 10 000
occupied bed days

0.15

0.10

0.05

0.00
Public Private
Hospital type

2005-06 2006-07 2007-08

a Excluding same-day separations. VRE infections derived from the VAED only include separations that had
a diagnosis code indicating a ‘C-prefix’ infection for B95.2 (Streptococcus group D, as the cause of diseases
classified to other chapters) and Z06.41 (vancomycin-resistant agent) indicating the Enterococci or Group D
Streptococci infection is vancomycin resistant. This excludes Group D Streptococcus infections identified by a
combined ‘infection site and organism code’ specifying both (a) that there is an infection, and (b) the organism
is Group D Streptococcus. As a result, the number of VRE infections may be underestimated. There were no
VRE infections in any rural private hospitals between 2005-06 and 2007-08.
Source: Victorian Department of Health (unpublished VAED data).

Table F.4 Hospital-acquired MRSA infections in Victoria by region

and ICU status, 2005-06 to 2007-08
Public Private

Metropolitan Rural Metropolitan Rural

No. of MRSA infectionsa 523 159 154 16
No. of ICU MRSA infectionsb 184 38 53 4
Acute occupied bed daysc 7 979 017 3 262 177 3 907 202 592 865
Infection rate (per 10 000 acute
occupied bed days) 0.66 0.49 0.39 0.27
ICU infection rate (per 10 000 acute
occupied bed days)d 0.23 0.12 0.14 0.07
a Excluding same-day separations. MRSA infections derived from the VAED only include separations that had
a diagnosis code indicating a ‘C-prefix’ infection for B95.6 (Staphylococcus aureus as the cause of diseases
classified to other chapters) and Z06.32 (Methicillin-resistant agent) indicating the Staphylococcus aureus
infection is methicillin resistant. This excludes Staphylococcus aureus infections identified by a combined
‘infection site and organism code’ specifying both (a) that there is an infection, and (b) the organism is
Staphylococcus aureus. As a result, the number of MRSA infections may be underestimated. b This includes
all patients who had hospital acquired MRSA infections and spent time in an ICU. c Excludes same-day
separations. d Separations where the patient spent at least part of the episode in an ICU.
Source: Victorian Department of Health (unpublished VAED data).

384 PUBLIC AND PRIVATE

HOSPITALS
Table F.5 Hospital-acquired VRE infections in Victoria by region and
ICU status, 2005-06 to 2007-08
Public Private

Metropolitan Rural Metropolitan Rural

No. of VRE infectionsa 155 8 12 –
No. of ICU VRE infectionsb 65 2 6 –
Acute occupied bed daysc 7 979 017 3 262 177 3 907 202 592 865
Infection rate (per 10 000 acute
occupied bed days) 0.19 0.02 0.03 –
ICU infection rate (per 10 000 acute
occupied bed days)d 0.08 0.01 0.02 –
a Excluding same-day separations. VRE infections derived from the VAED only include separations that had a
diagnosis code indicating a ‘C-prefix’ infection for B95.2 (Streptococcus group D, as the cause of diseases
classified to other chapters) and Z06.41 (Vancomycin-resistant agent) indicating the Enterococci or Group D
Streptococci infection is Vancomycin resistant. This excludes Group D Streptococcus infections identified by a
combined ‘infection site and organism code’ specifying both (a) that there is an infection, and (b) the organism
is Group D Streptococcus. As a result, the number of VRE infections may be underestimated. b This includes
all patients who had hospital acquired VRE infections and spent time in an ICU. c Excludes same-day
separations. d Separations where the patient spent at least part of the episode in an ICU unit. – Nil or rounded
to zero.
Source: Victorian Department of Health (unpublished VAED data).

F.2 Queensland
There are two key sources of infections data in Queensland:
• Health Quality and Complaints Commission (HQCC)
• Centre for Healthcare Related Infection Surveillance and Prevention (CHRISP).

Health Quality and Complaints Commission

The HQCC was established in July 2006 as an independent body to monitor and
improve the quality of health services in Queensland, and to manage health
complaints. It introduced standards for healthcare providers in July 2007, with a
staged approach to implementation (HQCC 2009). Acute hospitals and day
surgeries were the first group required to report their compliance with the standards,
and first reports were submitted to the HQCC in October 2007. The reporting of
infections data began in March 2008.

Hospitals have the option to advise the HQCC that they are unable to provide data,
although the HQCC advised the Productivity Commission that this has become less
of a problem over time. The HQCC provided this study with unpublished data it had
collected on SSIs and SAB BSIs for the six-month period from July to

STATE-LEVEL 385
INFECTIONS DATA
December 2008. This period has the most complete set of infections data collected
by the HQCC to date. The data show that average infection rates were lower in
private hospitals (table F.6), but this result needs to be highly qualified. The HQCC
cautioned that the data have a number of limitations because:
• the data are not risk adjusted
• the reporting arrangements are designed to enable healthcare providers to
measure their own quality improvements over time, rather than compare
themselves with other providers
• responsibility for data accuracy rests with reporting healthcare providers, as the
HQCC does not have a systematic process to verify all submitted data
• different healthcare providers employ different sampling methods and sizes.
These may not have been randomised or be representative of the provider’s
casemix. The HQCC has published guidance on appropriate sample sizes but
these have not always been followed. In addition, the Productivity Commission
understands that not all providers advise the HQCC about the methodology they
use
• providers have employed a mix of medical chart, observational and
administrative data audits to obtain the data
• differences in the casemix of individual providers may result in different
infection rates. Casemix differences are particularly relevant when comparing
the public and private sectors.

Table F.6 Selected hospital-acquired infections in Queensland,

July–December 2008a
Staphylococcus aureus bacteraemia Surgical-site infections
Infection rate Infection rate
Reporting Averageb Inter- Reporting Averageb Inter-
hospitals quartile hospitals quartile
Sector rangec rangec
no. per per no. per 100 per 100
100 000 100 000 surgical surgical
occupied occupied patients patients
bed days bed days
Public 103 8.27 0–0.89 37 2.30 0–1.95
Private 53 6.03 0–5.55 36 0.26 0–0.86
Total 156 7.41 0–3.50 73 0.76 0–2.33
a Excludes same-day facilities. b Aggregated average calculated by dividing the total number of infections
across all reporting hospitals by the total number of occupied bed days/surgical patients across all reporting
hospitals. c The range between the first and third quartiles.
Source: HQCC (unpublished data).

386 PUBLIC AND PRIVATE

HOSPITALS
Centre for Healthcare Related Infection Surveillance and Prevention

Twenty-four public hospitals in Queensland voluntarily submit surveillance data to

CHRISP, which is part of the Queensland Department of Health. Private hospitals
do not submit data to CHRISP.

The data aggregated and analysed by CHRISP include inpatient SSIs for
16 indicator procedures, healthcare-associated BSIs including SAB, and significant
organisms including MRSA and Clostridium difficile (CHRISP 2009; Queensland
Health, sub. 27). SAB data are collected for inpatients and non-inpatients, and can
be stratified into three hospital types based on the services that they provide. This is
a new classification system based on work undertaken by CHRISP that showed a
correlation between BSIs and particular services (Tong et al. 2009).

Definitions used by CHRISP are based on the Health Care Associated Infection
Surveillance Definitions from the Australian Infection Control Association (AICA)
and the ACSQHC. The risk-adjustment method used for SSIs is based on that
developed in the United States by the US National Healthcare Safety Network
(NHSN) (formerly the US National Nosocomial Infections Surveillance — NNIS).
Recent CHRISP data for SSIs are shown in table F.7.

CHRISP provides feedback to individual hospitals in six-monthly reports that

compare the hospital’s infection rates with statewide control limits. The control
limits are based on statewide averages for the relevant hospital type, and the
methodology is based on funnel plots (Spiegelhalter 2004).

STATE-LEVEL 387
INFECTIONS DATA
Table F.7 SSI rates for Queensland public hospitals by surgical
procedure, 2004–2008a
Risk categoryb
Surgical procedure 0 1 2
Total hip replacement 0.71 (0.45–1.07) 0.97 (0.58–1.51) 2.80 (1.35–5.09)
Revision total hip
replacement 3.10 (1.49–5.62) 3.36 (1.62–6.08) 8.33 (2.29–19.98)
Total knee replacement 0.68 (0.46–0.97) 0.91 (0.6–1.33) 0.61 (0.12–1.77)
Revision total knee
replacement 1.52 (0.31–4.38) 2.07 (0.56–5.22) 5.26 (0.6–17.74)
Femoro-popliteal bypass 7.14 (2.91–14.16) 5.88 (3.88–8.50) 8.43 (4.68–13.75)
Elective lower segment
caesarean section 0.45 (0.34–0.58) 0.99 (0.71–1.35) 3.61 (0.73–10.2)
Emergency lower segment
caesarean section 0.95 (0.80–1.12) 1.38 (1.09–1.73) 1.18 (0.13–4.21)
Mastectomy (simple) 0.69 (0.19–1.76) 1.59 (0.52–3.68) – ..
Mastectomy (radical) 0.31 (0.00–1.70) 0.60 (0.01–3.29) – ..
Total abdominal
hysterectomy 0.96 (0.61–1.42) 3.04 (2.08–4.29) 4.88 (1.33–12.02)
Cardiac valve replacement 1.05 (0.12–3.75) 3.03 (0.34–10.52) – ..
CABG with graft site
(sternal wound)c 1.27 (0.66–2.21) 1.46 (1.15–1.83) 2.82 (2.04–3.79)
CABG with graft site (graft
wound)c 1.65 (1.36–2.00) 3.40 (2.69–4.24) – ..
CABG with no separate
graft sitec 0.86 (0.17–2.51) 1.54 (0.17–5.44) – ..
a SSI rates are expressed in terms of infections per 100 procedures. b Risk categories are based on the
NHSN/NNIS risk index for SSIs (detailed in box 6.1). Numbers in parentheses are 95 per cent confidence
intervals. c CABG refers to coronary artery bypass graft. – Nil or rounded to zero. .. Not applicable.
Source: Queensland Department of Health (unpublished data).

F.3 South Australia

The Infection Control Service (ICS) within the SA Department of Health collects
data on BSIs, multi-resistant organisms (MROs) and Clostridium difficile. A wide
range of MROs are reported, including MRSA and extended spectrum

388 PUBLIC AND PRIVATE

HOSPITALS
beta-lactamase producing gram negatives.6 South Australia is the only Australian
state that conducts statewide surveillance of multiresistant gram-negative bacteria
infections (Christiansen et al. 2008). In many cases, the SA infections data can be
stratified into different risk groups, such as ICU/non-ICU, specialty and
inpatient/non-inpatient (SA Department of Health 2005a, 2005b, 2009a).

Reporting is voluntary and there are currently 17 participating hospitals (eight

public and nine private). The Commission understands that there is a high
participation rate among metropolitan hospitals, and that they account for the
majority of reporting establishments. Participating hospitals receive regular reports
from the ICS with statewide aggregates and the participant’s data. The ICS also
releases public reports, but these do not disaggregate data between public and
private hospitals.

The Commission obtained ICS data for eight public hospitals and eight private
hospitals, disaggregated by sector. These data show that from 2003 to 2008, rates of
hospital-acquired BSI were lower in private hospitals than in public hospitals
(figure F.3).

Figure F.3 Hospital-acquired BSIs in South Australia by sector,

2003–2008

8
7
Infection rate per 10 000
occupied bed days

6
5
4
3
2
1
0
Public hospitals Private hospitals

2003 2004 2005 2006 2007 2008

Source: SA Department of Health (unpublished data).

6 Targeted MROs are MRSA, VRE, Staphylococcus aureus with reduced susceptibility to
vancomycin (VISA), Staphylococcus aureus resistant to vancomycin (VRSA), multi-resistant
Pseudomonas aeruginosa, extended spectrum beta-lactamase producers (ESBL, including
gram-negative organisms, carbapenem-resistant Enterobacteriaceae and Acinetobacter species.
MRO definitions are based on those developed by the AICA (SA Department of Health 2005a).
STATE-LEVEL 389
INFECTIONS DATA
Likewise, MRSA infection rates for patients who were not admitted to an ICU were
lower in private hospitals (figure F.4). MRSA infection rates for patients admitted
to an ICU were marginally higher in private hospitals in 2003–05 but were similar
in both public and private hospitals over the period 2006–08. However, caution
should be exercised in interpreting infection rates for patients admitted to an ICU,
as the small number of affected patients means that one additional case can cause a
significant change in infection rates.

Figure F.4 Hospital-acquired MRSA infections in South Australia by

sector and ICU status, 2003–2008

10
9
Infection rate per 10 000

8
occupied bed days

7
6
5
4
3
2
1
0
Public hospitals - Private hospitals - Public hospitals - Private hospitals -
non-ICU patients non-ICU patients ICU patients ICU patients

2003-2005 2006-2008

Source: SA Department of Health (unpublished data).

F.4 Western Australia

In 2005, public and private hospitals in Western Australia began reporting infection
rates on a voluntary basis to the Healthcare Infection Surveillance Western
Australia (HISWA) program. The HISWA program is managed by the Health Care
Associated Infection Unit (HCAIU) within the WA Department of Health.
Reporting of some of the HISWA infection indicators was made mandatory in 2007
for public hospitals and private hospitals that provide services for public patients.
Private hospitals treating only private patients continue to report data voluntarily.

The HISWA program currently collects data on six different infection rates:
1. healthcare-associated MRSA
2. SSIs following elective hip and knee arthroplasty (surgical joint repair)
3. healthcare-associated SAB
390 PUBLIC AND PRIVATE
HOSPITALS
4. central-line-associated BSIs in an ICU
5. central-line-associated BSIs in haematology/oncology/outpatient intravenous
therapy units
6. haemodialysis-associated BSIs from access devices.

All but one of these indicators is, or will soon be, mandatory for public hospitals
and private hospitals that provide services for public patients (WA Department of
Health 2009a). The one exception is central-line-associated BSIs in
haematology/oncology/outpatient intravenous therapy units. Data on
healthcare-associated Clostridium difficile will be collected from January 2010.

The Australian Health Insurance Association (AHIA) claimed that the WA

mandatory reporting regime had helped to keep MRSA infection rates relatively
low:
It is the strongly held view of the AHIA that it is not coincidental that the lowest rate of
MRSA infection in Australia is in Western Australia, which is the only state or territory
where notification of MRSA infection is mandatory. (sub. 18, p. 6)

The HISWA indicators are based on nationally- and internationally-recommended

surveillance definitions. Results are collated and analysed by the HCAIU.
Individual hospital and aggregate reports are generated quarterly and more detailed
reports are published annually. Infection rates are risk adjusted where possible to
better reflect differences in clinical casemix between participating hospitals (WA
Department of Health 2008). Nevertheless, the HCAIU stressed that:
… the prime purpose of the HISWA surveillance program is to support internal
improvement, rather than performance comparison. This implies an emphasis on
collecting data over time to monitor progress, and internal validity within a facility.
(sub. 38, p. 3)

The published HCAIU reports have only a limited amount of information about the
relative performance of public and private hospitals. That information suggests that,
after using the NHSN/NNIS risk index to stratify data by risk groups, private
hospitals had lower SSI rates for hip and knee arthroplasty than public hospitals
during the period 2002–08 (WA Department of Health 2009b). The difference was
considered to be statistically significant, but the HCAIU cautioned that the
NHSN/NNIS risk index may not control for all risk differences between hospitals:
The reasons behind this variation may relate to a variety of practices and procedures
that are in place at these hospitals; however there is also likely to be differences in the
prevalence of risk factors for SSI such as smoking, obesity, diabetes and other
co-morbidities between institutions that are not incorporated into the risk adjustment.
Comparison therefore must be made carefully, and many factors will not necessarily be
modifiable by the hospitals involved. (WA Department of Health 2009b, p. 17)

STATE-LEVEL 391
INFECTIONS DATA
The published data also suggest that WA private hospitals tend to have lower rates
of hospital-acquired MRSA infections than public hospitals.7 However, the
difference may be largely due to private hospitals tending to have lower-risk
procedures, treatments and patients.

The Commission obtained unpublished data from the HCAIU on SAB BSIs,
MRSA, and SSIs following elective hip and knee arthroplasty. The data show that
private hospitals had lower rates of hospital-acquired MRSA infections than public
hospitals from 2006 to 2008 (table F.8). However, this difference was only
statistically significant in 2007. Furthermore, the HCAIU cautioned that:
This [MRSA infection] rate will depend on both the risk of a healthcare-associated
infection (which varies according to casemix as well as aspects of the quality of care
provided); and the risk of that infection being due to MRSA (which reflects endemic
MRSA rates in the patient population and the risk of acquiring MRSA in the hospital).
Comparison must therefore be made considering differences in case mix and MRSA
rates in the admitted patient population before associating differences in rates to
variation in the quality of care provided. Public hospitals may have both a more
complex patient case mix with an inherently higher risk of developing an HAI
[hospital-acquired infection] and a higher prevalence of MRSA carriage on admission
to hospital. (sub. 38, p. 5)

Table F.8 Hospital-acquired MRSA in Western Australia by sector,

2006–2008a
No. of events Occupied bed days Infection rateb

2006
Public 100 836 463 1.20 (0.98–1.46)
Private 40 482 633 0.83 (0.61–1.13)
Total 140 1 319 096 1.06 (0.90–1.25)
2007
Public 80 875 396 0.91 (0.73–1.14)
Private 23 508 023 0.45 (0.30–0.68)
Total 103 1 383 419 0.74 (0.61–0.90)
2008
Public 115 895 890 1.28 (1.07–1.54)
Private 43 521 618 0.82 (0.61–1.11)
Total 158 1 417 508 1.11 (0.95–1.30)
a Inpatient events only. b Infections per 10 000 occupied bed days. Numbers in parentheses are 95 per cent
confidence intervals.

7 In 2007-08, reporting private hospitals had an MRSA infection rate of 0.68 per 10 000 bed days
(95 per cent confidence interval of 0.48–0.95). This was compared to four (public) area health
services, which had rates that ranged from 0.19 (0.00–1.19) to 1.26 (0.92–1.71). The only area
health service with a lower rate than private hospitals was the Child and Adolescent Health
Service, which reported just one MRSA case in 2007-08.
392 PUBLIC AND PRIVATE
HOSPITALS
Source: HCAIU (unpublished data).

The unpublished data show that private hospitals also had lower rates of
hospital-acquired SAB BSIs than public hospitals (table F.9). However, this
difference was only statistically significant in 2008. Furthermore, the HCAIU
cautioned that:
The risk of an individual patient acquiring a Staphylococcus aureus bacteraemia is
related to their underlying medical condition, complexity of care and the invasive
procedures they are subject to, as well as the quality of care provided. (sub. 38, p. 5)

Table F.9 Hospital-acquired SAB BSIs in Western Australia by

sector, 2007–2008a
No. of events Occupied bed days Infection rateb

2007
Public 28 875 396 0.32 (0.22–0.47)
Private 7 508 023 0.14 (0.06–0.29)
Total 35 1 383 419 0.25 (0.18–0.35)
2008
Public 113 895 890 1.26 (1.05–1.52)
Private 35 521 618 0.67 (0.48–0.94)
Total 148 1 417 508 1.04 (0.89–1.23)
a Inpatient events only. b Infections per 10 000 occupied bed days. Numbers in parentheses are 95 per cent
confidence intervals.
Source: HCAIU (unpublished data).

The unpublished data suggest that public hospitals had higher SSI rates across all
risk categories for both hip and knee procedures (figures F.5 and F.6). However,
this difference was only statistically significant for hip procedures in risk category
zero. Furthermore, the HCAIU cautioned that while WA data on SSI rates are risk
adjusted using the NHSN/NNIS methodology, this does not control for all risk
factors:
SSI rates that are risk-adjusted using NHSN stratification do not account for systematic
differences in patient, operator and unit characteristics that raise the inherent or
underlying SSI risk of public hospitals. They are subject to bias, and while useful, must
be interpreted with this understanding. (sub. 38, p. 4)

STATE-LEVEL 393
INFECTIONS DATA
Figure F.5 Hip SSIs in Western Australia by risk category and sector,
2006–2008a

18
Infection rate per 100

16
14
procedures

12
10
8
6
4
2
0
Risk 0 Risk 1 Risk 2
Risk Category

Private hospitals Public hospitals

a Risk categories are based on the NNIS risk index. The vertical lines for each risk category indicate the
95 per cent confidence interval.
Source: HCAIU (unpublished data).

Figure F.6 Knee SSIs in Western Australia by risk category and

sector, 2006–2008a

14
Infection rate per 100

12
10
procedures

8
6
4
2
0
Risk 0 Risk 1 Risk 2
Risk Category

Private hospitals Public hospitals

a Risk categories are based on the NNIS risk index. The vertical lines for each risk category indicate the
95 per cent confidence interval.
Source: HCAIU (unpublished data).

394 PUBLIC AND PRIVATE

HOSPITALS
F.5 Tasmania
In 2008, the Tasmanian Infection Prevention and Control Unit (TIPCU) was
established by the Department of Health and Human Services to manage a
surveillance program for hospital-acquired infections. Four infection rates are
currently monitored — SAB BSIs, MRSA, Clostridium difficile and VRE
(TIPCU 2009). The definitions used for these indicators are based on those
recommended by the ACSQHC.

VRE has been a notifiable disease in Tasmania since 2000, and SAB BSIs since
December 2008. Thus, VRE and SAB reporting is mandatory for both public and
private hospitals. Private hospitals have volunteered to also report the MRSA and
Clostridium difficile indicators along with public hospitals. The Commission
understands that data are collected from four public hospitals and five private
hospitals.

TIPCU provides confidential reports back to all reporting hospitals. Its first public
report was released in March 2009. To date, only data for public hospitals have
been published (summarised in table F.10). On the basis of the published data,
TIPCU (2009) concluded that Tasmanian acute public hospitals have similar
infection rates for MRSA and SAB BSIs as public hospitals in other states. The rate
of Clostridium difficile in Tasmanian public hospitals was considered to be slightly
higher than that reported in other states, but there is limited data with which to make
comparisons.

Table F.10 Rate of hospital-acquired infections in Tasmanian public

hospitals by organism, 2005–2008a
2006 2007 2008
Staphylococcus aureus bacteraemia (SAB)b 0.92 1.11 1.07
SAB caused by methicillin-resistant Staphylococcus aureus (MRSA)b 0.14 0.04 0.21
Clostridium difficilec 2.20 1.80 3.20
a Infection rates are expressed as infections per 10 000 occupied bed days. b Based on six months of data for
2008. c Based on six months of data for both 2006 and 2008.
Source: TIPCU (2009).

The Commission requested comparative infections data for public and private
hospitals but the Tasmanian Department of Health and Human Services was unable
to provide such information in time for this report.

STATE-LEVEL 395
INFECTIONS DATA
G Referee reports on modelling

G.1 Report from Adjunct Professor Tim Coelli

This study of hospital performance in Australia sets a new benchmark in terms of
the sample coverage and the range of input, output and control variables included in
the econometric model. The Productivity Commission team has worked hard at
producing the best empirical model possible, in the face of challenging time and
data constraints. However, no empirical study is perfect, so in my brief discussion
below I provide my assessment of the analysis, pointing out what is to be
commended and what can perhaps be improved.

Data sample: The sample size of 508 is more than sufficient to allow the
Productivity Commission (PC) to estimate an econometric model that involves a
flexible functional form and a number of important input, output and control
variables. The main concern with the sample, as it stands, relates to a high non-
response rate on the part of not-for-profit hospitals. Hence the private hospitals in
the sample are mostly for-profit hospitals, and the results obtained should be viewed
in this light. However, my experience with not-for-profit hospitals (mostly run by
church groups) is that they tend to put extra resources into non-medical services and
hence I do not expect them to normally have much influence on the position of the
best-practice frontier. Hence, I expect that their low sample representation is
unlikely to significantly affect the efficiency scores of the remaining hospitals in the
sample.

Frontier methodology: There are two frontier estimation methods that are
commonly used in the literature: data envelopment analysis (DEA) and stochastic
frontier analysis (SFA). DEA is a linear programming method that has the
advantage that no particular functional form needs to be specified. However, SFA is
an econometric method that is less susceptible to the effects of data noise and
outliers and which also allows one to easily incorporate control variables that
involve categorical and ratio data. Hence the choice of SFA is appropriate for this
study.

REFEREE REPORTS 397

Functional form: The translog function form is a flexible second-order functional
form that can accommodate a range of scale and substitution possibilities, and hence
is a good choice in my assessment.

Output measures: The output measures involve a number of categories of admitted

and non-admitted separations, with the former casemix-adjusted. The level of detail
is substantially better than many past studies of hospital efficiency. The authors
emphasize the point that these are measures of intermediate outputs rather than
incremental health benefits derived from the services. However, this is standard
practice in this literature, given the very substantial challenges that would be
involved in attempting to derive these latter output measures.

Input measures: The input measures include three categories of staff members
(nursing, diagnostic and other), three monetary measures of non-staff variable
inputs (drugs, medical and surgical supplies and other) along with the number of
beds. This group of input measures is better than that used in the majority of past
studies, but can still be improved upon (given access to better data). In particular,
the beds measure treats an intensive care bed no differently to a standard bed, and
the staff measures exclude doctors. These issues could introduce some biases in
efficiency estimates if the casemix weights (used to define the output measures)
include allowances for the extra capital costs associated with complex cases, and if
there are differences among hospitals in the degree to which doctors versus nurses
undertake certain “grey area” tasks.

Quality measures: Quality issues have been often overlooked in past studies of
health sector efficiency. The PC is to be commended for their efforts in this regard.
The inclusion of a mortality rate measure that is adjusted for patient risk
characteristics is not a perfect measure, but should go a long way to capturing any
notable variations in the effects of service quality upon efficiency potentials.

Control measures: The PC has considered a wide range of exogenous control

measures that could potentially be affecting efficiency potentials, including network
membership, accident and emergency rates, and so on. These measures help the
analyst to avoid labelling a hospital as being “inefficient” when they may be using
more resources per unit output because they face different operating conditions
relative to other members of the sample.

Finally, I should emphasize a number of points. First, most if not all of the
comments made above are also mentioned in the main report. Second, time and data
constraints have clearly placed limits on the empirical analysis in this report. Third,
I look forward to seeing what is produced in the supplementary report that is due to
be released in March 2010, which will involve data from additional years and will

398 PUBLIC AND PRIVATE

HOSPITALS
also involve the investigation of some alternative models and a more detailed
investigation of the effects of scale on hospital performance.

G.2 Report from Professor Jim Butler

The estimation of hospital production functions and hospital cost functions is a
complicated exercise. These complications arise not just because hospitals are
multi-product organisations, but because of the large range and diversity of the
outputs they produce. The ‘treated patient’ is not a homogeneous unit of output but
differs according to the illness or illnesses with which they present, the severity of
those illnesses, the range of treatments available and which are selected, and patient
characteristics such as age, sex and frailty. The econometric modeller then faces a
quandary. Working with a sufficiently large number of output categories to
minimise heterogeneity within those categories will lead to a large number of
parameters to be estimated, especially if the specification involves a flexible
functional form. However, the pursuit of parameter parsimony, which requires a
smaller number of output categories, introduces more heterogeneity into the output
categories.

In addition to this type of conceptual difficulty, there are difficulties arising out of
the institutional arrangements for the provision of hospital services in Australia (e.g.
the inclusion of medical service costs in hospital costs for public hospitals but not
private hospitals), the lack of a ‘pure’ separation between type of hospital
ownership and the funding status of patients, and the paucity of data on capital
costs.

Notwithstanding these difficulties, and the tight timelines within which the
Commission was working, it has produced a high quality and interesting piece of
work comparing technical efficiency in public and private hospitals using data on
508 hospitals for 2006-07. Using stochastic frontier analysis and several functional
forms for a production function, the analysis concludes that the technical efficiency
of public hospitals and private hospitals is similar. Across all hospital size
groupings, public hospitals have a mean efficiency score of 0.797 and private
hospitals 0.750 suggesting slightly superior performance by public hospitals
(table 8.5 — there was virtually no difference between the scores for for-profit and
not-for-profit private hospitals). Given the difficulties associated with empirical
work in this area mentioned above (and that list is not exhaustive), the description
of these mean scores for public and private hospitals as ‘similar’ is a judicious call
of the results. The only size grouping where a more marked difference between the
scores for public and private hospitals emerges is the ‘small and very small’
hospitals category with mean public and private hospital scores of 0.788 and 0.641
REFEREE REPORTS 399
respectively (table 8.5). But with these results and others, one should bear in mind
the possibility of self-selection bias in the private hospital sample of 122 hospitals
as participation in the study by private hospitals was voluntary. The Report does
specifically mention this limitation.

An innovative aspect of this study is its distinction between public hospitals and
public contract hospitals. The latter are privately owned institutions whose caseload
comprises a large proportion of public patients treated under contract from
government (on average, public patients comprise 77.9% of the caseload of public
hospitals cf. 77.4% for public contract hospitals – see Table 8.2). There is virtually
no difference in the technical efficiency scores between these two types of
institution (0.797 for public hospitals, 0.800 for public contract hospitals —
table 8.5). While the sample size for public contract hospitals is not large (n=18),
this result is of some interest.

A result which has perhaps been somewhat underplayed is the absence of any
significant effect of risk-adjusted mortality ratios in the production model. The
Commission has constructed a predicted value of the mortality rate (proportion of
patients discharged dead) for each hospital using a Tobit regression with various
factors exogenous to the hospital as regressors. These predicted values are then used
to construct a risk-adjusted mortality ratio for each hospital which is used in the
production function to investigate possible quantity/quality trade-offs in hospital
production. The absence of a statistically significant effect here is potentially a
policy-significant result — there is no evidence that hospitals in this study attain
higher output levels or improve technical efficiency by allowing quality to
deteriorate.

In concluding, two important limitations of this study should be noted. First, it is

based upon data for only one year. Replication of the analyses using data from other
years may instil more confidence in the results. Second, the study investigates only
technical efficiency and not cost efficiency. As the Commission notes, a hospital’s
performance with respect to technical efficiency may differ from its performance
with respect to cost efficiency, so results on the latter would provide a more
complete picture of hospital performance. The Report indicates that both of these
limitations will be addressed in further analyses, the results of which will be
available in a supplementary report in March 2010. If those analyses are conducted
to the same standard as the analyses presented in this Report, they will undoubtedly
provide a very useful addition to the stock of knowledge in this field.

400 PUBLIC AND PRIVATE

HOSPITALS
References

ABS (Australian Bureau of Statistics) 2005, Australian Standard Geographical

Classification (ASGC), Cat. no. 1216.0, Canberra.
—— 2007, Private Hospitals, Cat. no. 4390.0, Canberra.
—— 2008a, An Introduction to Socio-Economic Indices for Areas (SEIFA), 2006,
Information paper, Cat. no. 2039.0, Canberra.
—— 2008b, Australian Demographic Statistics, Cat. no. 3101.0, Canberra.
—— 2008c, Average Weekly Earnings, Australia, Cat no. 6302.0, Canberra.
—— 2008d, Consumer Price Index, Australia, Cat. no. 6401.0, Canberra.
—— 2008e, Labour Price Index, Australia, Cat. no. 6345.0, Canberra.
—— 2008f (and previous issues), Private Hospitals, Cat. no. 4390.0, Canberra.
—— 2008g, Socio-Economic Indexes for Areas (SEIFA): Technical Paper 2006,
Cat. no. 2039.0.55.001, Canberra.
—— 2009a, Australian National Accounts: National Income, Expenditure and
Product, Cat. no. 5206.0, Canberra.
—— 2009b, National Health Survey, Australia, Cat. no. 4364.0, Canberra.
Access Economics 2009, Impact of Means Testing the PHI Rebate and Changing
MLS Parameters, Report for Catholic Health Australia, Canberra
www.accesseconomics.com.au/publicationsreports/getreport.php?report=203&id
=259 (accessed 16 September 2009).
ACEM (Australasian College for Emergency Medicine) 2000, ‘Policy on the
Australian Triage Scale’, Policy P06, www.acem.org.au/media/policies_and
_guidelines/P06_Aust_Triage_Scale_-_Nov_2000.pdf (accessed 29 July 2009).
ACHR (Australian Centre for Health Research) 2008, Measuring Quality in Private
Hospitals, Report by the National Health and Medical Research Council Centre
of Research Excellence in Patient Safety, Department of Epidemiology and
Preventive Medicine, Monash University, Melbourne.
ACHS (Australian Council on Healthcare Standards) 2007, Australasian Clinical
Indicator Report 1998–2006: Determining the Potential to Improve Quality of
Care, 8th edn, Sydney.

REFERENCES 401
—— 2008, Australasian Clinical Indicator Report 2001–2007: Determining the
Potential to Improve Quality of Care, 9th edn, Sydney.
ACSQHC (Australian Commission on Safety and Quality in Health Care) 2008,
Windows into Safety and Quality in Health Care 2008, Sydney.
ACT Revenue Office 2009a, Exempt Wages and Payments, Canberra,
www.revenue.act.gov.au/payroll_tax#Exempt_Wages_and_Payments (accessed
6 March 2009).
—— 2009b, Payroll Tax Thresholds and Rate, Canberra, www.revenue.act.gov.au/
payroll_tax#Payroll_Tax_Thresholds_and_Rate (accessed 6 March 2009).
Aigner, D.J., Lovell, C.A. Knox, and Schmidt, P. 1977, ‘Formulation and
estimation of stochastic frontier production function models’, Journal of
Econometrics, vol. 6, no. 1, pp. 21–37.
AHMAC (Australian Health Ministers’ Advisory Council) 2006, Aboriginal and
Torres Strait Islander Health Performance Framework Report 2006, Canberra.
AIHW (Australian Institute of Health and Welfare) 2004, Australian Hospital
Statistics 2002-03, Health Services Series no. 22, Cat. no. HSE 32, Canberra.
—— 2005, Australian Hospital Statistics 2003-04, Health Services Series no. 23,
Cat. no. HSE 37, Canberra.
—— 2006, Australian Hospital Statistics 2004-05, Health Services Series no. 26,
Cat. no. HSE 41, Canberra.
—— 2007a, Australian Hospital Statistics 2005-06, Health Services Series no. 30,
Cat. no. HSE 50, Canberra.
—— 2007b, The AIHW National Mortality Database Documentation,
www.aihw.gov.au/mortality/data_dictionary_latest.pdf (accessed 25 August
2009).
—— 2008a, A Set of Performance Indicators Across the Health and Aged Care
System, Canberra.
—— 2008b, Australian Hospital Statistics 2006-07, Health Services Series no. 31,
Cat. no. HSE 55, Canberra.
—— 2008c, Australia’s Health 2008, Cat. no. AUS 99, Canberra.
—— 2008d, (and previous issues), Australia’s Mothers and Babies 2006, Cat.
no. PER 46, Sydney.
—— 2008e, Elective Surgery in Australia: New Measures of Access,
Cat. no. HSE 57, Canberra.

402 PUBLIC AND PRIVATE

HOSPITALS
—— 2009a, Australian Hospital Statistics 2007-08, Health Services Series no. 33,
Cat. no. HSE 71, Canberra.
—— 2009b, Australian Hospital Statistics 2007-08, supplementary tables:
Australian Refined Diagnosis Related Groups for admitted patients, www.aihw.
gov.au/publications/hse/hse-71-10776/hse-71-10776-xx12.xls (accessed 5
November 2009).
—— 2009c, Health Expenditure Australia 2007-08, Health and Welfare
Expenditure Series no. 37, Cat. no. HWE 46, Canberra.
—— 2009d, Health Sector Reporting Relationships for Health Committees,
Canberra.
—— 2009e, Privacy of data. www.aihw.gov.au/dataonline/privacy_of_data.cfm,
Canberra.
—— 2009f, Sources of Funds for Admitted Patients in Australian Hospitals,
2005-06, Health and Welfare Expenditure Series, no. 39. Cat. no. HWE 44.
Canberra.
—— 2009g, Towards National Indicators of Safety and Quality in Health Care,
Cat. no. HSE 75, Canberra.
Al Shammari, M. 1999, ‘A multi-criteria data envelopment analysis model for
measuring the productive efficiency of hospitals’, International Journal of
Operations and Production Management, vol. 19, no. 9, pp. 879–90.
AMA (Australian Medical Association) Victoria 2001, Private hospital access fears
confirmed, Media release, Melbourne, 7 December.
Arocena, P. and Garcia-Prado, A. 2007, ‘Accounting for quality in the measurement
of hospital performance: evidence from Costa Rica’, Health Economics, vol. 16,
pp. 667-85.
ASX (Australian Securities Exchange) 2009, Health Care Sector Profile,
www.asx.com.au/research/pdf/health_care_sector_factsheet.pdf (accessed
29 July 2009).
ATO (Australian Tax Office) 2007a, Fringe Benefits Tax — Tax Basics for Non-
Profit Organisations, Canberra, www.ato.gov.au/nonprofit/content.asp?doc=/
content/33618.htm&page=10&H10=&pc=001/004/012/006/001&mnu=44780&
mfp=001&st=&cy=1 (accessed 4 March 2009).
Auditor-General for Western Australia 2007, First Do No Harm: Reducing Adverse
Events in Public Hospitals, Report 10, Perth.
Australian Government 2009, Budget Measures 2009-10, Budget Paper No. 2,
Canberra.

REFERENCES 403
Australian Senate Community Affairs Legislation Committee 2009, fairer private
health insurance incentives bill 2009 and two related bills,
www.aph.gov.au/senate/committee/clac_ctte/fairer_private_health_09/report/rep
ort.pdf (accessed 16 September 2009).
Baghurst, A., Place, M. and Posnett, J.W. 1999, ‘Dynamics of bed use in
accommodating emergency admission: stochastic simulation model’, British
Medical Journal, vol. 319, pp. 155–58.
Banker, R.D., Conrad, R.F. and Strauss, R.P. 1986, ‘A comparative application of
data envelopment analysis and translog methods: an illustrative study of hospital
production’, Management Science, vol. 39, no. 10, pp. 1265–73.
Bedard, J.C. and Wen, K.W. 1990, ‘A comparison of the efficiency effects of
prospective Reimbursement System’, Research in Governmental and Non-profit
Accounting, vol. 6, pp. 63–82.
Ben-Tovim, D., Woodman, R., Harrison, J.E., Pointer, S., Hakendorf, P. and
Henley, G. 2009, Measuring and Reporting Mortality in Hospital Patients,
AIHW, Cat. no. HSE 69, Canberra.
Biørn, E., Hagen, T.P, Iversen, T., and Magnussen, J. 2003, ‘The effect of activity
based financing on hospital efficiency: a panel data analysis of DEA efficiency
scores 1992-2000’, Health Care Management Science, vol. 6, pp. 271-83.
Borden, J.P. 1988, ‘An assessment of the impact of diagnosis-related group (DRG)-
based reimbursement on the technical efficiency of New Jersey hospitals using
data envelopment analysis’, Journal of Accounting and Public Policy, vol. 7,
pp. 77–96.
Breyer, F. 1987, ‘The specification of a hospital cost function – a comment on the
recent literature’, Journal of Health Economics, vol. 6, pp. 147–57.
Bridges, J., Haas, M. and Mazevska, D. 1999, A Qualitative Insight into Rural
Casemix Education, Centre for Health Economics Research and Evaluation
Project Report 10, Melbourne.
Brown, H.S. 2003, ‘Managed care and technical efficiency’, Health Economics,
vol. 12, pp. 149–58.
Bull, A., McGechie, D., Richards, M, Russo, P. and Worth, L. 2008, ‘Surgical Site
Infection’, in Cruickshank, M. and Ferguson, J. (eds), Reducing Harm to
Patients from Health Care Associated Infection: The Role of Surveillance,
Australian Commission on Safety and Quality in Health Care, Sydney,
pp. 91–107.
BUPA Australia 2004, Submission to the Productivity Commission's study into
Impacts of Advances in Medical Technology in Australia, Sydney.

404 PUBLIC AND PRIVATE

HOSPITALS
Burgess, J.F. and Wilson, P.W. 1995, ‘Decomposing hospital productivity changes,
1985–1988: a nonparametric malmquist approach’, Journal of Productivity
Analysis, vol. 6, pp. 343–63.
—— 1998, ‘Variation of inefficiency in US hospitals’, Canadian Journal of
Operational Research and Information, vol. 36, pp. 84–102.
Butler, 1988a, ‘Hospital costs and information theory case mix indexes: results for
Queensland’, Prometheus, vol. 6, no. 2, pp. 327–50.
—— 1988b, ‘Issues in hospital funding’, in CEDA (Committee for Economic
Development of Australia), The Economics of Health Care, APAIS, Canberra,
pp. 87–116.
—— 1995, Hospital Cost Analysis, Kluwer Academic, Dordrecht.
Cadwallader, H., Toohey, M., Linton, S., Dyson, A. and Riley, T. 2001,
‘A comparison of two methods for identifying surgical site infections following
orthopaedic surgery’, Journal of Hospital Infection, vol. 48, pp. 261–66.
Catholic Health Australia 2009a, Catholic Health Australia Clinical Placements
Survey — Interim Results, Canberra.
—— 2009b, The Sector. www.cha.org.au/site.php?id=24 (accessed 25 September
2009).
CHRISP (Centre for Healthcare Related Infection Surveillance and Prevention)
2003, CHRISP User Manual, Queensland Department of Health, Brisbane.
—— 2009, CHRISP Surveillance Manual, 3rd edn, Queensland Department of
Health, Brisbane.
Chambers, D.C., 2009, Letter in response to ‘Adverse outcomes of labour in public
and private hospitals in Australia’, Medical Journal of Australia, vol. 191,
no. 10, pp. 578–79.
Charlson, M.E., Pompei, P., Ales K.L. and MacKenzie, C.R. 1987, ‘A new method
of classifying prognostic comorbidity in longitudinal studies: development and
validation’, Journal of Chronic Diseases, vol. 40, no. 5, pp. 373–83.
Charnes, A., Cooper, WW, and Rhodes, E. 1978, ‘Measuring the efficiency of
decision-making units’, European Journal of Operational Research, vol. 2,
pp. 429–44.
Chen, S. N. 2006, ‘Productivity changes in Taiwanese hospitals and the national
health insurance’, Service Industries Journal, vol. 26 no. 4, pp. 459–77.
Christiansen, K., Coombs, G., Ferguson, J., Iredell, J., Marshall, C., Nimmo, G.,
Turnidge, J., Van Gessel, H. and Wilkinson, I. 2008, ‘Multiresistant organisms’,
in Cruickshank, M. and Ferguson, J. (eds), Reducing Harm to Patients From

REFERENCES 405
 Health Care Associated Infection: The Role of Surveillance, Australian
Commission on Safety and Quality in Health Care, Sydney, pp. 129–69.
Christiansen, K., Tibbett, P., Beresford, W., Pearman, J., Lee, R., Coombs, G., Kay,
I., O'Brien, F., Palladino, S., Douglas, C., Montgomery, P., Orrell, T., Peterson,
A., Kosaras, F., Flexman, J., Heath, C. and McCullough, C. 2004, ‘Eradication
of a large outbreak of a single strain of vanB vancomycin-resistant enterococcus
faecium at a major Australian teaching hospital’, Infection Control and Hospital
Epidemiology, vol. 25, no. 5, pp. 384–90.
Chua, C.L., Palangkaraya, A. and Yong, J. 2008, A Two-Stage Estimation of
Hospital Performance Using Mortality Outcome Measures: An Application
Using Victorian Hospital Data, Melbourne Institute Working paper no. 10/08,
The University of Melbourne.
—— 2009, Hospital Competition, Technical Efficiency and Quality, Melbourne
Institute Working paper no. 16/09, The University of Melbourne.
Clements, A., Tong, E., Morton, A. and Whitby, M. 2007, ‘Risk stratification for
surgical site infections in Australia: evaluation of the US National Nosocomial
Infection Surveillance risk index’, Journal of Hospital Infection, vol. 66, no. 2,
pp. 148–55.
Clement, J.P, Valdmanis, V.G., Bazzoli, G.J., Zhao, M., and Chukmaitov, A. 2008,
‘Is more better? An analysis of hospital outcomes and efficiency with a DEA
model of output congestion’, Health Care Management Science, vol. 11,
pp. 67–77.
Clezy, K., Cruickshank, M., Ferguson, J., Givney, R., Jorm, C., McLaws, M.,
Peterson, A. and Russo, P. 2008, ‘Surveillance and quality improvement’, in
Cruickshank, M. and Ferguson, J. (eds), Reducing Harm to Patients from Health
Care Associated Infection: The Role of Surveillance, Australian Commission on
Safety and Quality in Health Care, Sydney, pp. 27–49.
CCSC (Clinical Costing Standards Committee) 2006, clinical costing standard
number 7: exclusion of costs for the purpose of clinical costing, Melbourne,
www.ccsaa.com/standards/clinical-costing-standard-7-exclusion-of-costs-for-
the-purpose-of-clinical-costing.htm (accessed 27 August 2009).
COAG (Council of Australian Governments) 2007, Council of Australian
Governments’ Meeting, 20 December 2007, Communiqué, www.coag.gov.au/
coag_meeting_outcomes/2007-12-20/cooag20071220.pdf (accessed 31 July
2009).
—— 2008a, Council of Australian Governments’ Meeting, 29 November 2008,
Attachment A Health and Ageing, www.coag.gov.au/coag_meeting_outcomes
/2008-11-29/attachments.cfm#attachmenta (accessed 14 September 2009).
406 PUBLIC AND PRIVATE
HOSPITALS
—— 2008b, Council of Australian Governments’ Meeting, 29 November 2008,
Communiqué, Canberra, www.coag.gov.au/intergov_agreements/federal_financi
al_relations/index.cfm (accessed 14 September 2009).
—— 2008c, Intergovernmental Agreement on Federal Financial Relations,
www.coag.gov.au/intergov_agreements/federal_financial_relations/docs/IGA_fe
deral_financial_relations.pdf (accessed 26 May 2009).
—— 2008d, National Healthcare Agreement, Canberra, www.coag.gov.au/
intergov_agreements/federal_financial_relations/docs/IGA_FFR_ScheduleF_Nat
ional_Healthcare_Agreement.rtf (accessed 26 May 2009).
—— 2008e, National Partnership Agreement on Hospital and Health Workforce
Reform, www.coag.gov.au/intergov_agreements/federal_financial_relations/docs
/national_partnership/national_partnership_on_hospital_and_health_workforce_r
eform.pdf (accessed 14 September 2009).
Coelli, T. Rao, D.S.R, and G.E. Battese, 1998, An Introduction to Efficiency and
Productivity Analysis, Kluwer Academic.
Coelli, T. Rao, D.S.R, O’Donnell, C.J. and G.E. Battese, 2005, An introduction to
efficiency and productivity analysis, Springer, New York.
Collignon, P., Dreimanis, D., Ferguson, J., Van Gessel, H., Taylor, P., Wilkinson,
I., and Worth, L. 2008, ‘Bloodstream Infection’, in Cruickshank, M. and
Ferguson, J. (eds), Reducing Harm to Patients from Health Care Associated
Infection: The Role of Surveillance, Australian Commission on Safety and
Quality in Health Care, Sydney, pp. 53–89.
Collignon, P., Nimmo, G., Gottlieb, T. and Gosbell I. 2005, ‘Staphylococcus aureus
bacteremia, Australia’, Emerging Infectious Diseases, vol. 11, no. 4,
pp. 554–61.
Collignon, P., Wilkinson, I., Gilbert, G., Grayson, M. and Whitby, R. 2006, ‘Health
care-associated Staphylococcus aureus bloodstream infections: a clinical quality
indicator for all hospitals’, Medical Journal of Australia, vol. 184, no. 8,
pp. 404–06.
Coory, M.D., Koh, G.T., Flenady, V. and Kamp, M. 2009, Letter in response to
‘Adverse outcomes of labour in public and private hospitals in Australia’,
Medical Journal of Australia, vol. 191, no. 10, p. 578.
Cosgrove, S., Sakoulas, G., Perencevich, E., Schwaber, M., Karchmer, A. and
Carmeli, Y. 2003, ‘Comparison of mortality associated with methicillin-resistant
and methicillin-susceptible Staphylococcus aureus bacteremia: a meta-analysis’,
Clinical Infectious Diseases, vol. 36, no. 1, pp. 53–59.

REFERENCES 407
Crotty, B. J. 2005, ‘More students and less patients: the squeeze on medical
teaching resources’, Medical Journal of Australia, vol. 183, no. 9, pp. 444-45.
Cruickshank, M. and Ferguson, J. (eds) 2008, Reducing Harm to Patients from
Health Care Associated Infection: The Role of Surveillance, Australian
Commission on Safety and Quality in Health Care, Sydney.
Dailey, L. van Gessel, H. and Peterson, A. 2009, ‘Two years of surgical site
infection surveillance in Western Australia: analysing variation between
hospitals’, Healthcare Infection, vol. 14, pp. 51–60.
Deeble, J. 1965, ‘An economic analysis of hospital costs’, Medical Care, vol. 3,
no. 3, pp. 138–46.
Department of Health Victoria 2009, Nurses (Victorian Public Health Sector)
Multiple Business Agreement 2007-2011, Melbourne, www.health.vic.gov.au/__
data/assets/pdf_file/0008/356696/Nurses-Public-Sector-EBA-2004-2007-
_AG840794-2.pdf (accessed 20 November 2009).
Department of Finance and Administration 2006, Handbook of Cost-Benefit
Analysis, Canberra, www.finance.gov.au/publications/finance-circulars/2006/
docs/Handbook_of_CB_analysis.pdf (accessed 9 September 2009).
DHS (Victorian Department of Human Services) 2008, Sentinel Event Program:
Annual Report 2007-08, Building Foundations to Support Patient Safety,
Melbourne.
—— 2009 Elective surgery performance data July 2008 to June 2009, Melbourne,
http://serviceforip.webcentral.com.au/yourhospitals/performance.asp.
Dietz, H.P. and Peek, M.J. 2004, ‘Will there ever be an end to the Caesarean section
rate debate?’, Australian and New Zealand Journal of Obstetrics and
Gynaecology, vol. 44, pp. 103–06.
DOHA (Department of Health and Ageing) 2004, Australian Refined Diagnosis
Related Groups Version 5.1: Definitions Manual, Canberra.
—— 2008a, A Users Guide for the Collection of Hospital Casemix Protocol (HCP)
And Private Hospital Data Bureau (PHDB), Canberra, www.health.gov.au/
internet/main/publishing.nsf/Content/303919F7B92DC9B5CA25701C00007E3
B/$File/Data%20Definitions%20Manual.pdf (accessed 17 August 2009).
—— 2008b, Cost Report 11 (2006-07), Canberra, www.health.gov.au/internet/
main/publishing.nsf/Content/5F8B6BE822DC75B3CA25748300164037/$File/R
11CostReport_Final.pdf (accessed 24 April 2009).
—— 2008c, National Hospital Cost Data Collection: Reference Manual Round 11
(2006-07), Canberra, www.health.gov.au/internet/main/publishing.nsf/Content/

408 PUBLIC AND PRIVATE

HOSPITALS
 90F1A1681D35BCA4CA2570BC0015B1FC/$File/Round%2011.pdf (accessed
18 September 2009).
—— 2008d, Private Health Insurance, Canberra, www.health.gov.au/internet/
main/publishing.nsf/Content/health-privatehealth-outofpocket-expenses
(accessed 8 July 2009).
—— 2009a, Cost Report Round 12 (2007-08), Canberra, www.health.gov.au/
internet/main/publishing.nsf/Content/0BF59B7DB88A427FCA257609001FCD3
D/$File/1_R12CostReport.pdf (accessed 18 September 2009).
—— 2009b, List of caveats for the round 12 National Hospital Cost Data Collection
(NHCDC) Private Cost Weight Tables, Canberra, www.health.gov.au/
internet/main/publishing.nsf/Content/0BF59B7DB88A427FCA257609001FCD3
D/$File/8_PrivateCWEst51.xls (accessed 25 September 2009).
—— 2009c, Private Hospitals @ 9 September 2009, www.health.gov.au/internet/
main/publishing.nsf/Content/0476BA161BDC3A0BCA257376001DA7D5/$File
/Private%20Hospitals%20@%2009.09.09.xls (accessed 24 September 2009).
—— 2009d, Prostheses List — Excel version, www.health.gov.au/internet/
main/Publishing.nsf/Content/prostheses-list-excel.htm (accessed 5 November
2009).
—— 2009e, Public Hospitals @ 9 September 2009, www.health.gov.au/internet/
main/publishing.nsf/Content/0476BA161BDC3A0BCA257376001DA7D5/$File
/Public%20Hospitals%20@%2009.09.09.xls (accessed 24 September 2009).
—— 2009f, State of Our Public Hospitals, Canberra, www.health.gov.auinternet/
main/publishing.nsf/Content/1186FF4200B2B2E6CA2575D9000F73B8/$File/S
oOPH%202009.pdf (accessed 3 August 2009).
—— 2009g, The Expanded Specialist Training Program, www.health.gov.au/
internet/main/publishing.nsf/Content/work-pr-estp-infopro (accessed 31 August
2009).
DOH SA (SA Department of Health) 2006, South Australian Safety and Quality
Framework & Strategy 2007-2011: Project Report and Resource Documents a
System Wide Approach to Sustainable Safety and Quality Care and Services,
Adelaide.
Department of Treasury and Finance (Tasmania) 2008a, Pay-roll Tax Rates and
Thresholds in Tasmania, Hobart, www.treasury.tas.gov.au/domino/dtf/dtf.nsf/5f
a9db889309e2314a256819001d8eb2/cea6079c368bfc56ca25693f00032739?Ope
nDocument (accessed 6 March 2009).

REFERENCES 409
—— 2008b, Employers Guide to Payroll Tax Payroll Tax Act 2008, Hobart,
www.treasury.tas.gov.au/domino/dtf/dtf.nsf/v-sro/F2078DA7D1560837CA257
02D001AF8D0/$file/EmpGuideToPTX.pdf (accessed 6 March 2009).
Dor, A. and Farley, D.E. 1996, ‘Payment source and the cost of hospital care:
evidence from a multiproduct cost function with multiple payers’, Journal of
Health Economics, vol. 15, pp. 1–21.
Dormont, B. and Milcent, C. 2004, ‘The sources of hospital cost variability’, Health
Economics, vol. 13, pp. 927–39.
Eckermann, S. and Coelli, T. 2008, ‘Including Quality Attributes in a Model of
Health Care Efficiency: A Net Benefits Approach’, Centre for Efficiency and
Productivity Analysis, Working paper no. WP03/2008 University of
Queensland.
Erlandsen, E., 2008, ‘Improving the efficiency of health care spending: what can be
learnt from partial and selected analyses of hospital performance?’ OECD
Journal of Economic Studies, vol. 2008, no. 1, pp. 49–69.
Evans, N., Malcolm, G. and Gordon, A. 2009, Letter in response to ‘Adverse
outcomes of labour in public and private hospitals in Australia’, Medical Journal
of Australia, vol. 191, no. 10, p. 579.
Evans, R.G. and Walker, H. 1972, ‘Information theory and the analysis of hospital
cost structure’, Canadian Journal of Economics, vol. 5, no. 3, pp. 398–418.
Färe, R., Grosskopf, S., Lovell, C.A.K. and Pasurka, C. 1989, ‘Multilateral
productivity comparisons when some outputs are undesirable: a nonparametric
approach’, Review of Economics and Statistics, vol. 71, no. 1, pp. 90–98.
Färe, R., Grosskopf, S., and Valdmanis, V. 1989, ‘Capacity, competition and
efficiency in hospitals: a non-parametric approach’, Journal of Productivity
Analysis, vol. 1, pp. 123–39.
Färe, R., and Primont, D. 1990, ‘A distance function approach to multi-output
technologies’, Southern Economic Journal, vol. 56, pp. 879–91.
—— 1995, Multi-Output Production and Duality: Theory and Applications, Kluwer
Academic, Boston.
Färe, R., Grosskopf, S., Lindgren, B. and Poullier, J. P. 1997, ‘Productivity growth
in health care delivery’, Medical Care, vol. 35, no. 4, pp. 354–66.
Farrell, M.J. 1957, ‘The measurement of productive efficiency’, Journal of the
Royal Statistical Society, vol. 120, no. 3, pp. 253–90.
Feldstein, M.S. 1967, Economic Analysis for Health Service Efficiency, North
Holland, Amsterdam.

410 PUBLIC AND PRIVATE

HOSPITALS
Ferrari, A. 2006, ‘The internal market and hospital efficiency: a stochastic distance
function approach’, Applied Economics, vol. 38, pp. 2121–30.
Ferrier, G. and Valdmanis, V. 1996, ‘Rural hospital performance and its correlates’,
Journal of Productivity Analysis, vol. 7, pp. 63–80.
Folland, S.T. and Hofler, R.A. 2001, ‘How reliable are hospital efficiency
estimates? Exploiting the dual to homothetic production’, Health Economics,
vol. 10, pp. 682–98.
Forero, R. and Hillman, K. 2008, Access Block and Overcrowding: A Literature
Review, Prepared for the Australasian College for Emergency Medicine,
Simpson Centre for Health Services Research, South Western Sydney Clinical
School, University of New South Wales.
Friedman, N., Bull, A., Russo, P., Gurrin, L. and Richards, M. 2007, ‘Performance
of the National Nosocomial Infections Surveillance risk index in predicting
surgical site infection in Australia’, Infection Control and Hospital
Epidemiology, vol. 28, no. 1, pp. 55–59.
Gabbitas, O. and Jeffs, C. 2008, ‘Assessing productivity in the delivery of public
hospital services: some experimental estimates’, Paper presented to the
Australian Health Economics Conference, Adelaide, 2–3 October 2008.
Garling Review (Special Commission of Inquiry into Acute Care Services in NSW
Public Hospitals) 2008, Final Report of the Special Commission of Inquiry Acute
Care Services in NSW Public Hospitals: Overview, Sydney.
Gerdtham, U.G., Löthgren, M., Tambour, M. and Rehnberg, C. 1999, ‘Internal
markets and health care efficiency: a multiple-output stochastic frontier
analysis’, Health Economics, vol. 8, pp. 151–64.
Goss, J. 2008, Projection of Australian Health Care Expenditure by Disease: 2003
to 2033, Australian Institute of Health and Welfare, Cat. no. HWE 43, Canberra.
Granneman, T.W., Brown, R.S. and Pauly, M.V. 1986, ‘Estimating hospital costs: a
multiple-output analysis’, Journal of Health Economics, vol. 5, no. 2,
pp. 107-27.
Graves, N., Halton, K. and Robertus, L. 2008, ‘Costs of Health Care Associated
Infection’, in Cruickshank, M. and Ferguson, J. (eds), Reducing Harm to
Patients from Health Care Associated Infection: The Role of Surveillance,
Australian Commission on Safety and Quality in Health Care, Sydney,
pp. 307–35.
Greene, W.H. 1997, ‘Frontier Production Functions’, in Pesaran, H.M and
Schmidt, P. (eds), Handbook of Applied Econometrics, Volume II,
Microeconomics, Blackwell, Oxford, pp. 81–166.

REFERENCES 411
Grosskopf, S., Margaritis, D., Valdmanis, V., 1995, ‘Estimating output
substitutability of hospital services: a distance function approach’, European
Journal of Operational Research, vol. 80, pp. 575–87.
Grosskopf, S., Hayes, K., Taylor, L. and Weber, W. 1997, ‘Budget constrained
frontier estimation of fiscal equality and efficiency in schooling’, Review of
Economics and Statistics, vol. 79, pp. 116–24.
Harper, R.W. 2007, ‘Drug-eluting coronary stents — a note of caution’, Medical
Journal of Australia, vol. 186, no. 5, pp. 253–55.
Harrison, J.P, Coppola, M.N. and Wakefield, M. 2004, ‘Efficiency of federal
hospitals in the United States’, Journal of Medical Systems, vol. 28, no. 5,
pp. 411–22.
Harrison, J.P. and Sexton, C. 2006, ‘The improving efficiency frontier of not-for-
profit religious hospitals’, Hospital Topics, vol. 84, no. 1, pp. 2–10.
Hasan, M. 2001, ‘Readmission of patients to hospital: still ill defined and poorly
understood: counterpoint’, International Journal for Quality in Health Care,
vol. 13, no. 3, pp. 177–79.
HQCC (Health Quality and Complaints Commission) 2009, Quality of Health
Services — Duty of Provider, Version: 1.1, Brisbane.
Healthscope Limited 2007, Annual Report, Melbourne, www.healthscope.com.au/
info/general/Document/get/329/documentId/ (accessed 8 September 2009).
—— 2009a, About Us, Melbourne, www.healthscopehospitals.com.au/info/general/
Content/get/205/itemId/ (accessed 25 September 2009).
—— 2009b, Careers at Healthscope, www.healthscope.com.au/info/general/
Content/get/221/itemId/ (accessed 6 August 2009).
Heckman, J. 1976, ‘The common structure of statistical models of truncation,
sample selection and limited dependent variables and a simple estimator for such
models,’ Annals of Economic and Social Measurement, vol. 5, no. 4,
pp. 475–92.
Herr, A., 2008, ‘Cost and technical efficiency of German hospitals: does ownership
matter?’, Health Economics, vol. 17, pp. 1057–71.
Hofmarcher, M.M., Paterson, I. and Riedel, M. 2002, ‘Measuring hospital
efficiency in Austria – A DEA Approach’, Health Care Management Science,
vol. 5, pp. 7–14.
Hollingsworth, B. 2008, ‘The measurement of efficiency and productivity of health
care delivery’, Health Economics, vol. 17, pp. 1107–28.

412 PUBLIC AND PRIVATE

HOSPITALS
Hollingsworth, B. and Peacock, S. 2008, Efficiency Measurement in Health and
Health Care, Routledge, London.
Hurley, E., McRae, I., Bigg, I., Stackhouse, L., Boxall, A. and Broadhead, P. 2009,
The Australian Health Care System: The Potential for Efficiency Gains — A
Review of the Literature, Background paper prepared for the National Health and
Hospitals Reform Commission, Canberra, www.nhhrc.org.au
/internet/nhhrc/publishing.nsf/Content/A5665B8B9EAB34B2CA2575CB00184
FB9/$File/Potential%20Efficiency%20Gains%20-%20NHHRC%20Background
%20Paper.pdf (accessed 27 July 2009).
Ipsos Australia 2005, Consumer Survey – Informed Financial Consent,
Melbourne, www.health.gov.au/internet/main/publishing.nsf/Content/6E7A49A
4305C2B72CA25743B000C12BF/$File/consumer_survey.pdf (accessed 30
April 2009).
—— 2007a, Consumer Survey – Informed Financial Consent, Melbourne,
www.health.gov.au/internet/main/publishing.nsf/Content/6E7A49A4305C2B72
CA25743B000C12BF/$File/FINAL%20REPORT_IFC%20survey%202006%20
%2020070313.pdf (access 30 April 2009).
—— 2007b, Ipsos Healthcare and Insurance Australia Report, Melbourne.
—— 2008, Consumer Survey – Informed Financial Consent 2007, Melbourne,
www.health.gov.au/internet/main/publishing.nsf/Content/6E7A49A4305C2B72
CA25743B000C12BF/$File/IFC%20Report%202007%20-20Ipsos%20Final.pdf
(accessed 14 April 2009).
Jackson, T. 2008, ‘Hospital Acquired Diagnoses: At What Cost?’ Presentation to
the National Forum on Safety and Quality in Health Care, Adelaide, 30 October,
www.achs.org.au/pdf/thur.plen4.jackson.pdf (accessed 14 September 2009).
Jackson, T.J., Michel, J.L., Roberts, R.F., Jorm, C.M. and Wakefield, J.G. 2009, ‘A
classification of hospital-acquired diagnoses for use with routine data’, Medical
Journal of Australia, vol. 191, no. 10, pp. 544–8.
Jackson, T.J., Ngheim H.S., Rowell, D., Jorm, C. and Wakefield, J. 2009, ‘Setting
economic priorities for patient safety programs and patient safety research using
casemix costing data’, BMC Health Services Research, vol. 9, Supplement 1, p.
A4.
Jacobs, R. 2001, ‘Alternative methods to examine hospital efficiency: data
envelopment analysis and stochastic frontier analysis’, Health Care
Management Science, vol. 4, no. 2, pp. 103–15.

REFERENCES 413
Jarvis, W. 2003, ‘Benchmarking for prevention: The Centres for Disease Control
and Prevention’s National Nosocomial Infections Surveillance (NNIS) system
experience’, Infection, vol. 31, Supplement 2, pp. 44–8.
Jensen, P.H, Webster, E. and Witt, J. 2007, ‘Hospital Type and Patient Outcomes:
An Empirical Examination Using AMI Re-admission and Mortality Records’,
Melbourne Institute Working paper no. 31/07, Melbourne.
Katz, J.N., Losina, E., Barrett, J., Phillips, C.B., Mahomed, N.N., Lew, R.A.,
Guadagnoli, E., Harris, W.H., Poss, R. and Baron, J.A. 2001, ‘Association
between hospital and surgeon procedure volume and outcomes of total hip
replacement in the United States Medicare population’, Journal of Joint and
Bone Surgery, vol. 83, pp. 1622–9.
KPMG (2009), Report for Catholic Health Australia on Concessional Tax Benefits
for Not-for-profit Hospitals and Aged Care Services, Attachment to submission
made to Contribution of the Not for Profit Sector study,
www.pc.gov.au/__data/assets/pdf_file/0006/92607/subdr198-attachment.pdf
(accessed 24 November 2009).
Kumbhakar, S.C. and Lovell, C.A. Knox 2000, Stochastic Frontier Analysis,
Cambridge University Press, Cambridge.
Lave, J.R. 1966, ‘A Review of the Methods Used to Study Hospital Costs’, Inquiry,
vol. 3, pp. 57–81.
LAOS (Limited Adverse Occurrence Screening) Victoria 2008, Limited Adverse
Occurrence Screening (LAOS): Annual Report for 2007-08, Department of
Human Services, Melbourne, www.health.vic.gov.au/clinrisk/downloads/laos-
prog-report.pdf (accessed 2 August 2009).
Linna, M. 1998, ‘Measuring hospital cost efficiency with panel data models’,
Health Economics, vol. 7, no. 5, pp. 415–27.
Löthgren, M. 2000, ‘Specification and estimation of stochastic multiple-output
production and technical inefficiency’, Applied Economics, vol. 32
pp. 1533–40.
Mangano, M. 2003, A Stochastic Frontier Examination of Victorian Public
Hospitals, Paper presented at 2003 PhD Conference in Economics and Business,
Perth, November.
—— 2006, Frontier Methods for Comparing Public Hospital Efficiency: The Effect
of Casemix Funding in Victoria, PhD Thesis, School of Economics and Finance,
Curtin University of Technology, Western Australia.

414 PUBLIC AND PRIVATE

HOSPITALS
Maniadakis, N. and Thanassoulis, E. 2000, ‘Assessing productivity changes in UK
hospitals reflecting technology and input prices’, Applied Economics, vol. 32,
pp. 1575-89.
Mascinie, E., and Bonten, M. 2005, ‘Vancomycin resistant enterococci:
Consequences for therapy and infection control’, Clinical Microbiology
Infection, Supp, vol. 11, supplement 4, pp. 43–56.
McClure, B. 2004, EV gets into Gear, Investopedia, Edmonton, www.investopedia.
com/articles/fundamental/04/031004.asp (accessed 30 September 2009).
McGregor, A., Riley, T. and Van Gessel, H. 2008, ‘Clostridium difficile associated
disease’, in Cruickshank, M. and Ferguson, J. (eds), Reducing Harm to Patients
from Health Care Associated Infection: The Role of Surveillance, Australian
Commission on Safety and Quality in Health Care, Sydney, pp. 171–87.
McLean, S.M. and Clark, D.J. 2008, ‘A shift in the use of drug-eluting stents in
Australian private hospitals’, Medical Journal of Australia, vol. 188, no. 2,
p. 127.
Medew, J. 2009, ‘Inquiry told of delays in critical care’, The Age, Wednesday
19 August 2009, p. 3.
Medibank Private 2009, Private patients highly satisfied with hospital care, Media
release, 15 September.
Meeusen, W. and van den Broeck, J. 1977, ‘Efficiency estimation from Cobb–
Douglas production functions with composed error’, International Economic
Review, vol. 18, no. 2, pp. 435–44.
Morey, R.C. and Dittman, D.A. 1996, ‘Cost pass-through reimbursement to
hospitals and their impact on operating efficiencies’, Annals of Operations
Research, vol. 67, pp. 117–39.
Mortimer, D. 2002, Competing Methods for Efficiency Measurement: A Systematic
Review of Direct DEA vs SFA/DFA Comparisons, Centre for Health Program
Evaluation, Working paper no. 136, Monash University, Melbourne.
Nayar, P. and Ozcan, Y. 2008, ‘Data envelopment analysis comparison of hospital
efficiency and quality’, Journal of Medical Systems, vol. 32, pp. 198–9.
NHHRC (National Health and Hospitals Reform Commission) 2008, A Healthier
Future For All Australians – Interim Report December 2008, Canberra,
www.nhhrc.org.au/internet/nhhrc/publishing.nsf/Content/BA7D3EF4EC7A1F2
BCA25755B001817EC/$File/NHHRC.pdf (accessed 28 July 2009).
—— 2009, A Healthier Future for All Australians – Final Report June
2009, Canberra, www.nhhrc.org.au/internet/nhhrc/publishing.nsf/Content/1AFD

REFERENCES 415
 EAF1FB76A1D8CA257600000B5BE2/$File/Final_Report_of_the%20nhhrc_Ju
ne_2009.pdf (accessed 28 July 2009).
NHPC (National Health Performance Committee) 2001, National Health
Performance Framework Report, Report to the Australian Health Ministers’
Conference, Queensland Health, Brisbane.
—— 2002, National Report on Health Sector Performance Indicators 2001,
Queensland Health, Brisbane.
—— 2004, National Report on Health Sector Performance Indicators 2003,
Australian Institute of Health and Welfare, Cat. No. HWI 78, Canberra.
NMHWG (National Mental Health Working Group) 2005, Key Performance
Indicators for Australian Public Mental Health Services, ISC Discussion Paper
no.6, Australian Government Department of Health and Ageing, Canberra.
Newhouse, J. 1994, ‘Frontier estimation: how useful a tool for health economics?’,
Journal of Health Economics, vol. 13, pp. 317–22.
Nguyen, K. and Coelli, T. 2009, Quantifying the effects of modelling choices on
hospital efficiency measures: A meta-regression analysis, Centre for Efficiency
and Productivity Analysis, WP07/2009, University of Queensland,
www.uq.edu.au/economics/cepa/docs/WP/WP072009.pdf (accessed 12
November 2009).
Nimmo, G., Coombs, G., Pearson. J., O'Brien, F., Christiansen, K., Turnidge, J.,
Gosbell, I., Collignon, P. and McLaws, M. 2006, ‘Methicillin-resistant
Staphylococcus aureus in the Australian community: an evolving epidemic’,
Medical Journal of Australia, vol. 184, no. 8, pp. 384–8.
NSW Department of Health nd, Monthly Hospital Performance Reports,
www.health.nsw.gov.au/reports/reports.asp, Sydney (accessed 28 July 2009).
—— 2005a, Infection Control Program Quality Monitoring, Policy Directive
PD2005_414, Sydney.
—— 2005b, NSW Patient Safety and Clinical Quality Program, NSW Department
of Health, Sydney, www.health.nsw.gov.au/policies/pd/2005/pdf/PD2005_608.
pdf (accessed 16 September 2009).
—— 2008, Healthcare Associated Infection: Clinical Indicator Manual:
Version 2.0, Sydney.
NSW Nurses’ Association 2009, Ramsay members vote yes!, www.nswnurses.asn.
au/news/19653.html (accessed 5 August 2009).
OSQHC (Office of Safety and Quality in Health Care, Western Australia) nd, Office
of Safety and Quality in Health Care, WA Department of Health,

416 PUBLIC AND PRIVATE

HOSPITALS
 www.safetyandquality.health.wa.gov.au/home/index.cfm (accessed 8 September
2009).
—— 2006, ‘Report on WA data collected by the Australian Incident Monitoring
System (AIMS): 1 October 2005 to 30 December 2005’, WA Department of
Health, Perth www.safetyandquality.health.wa.gov.au/docs/aims/AIMS_
Quarterly_Report_(Oct-Dec05).pdf (accessed 8 September 2009).
Office of the Privacy Commissioner 2009a, Privacy Law, Canberra,
www.privacy.gov.au/law (accessed 12 November 2009).
—– 2009b, The Australian Bureau of Statistics (ABS) has asked me to complete a
survey. Do I have to answer the questions in the survey?, Canberra,
www.privacy.gov.au/faq/individuals/q16 (accessed 12 November 2009).
Office of State Revenue (New South Wales) 2007, NSW Payroll Tax Information
for Employers. Sydney, www.osr.nsw.gov.au/lib/doc/factsheets/ofs_pt3_2007.
pdf (accessed 6 March 2009).
—— 2008, Rates and Thresholds, Sydney, www.osr.nsw.gov.au/taxes/payroll/
general/rates/ (accessed 6 March 2009).
Office of State Revenue (Queensland) 2009a, Deductions, Brisbane,
www.osr.qld.gov.au/payroll-tax/payroll-tax-returns/deductions.shtml (accessed 6
March 2009).
—— 2009b, Exempt Institutions and Exempt Charitable Institutions,.
Brisbane. www.osr.qld.gov.au/legislation-rulings/public-rulings/general/gen003
-1.shtml (accessed 6 March 2009).
—— 2009c, Previous Rates and Thresholds, Brisbane, www.osr.qld.gov.au/
historical-information/previous-rates-thresh/index.shtml (accessed 6 March
2009).
Office of State Revenue (Western Australia) 2007, Payroll Tax – General
Information, Perth, www.dtf.wa.gov.au/cms/uploadedFiles/General_Information
_July_2007.pdf (accessed 6 March 2009).
OECD (Organisation for Economic Co-operation and Development) 2003, Private
Health Insurance in Australia: A Case Study, Paris.
—— 2009a, OECD Health Data 2009: How Does Australia Compare, Paris.
O’Neill, L. 1998, ‘Multifactor efficiency in data envelopment analysis with an
application to urban hospitals’, Health Care Management Science, vol. no. 1, pp.
19–27.

REFERENCES 417
O’Neill, L., Rauner, M., Heidenberger, K., and Kraus, M. 2008, ‘A cross-national
comparison and taxonomy of DEA-Based hospital efficiency studies’,
Socio-Economic Planning Sciences, vol. 42, pp. 158–89.
Paul, C.J.M. 2002, ‘Productive structure and efficiency of public hospitals’, in Fox,
K. J. (ed), Efficiency in the Public Sector, Kluwer Academic Publishers, Boston.
Peacock, S., Chan, C., Mangolini, M. and Johansen, D. 2001, ‘Techniques for
Measuring Efficiency in Health Services’, Staff Working paper, Productivity
Commission, July.
Pesce, A.F. 2009, ‘Private obstetric intervention: good, bad or whatever?, Medical
Journal of Australia, vol. 90, no. 9, pp. 467–468.
Preen D.B., Holman C.D.J., Semmens J.B., Spilsbury K., Brameld K.J. 2006,
‘Length of comorbidity lookback period affected regression model performance
of administrative health data’, Journal of Clinical Epidemiology, vol. 59, no. 9,
pp. 940–6.
PHIAC (Private Health Insurance Administration Council) 2003, Operations of the
Registered Health Benefits Organisations Annual Report 2002-03, Canberra.
—— 2004, Operations of the Registered Health Benefits Organisations Annual
Report 2003-04, Canberra.
—— 2007a, Operations of the Private Health Insurers Annual Report 2006-07,
Canberra.
—— 2007b, Quarterly Gap Payment & Medical Benefit Expenses – June 2007,
Canberra, www.phiac.gov.au/resources/file/gaprptJun2007.pdf (accessed 16 July
2009).
—— 2008, Operations of the Private Health Insurers Annual Report 2007-08,
Canberra.
—— 2009a, (and previous issues), Quarterly Gap Payment & Medical Benefit
Expenses – September 2009, Canberra, www.phiac.gov.au/resources/file/P1Gap-
Sep09.pdf (accessed 25 November 2009).
—— 2009b, (and previous issues), PHIAC A Report – June 2009, Canberra,
www.phiac.gov.au/resources/file/phiacadata/alljun09.pdf (accessed 25
September 2009).
—— 2009c, Quarterly Statistics: March 2009, Canberra.
PC (Productivity Commission) 1999a, An Assessment of the Performance of
Australian Railways: 1990 to 1998, Supplement to Inquiry Report on Progress in
Rail Reform, AusInfo, Canberra.

418 PUBLIC AND PRIVATE

HOSPITALS
—— 1999b, Private Hospitals in Australia, Commission Research paper, AusInfo,
Canberra.
—— 2005, Economic Implications of an Ageing, Research Report, Canberra.
—— 2006a, Potential Benefits of the National Reform Agenda, Report to the
Council of Australian Governments, Canberra.
—— 2006b, Waste Management, Inquiry Report, Canberra.
—— 2008a, Assessing Local Government Revenue Raising Capacity, Research
Report, Canberra.
—— 2008b, Chemicals and Plastics Regulation, Research Report, Canberra.
—— 2009, Contribution of the Not-for-Profit Sector, Draft Research Report,
Canberra.
Prior, D. 2006, ‘Efficiency and total quality management in health care
organizations: a dynamic frontier approach’, Annals of Operations Research,
vol. 145, pp. 281–299.
Private Health Insurance Ombudsman 2009, Glossary, www.privatehealth.gov.
au/glossary.htm (accessed 16 July 2009).
Quan, H., Sundarajan, V., Halfon, P., Fong, A., Burnand, B., Luthi, J-C, Saunder,
L.D., Beck, C.A., Feasby, T.E. and W.A. Ghali 2005, ‘Coding algorithms for
defining comorbidities in ICD-9-CM and ICD-10 administrative data’, Medical
Care, vol. 43 no. 1, pp. 1130-9.
Queensland Department of Health 2004, Measured Quality Service, Report to
Board of Management, Brisbane.
RACS (Royal Australasian College of Surgeons) 2008, Surgical Audit and Peer
Review: A Guide by the Royal Australasian College of Surgeons, 3rd edn,
Melbourne.
Ramsay Health Care Limited 2007, Annual Report, Sydney, www.ramsayhealth.
com.au/annualreport2007/Annual%20Report%202007%20web.pdf (accessed 8
September 2009).
—— 2008, Annual Report, Sydney, www.ramsayhealth.com.au/annualreport2008
/Annual%20Report%202008%20web.pdf (accessed 14 September 2009).
—— 2009a, About Ramsay Health Care, Sydney, www.ramsayhealth.com.au
/profile/profile.asp (accessed 9 September 2009).
—— 2009b, Opportunities for Doctors, www.opps4docs.com.au/specialties/
specialists.asp (accessed 5 August 2009).

REFERENCES 419
—— 2009c, The Ramsay Way, www.ramsayhealth.com.au/news/documents/TRW
_09_Sept.pdf (accessed 20 November 2009).
Revenue SA 2008, Payroll Tax – A guide to legislation, Adelaide,
www.revenuesa.sa.gov.au/circulars/prtguidec303.pdf (accessed 10 March 2009)
—— 2009, Payroll Tax, Adelaide, www.revenuesa.sa.gov.au/index.html? www.
revenuesa.sa.gov.au/taxes/payroll.html (accessed 6 March 2009).
Richardson, J. (2005), Priorities of health policy: cost shifting or population health,
Australia and New Zealand Health Policy, vol. 2, issue 1.
Richards, M. and Russo, P. 2007, ‘Surveillance of hospital-acquired infections in
Australia — one nation, many states’, Journal of Hospital Infection, vol. 65,
no. 52, pp. 174–81.
Robinson, M. 2008, ‘Standing Committee on Economics’, Debates, 21 July.
Robson, S.J., Laws, P. and Sullivan, E.A., 2009, ‘Adverse outcomes of labour in
public and private hospitals in Australia’, Medical Journal of Australia, vol.
190, no. 9, p. 474-7.
Rosko, M.D. and Chilingerian, J.A. 1999, ‘Estimating hospital inefficiency: does
casemix matter?’, Journal of Medical Systems, vol. 23, no. 1, pp. 57–71.
Royal College of Pathologists Australasia (RCPA) 2009, Manual of Use and
Interpretation of Pathology Tests, 5th edn, www.rcpamanual.edu.au/sections
/clinicalproblem.asp?s=25&i=109 (accessed 24 August 2009).
SA Department of Health 2005a, MRO Surveillance Definitions, Infection Control
Service, Adelaide.
—— 2005b, South Australian Statewide Nosocomial Bloodstream Infection Report
2005, Infection Control Service, Adelaide.
—— 2006, Improving the System: South Australian Patient Safety Report 2003-04,
Adelaide.
—— 2008, Methicillin Resistant Staphylococcus Aureus (MRSA): Frequently Asked
Questions, Adelaide.
—— 2009a, Statewide MRSA Report: All Contributors: March 2009, Infection
Control Service, Adelaide.
—— 2009b, Vancomycin-Resistant Enterococci (VRE): Frequently Asked
Questions, Adelaide.
Sammut, J. 2009, Why Public Hospitals are Overcrowded: Ten Points for
Policymakers, Policy Monograph no. 99, Centre for Independent Studies,
Sydney.

420 PUBLIC AND PRIVATE

HOSPITALS
Schmidt, P. and Sickles, R.C. 1984, ‘Production frontiers and panel data’, Journal
of Business and Economic Statistics, vol. 2, pp. 367–74.
Scott, A. and Parkin, D. 1995, ‘Investigating hospital efficiency in the new NHS:
the role of the translog cost function’, Health Economics, vol. 4, pp. 467–78.
SCRGSP (Steering Committee for the Review of Government Service Provision)
2004, Report on Government Services 2004, Productivity Commission,
Canberra.
—— 2005, Report on Government Services 2005, Productivity Commission,
Canberra.
—— 2009, Report on Government Services 2009, Productivity Commission,
Canberra.
SCRCSSP (Steering Committee for the Review of Commonwealth/State Service
Provision) 1997, Data Envelopment Analysis, A Technique for Measuring the
Efficiency of Government Service Delivery, AGPS, Canberra.
—— 2001, Asset Measurement In The Costing of Government Services,
Productivity Commission, Canberra.
Sherman, H.D. 1984, ‘Hospital efficiency measurement and evaluation: empirical
test of a new technique’, Medical Care, vol. 22, no. 10, pp. 922–38.
Siciliani, L. 2006, ‘Estimating technical efficiency in the hospital sector with panel
data: a comparison of parametric and non-parametric techniques’, Applied
Health Economics and Health Policy, vol. 5, no. 2, pp. 99–116.
Skinner, J. 1994, ‘What do stochastic frontier functions tell us about inefficiency?’,
Journal of Health Economics, vol 13, pp. 317–22.
Solà, M. and Prior, D. 2001, ‘Measuring productivity and quality changes using
data envelopment analysis: an application to Catalan hospitals’, Financial
Accountability and Management, vol. 17, no. 3, pp. 219–45.
Spiegelhalter, D. 2004, ‘Funnel plots for comparing institutional performance’,
Statistics in Medicine, vol. 24, no. 8, pp. 1185–202.
Sportsmed SA 2009, Orthopaedics, Adelaide, www.sportsmed.com.au/AboutUs/
OURSERVICES/ORTHOPAEDICS/tabid/68/Default.aspx (accessed 14 April
2009).
Sprivulis, P.C., Da Silva, J-A., Jacobs, I.G., Frazer, A.R.L. and Jelinek, G.A. 2006,
‘The association between hospital overcrowding and mortality among patients
admitted via Western Australian emergency departments’, Medical Journal of
Australia, vol. 184, no. 5, pp. 208–12.

REFERENCES 421
State Revenue Office (Victoria) 2009a, Exemptions / Concessions, Melbourne,
www.sro.vic.gov.au/sro/srowebsite.nsf/taxes_exemptions.htm#3 (accessed 6
March 2009).
—— 2009b, Payroll Tax Historical Rates - From 1 July 1997 to 30 June
2006, Melbourne, www.sro.vic.gov.au/sro/srowebsite.nsf/taxes_payroll_hist_rat
es.htm (accessed 6 March 2009).
Street, A. 2003, ‘How much confidence should we place in efficiency estimates?’
Health Economics, vol. 12, pp. 895–907.
Sundarajan, V., Henderson, T, Perry, C., Muggivan, A., Quan, H. and Ghali, W.A.
2004, New ICD version of the Charlson Index predicted in-hospital mortality,
Journal of Clinical Epidemiology, vol. 57, pp. 1288-94.
Territory Revenue Office (Northern Territory) 2008a, Employers' Guide to Payroll
Tax in the Northern Territory, Darwin, www.nt.gov.au/ntt/revenue/pdf/I-PRT-
001.pdf (accessed 6 March 2009).
—— 2008b, Rates and Thresholds, Darwin, www.nt.gov.au/ntt/revenue/rates
_thresholds.shtml (accessed 6 March 2009).
Thiel, H. 1967, Economics and Information Theory, North-Holland, Amsterdam.
Thomas, C. and Riley, T. 2003, ‘Restriction of Third Generation Cephalosporin Use
Reduces the Incidence of Clostridium Difficile-Associated Diarrhoea in
Hospitalised Patients’, Communicable Diseases Intelligence, vol. 27,
Supplement: Antimicrobial Resistance in Australia, pp. S28–S31.
TIPCU (Tasmanian Infection Prevention and Control Unit) 2009, Tasmanian Acute
Public Hospitals Healthcare Associated Infection Surveillance Report: March
2009, Department of Health and Human Services, Hobart.
Tong, E., Clements, A., Haynes, M., Jones, M., Morton, A. and Whitby, M. 2009,
Improved hospital-level risk adjustment for surveillance of healthcare-associated
bloodstream infections: a retrospective cohort study, BMC Infectious Diseases,
vol. 9, pp. 145–53.
Tracy, S.K., Welsh, A.W., Dahlen, H.G. and Tracy, M.B. 2009, Letter in response
to ‘Adverse outcomes of labour in public and private hospitals in Australia’,
Medical Journal of Australia, vol. 191, no. 10, pp. 579–80.
Treasury 2007, Intergenerational Report 2007, Canberra.
—— 2008a, Australia’s Future Tax System: Consultation paper, Canberra.
—— 2008b, Budget Measures 2008-09, Budget paper no. 2, Canberra.
—— 2008c, Tax Expenditure Statement, Canberra.
—— 2009, Budget Overview 12 May 2009, Canberra.
422 PUBLIC AND PRIVATE
HOSPITALS
Turrell, G., Stanley, L., de Looper, M. and Oldenburg, B. 2006. Health Inequalities
in Australia: Morbidity, Health Behaviours, Risk Factors and Health Service
Use. Health Inequalities Monitoring Series No. 2, AIHW, Cat. No. PHE 72.
Queensland University of Technology and the Australian Institute of Health and
Welfare, Canberra.
Victorian Auditor-General 2009, Access to Public Hospitals: Measuring
Performance, 2008-09:18, Melbourne.
Victorian Department of Human Services 2008a, Type 1 Surveillance Manual,
Version 6, VICNISS Coordinating Centre, Melbourne.
—— 2008b, Hospital Acquired Infection Project Year 6 Report — August 2008,
Statewide Quality Branch, Rural and Regional Health and Aged Care Services,
Melbourne, www.vicniss.org.au/Resources/VICNISSAnnualReport0808.pdf
(accessed 11 August 2009).
—— 2009, Victoria: The 2008-09 Statement of Priorities and Performance
Framework Business Rules, Melbourne, www.health.vic.gov.au/hospital-
performance/busrule09.pdf (accessed 11 August 2009).
Vitikainen, K., Street, A., Linna, M. 2009, ‘Estimation of hospital efficiency-Do
different definitions and casemix measures for hospital outputs affect the
results?’, Health Policy, vol. 89, pp. 149–59.
Vitiliano D.F., and Toren, M., 1994, ‘Cost and efficiency in nursing homes: a
stochastic frontier approach’, Journal of Health Economics, vol. 13, pp. 218–
300.
WA Department of Health 2008, Healthcare Infection Surveillance Western
Australia Surveillance Manual, Version 2, Healthcare Associated Infection Unit,
Perth.
—— 2009a, Healthcare-Associated Infection Surveillance In Western Australia,
Healthcare Associated Infection Unit, Operational Directive OD 0197/09, Perth.
—— 2009b, Healthcare Infection Surveillance Western Australia Annual Report
2007–2008, Healthcare Associated Infection Unit, Perth.
Wang, J. and Mahmood, A. 2000a, ‘Efficiency of the NSW Public Acute Hospitals:
An Application of the Data Envelopment Analysis’, in Bridges, J. (ed),
Economics and Health: 2000 – Proceedings of the Twenty-second Australian
Conference of Health Economists, University of New South Wales School of
Health Services Management, Sydney.
—— 2000b, ‘Relative Efficiency of NSW Public Acute Hospitals: A Stochastic
Frontier Cost Function Analysis’, in Bridges, J. (ed), Economics and Health:
2000 – Proceedings of the Twenty-second Australian Conference of Health

REFERENCES 423
 Economists, School of Health Services Management, University of New South
Wales, Sydney.
Webster, R., Kennedy, S. and Johnson, L. 1998, Comparing Techniques for
Measuring the Efficiency and Productivity of Australian Private Hospitals,
Working paper no. 98/3, Australian Bureau of Statistics, Cat. no. 1351.0,
Canberra.
Weng, S-J., Wu, T., Blackhurst, J. and Mackulak, G. 2009, ‘An extended DEA
model for hospital performance evaluation and improvement’, Health Services
and Outcomes Research Methodology, vol. 9, no. 1, pp. 39–53.
Wilson, R.M., Runciman, W.B., Gibberd, R.W., Harrison, B.T., Newby, L. and
Hamilton, J.D. 1995, ‘The quality in Australian health care study’, Medical
Journal of Australia, vol. 163, no. 9, pp. 458–71.
Women’s Hospitals Australasia 2007, Supporting Excellence in Maternity Care:
The Core Maternity Indicators Project, www.safetyandquality.gov.au/
internet/safety/publishing.nsf/Content/3A59DB5FECB57A99CA25753C001B50
ED/$File/CMI-Report-Jan2007.pdf (accessed 14 September 2009).
Wooldridge, M. 1996, (Minister for Health and Aged Care), Medicare Levy
Amendment Bill (No. 2) 1996 – Second Reading Speech, Debates,
13 December.
Wooldridge, M. (Minister for Health and Family Services) 1997, Legislation
Introduced to Reform Private Health Insurance, Media release no. 140/97,
27 November.
Workplace Authority 2008a, ANF (Vic. Branch) and HSU (Vic. Branch 1) and The
Healthscope Nurses Agreement, Melbourne. www.ca.workplaceauthority.
gov.au/docs/CAs/CAUN085186714.pdf (accessed 20 November 2009).
—— 2008b, Epworth and Australian Nursing Federation Enterprise Agreement
2008-2012, Melbourne, www.ca.workplaceauthority.gov.au/docs/CAs/
CAUN085284474.pdf (accessed 20 November 2009).
Worthington, A. 2004, ‘Frontier efficiency measurement in health care: a review of
empirical techniques and selected applications’, Medical Care Research Review,
vol. 61, pp. 135–70.
Yaisarwang, S. and Burgess, J.F. 2006, ‘Performance-based budgeting in the public
sector: an illustration from the VA health care system’, Health Economics,
vol. 15, pp. 295–310.
Yong, K. and Harris, A. 1999, Efficiency of Hospitals in Victoria Under Casemix
Funding: A Stochastic Frontier Approach, Centre for Health Program
Evaluation, Working paper no. 92, Monash University, Melbourne.

424 PUBLIC AND PRIVATE

HOSPITALS
Zuckerman, S., Hadley, J. and Iezzoni, L. 1994, ‘Measuring hospital efficiency with
frontier cost functions’, Journal of Health Economics, vol. 13, pp. 255–80.

REFERENCES 425