Open Access Research

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
The impact of chronic kidney disease
and cardiovascular comorbidity
on mortality in a multiethnic population:
a retrospective cohort study
Mark Jesky,1,2 Amanda Lambert,3 A C Felix Burden,4 Paul Cockwell1,2

To cite: Jesky M, Lambert A, ABSTRACT

Burden ACF, et al. The impact Strengths and limitations of this study
Objective: To assess the impact of chronic kidney
of chronic kidney disease
disease (CKD) and cardiovascular comorbidity on ▪ Sample size with inclusion of many practices.
and cardiovascular
comorbidity on mortality in a
mortality in a multiethnic primary care population. ▪ Ethnicity data self-reported and well recorded
multiethnic population: Design: Retrospective cohort study. (>80%).
a retrospective cohort study. Setting: Inner-city primary care trust in West Midlands, ▪ Individuals of white ethnicity relatively under-
BMJ Open 2013;3:e003458. UK. represented.
doi:10.1136/bmjopen-2013- Participants: Individuals aged 40 years and older, of
003458 South Asian, black or white ethnicity, registered with a
general practice and with their kidney function checked lower mortality risk than people of white
▸ Prepublication history and within the last 12 months (n=31 254). ethnicity.9 10 An increased risk of death is
additional material for this Outcome measure: All-cause mortality. also associated with other comorbidities,
paper is available online. To including hypertension, diabetes and
Results: Reduced estimated glomerular filtration rate,
view these files please visit
the journal online
higher albuminuria, older age, white ethnicity (vs South cardiovascular (CV) disease.11–16
(http://dx.doi.org/10.1136/ Asian or black ethnicity) and increasing cardiovascular While previous studies have indicated sur-
bmjopen-2013-003458). comorbidities were independent determinants of a vival differences between ethnic groups,8 17–21
higher mortality risk. In the multivariate model including there has been limited reporting in these
Received 21 June 2013 comorbidities and kidney function, the HR for mortality studies on the relative impact of comorbid-
Revised 4 September 2013 for South Asians was 0.697 (95% CI 0.56 to 0.868,
Accepted 23 October 2013
ities including kidney function on a popula-
p=0.001) and for blacks it was 0.533 (95% CI 0.403 to
tion basis. This paucity of data reflects a
0.704, p<0.001) compared to whites.
shortfall in the availability of population-
Conclusions: The HR for death is lower for South
Asian and black individuals compared to white based primary care databases linked to esti-
individuals. This is, in part, independent of age, mated glomerular filtration rate (eGFR) and
gender, socioeconomic status, kidney function and albuminuria reporting and traceable to mor-
comorbidities. Risk of death is higher in individuals with tality. Furthermore, there is minimal com-
CKD and with a higher cumulative cardiovascular parative data on people of South Asian
comorbidity. ethnicity; comparative studies usually report
data on Chinese-Asians.5
1
Department of Renal In the UK, there has been a systematic
Medicine, Queen Elizabeth improvement in chronic disease recognition
Hospital Birmingham, through a primary care pay for performance
Birmingham, UK
2
Division of Infection and
INTRODUCTION system, the Quality and outcomes framework
Immunity, University of Chronic kidney disease (CKD) is a risk factor (QOF).22 23 This system utilises chronic
Birmingham, Birmingham, for increased mortality,1 with an increased disease registers for the identification, moni-
UK risk of death associated with declining excre- toring and management of patients with
3
Public Health Intelligence, tory renal function and albuminuria.2–4 CKD known comorbidities; a component of this
Birmingham City Council,
Birmingham, UK
prevalence and the risk imparted by CKD monitoring involves measuring and docu-
4
Sandwell and West may vary by ethnicity; for example, some menting renal function. These disease regis-
Birmingham Clinical studies indicate that CKD is more common ters can be combined with laboratory results
Commissioning Group, in people of white ethnicity,5 6 but non-white and linked with demographic and mortality
Birmingham, UK ethnic groups have a faster progression to data to better identify determinants of
Correspondence to end-stage kidney disease.7 8 Paradoxically, outcomes.
Mark Jesky; when treated with chronic dialysis treatment, We have therefore utilised chronic disease
[email protected] people of non-white ethnicity have a registers to perform a retrospective cohort

Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458 1

Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
study of the relationship between CKD, CV comorbidity A standardised isotope dilution mass spectrometry
and mortality within a deprived, inner-city multiethnic (IDMS) MDRD eGFR28 was reported from one of three
population. Our study hypotheses were local biochemistry laboratories; however, eGFR reporting
1. There are differences in mortality between different was not universally recorded on primary care systems in
ethnic groups. 2008 and if this was not available, the eGFR was calcu-
2. These differences in mortality are explained by lated by utilising laboratory provided correction factors
known risk factors including comorbidities, renal for the creatinine to generate IDMS-traceable MDRD
function, demographic and socioeconomic factors. eGFR. One general practice in the catchment area was
This study incorporated all stages of kidney function excluded as IDMS-traceable creatinine was not available
except stage 5 CKD (an eGFR below 15 mL/min/ from a fourth laboratory that provided blood tests specif-
1.73 m2) in patients with known CV comorbidities and ically for that catchment area.
focused on three ethnic groups: South Asian (including SES was assessed using the Index of Multiple
individuals of Bangladeshi, Indian and Pakistani Deprivation29; this utilises the postcode from an indivi-
descent), black (individuals from, or who have ancestors dual’s address to identify the lower layer super output
from, Africa or the Caribbean) and white. area (LSOA) where the individual resides. Each of the
32 482 LSOAs in England are assigned a score and rank
for the IMD 2007, with lower ranks corresponding to the
METHODS most deprived areas. The IMD has been validated as
Ethics superior to traditional deprivation indexes such as the
The data were fully anonymised and were available as a Townsend score,30 due to its use of multiple domains
component of an ongoing clinical development pro- reflective of socioeconomic deprivation.31 The IMD 2007
gramme. The responsible National Health Service score incorporates seven areas of deprivation: income
(NHS) R&D Consortium stated that this study did not deprivation; employment deprivation; health deprivation
require ethical submission to an NHS research ethics and disability; education; skills and training deprivation;
committee as it represented an evaluation of part of an barriers to housing and services; living environment
ongoing primary care trust (PCT) programme. For PCT deprivation and crime. For the analyses presented,
data extraction, the PCT professional executive commit- deprivation was divided into national quintiles, with the
tee and general practitioner (GP) locality leads provided most deprived quintile as the reference population (ie,
approval for the programme, including evaluation and how mortality in less deprived quintiles compared to the
publication. most deprived quintile).
Mortality data were obtained from the Primary Care
Cohort identification Mortality Database,32 a resource developed by The NHS
The cohort was derived from Heart of Birmingham Information Centre in partnership with the Office for
(Teaching) Primary Care Trust (HoB PCT), which had a National Statistics (ONS). Data obtained from ONS
registered population of 312 070 (September 2008). The records are linked to the general practice where the indi-
majority of the population (62%) were non-white.24 vidual was registered and therefore allow data to be
Sixty-nine per cent of the population were below 40 years extracted for specific general practices (ie, those within
of age. Data were collected centrally, utilising software HoB PCT). Individuals included in this analysis were
able to identify comorbidities through their classification either still registered with an HoB PCT GP at the end of
on chronic disease registers (Enhanced Healthcare Services, the follow-up period or had died while still registered at
Essex, UK). Complete sets of anonymised data were avail- the practice. 9,907 (11.1%) individuals who had left, and
able for 63 of 73 general practices within HoB PCT, com- therefore were no longer registered with their practices,
prising a population of 285 221, and these were extracted were excluded from the analysis. The follow-up period
from electronic downloads. Figure 1 illustrates the selec- was 23 months from May 2008 until February 2011.
tion process for inclusion in the study.
The inclusion criteria comprised individuals aged
40 years and over who had kidney function testing per- Statistical analyses
formed within the previous 12 months as recommended All analyses were performed using PASW statistics 18 for
by national guidelines.25 Data for the following variables Windows (IBM, Chicago, Illinois, USA).
were collected: age, gender, ethnicity, current smoking Measurements for kidney function were divided into cat-
status, socioeconomic status (SES), eGFR and/or cre- egories; eGFR into six categories (15–29, 30–44, 45–59,
atinine, urinary albumin : creatinine ratio (ACR) and vas- 60–89, 90–119 and ≥120 mL/min) with the eGFR range
cular comorbidity (atrial fibrillation, CKD, diabetes between 90 and 119 mL/min as the reference population.
mellitus, heart failure, hypertension, ischaemic heart Individuals with an eGFR <15 mL/min were excluded
disease and stroke) as defined by a relevant clinical from the analysis. ACR was divided into five categories
(Read) code specified by the UK pay for performance (<1.1 mg/mmol ‘optimal’, 1.1–2.99 ‘high normal’,
(QOF) business rules.26 Ethnicity was self-reported, con- 3–29.99 ‘high’, 30–199.99 ‘very high’ and ≥200 ‘neph-
sidered the ‘gold standard’ for classification.27 rotic’) in line with the KDIGO consensus conference.33

2 Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458

 Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
Figure 1 Flow diagram indicating selection process for inclusion in the analyses (ACR, albumin : creatinine ratio; eGFR,
estimated glomerular filtration rate; IDMS, isotope dilution mass spectrometry; GP, general practitioner; HoB, Heart of
Birmingham).

The relationship between age and mortality was not linear. and then presented as three models. The choice of model
Therefore, age was divided into six categories (50 years and variables was determined by the availability in the dataset
under, 51–60, 61–70, 71–80, 81–90, greater than 90 years), of demographic and clinical risk factors consistent with
with the youngest group serving as comparator. those utilised by other investigators in previous work in
Descriptive statistics are presented as mean with SD or similar populations,34 35 where the variable was available in
median with IQR depending on distribution. our target population. Model 1 incorporates the number
Continuous variables were compared using analysis of of identified vascular comorbidities (0–7), ethnicity, age,
variance (normal distribution) with post hoc Bonferroni gender, smoking status and SES. Model 2 includes the
analysis or Kruskal-Wallis (non-parametric distribution) eGFR level with removal of CKD from the comorbidity
tests. χ2Tests were used to compare categorical variables. score (the possible scores are therefore 0–6) in order to
Cox regression survival analysis was used to evaluate avoid the association between declining renal function
the association of ethnicity and mortality, both before and the likelihood of being on the CKD register. Model 3
and after adjusting for covariates. Data are presented added ACR to the variables in model 2.
using survival plots, HRs with 95% CI and p values. Both A complete case model was used in the analyses. All
univariate (unadjusted) and multivariate (adjusted) data were complete with the exception of ACR.
regression analyses are presented. The proportionality Therefore, data were analysed for all individuals identi-
hazard assumption, assessed using log(−log(survival fied (unadjusted, models 1 and 2) and then repeated
function)) plots, was met for all covariates. for individuals who had an ACR recorded (unadjusted
The association between comorbidity, ethnicity and mor- and models 1–3). An ‘enter’ technique was used for the
tality was assessed by univariate analyses for all risk factors regression analysis.

Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458 3

Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
RESULTS death of 0.421 (95% CI 0.376 to 0.471, p<0.001) for
Complete cohort people of South Asian ethnicity and 0.522 (95% CI
At inception (May 2008), 31 254 individuals fulfilled the 0.447 to 0.609, p<0.001) for people of black ethnicity
inclusion criteria for analysis. People of South Asian eth- compared to people of white ethnicity. The mortality
nicity formed the largest ethnic group (16 724, 53.4%), rate increased exponentially with age and a higher HR
followed by people of white ethnicity (9146, 29.3%) and was observed for male gender, current smokers and total
black ethnicity (5384, 17.2%). Baseline characteristics of number of comorbidities. No difference in mortality was
the study population are shown in table 1. The age dis- found between deprivation quintiles. Using an eGFR of
tribution differed between groups with South Asians sig- 90–119 mL/min as the reference, a J-shaped relationship
nificantly younger than the other two ethnic groups. was observed with a higher risk of death seen for higher
There was no significant difference in gender between and lower eGFR values. The HR for death increased pro-
the three ethnic groups. Smoking was least common in gressively by stage of CKD with an eGFR<90 mL/min.
the South Asian group. The majority of all three ethnic The univariate analysis was repeated for those indivi-
groups resided in the most deprived quintile, with a duals who had their ACR reported (table 3b) with
higher proportion of people of South Asian and black similar trends identified in the whole population analysis
ethnicity in this quintile than people of white ethnicity. with the exception of no observed difference between
The number of vascular comorbidities was similar individuals with an eGFR of ≥120 compared to 90–
between groups, with 11–13% of each ethnic group 119 mL/min. A progressive increase in HR for death was
having three or more comorbidities. Prevalence of dif- seen with each increasing category for ACR.
ferent vascular comorbidities varied between groups: the
white group had a lower reported prevalence of diabetes Multivariate analysis
but a higher prevalence of CKD, atrial fibrillation, heart Following adjustment for covariates, the differences in
failure and stroke. ethnicity remained; people of South Asian and black
Median eGFR (corrected for ethnicity as appropriate) ethnicities had a lower HR for death in all analyses.
was 80.2 mL/min and was lowest in the white group Model 1 (complete cohort, incorporating the number
(74.9 mL/min compared to 81.3 mL/min for South of identified comorbidities, see online supplementary
Asian individuals and 85.5 mL/min for those of black table I) analysed the complete cohort and showed an
ethnicity; p<0.001). 21.5% of white, 13.8% of South Asian adjusted HR for death of 0.673 (95% CI 0.595 to 0.761,
and 11.5% of black individuals had an eGFR between p<0.001) for people of South Asian ethnicity and 0.592
15 and 59 mL/min, consistent with stages 3–4 CKD. (95% CI 0.504 to 0.696, p<0.001) for people of black eth-
At the end of the study period, a higher proportion of nicity compared to people of white ethnicity. When the
white individuals had died (7.4%) compared to the two analysis was restricted to the cohort with ACR tests avail-
other ethnic groups (South Asian 3.2%, black 4%; able, the HR for death was 0.757 (95% CI 0.61 to 0.939,
p<0.001). p=0.011) for people of South Asian ethnicity and 0.526
for people of black ethnicity (95% CI 0.4 to 0.692,
Albumin creatinine ratio cohort p<0.001) compared to people of white ethnicity. For the
An ACR cohort had been tested in 7022 (42%), 2275 complete cohort, mortality risk was lower in IMD quin-
(24.9%) and 1908 (20.9%) South Asian, black and white tiles 3 and 4 (compared to the most deprived quintile 5).
individuals, respectively. Table 2 lists the baseline charac- No significant difference between IMD quintiles was
teristics for this subgroup. The median ACR was 1.1 mg/ identified in the ACR cohort. Increasing age (51 and
mmol and was highest in the South Asian group (1.2 over in the complete cohort, 61 and over in the ACR
compared to 1.0 mg/mmol for both white and black cohort), smoking status and male gender was
individuals; p<0.001). There were similar trends to the significant in analyses for both cohorts. An increased HR
whole cohort for age distribution, eGFR, smoking status for death was observed for two or more comorbidities,
and deprivation. with the HR increasing as the number of comorbidities
Those with an ACR tested were more likely to have a increased.
greater vascular comorbid burden (18–20% having Kidney function (eGFR) was incorporated into model
three or more comorbidities). A higher proportion of 2 (with the removal of CKD from the comorbidity score,
individuals of South Asian descent, male gender and see online supplementary table I), and in the complete
with diabetes had their ACR tested. cohort HR for people of South Asian ethnicity was 0.678
In concordance to the whole group analyses, deaths in (95% CI 0.6 to 0.767, p<0.001) and for people of black
the ACR cohort were highest among white individuals ethnicity it was 0.614 (95% CI 0.522 to 0.722, p<0.001)
(7.8%) compared to the South Asian (3.6%) and black compared to people of white ethnicity. Similarly, when
individuals (3.7%; p<0.001). the analysis was restricted to the cohort of patients with
ACR tests available, people of South Asian and black eth-
Univariate analysis nicity had a lower proportion of deaths compared to
The univariate (unadjusted) analysis for the complete people of white ethnicity with HRs of 0.789 (95% CI
cohort (table 3a) demonstrated unadjusted HRs for 0.635 to 0.98, p=0.032) and 0.575 (95% CI 0.435 to

4 Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458

 Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
Table 1 Baseline characteristics by ethnicity: complete cohort
All White South Asian Black p Value
Number
n (%) 31 254 (100) 9146 (29.3) 16 724 (53.4) 5384 (17.2)
Age
Median (lower, upper quartile) 59.0 (50.0, 71.0) 65.0 (55.0, 75.0) 56.0 (49.0, 68.0) 61.0 (48.0, 73.0) <0.001
50 and under (%) 8421 (26.9) 1515 (16.6) 5124 (30.6) 1782 (33.1) <0.001
51–60 (%) 8017 (25.7) 1948 (21.3) 5170 (30.9) 899 (16.7)
61–70 (%) 6650 (21.3) 2459 (26.9) 3206 (19.2) 985 (18.3)
71–80 (%) 6006 (19.2) 2109 (23.1) 2568 (15.4) 1329 (24.7)
81–90 (%) 1974 (6.3) 1008 (11.0) 604 (3.6) 362 (6.7)
>90 (%) 186 (0.6) 107 (1.2) 52 (0.3) 27 (0.5)
Gender
Female (%) 15 248(48.8) 4384 (47.9) 8184 (48.9) 2680 (49.8) 0.085
Smoking
n (%) 5150 (16.5) 2285 (25.0) 1812 (10.8) 1053 (19.6) <0.001
IMD rank
Quintile 1 (least deprived) (%) 152 (0.5) 59 (0.6) 92 (0.6) 1 (0.0) <0.001
Quintile 2 (%) 316 (1.0) 132 (1.4) 173 (1.0) 11 (0.2)
Quintile 3 (%) 3348 (10.7) 1860 (20.3) 1255 (7.5) 233 (4.3)
Quintile 4 (%) 5144 (16.5) 2243 (24.5) 2238 (13.4) 663 (12.3)
Quintile 5 (most deprived) (%) 22 294 (71.3) 4852 (53.1) 12 966 (77.5) 4476 (83.1)
AF
n (%) 807 (2.6) 515 (5.6) 212 (1.3) 80 (1.5) <0.001
CKD
n (%) 3648 (11.7) 1318 (14.4) 1691 (10.1) 639 (11.9) <0.001
Diabetes
n (%) 9931 (31.8) 1771 (19.4) 6415 (38.4) 1745 (32.4) <0.001
Heart failure
n (%) 822 (2.6) 308 (3.4) 385 (2.3) 129 (2.4) <0.001
Hypertension
n (%) 16 505(52.8) 5181 (56.6) 8063 (48.2) 3261 (60.6) <0.001
IHD
n (%) 4226 (13.5) 1417 (15.5) 2386 (14.3) 423 (7.9) <0.001
Stroke
n (%) 1476 (4.7) 570 (6.2) 673 (4.0) 233 (4.4) <0.001
Comorbidities
Median (lower, upper quartile) 1.0 (0.0, 2.0) 1.0 (0.0, 2.0) 1.0 (0.0, 2.0) 1.0 (0.0, 2.0) 0.075
0 (%) 9879 (31.6) 2829 (30.9) 5459 (32.6) 1591 (29.6) <0.001
1 (%) 10 707 (34.3) 3253 (35.6) 5524 (33) 1930 (35.8)
2 (%) 6845 (21.9) 1898 (20.8) 3694 (22.1) 1253 (23.3)
3 (%) 2667 (8.5) 785 (8.6) 1451 (8.7) 431 (8)
4 (%) 828 (2.6) 254 (2.8) 447 (2.7) 127 (2.4)
5 (%) 268 (0.9) 103 (1.1) 124 (0.7) 41 (0.8)
6 (%) 55 (0.2) 23 (0.3) 23 (0.1) 9 (0.2)
7 (%) 5 (<0.1) 1 (<0.1) 2 (<0.1) 2 (<0.1)
Creatinine (µmol/L)
Mean (SD) 87.0 (25.8) 88.2 (24.7) 84.6 (25.4) 92.3 (28) <0.001
eGFR (mL/min)
Median (lower, upper quartile) 80.2 (66.7, 94.3) 74.9 (62.3, 88.8) 81.3 (68.1, 95.3) 85.5 (72.3, 100.1) <0.001
>120 (%) 1473 (4.7) 264 (2.9) 802 (4.8) 407 (7.6) <0.001
90–120 (%) 8523 (27.3) 1842 (20.1) 4841 (28.9) 1840 (34.2)
60–89 (%) 16 373 (52.4) 5077 (55.5) 8776 (52.5) 2520 (46.8)
45–59 (%) 3447 (11.0) 1389 (15.2) 1627 (9.7) 431 (8.0)
30–44 (%) 1134 (3.6) 466 (5.1) 517 (3.1) 151 (2.8)
15–29 (%) 304 (1.0) 108 (1.2) 161 (1.0) 35 (0.7)
Died
n (%) 1435 (4.6) 681 (7.4) 541 (3.2) 213 (4.0) <0.001
ACR, albumin : creatinine ratio; AF, atrial fibrillation; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; IHD, ischaemic
heart disease; IMD, Index of Multiple Deprivation.

Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458 5

Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
Table 2 Baseline characteristics by ethnicity: ACR-tested cohort
All White South Asian Black p Value
Number
n (%) 11 205 (100) 1908 (17) 7022 (62.7) 2275 (20.3)
Age (years)
Median (lower, upper quartile) 59.0 (50.0, 71.0) 65.0 (55.0, 75.0) 57.0 (50.0, 68.0) 65.0 (49.0, 74.0) <0.001
50 and under (%) 1900 (25.9) 304 (15.9) 1961 (27.9) 635 (27.9) <0.001
51–60 (%) 3024 (27.0) 413 (21.6) 2239 (31.9) 372 (16.4)
61–70 (%) 2370 (21.2) 496 (26.0) 1423 (20.3) 451 (19.8)
71–80 (%) 2251 (20.1) 456 (23.9) 1152 (16.2) 643 (28.3)
81–90 (%) 611 (5.5) 222 (11.6) 226 (3.2) 163 (7.2)
>90 (%) 49 (0.4) 17 (0.9) 21 (0.3) 11 (0.5)
Gender
Female (%) 4348 (38.8) 682 (35.7) 2754 (39.2) 912 (40.1) 0.008
Smoking
n (%) 1869 (16.7) 518 (27.1) 872 (12.4) 479 (21.1) <0.001
IMD rank
Quintile 1 (least deprived) (%) 30 (0.3) 4 (0.2) 25 (0.4) 1 (0.0) <0.001
Quintile 2 (%) 84 (0.7) 19 (1.0) 60 (0.9) 5 (0.2)
Quintile 3 (%) 712 (6.4) 233 (12.2) 540 (5.7) 78 (3.4)
Quintile 4 (%) 1458 (13.0) 339 (17.8) 876 (12.5) 243 (10.7)
Quintile 5 (most deprived) (%) 8921 (79.6) 1313 (68.8) 5660 (80.6) 1948 (85.6)
AF
n (%) 233 (2.1) 113 (5.9) 91 (1.3) 29 (1.3) <0.001
CKD
n (%) 1637 (14.6) 356 (18.7) 921 (13.1) 360 (15.8) <0.001
Diabetes
n (%) 6828 (60.9) 990 (51.9) 4505 (62.4) 1333 (58.6) <0.001
Heart failure
n (%) 310 (2.8) 74 (3.9) 175 (2.5) 61 (2.7) 0.005
Hypertension
n (%) 6189 (55.2) 1092 (57.2) 3679 (52.4) 1418 (62.3) <0.001
IHD
n (%) 1556 (13.9) 281 (14.7) 1071 (15.3) 201 (8.8) <0.001
Stroke
n (%) 480 (4.3) 97 (5.1) 283 (4.0) 100 (4.4) 0.126
Comorbidities
Median (lower, upper quartile) 1.0 (1.0, 2.0) 2.0 (1.0, 2.0) 1.0 (1.0, 2.0) 2.0 (1.0, 2.0) 0.818
0 (%) 2510 (22.4) 472 (24.7) 1514 (21.6) 524 (23.0) <0.001
1 (%) 3139 (28.0) 466 (24.4) 2103 (29.9) 870 (25.1)
2 (%) 3438 (30.7) 574 (30.1) 2093 (29.8) 771 (33.9)
3 (%) 1481 (13.2) 261 (13.7) 928 (13.2) 292 (12.8)
4 (%) 448 (4.0) 79 (4.1) 284 (4.0) 85 (3.7)
5 (%) 154 (1.4) 46 (2.4) 83 (1.2) 25 (1.1)
6 (%) 32 (0.3) 10 (0.5) 15 (0.2) 7 (0.3)
7 (%) 3 (<0.1) 0 (<0.1) 2 (<0.1) 1 (<0.1)
Creatinine (µmol/L)
Mean (SD) 89.1 (27.6) 91.8 (26.2) 86.2 (26.8) 95.8 (29.6) <0.001
eGFR (mL/min)
Median (lower, upper quartile) 81.1 (66.3, 95.9) 74.3 (59.7, 89.8) 82 (67.4, 89.8) 84.2 (70.0, 98.9) <0.001
>120 (%) 611 (5.5) 67 (3.5) 380 (5.4) 164 (7.2) <0.001
90–120 (%) 3234 (28.9) 404 (21.2) 2091 (29.8) 739 (32.5)
60–89 (%) 5451 (48.6) 953 (49.9) 3453 (49.2) 1045 (45.9)
45–59 (%) 1300 (11.6) 323 (16.9) 750 (10.7) 227 (10.0)
30–44 (%) 487 (4.3) 131 (6.9) 274 (3.9) 82 (3.6)
15–29 (%) 122 (1.1) 30 (1.6) 74 (1.1) 18 (0.8)
ACR (mg/mmol)
Median (lower, upper quartile) 1.1 (0.4, 3.4) 1.0 (1.4, 2.8) 1.2 (0.5, 3.8) 1.0 (0.3, 2.9) <0.001
Optimal (<1.1) (%) 5641 (50.3) 1026 (53.8) 3400 (48.4) 1214 (53.4) <0.001
High normal (1.1–2.99) (%) 2485 (22.2) 426 (22.3) 1560 (22.2) 499 (21.9)
Continued

6 Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458

 Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
Table 2 Continued
All White South Asian Black p Value
High (3.0–29.99) (%) 2594 (23.2) 402 (21.1) 1717 (24.4) 475 (20.9)
Very high (30–200) (%) 413 (3.7) 49 (2.6) 287 (4.1) 77 (3.4)
Nephrotic (>200) (%) 73 (0.7) 5 (0.3) 58 (0.8) 10 (0.4)
Died
n (%) 484 (4.3) 149 (7.8) 250 (3.6) 85 (3.7) <0.001
ACR, albumin : creatinine ratio; AF, atrial fibrillation; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; IHD, ischaemic
heart disease; IMD, Index of Multiple Deprivation.

0.759, p<0.001), respectively. In the complete cohort, comorbidity scores such as the Charlson Comorbidity
mortality risk was lower in the IMD quintile 4. More Index36 are difficult to calculate accurately in a large
than two comorbidities were associated with an increas- primary care setting, our study demonstrates that a
ing HR and an increased HR of death compared to the simple cumulative score provides prognostic informa-
reference eGFR range (90–119 mL/min) was seen with tion. Individual comorbidities were present in varying
an eGFR ≥120 and ≥45 mL/min. An eGFR of 60– frequencies within different ethnic groups, a finding
89 mL/min was associated with a lower HR. In the ana- consistent with that found in other ethnically diverse
lysis of those with ACR tested, an eGFR <60 mL/min was populations.37 While individual comorbidities were
associated with progressively higher HR by CKD stage. associated with different mortality risks, we found that
In model 3 (all vascular comorbidities except CKD the cumulative effect of comorbidities conveyed the
and the addition of eGFR and ACR, table 4), the HR for greatest prognostic implication. A similar approach, but
death for people of South Asian ethnicity was 0.697 also including non-CV risk factors, has recently been
(95% CI 0.56 to 0.868, p=0.001) and for people of black described.38 Our study suggests that routinely collected
ethnicity it was 0.533 (95% CI 0.403 to 0.704, p<0.001) clinical data concerning cumulative comorbidity
compared to people of white ethnicity (figure 2). Older may be utilised to quantify risk; however, further work
age, male gender, being a current smoker and increas- would be required to validate this as a tool for use in
ing comorbidity (two or more) were associated with an clinical care.
increased HR of death (figure 3). An ACR of ‘high’ or SES was measured by the IMD 2007 score, a cumulative
greater (ie, ≥3.0 mg/mmol) and an eGFR <45 mL/min deprivation index score incorporating seven areas of
was also associated with an increased HR for death. No deprivation which has been validated as superior to
significant differences in HRs were observed between other deprivation scores.31 One notable finding is that we
deprivation quintiles. did not demonstrate any association between mortality
when corrected for all other factors including comorbid-
ity and ethnicity. This is not consistent with several other
DISCUSSION studies, which have shown that there is an independent
This study utilised routinely available clinical and labora- relationship between SES and mortality across disease
tory data, including kidney function assessed by eGFR states and ethnic groups within the UK.39–42 This relation-
and ACR, from a large primary care population. We ship varies by population group studied43 and there have
included in the analysis detailed SES and, importantly, been limited studies investigating health disparities in
studied three ethnic groups, South Asian, black and similar inner-city populations. While we studied a health-
white. Prior to this research, there has been uncertainty care system that is free at the point of care, limiting pos-
about the impact of ethnicity and SES on clinical out- sible health access issues, the majority of individuals were
comes in people with significant comorbidities including from the most deprived national quintile and our study
CKD. The comprehensive nature of the dataset coupled may therefore underestimate the influence of the com-
with the ability to utilise the Primary Care Mortality plete spectrum of SES on mortality. To attempt to correct
Database has allowed us to assess the relative impact of for this, we reran the analyses dividing the cohort into
these factors on survival. equal quintiles. All analyses continued to indicate the
We found that previous associations between lower effect of ethnicity and the importance of CV comorbidity
eGFR and higher ACR and increased mortality applied and renal function. The univariate analysis (see online
to this population. Furthermore, these associations supplementary table II) and the most comprehensive
remained significant when adjusted for ethnicity, age, multivariate analysis (model 3, see online supplementary
gender, CV risk factors and SES. These results add table III) did not show any differences between the most
weight to the risk stratification benefit of measuring and least deprived quintiles.
ACR has in high-risk groups. One of the seven areas included in the IMD is health
A strong cumulative impact of comorbidity on CKD deprivation, raising the possibility of an inbuilt relation-
and ethnicity was shown. Whereas traditional ship between deprivation and health even before

Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458 7

Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
Table 3 Cox proportional hazard regression analysis: univariate (unadjusted) analyses
Complete cohort (3a) ACR tested cohort (3b)
HR (95% CI) p Value HR (95% CI) p Value
Ethnicity
White 1 (<0.001*) 1 (<0.001*)
South Asian 0.421 (0.376 to 0.471) <0.001 0.444 (0.362 to 0.545) <0.001
Black 0.522 (0.447 to 0.609) <0.001 0.467 (0.357 to 0.611) <0.001
Age (years)
50 and under 1 (<0.001*) 1 (<0.001*)
51–60 2.127 (1.553 to 2.914) <0.001 1.757 (1.057 to 2.921) 0.03
61–70 5.429 (4.078 to 7.228) <0.001 4.646 (2.926 to 7.345) <0.001
71–80 12.971 (9.887 to 17.016) <0.001 11.363 (7.376 to 17.505) <0.001
81–90 32.86 (29.952 to 43.275) <0.001 24.725 (15.769 to 38.767) <0.001
>90 90.904 (65.097 to 126.943) <0.001 82.731 (46.684 to 146.612) <0.001
Gender
Female as reference 1.375 (1.238 to 1.529) <0.001 1.401 (1.155 to 1.699) 0.001
Smoker
Non-smoker as reference 1.154 (1.009 to 1.317) 0.036 1.259 (1.006 to 1.574) 0.044
IMD rank
Quintile 1 (least deprived) 0.860 (0.385 to 1.919) 0.713 <0.001 (<0.001—>10^5) 0.939
Quintile 2 0.822 (0.465 to 1.453) 0.501 <0.001 (<0.001—>10^5) 0.897
Quintile 3 1.002 (0.846 to 1.186) 0.983 1.151 (0.818 to 1.619) 0.419
Quintile 4 0.925 (0.800 to 1.070) 0.297 0.774 (0.577 to 1.039) 0.088
Quintile 5 (most deprived) 1 (0.802*) 1 (0.42*)
AF 5.588 (4.757 to 6.565) <0.001 6.123 (4.568 to 8.207) <0.001
CKD 3.442 (3.074 to 3.854) <0.001 3.498 (2.904 to 4.213) <0.001
Diabetes 1.346 (1.209 to 1.498) <0.001 1.939 (1.577 to 2.385) <0.001
Heart failure 7.622 (6.595 to 8.804) <0.001 7.279 (5.681 to 9.327) <0.001
Hypertension 2.079 (1.857 to 2.325) <0.001 2.05 (1.681 to 2.499) <0.001
IHD 2.796 (2.495 to 3.132) <0.001 3.136 (2.592 to 3.795) <0.001
Stroke 3.654 (3.154 to 4.233) <0.001 3.709 (2.855 to 4.817) <0.001
Comorbidities
0 1 (<0.001*) 1 (<0.001*)
1 1.775 (1.487 to 2.118) <0.001 1.630 (1.094 to 2.430) 0.016
2 2.930 (2.458 to 3.493) <0.001 2.917 (2.023 to 4.205) <0.001
3 5.486 (4.550 to 6.615) <0.001 5.580 (3.837 to 8.113) <0.001
4 9.584 (7.691 to 11.942) <0.001 9.855 (6.511 to 14.917) <0.001
5 17.591 (13.490 to 22.939) <0.001 21.091 (13.479 to 33.001) <0.001
6 28.391 (18.411 to 43.782) <0.001 33.673 (17.519 to 64.722) <0.001
7 11.873 (1.664 to 84.728) 0.014 29.402 (4.031 to 214.462) 0.001
eGFR (mL/min)
>120 1.492 (1.110 to 2.007) 0.008 1.072 (0.603 to 1.903) 0.813
90–120 1 (<0.001*) 1 (<0.001*)
60–89 1.360 (1.162 to 1.591) <0.001 1.504 (1.138 to 1.987) 0.04
45–59 3.849 (3.239 to 4.573) <0.001 4.255 (3.155 to 5.737) <0.001
30–44 6.590 (5.401 to 8.041) <0.001 7.715 (5.564 to 10.699) <0.001
15–29 14.465 (11.341 to 18.450) <0.001 15.054 (9.942 to 22.796) <0.001
ACR (mg/mmol)
Optimal (<1.1) 1 (<0.001*)
High normal (1.1–2.99) 1.363 (1.038 to 1.788) 0.026
High (3.0–29.99) 2.967 (2.381 to 3.697) <0.001
Very high (30–200) 6.253 (4.493 to 14.005) <0.001
Nephrotic (>200) 7.932 (4.493 to 14.005) <0.001
*p Value for overall effect.
ACR, albumin : creatinine ratio; AF, atrial fibrillation; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; IHD, ischaemic
heart disease; IMD, Index of Multiple Deprivation.

analyses are undertaken. The possible implication of this and found that its removal had little practical effect.
was investigated by Adams and White44 who analysed This suggests that the presence of the health domain is
data having removed the health domain from IMD 2004 unlikely to influence our result.

8 Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458

 Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
Table 4 Cox proportional hazard regression analysis: multivariate (adjusted) analyses (model 3)
ACR tested cohort
HR (95% CI) p Value
Ethnicity
White 1 (<0.001*)
South Asian 0.697 (0.56 to 0.868) 0.001
Black 0.533 (0.403 to 0.704) <0.001
Age (years)
50 and under 1 (<0.001*)
51–60 1.519 (0.907 to 2.546) 0.112
61–70 3.521 (2.17 to 5.712) <0.001
71–80 7.381 (4.61 to 11.818) <0.001
81–90 15.721 (9.534 to 25.922) <0.001
>90 51.641 (27.889 to 95.621) <0.001
Gender
Female as reference 1.782 (1.46 to 2.176) <0.001
Smoker
Non-smoker as reference 1.886 (1.488 to 2.392) <0.001
IMD rank
Quintile 1 (least deprived) <0.001 (<0.001 to >10^5) 0.952
Quintile 2 <0.001 (<0.001 to >10^5) 0.913
Quintile 3 0.978 (0.68 to 1.387) 0.902
Quintile 4 0.788 (0.585 to 1.062) 0.118
Quintile 5 (most deprived) 1 (0.65*)
Comorbidities
0 1 (<0.001*)
1 1.371 (0.932 to 2.016) 0.109
2 1.486 (1.019 to 2.166) 0.039
3 2.29 (1.53 to 3.428) <0.001
4 3.153 (2.002 to 4.964) <0.001
5 5.141 (2.869 to 9.212) <0.001
6 10.54 (2.52 to 44.084) 0.001
eGFR (mL/min)
>120 1.396 (0.782 to 2.492) 0.26
90–120 1 (<0.001*)
60–89 0.907 (0.982 to 1.207) 0.505
45–59 1.282 (0.932 to 1.763) 0.126
30–44 1.566 (1.095 to 2.239) 0.014
15–29 2.073 (1.315 to 3.268) 0.002
ACR (mg/mmol)
Optimal (<1.1) 1 (<0.001*)
High normal (1.1–2.99) 1.032 (0.784 to 1.359) 0.821
High (3.0–29.99) 1.837 (1.464 to 2.305) <0.001
Very high (30–200) 2.956 (2.132 to 4.099) <0.001
Nephrotic (>200) 3.838 (2.108 to 6.985) <0.001
*p Value for overall effect.
ACR, albumin : creatinine ratio; eGFR, estimated glomerular filtration rate; IMD, Index of Multiple Deprivation.

We found that the risk of death was lower for people requires further work. Variables such as health promo-
of South Asian and black ethnicity compared to people tion targeted at specific groups, differences in medica-
of white ethnicity, and this remained in all analyses tion usage or factors related to genetic diversity may
(adjusted and unadjusted) performed. Previous studies offer potential explanations for this variation.45 46
comparing the outcomes of different ethnic groups have A major strength in this study is the sample size, which
been limited in their generalisability. They have either included 62 practices of varying list size and a number
looked at disease-specific mortality8 18 20 21 or have been of practitioners. Ethnicity was documented in over 80%
based in populations that do not have access to free- of the population studied; this is much higher than is
comprehensive healthcare. The finding that differences normally found in primary care records.47 Self-reporting
in mortality risk between ethnic groups is independent is considered the ‘gold standard’ method of assessing
of age, gender, SES, kidney function and comorbidities ethnicity,27 taking into account an individual’s culture

Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458 9

 Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
electronic downloads. These downloads indicate who is
on a specific CV risk register and therefore may not clas-
sify people correctly. There is a relative paucity of pub-
lished literature regarding the correct identification of
people onto the correct risk registers.23 48–50 Surrogate
measures of accuracy of the data include previous
studies looking at gaming for QOF points (falsely classi-
fying people with conditions they do not have, thereby
increasing revenue) or exception reporting (excluding
individuals who have not had the appropriate monitor-
ing completed) suggesting that both these are
rare.23 51 52
When comparing the breakdown of the population
studied in these analyses with the source population, it
is important to highlight two key differences. First,
there is a relative underrepresentation of individuals of
white ethnicity, consistent with previous research.53 This
is most marked in those who had their ACR measured;
a higher number of men and individuals with diabetes
or of South Asian ethnicity had their ACR measured.
Comparing the whole cohort to those who had their
Figure 2 Cox regression survival plot indicating cumulative
survival between ethnicities in model 3 (comorbidities,
ACR reported showed similar trends for mortality in
estimated glomerular filtration rate and albumin : creatinine respect of age, eGFR, smoking status and SES, suggest-
ratio). Table below survival plot demonstrates the number of ing a generalisability of results. Second, one criterion
individuals who remained in follow-up at each time point. for inclusion was the recording of renal function within
the previous 12 months. This is likely to have resulted
in an overrepresentation of comorbidity as people with
and self-identity. Renal function was described in terms CV conditions would be more likely to have their renal
of eGFR and ACR, the latter assuming increased promin- function checked. A further consideration is that the
ence in the stratification of CV risk. accuracy and applicability of creatinine-based eGFR
Our analyses have used data from primary care coding equations, such as the formula used in this analysis, in
and recording systems, which formed part of the non-white ethnic groups is a subject of ongoing
research.54–56 Cystatin-based equations may be more
accurate,57 but are not routinely measured in clinical
practice.
In summary, we have shown that the determinants of
mortality were multifactorial in a high-risk population
and that ethnicity should be considered as a non-
traditional risk factor for mortality; the HR for death was
lower for South Asian and black individuals compared
to white individuals, which was, in part, independent of
age, gender, SES, renal function and comorbidities.
Furthermore, a simple cumulative comorbidity system
may have prognostic utility. Renal function (eGFR and
ACR) provides additional information and gender, while
age and smoking status remain significant risk factors
for mortality.

Contributors MJ participated in the study design, data analysis and

preparation of the manuscript. AL participated in the study design, data
acquisition and analysis and review of the manuscript. ACFB participated in
the study design, data acquisition and review of the manuscript. PC
participated in the study design, data analysis, preparation and review of the
manuscript.
Funding This research received no specific grant from any funding agency in
the public, commercial or not-for-profit sectors.
Competing interests MJ has received funding from the JABBS Foundation.
Figure 3 HR for death by number of comorbidities. ACFB has advised and received honoraria from Enhanced Healthcare Services
Multivariate (adjusted) analysis: model 3. Ltd.

10 Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458

 Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
Ethics approval The data were fully anonymised and were available as a 18. Balarajan R. Ethnic differences in mortality from ischaemic heart
component of an ongoing clinical development programme. The responsible disease and cerebrovascular disease in England and Wales. BMJ
NHS R&D Consortium stated that this study did not require ethical 1991;302:560–4.
19. Cappuccio FP. Ethnicity and cardiovascular risk: variations in people
submission to an NHS research ethics committee as it represented an of African ancestry and South Asian origin. J Hum Hypertens
evaluation of part of an ongoing primary care trust (PCT) programme. For 1997;11:571–6.
PCT data extraction, the PCT professional executive committee and GP locality 20. Khan NA, Wang H, Anand S, et al. Ethnicity and sex affect diabetes
leads provided approval for the programme, including evaluation and incidence and outcomes. Diabetes Care 2011;34:96–101.
publication. 21. Chaturvedi N, Jarrett J, Morrish N, et al. Differences in mortality and
morbidity in African Caribbean and European people with
Provenance and peer review Not commissioned; externally peer reviewed. non-insulin dependent diabetes mellitus: results of 20 year follow
up of a London cohort of a multinational study. BMJ
Data sharing statement No additional data are available. 1996;313:848–52.
22. Shekelle P. New contract for general practitioners. BMJ
Open Access This is an Open Access article distributed in accordance with 2003;326:457–8.
the Creative Commons Attribution Non Commercial (CC BY-NC 3.0) license, 23. Calvert M, Shankar A, McManus RJ, et al. Effect of the quality and
which permits others to distribute, remix, adapt, build upon this work non- outcomes framework on diabetes care in the United Kingdom:
commercially, and license their derivative works on different terms, provided retrospective cohort study. BMJ 2009;338:b1870.
the original work is properly cited and the use is non-commercial. See: http:// 24. PEEG, Current Estimates, Population Estimates by Ethnic Group
Mid-2002-2008 for Primary Care Organisations (experimental).
creativecommons.org/licenses/by-nc/3.0/
Secondary PEEG, Current Estimates, Population Estimates by
Ethnic Group Mid-2002-2008 for Primary Care Organisations
(experimental) 02/03/2009. 2008. http://www.ons.gov.uk/ons/
taxonomy/index.html?nscl=Population+Estimates+by+Ethnic+Group
25. National Institute for Health and Clinical Excellence: CG73: Chronic
Kidney Disease. Secondary National Institute for Health and Clinical
REFERENCES Excellence: CG73: Chronic Kidney Disease 2008. http://www.nice.
1. Go AS, Chertow GM, Fan D, et al. Chronic kidney disease and the org.uk/CG73
risks of death, cardiovascular events, and hospitalization. N Engl J 26. Centre HaSCI. Quality and Outcomes Framework. Secondary
Med 2004;351:1296–305. Quality and Outcomes Framework. http://www.hscic.gov.uk/qof
2. Matsushita K, van der Velde M, Astor BC, et al. Association of 27. Saunders CL, Abel GA, El Turabi A, et al. Accuracy of routinely
estimated glomerular filtration rate and albuminuria with all-cause recorded ethnic group information compared with self-reported
and cardiovascular mortality in general population cohorts: a ethnicity: evidence from the English Cancer Patient Experience
collaborative meta-analysis. Lancet 2010;375:2073–81. survey. BMJ Open 2013;3:pii: e002882.
3. Astor BC, Matsushita K, Gansevoort RT, et al. Lower estimated 28. Levey AS, Bosch JP, Lewis JB, et al. A more accurate method to
glomerular filtration rate and higher albuminuria are associated with estimate glomerular filtration rate from serum creatinine: a new
mortality and end-stage renal disease. A collaborative meta-analysis prediction equation. Modification of Diet in Renal Disease Study
of kidney disease population cohorts. Kidney Int 2011;79:1331–40. Group. Ann Intern Med 1999;130:461–70.
4. Gansevoort RT, Matsushita K, van der Velde M, et al. Lower 29. Index of Multiple Deprivation (IMD) 2007. Secondary Index of
estimated GFR and higher albuminuria are associated with adverse Multiple Deprivation (IMD) 2007 26/1/2013 2007. http://data.gov.uk/
kidney outcomes. A collaborative meta-analysis of general and dataset/index_of_multiple_deprivation_imd_2007
high-risk population cohorts. Kidney Int 2011;80:93–104. 30. Townsend P, Phillimore P, Beattie A. Health and deprivation:
5. Zhang QL, Rothenbacher D. Prevalence of chronic kidney disease in inequality and the north. London: Croom Helm, 1988.
population-based studies: systematic review. BMC Public Health 31. Jordan H, Roderick P, Martin D. The Index of Multiple Deprivation
2008;8:117. 2000 and accessibility effects on health. J Epidemiol Community
6. Coresh J, Selvin E, Stevens LA, et al. Prevalence of chronic kidney Health 2004;58:250–7.
disease in the United States. JAMA 2007;298:2038–47. 32. Centre HSCI. Primary Care Mortality Database. Secondary Primary
7. Peralta CA, Katz R, Deboer I, et al. Racial and ethnic differences in Care Mortality Database. http://www.hscic.gov.uk/pcmdatabase
kidney function decline among persons without chronic kidney 33. Levey AS, de Jong PE, Coresh J, et al. The definition, classification,
disease. J Am Soc Nephrol 2011;22:1327–34. and prognosis of chronic kidney disease: a KDIGO Controversies
8. Barbour SJ, Er L, Djurdjev O, et al. Differences in progression of Conference report. Kidney Int 2011;80:17–28.
CKD and mortality amongst Caucasian, Oriental Asian and South 34. Abbasi A, Peelen LM, Corpeleijn E, et al. Prediction models for risk
Asian CKD patients. Nephrol Dial Transplant 2010;25:3663–72. of developing type 2 diabetes: systematic literature search and
9. Choi AI, Rodriguez RA, Bacchetti P, et al. White/black racial independent external validation study. BMJ 2012;345:e5900.
differences in risk of end-stage renal disease and death. Am J Med 35. Smith S, Waterall J, Burden AC. An evaluation of the performance of
2009;122:672–8. the NHS Health Check programme in identifying people at high risk
10. Crews DC, Sozio SM, Liu Y, et al. Inflammation and the paradox of developing type 2 diabetes. BMJ Open 2013;3:pii: e002219.
of racial differences in dialysis survival. J Am Soc Nephrol 36. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying
2011;22:2279–86. prognostic comorbidity in longitudinal studies: development and
11. Hallan SI, Matsushita K, Sang Y, et al. Age and association of validation. J Chronic Dis 1987;40:373–83.
kidney measures with mortality and end-stage renal disease. JAMA 37. Khunti K, Morris DH, Weston CL, et al. Joint prevalence of diabetes,
2012;308:2349–60. impaired glucose regulation, cardiovascular disease risk and chronic
12. Herzog CA, Asinger RW, Berger AK, et al. Cardiovascular disease kidney disease in South Asians and White Europeans. PLoS ONE
in chronic kidney disease. A clinical update from Kidney Disease: 2013;8:e55580.
Improving Global Outcomes (KDIGO). Kidney Int 2011;80:572–86. 38. Carey IM, Shah SM, Harris T, et al. A new simple primary care
13. Mahmoodi BK, Matsushita K, Woodward M, et al. Associations of morbidity score predicted mortality and better explains between
kidney disease measures with mortality and end-stage renal disease practice variations than the Charlson index. J Clin Epidemiol
in individuals with and without hypertension: a meta-analysis. Lancet 2013;66:436–44.
2012;380:1649–61. 39. Johnson-Lawrence V, Kaplan G, Galea S. Socioeconomic mobility in
14. Benjamin EJ, Wolf PA, D’Agostino RB, et al. Impact of atrial adulthood and cardiovascular disease mortality. Ann Epidemiol
fibrillation on the risk of death: the Framingham Heart Study. 2013;23:167–71.
Circulation 1998;98:946–52. 40. Lazzarino AI, Hamer M, Stamatakis E, et al. The combined
15. Ford ES. Trends in mortality from all causes and cardiovascular association of psychological distress and socioeconomic status with
disease among hypertensive and nonhypertensive adults in the all-cause mortality: a national cohort study. JAMA Intern Med
United States. Circulation 2011;123:1737–44. 2013;173:22–7.
16. Afkarian M, Sachs MC, Kestenbaum B, et al. Kidney disease and 41. Jackson CA, Jones NR, Walker JJ, et al. Area-based socioeconomic
increased mortality risk in type 2 diabetes. J Am Soc Nephrol status, type 2 diabetes and cardiovascular mortality in Scotland.
2013;24:302–8. Diabetologia 2012;55:2938–45.
17. Bahrami H, Kronmal R, Bluemke DA, et al. Differences in the 42. Gallo V, Mackenbach JP, Ezzati M, et al. Social inequalities and
incidence of congestive heart failure by ethnicity: the multi-ethnic mortality in Europe—results from a large multi-national cohort. PLoS
study of atherosclerosis. Arch Intern Med 2008;168:2138–45. ONE 2012;7:e39013.

Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458 11

 Open Access

BMJ Open: first published as 10.1136/bmjopen-2013-003458 on 3 December 2013. Downloaded from http://bmjopen.bmj.com/ on March 19, 2024 by guest. Protected by copyright.
43. Levene LS, Bankart J, Khunti K, et al. Association of primary care 51. Doran T, Fullwood C, Reeves D, et al. Exclusion of patients from
characteristics with variations in mortality rates in England: an pay-for-performance targets by English physicians. N Engl J Med
observational study. PLoS ONE 2012;7:e47800. 2008;359:274–84.
44. Adams J, White M. Removing the health domain from the Index of 52. Campbell S, Hannon K, Lester H. Exception reporting in the Quality
Multiple Deprivation 2004—effect on measured inequalities in and Outcomes Framework: views of practice staff—a qualitative
census measure of health. J Public Health (Oxf ) 2006;28:379–83. study. Br J Gen Pract 2011;61:183–9.
45. Conley J, Tonelli M, Quan H, et al. Association between GFR, 53. de Lusignan S, Nitsch D, Belsey J, et al. Disparities in testing for
proteinuria, and adverse outcomes among white, Chinese, and renal function in UK primary care: cross-sectional study. Fam Pract
South Asian individuals in Canada. Am J Kidney Dis 2012;59:390–9. 2012;28:638–46.
46. Buckalew VM Jr, Freedman BI. Reappraisal of the impact of race on 54. Delanaye P, Cavalier E, Mariat C, et al. Estimating glomerular
survival in patients on dialysis. Am J Kidney Dis 2010;55:1102–10. filtration rate in Asian subjects: where do we stand? Kidney Int
47. Raleigh VS. Collection of data on ethnic origin in England. BMJ 2011;80:439–40.
2008;337:a1107. 55. Delanaye P, Mariat C, Maillard N, et al. Are the creatinine-
48. Lambert AM, Burden AC, Chambers J, et al. Cardiovascular based equations accurate to estimate glomerular filtration rate in
screening for men at high risk in Heart of Birmingham Teaching African American populations? Clin J Am Soc Nephrol
Primary Care Trust: the ‘Deadly Trio’ programme. J Public Health 2011;6:906–12.
(Oxf ) 2012;34:73–82. 56. Matsushita K, Mahmoodi BK, Woodward M, et al. Comparison of risk
49. Humphreys J, Harvey G, Coleiro M, et al. A collaborative project to prediction using the CKD-EPI equation and the MDRD study
improve identification and management of patients with chronic equation for estimated glomerular filtration rate. JAMA
kidney disease in a primary care setting in Greater Manchester. BMJ 2012;307:1941–51.
Qual Saf 2012;21:700–8. 57. Teo BW, Xu H, Wang D, et al. Estimating glomerular filtration rates
50. Martin D, Wright JA. Disease prevalence in the English population: by use of both cystatin C and standardized serum creatinine avoids
a comparison of primary care registers and prevalence models. Soc ethnicity coefficients in Asian patients with chronic kidney disease.
Sci Med 2009;68:266–74. Clin Chem 2012;58:450–7.

12 Jesky M, Lambert A, Burden ACF, et al. BMJ Open 2013;3:e003458. doi:10.1136/bmjopen-2013-003458