Annals of Internal Medicine Article

Liability Claims and Costs Before and After Implementation of a

Medical Error Disclosure Program
Allen Kachalia, MD, JD; Samuel R. Kaufman, MA; Richard Boothman, JD; Susan Anderson, MBA, MSN; Kathleen Welch, MS, MPH;
Sanjay Saint, MD, MPH; and Mary A.M. Rogers, PhD

Background: Since 2001, the University of Michigan Health System (RR, 0.35 [CI, 0.22 to 0.58]). Median time from claim report-
(UMHS) has fully disclosed and offered compensation to patients ing to resolution decreased from 1.36 to 0.95 years. Average
for medical errors. monthly cost rates decreased for total liability (RR, 0.41 [CI,
0.26 to 0.66]), patient compensation (RR, 0.41 [CI, 0.26 to
Objective: To compare liability claims and costs before and after 0.67]), and non– compensation-related legal costs (RR, 0.39
implementation of the UMHS disclosure-with-offer program. [CI, 0.22 to 0.67]).
Design: Retrospective before–after analysis from 1995 to 2007. Limitations: The study design cannot establish causality. Malprac-
Setting: Public academic medical center and health system. tice claims generally declined in Michigan during the latter part of
the study period. The findings might not apply to other health
Patients: Inpatients and outpatients involved in claims made to systems, given that UMHS has a closed staff model covered by a
UMHS. captive insurance company and often assumes legal responsibility.

Measurements: Number of new claims for compensation, number Conclusion: The UMHS implemented a program of full disclosure
of claims compensated, time to claim resolution, and claims-related of medical errors with offers of compensation without increasing its
costs. total claims and liability costs.

Results: After full implementation of a disclosure-with-offer pro- Primary Funding Source: Blue Cross Blue Shield of Michigan
gram, the average monthly rate of new claims decreased from 7.03 Foundation.
to 4.52 per 100 000 patient encounters (rate ratio [RR], 0.64 [95%
CI, 0.44 to 0.95]). The average monthly rate of lawsuits de- Ann Intern Med. 2010;153:213-221. www.annals.org
creased from 2.13 to 0.75 per 100 000 patient encounters For author affiliations, see end of text.

E thical obligations and patient safety principles support

prompt disclosure of harmful medical errors (1– 4).
Disclosure can strengthen trust in the patient–physician
program on liability-related performance (Appendix, avail-
able at www.annals.org).

relationship and is widely acknowledged as the “right METHODS

thing” for hospitals and physicians to do (5– 8). However, The Claims Model
fears that disclosure will invite new claims or complicate
The current UMHS claims management program has
subsequent litigation can inhibit the impulse to disclose
been described in detail (17). In brief, before 2001,
(9 –11). In practice, disclosure may not occur as frequently
UMHS pursued a traditional approach to risk and claims
as we might hope (10 –15).
management. Once received, claims for compensation
Whether more disclosure will increase or decrease
were typically assigned to a defense counsel. A claims man-
liability remains unclear. Some physicians and risk manag-
agement committee would ultimately review all claims and
ers worry that admitting a medical error may amount to
advise on settling or going to trial.
handing over a “blank check” and invite lawsuits and dis-
In July 2001, UMHS began responding to all open
putes about compensation amounts (16). Others counter
and new malpractice claims by admitting fault and offering
that prompt disclosure may actually reduce liability be-
cause patients primarily seek the facts, a sincere apology, a
commitment to prevent the error from recurring, and fair
compensation (17, 18). The debate continues amid a lack See also:
of widely generalizable data on disclosure’s effect on liabil-
Print
ity (19, 20).
Editors’ Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
In 2001, the University of Michigan Health System
Editorial comment. . . . . . . . . . . . . . . . . . . . . . . . . . 266
(UMHS) launched a comprehensive claims management
Summary for Patients. . . . . . . . . . . . . . . . . . . . . . . I-28
model with disclosure as its centerpiece (17). Emphasizing
transparent communication, the UMHS program has re- Web-Only
ceived national attention for its process of disclosure with Appendix
offer of compensation for harmful medical errors (1, 8). To Appendix Figure
better understand the relationship between disclosure and Conversion of graphics into slides
malpractice liability, we evaluated the effect of the UMHS
© 2010 American College of Physicians 213
Article Liability Claims and Costs With a Medical Error Disclosure Program

Context
tional review board from the University of Michigan ap-
proved the study protocol.
The University of Michigan Health System performs active
surveillance for medical errors, fully discloses found errors Data Sources and Measures
to patients, and offers compensation when it is at fault. We linked 2 data sets, the UMHS risk management
database (which contains claims-related performance data,
Contribution
such as injury and disposition dates, disposition status, and
This analysis found a decrease in new legal claims, number liability costs) and the Clinical Information & Decision
of lawsuits per month, time to claim resolution, and costs Support Services database, to assess 4 primary study mea-
after implementation of the program of disclosure with sures: number of new claims, number of claims receiving
offer of compensation.
compensation, time to claim resolution, and claims-related
Caution costs.
Similar findings were reported in Michigan generally We defined a claim as any request for compensation
through the latter part of the study period. for an unanticipated medical outcome whether initiated by
the patient (or a family member or attorney) or by disclo-
Implication sure. We were unable to categorize the source of the claim
A disclosure-with-offer approach to medical errors did (patient- or family-initiated vs. UMHS-initiated) because
not increase legal claims and costs at a large U.S. health UMHS opted not to distinguish between the sources upon
system during the past 10 years. initiation of its program. We attempted to determine the
number of lawsuits that had resulted from disclosure but
—The Editors could not reliably do so. Claims counts are by occurrence
regardless of the number of caregivers named. We excluded
service recovery claims (those brought by patients without
compensation when an internal investigation reveals med- an attorney and compensated for less than $5000) because
ical error. If an investigation reveals no error, UMHS pro- payments for those claims primarily serve to restore patient
vides the reasons for its conclusion and vigorously defends relations rather than redress a harmful error. We also ex-
a claim, if necessary. In April 2002, UMHS began linking cluded off-site claims (those involving a UMHS physician
the investigation process with peer review and quality im- at a non-UMHS facility) because off-site locations fre-
provement efforts. quently lacked a formal disclosure program.
By February 2003, the disclosure program was fully We calculated the monthly rate of new claims by using
integrated with patient safety efforts. The program now the number of claims in a month (based on report date) for
identifies patient injuries through various means, including the numerator and the number of patient encounters in
reporting by employees, patients or family members, or that month as the denominator. Patient encounters were
patients’ attorneys. It uses experienced risk managers with defined as the sum of hospital discharges and outpatient
clinical backgrounds to lead investigations and mediate pa- visits for that month.
tient concerns as facts are collected, care quality is evalu- We defined total liability costs as the sum of all patient
ated, and conclusions are disclosed. The UMHS empha- compensation and legal costs incurred by UMHS. Patient
sizes honesty and transparency with patients and staff, compensation costs included amounts paid to the patients
regardless of whether events resulted from error, and en- (or families) and lien holders. Total legal costs were pri-
courages staff to enlist risk management in the disclosure marily defense attorney and expert fees but also included
process. lawsuit-associated items, such as filing fees. Because costs
Settlements, if made, generally occur in the institu- are presented here as a portion of clinical operating reve-
tion’s name, in line with common practice at many insti- nue, we adjusted them to 2007 U.S. dollars by using the
tutions with closed medical staffs (Appendix). Conse- cost of medical care from the U.S. Bureau of Labor Statis-
quently, reporting of individual caregivers in medical tics. We calculated the monthly liability cost rate by using
malpractice claims in the National Practitioner Data Bank total costs (by claim report date) in a given month as the
is rare. However, full claims histories are maintained and numerator and total operating revenue in that month as
reported for each involved caregiver, as required. the denominator.
Design Statistical Analysis
We used a before–after approach to evaluate the To evaluate differences in the rate of claims in the
UMHS program. The study period included claims re- periods before and after disclosure, we used negative bino-
ported to risk management from 1 July 1995 to 30 Sep- mial generalized linear models (GLMs) with a log link.
tember 2007, with 1 July 2001 as the date of initial im- Rate ratios (RRs) were calculated by comparing the rate of
plementation of the disclosure program and 1 February claims before implementation of the disclosure program
2003 as the date of full implementation. We categorized with that after full implementation. For the main analyses,
claims on the basis of their date of reporting. The institu- only those claims that were reported and closed before the
214 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 www.annals.org
 Liability Claims and Costs With a Medical Error Disclosure Program Article

initial implementation of the policy (1 July 2001) were design, conduct, analysis, or decision to submit this manu-
included in the period before the program (reference group). script for publication.
We included claims that were reported after the full imple-
mentation of the policy (1 February 2003) in the period
after disclosure. We used a locally weighted, scatterplot RESULTS
smoother (lowess) to display trends over time. Piecewise Total Claims
regression with linear splines was used to assess the differ- The UMHS risk management claims database con-
ence in trends before and after initial implementation of tained 1131 claims for the study period after excluding
the disclosure program (1 July 2001). service recovery (n ⫽ 33) and off-site cases (n ⫽ 145).
We conducted survival analyses to assess the time to Mean patient age was 40.4 years (SD, 20.7) at the time of
claim resolution. Kaplan–Meier estimators of the survivor- injury; 87% of patients were white, 52% were women, and
ship function were plotted for claims reported before the 55% were inpatients.
initial implementation of the disclosure program (1 July Of the 1131 total claims, 633 were asserted before and
2001) and those reported afterward. A Cox proportional 498 after implementation of the disclosure-with-offer pro-
hazards regression model was used to calculate hazard ra- gram. Of the claims made before implementation, 632
tios for the difference in resolution rates after versus before were closed as of 31 March 2008 and 319 (50.5% [95%
(reference group) implementation of the program; visual CI, 46.5% to 54.4%]) were compensated, compared with
inspection of the survival curves suggested no violation of 463 closed and 198 (42.8% [CI, 38.2% to 47.4%]) com-
the proportional hazards assumption. We adjusted for pa- pensated after program implementation (P ⫽ 0.012). This
tient age, sex, inpatient status, and whether the claim went averaged 53.2 paid claims per year before and 31.7 after
to trial. the program began.
To assess the differences in costs of the program before Claims Rates
and after disclosure, we used GLM with a gamma distri- The monthly rate of new claims decreased from 7.03
bution and log link. At the time of analysis, 36 claims that (CI, 5.98 to 8.08) per 100 000 patient encounters before
had been filed during the study period remained open (that initial program implementation to 4.52 (CI, 3.96 to 5.08)
is, cases that had not been closed as of 31 March 2008); we after full implementation (RR, 0.64 [CI, 0.44 to 0.95])
included the costs for these open claims in the statistical (Table 1). The trend in monthly rate was stable before
analyses, as incurred as of 31 March 2008. (⫺0.002 [CI, ⫺0.050 to 0.046]; P ⫽ 0.935) but decreased
Two sensitivity analyses were conducted for the cost after the program was initially implemented in July 2001
data. The first excluded outliers in the GLM. Outliers were (⫺0.061 [CI, ⫺0.082 to ⫺0.040]; P ⬍ 0.001), which was
determined by calculating costs that exceeded the product a statistically significant before–after difference (⫺0.059
of 1.5 and the limits of the interquartile range (IQR). The [CI, ⫺0.110 to ⫺0.008]; P ⫽ 0.023) (Figure 1).
second sensitivity analysis used the date of the initial im- Changes in rates of claims before and after program
plementation of the disclosure program (1 July 2001) as implementation were statistically significant only for claims
the cut point for the before-versus-after comparison. In that resulted in a lawsuit. The UMHS experienced 232
addition, we included all claims on the basis of the date of lawsuits (38.7 per year) before and 106 (17.0 per year)
the reported claim, regardless of disposition status. after program implementation. A decrease was still evident,
All analyses were 2-tailed, with an ␣ level of 0.05. We assuming all cases that were open at the end of the obser-
performed analyses in Stata SE, version 10.0 (StataCorp, vation period (1 before and 35 after implementation) re-
College Station, Texas). sulted in lawsuits, with 233 lawsuits (38.8 per year) before
Role of the Funding Source and 141 lawsuits (22.6 per year) after program implemen-
Blue Cross Blue Shield of Michigan Foundation tation. Monthly lawsuit rates decreased from 2.13 (CI,
funded this study. The funding source had no role in the 1.58 to 2.67) per 100 000 patient encounters before initial

Table 1. Monthly Rates and Incidence Rate Ratios of New Claims Before and After Full Implementation of the University of
Michigan Health System Disclosure-With-Offer Program

Variable Mean Monthly Rate (95% CI)* Rate Ratio (95% CI)† P Value

Before After
Total claims 7.03 (5.98–8.08) 4.52 (3.96–5.08) 0.64 (0.44–0.95) 0.025
Lawsuits 2.13 (1.58–2.67) 0.75 (0.47–1.03) 0.35 (0.22–0.58) ⬍0.001
All other claims 4.90 (4.17–5.63) 3.77 (3.27–4.26) 0.77 (0.52–1.14) 0.191

* Number of claims in 1 mo per 100 000 patient encounters.

† Rate after full implementation compared with rate before program began (reference group).

www.annals.org 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 215
Article Liability Claims and Costs With a Medical Error Disclosure Program

Figure 1. Monthly rates of new claims before and after implementation of the University of Michigan Health System
disclosure-with-offer program.

25
Total Claims per Month (per 100 000 Patient Encounters)

20

Initial implementation
Full implementation

15

10

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Year

12
(per 100 000 Patient Encounters)

10
Lawsuits per Month

16
(per 100 000 Patient Encounters)
Nonlawsuit Claims per Month

14
12
10
8
6
4
2
0

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Year

216 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 www.annals.org
 Liability Claims and Costs With a Medical Error Disclosure Program Article

implementation to 0.75 (CI, 0.47 to 1.03) per 100 000

Figure 2. Time to claim resolution before and after
patient encounters after full implementation (RR, 0.35
implementation of the University of Michigan Health
[CI, 0.22 to 0.58]) (Table 1). The trend in monthly
System disclosure-with-offer program.
rates of lawsuits before and after initial implementation
of the program demonstrated a significant change (dif-
ference in trend, ⫺0.028 [CI, ⫺0.054 to ⫺0.001]; P ⫽ 1.00

Proportion Unresolved
0.04) (Figure 1). Before
0.75
In contrast, there was no change in the rate of claims After
that did not result in a lawsuit after the program was fully
0.50
implemented (RR, 0.77 [CI, 0.52 to 1.14]) (Table 1), with
no significant change in trend after initial implementation
(difference in trend, ⫺0.031 [CI, ⫺0.070 to 0.007]; P ⫽ 0.25

0.108).
0.00
Our primary analyses counted only claims reported
0 2 4 6 8
and closed before initial program implementation as “be-
Years
fore” claims, but findings were similar when we compared Unresolved claims, n
rates of all claims reported before and after initial program Before 633 220 40 6 0
implementation, independent of time of disposition (RR After 498 118 13 0 0
for all claims, 0.59 [CI, 0.41 to 0.83]; RR for claims re-
sulting in lawsuits, 0.34 [CI, 0.23 to 0.51]; and RR for
claims not resulting in lawsuits, 0.73 [CI, 0.51 to 1.04]).
⫺0.050]; P ⫽ 0.024) costs (Figure 3). Although the total
Time to Resolution costs associated with lawsuits decreased after full imple-
Median time to claim resolution was 1.36 years (IQR, mentation (RR, 0.27 [CI, 0.13 to 0.54]), the total costs for
0.72 to 2.44 years) before initial implementation and 0.95 nonlawsuit claims did not (RR, 0.81 [CI, 0.47 to 1.38])
year (IQR, 0.55 to 1.96 years) after initial program imple- (Table 2).
mentation (Figure 2); the rate of resolution increased after In a sensitivity analysis excluding outliers, results were
program implementation with an adjusted hazard ratio qualitatively similar for claims overall (RR, 0.66 [CI, 0.44
of 1.27 (CI, 1.11 to 1.45; P ⬍ 0.001). No effect mod- to 0.98]), type of claim (RR for lawsuits, 0.35 [CI, 0.21 to
ification by site (inpatient vs. outpatient) occurred. 0.58]; RR for nonlawsuits, 0.77 [CI, 0.52 to 1.14]), and
Liability Costs type of costs (RR for legal costs, 0.43 [CI, 0.26 to 0.70];
Median and mean total liability costs decreased after RR for patient compensation, 0.45 [CI, 0.31 to 0.67]). We
full program implementation (RR for mean costs, 0.41 also found similar results in a sensitivity analysis by using
[CI, 0.26 to 0.66]; P ⬍ 0.001), attributable to decreases in the date of initial rather than full program implementation
both legal and patient compensation costs (Table 2). After as the before–after marker (data not shown).
initial program implementation, total cost rates signifi- The average cost per lawsuit significantly decreased
cantly decreased (difference in trend, ⫺0.449 [CI, ⫺0.806 from $405 921 before to $228 308 after initial program
to ⫺0.092]; P ⫽ 0.014) as did legal (difference in trend, implementation (RR, 0.40 [CI, 0.24 to 0.68]; P ⫽ 0.001).
⫺0.066 [CI, ⫺0.111 to ⫺0.022]; P ⫽ 0.004) and patient Costs did not change for nonlawsuits (RR, 1.27 [CI, 0.73
compensation (difference in trend, ⫺0.383 [CI, ⫺0.715 to to 2.23]; P ⫽ 0.397). This pattern was similar when time

Table 2. Monthly Rates of Liability Costs Before and After Full Implementation of the University of Michigan Health System
Disclosure-With-Offer Program

Category Median Cost Rate (IQR)* Mean Cost Rate (95% CI)* Rate Ratio (95% CI)†

Before After Before After

All liability costs 8.48 (3.24–21.48) 4.00 (1.16–9.31) 18.91 (12.61–25.21) 7.78 (5.14–10.42) 0.41 (0.26–0.66)
Type of claim
Lawsuit 4.06 (0.02–13.95) 0 (0–3.65) 13.85 (8.26–19.43) 3.71 (1.46–5.95) 0.27 (0.13–0.54)
Nonlawsuit 1.02 (0.13–5.95) 2.45 (0.29–5.61) 5.06 (3.07–7.04) 4.07 (2.55–5.60) 0.81 (0.47–1.38)
Type of costs
Patient compensation 7.88 (2.11–19.09) 3.56 (1.07–8.00) 16.64 (10.90–22.38) 6.90 (4.51–9.30) 0.41 (0.26–0.67)
Legal 0.95 (0.18–2.87) 0.19 (0.01–0.89) 2.26 (1.46–3.06) 0.88 (0.48–1.27) 0.39 (0.22–0.67)

IQR ⫽ interquartile range.

* Costs in 1 mo per $1000 operating revenue.
† Generalized linear models comparing average monthly costs after full implementation with average monthly costs before implementation (values ⬍1.0 represent a decrease
in costs).

www.annals.org 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 217
Article Liability Claims and Costs With a Medical Error Disclosure Program

Figure 3. Monthly rates of legal and patient compensation costs before and after implementation of the University of Michigan
Health System disclosure-with-offer program.

125

100
(per $1000 Operating Revenue), $
Total Liability Costs per Month

Initial implementation
Full implementation
75

50

25

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Year

14
(per $1000 Operating Revenue), $

12
Legal Costs per Month

10

0
Patient Compensation Costs per Month

125
(per $1000 Operating Revenue), $

100

75

50

25

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Year

218 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 www.annals.org
 Liability Claims and Costs With a Medical Error Disclosure Program Article

trends were considered, with a significant decrease for law- erned by the Federal Tort Claims Act and serves a popu-
suit costs (before–after change in slope, ⫺7848.73 [CI, lation largely restricted to military veterans.
⫺14035.56 to ⫺1661.90]; P ⫽ 0.013) but not nonlaw- The near-absence of data on disclosure’s direct effect
suits (before–after change in slope, ⫺1119.33 [CI, on liability risk has led investigators to examine other in-
⫺3831.10 to 1592.45]; P ⫽ 0.416) (Figure 4). dicators. For example, researchers have generated a predic-
tive program and concluded that liability costs may actu-
DISCUSSION ally increase with disclosure (20). Others argue that
In this analysis of changes in liability claims and costs disclosure reduces lawsuits, citing surveys that suggest pa-
with the introduction of a comprehensive disclosure-with- tients may be more likely to sue if they sense a lack of
offer program at the UMHS, we detected a reduced rate of transparency (23–25). We provide empirical information
claims, primarily driven by a decrease in the number of on the direct liability-related consequences of a disclosure-
lawsuits; lower liability costs; and shorter time to resolu- with-offer program.
tion after the program was started. These findings demon- The UMHS program was designed to expedite com-
strate that it is possible to implement a disclosure-with- pensation and claim resolution. Two frequent criticisms
offer program without increasing liability claims and costs. levied on the tort liability system are that only a small
Other providers and insurers have undertaken disclo- proportion of patients are ever compensated for negligent
sure initiatives or programs, but only 1 organization with a injury and that the time to obtaining compensation is ex-
comprehensive disclosure-with-offer program, the Veterans cessively long (26). Our finding that time to claim resolu-
Affairs Medical Center (VAMC) in Lexington, Kentucky, tion was shorter with the disclosure program suggests that
has reported its experience (21, 22). Assessment of total the program seems to address the latter criticism. Quicker
malpractice payments 9 years after initiation of the pro- resolution can be important, especially for patients sustain-
gram demonstrated that the medical center had moved ing disabling injuries.
from the top to the bottom quartile in its peer group (21). Our finding that fewer patients were compensated
Despite these results, widespread adoption of disclosure- during the disclosure period may raise concerns that dis-
with-offer programs has been limited, perhaps because the closure is not practiced with every case of error. This anal-
Lexington VAMC program is set in a medical center gov- ysis, however, did not identify the specific factors that

Figure 4. Mean costs per claim before and after implementation of the University of Michigan Health System disclosure-with-offer
program.

3 000 000
Initial implementation
Mean Cost per Lawsuit, $

Full implementation

2 000 000

1 000 000

0
Mean Cost per Nonlawsuit Claim, $

3 000 000

2 000 000

1 000 000

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Year

www.annals.org 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 219
Article Liability Claims and Costs With a Medical Error Disclosure Program

might account for the finding. Plausible explanations in- demonstrated a claims rate that was more than 25% lower
clude a general decrease in claims for compensation, fewer after the disclosure program.
injuries as a result of patient safety efforts, or patient satis- State trends, through data submitted largely by commer-
faction with an apology and honesty. In light of the Uni- cial carriers insuring individual Michigan physicians, demon-
versity’s transparency, patients (and their lawyers) may also strated declining numbers of reported claims from 2000
be less likely to seek compensation if they believe they are through 2007, with about 4.5 years from claim opening to
getting the “real story” when UMHS denies that an error closure (36). The UMHS experience compared favorably dur-
occurred. The UMHS’s stance not to settle nuisance claims ing this period: decreasing claims but with shorter resolution
may also decrease the number of paid claims. times. Aggregated national data from 20 physician insurer
A program of disclosure with offer of compensation companies of claims closed from 2001 to 2008 revealed a
may also address another criticism of the malpractice sys- relatively stable percentage of claims receiving payment (24%
tem, namely its high administrative expenses (26, 27). Af- to 32%) and compensation costs, whereas legal expenses in-
ter implementation, mean legal expenses for UMHS de- creased by about 28% (37). For claims opened during this
creased by about 61%. Part of these savings were probably period, UMHS compensated approximately 43% and had de-
offset by the increase in the UMHS risk management bud- creasing compensation and legal costs.
get needed to more proactively address claims internally, In addition, the UMHS approach of accepting
but the risk management expenses were more reflective of systems-level responsibility with regard to the National
greater resources dedicated to the improvement of patient Practitioner Data Bank reporting may affect applicability.
safety rather than administration of the disclosure pro- For disclosure programs that do not adopt this approach,
gram. Moreover, decreases in transactional costs for pa- the willingness of physicians to settle may be limited. Fi-
tients are also apparent. Not only can the shorter time to nally, because the UMHS program is one of disclosure
resolution translate to lower legal expenses for patients, with offer (and not disclosure alone), this analysis does not
many plaintiff attorneys now take cases on an hourly basis necessarily inform on the liability results for providers that
(as opposed to the more expensive contingency basis) in opt to disclose but not offer compensation.
claims in which the UHMS has admitted error. Limitations notwithstanding, the study results have
In addition to addressing some of the problems with important implications. First, a medical center can imple-
the malpractice system, the disclosure program is compat- ment a disclosure-with-offer program without increasing
ible with other patient safety needs. Experts have called for malpractice costs. Second, a disclosure program may ad-
greater reporting of errors as an important part of deliver- dress some of the main shortcomings of our current liabil-
ing safe and high-quality care (28 –30). Despite UMHS ity system, namely shortening long waits for compensation
directly linking its patient safety reporting systems to an and decreasing administrative expenses. Third, disclosure
open-disclosure risk management program (which could may actually reduce another inefficiency of the malpractice
theoretically discourage reporting), the number of reported system: preventing both meritorious and nonmeritorious
incidents increased tremendously (Appendix Figure, avail- claims from becoming expensive lawsuits. Fourth, the
able at www.annals.org). lower number of paid claims after implementation may
Our analysis has several limitations. The state of Mich- suggest that disclosure with offer may not always ensure
igan’s implementation of malpractice reform in 1994 (7 that injured patients receive compensation. This finding,
years before UMHS adopted its disclosure program) may however, may challenge past assumptions that everyone
have promoted a general decrease in liability claims and who had a harmful error expects compensation. The
costs in the state. The legislation included caps on noneco-
openness and accompanying patient safety efforts may
nomic damages, a 6-month compulsory presuit notice pe-
satisfy patients for whom litigation was formerly their
riod, and new expert witness foundation requirements
only alternative.
(31–34). However, Michigan was not considered immune
In an era of calls for greater transparency in health
to the most recent liability crisis (35).
care, disclosure is often cited as a practice necessary to
Our study did not include a concurrent control to
physician ethics and patient safety. The UMHS experience
allow the disentanglement from secular trends. However, it
demonstrates that disclosure with offer can be conducted—in
is possible to glean some limited insight about what may
a setting similar to many other centers in the United
have happened in the absence of a disclosure program. The
States—without exacerbating liability costs. We hope that
UMHS outperformed its own actuarial models (based on
this study will encourage further disclosure efforts, as well
external factors and UMHS claims experience before the
as the detailed evaluation of their effects.
disclosure program) by demonstrating a savings of approx-
imately 39% from predicted total costs from 2003 to From Brigham and Women’s Hospital, Boston, Massachusetts, and Ann
2008. The UMHS actuarial modeling also demonstrated a Arbor Veterans Affairs Medical Center, University of Michigan Health
relatively stable claims rate from 1995 to 2001 that was System, University of Michigan, and Center for Statistical Consultation
predicted to continue through 2008. Actual experience and Research, Ann Arbor, Michigan.
220 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 www.annals.org
 Liability Claims and Costs With a Medical Error Disclosure Program Article
Disclaimer: The content is solely the responsibility of the authors and ing disclosing errors to patients. Arch Intern Med. 2006;166:1605-11. [PMID:
does not necessarily represent the official views of the Department of 16908793]
Veterans Affairs or the National Institutes of Health. 13. Weissman JS, Annas CL, Epstein AM, Schneider EC, Clarridge B, Kirle L,
et al. Error reporting and disclosure systems: views from hospital leaders. JAMA.
2005;293:1359-66. [PMID: 15769969]
Acknowledgment: The authors thank Vinita Bahl, DMD, MPP; Ellen
14. Loren DJ, Klein EJ, Garbutt J, Krauss MJ, Fraser V, Dunagan WC, et al.
Bunting, MA; and Elaine Commiskey, MS, for their invaluable efforts in Medical error disclosure among pediatricians: choosing carefully what we might
supplying data essential to this study. say to parents. Arch Pediatr Adolesc Med. 2008;162:922-7. [PMID: 18838644]
15. Wu AW, Folkman S, McPhee SJ, Lo B. Do house officers learn from their
Grant Support: By Blue Cross Blue Shield of Michigan Foundation mistakes? JAMA. 1991;265:2089-94. [PMID: 2013929]
(1217.II). In addition, Dr. Saint was supported by an Advanced Career 16. Butcher L. Lawyers say ‘sorry’ may sink you in court. Physician Exec. 2006;
Development Award from the Health Services Research & Development 32:20-4. [PMID: 16615399]
Service of the Department of Veterans Affairs during a portion of the 17. Boothman RC, Blackwell AC, Campbell DA Jr, Commiskey E, Anderson
time this study was conducted. Dr. Saint is currently supported by S. A better approach to medical malpractice claims? The University of Michigan
experience. J Health Life Sci Law. 2009;2:125-59. [PMID: 19288891]
awards R21-DK078717 and R01-NR010700 from the National Insti-
18. Wojcieszak D, Banja J, Houk C. The Sorry Works! Coalition: making the
tutes of Health. case for full disclosure. Jt Comm J Qual Patient Saf. 2006;32:344-50. [PMID:
16776389]
Potential Conflicts of Interest: Disclosures can be viewed at www. 19. Kachalia A, Shojania KG, Hofer TP, Piotrowski M, Saint S. Does full
acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum⫽M09 disclosure of medical errors affect malpractice liability? The jury is still out. Jt
-1494. Comm J Qual Saf. 2003;29:503-11. [PMID: 14567259]
20. Studdert DM, Mello MM, Gawande AA, Brennan TA, Wang YC. Disclo-
Reproducible Research Statement: Study protocol, statistical code, and sure of medical injury to patients: an improbable risk management strategy.
data set: Not available. Health Aff (Millwood). 2007;26:215-26. [PMID: 17211031]
21. Kraman SS, Hamm G. Risk management: extreme honesty may be the best
policy. Ann Intern Med. 1999;131:963-7. [PMID: 10610649]
Requests for Single Reprints: Allen Kachalia, MD, JD, Brigham and 22. Peto RR, Tenerowicz LM, Benjamin EM, Morsi DS, Burger PK. One
Women’s Hospital, 75 Francis Street, Boston, MA 02115; e-mail, system’s journey in creating a disclosure and apology program. Jt Comm J Qual
[email protected]. Patient Saf. 2009;35:487-96. [PMID: 19886087]
23. Mazor KM, Simon SR, Yood RA, Martinson BC, Gunter MJ, Reed GW,
Current author addresses and author contributions are available at www et al. Health plan members’ views about disclosure of medical errors. Ann Intern
.annals.org. Med. 2004;140:409-18. [PMID: 15023706]
24. Vincent C, Young M, Phillips A. Why do people sue doctors? A study of
patients and relatives taking legal action. Lancet. 1994;343:1609-13. [PMID:
7911925]
References 25. Witman AB, Park DM, Hardin SB. How do patients want physicians to
1. Clinton HR, Obama B. Making patient safety the centerpiece of medical handle mistakes? A survey of internal medicine patients in an academic setting.
liability reform. N Engl J Med. 2006;354:2205-8. [PMID: 16723612] Arch Intern Med. 1996;156:2565-9. [PMID: 8951299]
2. Levinson W, Gallagher TH. Disclosing medical errors to patients: a status 26. Studdert DM, Mello MM, Brennan TA. Medical malpractice. N Engl J
report in 2007. CMAJ. 2007;177:265-7. [PMID: 17664451] Med. 2004;350:283-92. [PMID: 14724310]
3. Hospital Accreditation Standards. Oakbrook Terrace, IL: Joint Commission 27. Studdert DM, Mello MM, Gawande AA, Gandhi TK, Kachalia A, Yoon C,
on Accreditation of Healthcare Organizations; 2008. et al. Claims, errors, and compensation payments in medical malpractice litiga-
4. National Quality Forum. Safe practices for better healthcare 2009 update: a tion. N Engl J Med. 2006;354:2024-33. [PMID: 16687715]
consensus report. Washington, DC: National Quality Forum; 2009. 28. Kohn L, Corrigan J, Donaldson M. To Err is Human: Building a Safer
5. Blendon RJ, DesRoches CM, Brodie M, Benson JM, Rosen AB, Schneider Health System. Washington, DC: National Academies Pr; 2000.
E, et al. Views of practicing physicians and the public on medical errors. N Engl 29. Mello MM, Studdert DM, Kachalia AB, Brennan TA. “Health courts” and
J Med. 2002;347:1933-40. [PMID: 12477944] accountability for patient safety. Milbank Q. 2006;84:459-92. [PMID:
6. Snyder L, Leffler C; Ethics and Human Rights Committee, American Col- 16953807]
lege of Physicians. Ethics manual: fifth edition. Ann Intern Med. 2005;142:560- 30. Leape LL. Reporting of adverse events. N Engl J Med. 2002;347:1633-8.
82. [PMID: 15809467] [PMID: 12432059]
7. Mazor KM, Simon SR, Yood RA, Martinson BC, Gunter MJ, Reed GW, 31. Mich Comp Laws §600.2169.
et al. Health plan members’ views on forgiving medical errors. Am J Manag Care. 32. Mich Comp Laws §600.1483.
2005;11:49-52. [PMID: 15697100] 33. Mich Comp Laws §600.5838a.
8. Gallagher TH, Studdert D, Levinson W. Disclosing harmful medical errors to 34. Mich Comp Laws §600.5856.
patients. N Engl J Med. 2007;356:2713-9. [PMID: 17596606] 35. Xu X, Siefert KA, Jacobson PD, Lori JR, Ransom SB. The effects of medical
9. Kachalia A, Shojania KG, Hofer TP, Piotrowski M, Saint S. Does full liability on obstetric care supply in Michigan. Am J Obstet Gynecol. 2008;198:
disclosure of medical errors affect malpractice liability? The jury is still out. Jt 205.e1-9. [PMID: 17997388]
Comm J Qual Saf. 2003;29:503-11. [PMID: 14567259] 36. Office of Financial and Insurance Regulation. Evaluation of the Michigan
10. Lamb RM, Studdert DM, Bohmer RM, Berwick DM, Brennan TA. Hos- Medical Professional Liability Insurance Market. A Market Evaluation Study
pital disclosure practices: results of a national survey. Health Aff (Millwood). issued by Commissioner Ken Ross. Lansing, MI: State of Michigan Department of
2003;22:73-83. [PMID: 12674409] Energy, Labor and Economic Growth; October 2009. Accessed at www.michigan
11. Gallagher TH, Waterman AD, Ebers AG, Fraser VJ, Levinson W. Patients’ .gov/documents/dleg/Michigan_Medical_Liability_Ins_Rpt_297694_7.pdf on 1
and physicians’ attitudes regarding the disclosure of medical errors. JAMA. 2003; July 2010.
289:1001-7. [PMID: 12597752] 37. Physician Insurers Association of America. Claims Trend Analysis: A Com-
12. Gallagher TH, Waterman AD, Garbutt JM, Kapp JM, Chan DK, Du- prehensive Analysis of Medical Liability Data Reported to the PIAA Data Sharing
nagan WC, et al. US and Canadian physicians’ attitudes and experiences regard- Project. Rockville, MD: Physician Insurers Association of America; 2009.

www.annals.org 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 221
 Annals of Internal Medicine
Current Author Addresses: Dr. Kachalia: Brigham and Women’s Hos- APPENDIX
pital, 75 Francis Street, Boston, MA 02115. Background
Mr. Kaufman: Department of Internal Medicine, Division of Cardiovas-
The UMHS is a major public academic center located in
cular Medicine, The Blue Cross Blue Shield of Michigan Cardiovascular
Consortium, 2929 Plymouth Road, Suite 225, Ann Arbor, MI 48105- Ann Arbor, Michigan. With few exceptions, the UMHS has a
3206. closed medical staff comprising faculty members employed by the
Mr. Boothman: University of Michigan Health System, Med Inn Build- University of Michigan Medical School, which is governed by the
ing, C 201, SPC 5825, 1500 East Medical Center Drive, Ann Arbor, MI University’s Board of Regents. The system and its employees
48109.
have exclusive occurrence-based professional liability coverage
Ms. Anderson: University of Michigan Health System, Risk Manage-
ment Department, 300 North Ingalls Building, Room 8A06/SPC 5478, provided by an established captive insurance company. Medical
Ann Arbor, MI 48109-5478. staff and other employees can be sued individually in the circuit
Ms. Welch: Center for Statistical Consultation and Research, 3554 courts with juries, but as an agency of the state of Michigan, the
Rackham Building, 915 East Washington Street, Ann Arbor, MI 48109- Regents may only be sued in the state’s Court of Claims (which
1070.
does not allow juries). For judicial economy, lawsuits are typically
Dr. Saint: Ann Arbor Veterans Affairs Medical Center, Division of Gen-
eral Medicine, Department of Internal Medicine, University of Michi- joined administratively but are not consolidated; separate judg-
gan, 300 North Ingalls Building, Room 7E08, Ann Arbor, MI 48109. ments are obtained in each lawsuit that goes to verdict.
Dr. Rogers: Division of General Medicine, Department of Internal Med-
icine, University of Michigan, 300 North Ingalls Building, Room 7E07,
Ann Arbor, MI 48109.

Author Contributions: Conception and design: A. Kachalia, S. Saint,

S.R. Kaufman, M.A.M. Rogers, R. Boothman, S. Anderson.
Analysis and interpretation of the data: A. Kachalia, S. Saint, S.R. Kauf-
man, M.A.M. Rogers, R. Boothman, S. Anderson, K. Welch.
Drafting of the article: A. Kachalia, S.R. Kaufman, M.A.M. Rogers.
Critical revision of the article for important intellectual content: A.
Kachalia, S. Saint, S.R. Kaufman, M.A.M. Rogers, R. Boothman.
Final approval of the article: A. Kachalia, S. Saint, S.R. Kaufman,
M.A.M. Rogers, R. Boothman, S. Anderson.
Statistical expertise: S.R. Kaufman, M.A.M. Rogers, K. Welch.
Obtaining of funding: A. Kachalia, S.R. Kaufman, S. Saint, M.A.M.
Rogers.
Administrative, technical, or logistic support: A. Kachalia, R. Boothman,
S. Anderson.
Collection and assembly of data: A. Kachalia, S.R. Kaufman, R. Boothman,
S. Anderson.

Appendix Figure. Number of incidents reported to University of Michigan Health System risk management, by fiscal year.

18 000

16 000

14 000
Incidents Reported, n

12 000

10 000

8 000

6 000

4 000

2 000

0
FY 96 FY 97 FY 98 FY 99 FY 00 FY 01 FY 02 FY 03 FY 04 FY 05 FY 06 FY 07 FY 08
Incidents reported, n 636 578 360 1585 2622 2660 3007 3570 4705 9420 12 848 15 186 15 650

An incident is any event (whether involving injury, potential injury, or any concern) that was reported to risk management. FY ⫽ fiscal year.
W-72 17 August 2010 Annals of Internal Medicine Volume 153 • Number 4 www.annals.org
Copyright of Annals of Internal Medicine is the property of American College of Physicians and its content
may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express
written permission. However, users may print, download, or email articles for individual use.