The Role of Race and Poverty on Steps to Kidney Transplantation in the Southeastern United States


Rachel E. Patzer,


Racial disparities in access to renal transplantation exist, but the effects of race and socioeconomic status (SES) on early steps of renal transplantation have not been well explored. Adult patients referred for renal transplant evaluation at a single transplant center in the Southeastern United States from 2005 to 2007, followed through May 2010, were examined. Demographic and clinical data were obtained from patient's medical records and then linked with United States Renal Data System and American Community Survey Census data. Cox models examined the effect of race on referral, evaluation, waitlisting and organ receipt. Of 2291 patients, 64.9% were black, the mean age was 49.4 years and 33.6% lived in poor neighborhoods. Racial disparities were observed in access to referral, transplant evaluation, waitlisting and organ receipt. SES explained almost one-third of the lower rate of transplant among black versus white patients, but even after adjustment for demographic, clinical and SES factors, blacks had a 59% lower rate of transplant than whites (hazard ratio = 0.41; 95% confidence interval: 0.28–0.58). Results suggest that improving access to healthcare may reduce some, but not all, of the racial disparities in access to kidney transplantation.


body mass index


confidence interval


electronic medical records


end-stage renal disease


erythropoiesis-stimulating agent


Emory Transplant Center


hazard ratio


interquartile range


socioeconomic status


United Network for Organ Sharing


United States Renal Data System


The 1972 enactment of legislation declaring end-stage renal disease (ESRD) patients as disabled provided the ESRD population with near-universal entitlement to Medicare coverage for both dialysis and transplantation. However, provision of Medicare coverage for ESRD did not eliminate reduced access to renal transplantation among different socioeconomic and racial groups (1,2). The reasons for these disparities are multifactorial, and occur both inside and outside of the healthcare arena (3). Increasing evidence suggests that poverty plays a role, and that transplant access is influenced by the socioeconomic status (SES) of a patient's local environment (1,4–6) above and beyond individual-level SES. In the Southeastern United States (US), the interplay between race and SES is particularly apparent, where black ESRD patients living in the poorest neighborhoods have been documented as 67% less likely to be placed on the deceased donor waiting list compared to whites in poor neighborhoods (6).

Because the deceased donor transplant process starts months to years before a patient may be waitlisted, focusing on earlier steps of the transplant process, such as renal transplant referral and evaluation, may better inform intervention efforts to improve equity (7). National surveillance systems currently do not capture data on transplant referral and evaluation. It is unclear whether the racial disparities observed in waitlisting and transplant receipt are due to disparities that occur prior to referral and evaluation, in between referral and waitlisting, or both, and whether SES plays a role. Several regional studies have documented racial disparities in earlier steps of transplant access (8–12). However, not all of these studies examined the role of SES; those that did either lacked individual measures of SES or used zip code-level measures of neighborhood poverty, which may have inadequately estimated the effect of SES on racial disparities. Examining data from a single transplant center provides more granularity in SES data to address these prior limitations. In particular, since the source population of patients from the Southeastern United States has a higher concentration of poverty (13) and majority African American ESRD population (14), we have greater power to examine race and SES disparities in early steps of the kidney transplant process. The purpose of this study was to determine whether racial disparities exist in kidney transplant referral, evaluation, waitlisting and deceased donor transplantation among adult, ESRD patients referred to a single transplant center in the Southeastern United States, and to determine the role of SES barriers that affect the rate of completion of each transplant step.

Materials and Methods

Data sources

Basic demographic and clinical data were obtained from the Emory University Hospital's hardcopy and electronic medical records. The patient data were linked to the United States Renal Data System (USRDS) surveillance data to obtain patient demographic and clinical information at the time of ESRD start. Follow-up data on evaluation, waitlisting and receipt of renal transplant were obtained from patient records and the United Network for Organ Sharing (UNOS) files on waitlisting sequence and transplant. The patient residential address was geocoded and assigned a census tract using ArcGIS 9.2. Data on neighborhood poverty were obtained from the American Community Survey 2005–2009 by patient census tract (15,16).

Study population

A total of 2821, incident, adult (age >18 years) ESRD patients were referred to the ETC's kidney transplant program for evaluation from 2005 to 2007. Patients were excluded from the study if their home address was missing or listed as a P.O. Box (n = 6) or if they lived outside of the Southeast (Alabama, Florida, Georgia, North Carolina, South Carolina or Tennessee) region (n = 79). Due to limited sample size, this study was restricted to patients who reported race as either black or white; patients with ‘other’ race or ‘Hispanic’ ethnicity were excluded (n = 174). Patients with no USRDS record were excluded (n = 152) due to lack of information about baseline characteristics. Patients who were listed at other transplant centers prior to referral to ETC (n = 119) were excluded. A total of 2291 patients referred to the ETC were included in the final study population (Figure 1).

Figure 1.

Flow diagram of study inclusion criteria and follow-up status.

Outcome variables

The primary outcome was time (in days) from ESRD start to receipt of deceased donor renal transplant. We also examined several distinct steps in access to transplantation: (1) ESRD to referral, (2) referral to evaluation start, (3) evaluation start to completion and (4) waitlisting to receipt of deceased donor transplant. We defined evaluation completion as the date a patient completed standard evaluation requirements and the transplant team determined waitlisting eligibility. Waitlisting was defined as the date a patient was listed as either status 7 (inactive) or status 1 (active) on the UNOS waiting list.

Patient outcome data were ascertained from ETC through May 2010, and confirmed with USRDS data through September 30, 2009, the most recent data available. High agreement between ETC and USRDS data was observed among waitlisting (91.3%; Cohen's kappa = 0.82) and transplant (94.5%; Cohen's kappa = 0.84) outcomes. When discrepancies were observed, the earlier date for the outcome was used. For all time to event analyses, patients were censored at death (n = 214), living donor transplant (n = 115), or end of the study (May 31, 2010). Patients were also censored at the date they were listed at another center (n = 189), or when the patient was removed from the list due to deterioration in medical condition, transfer to another center, or other reasons (n = 15). Patients who were referred (n = 535), waitlisted (n = 160) or transplanted (n = 60) prior to starting dialysis were assigned a time of 1 day for the time to event of interest.

Primary explanatory variable

The primary exposure variable for all analyses was self-reported race (black or white), based on data collected from the ETC at the time of renal transplant evaluation.

Patient-level covariates

Patient characteristics at the time of transplant referral were obtained from patient EMRs and included sex, patient address, etiology of ESRD (diabetes, hypertension, glomerulonephritis, or other) and BMI > 35 kg/m2. To obtain information about a patient's health status at the start of ESRD and for patients who were referred but did not start the evaluation, patient data were obtained from the USRDS medical evidence form (CMS 2728 form), given to all patients at the initiation of dialysis. Demographic data included patient age at dialysis start and clinical variables included predialysis erythropoiesis-stimulating agent (ESA) use (yes/no), hemoglobin (<11 g/dL vs. ≥ 11 g/dL), serum albumin (< 3.5 g/dL vs. ≥ 3.5 g/dL), cardiovascular disease (defined as history of congestive heart failure, ischemic heart disease, cardiac arrest, myocardial infarction, cardiac dysrhythmia, pericarditis or cerebrovascular disease), tobacco use (yes/no), or history of cancer (yes/no) at dialysis initiation. Reasons for incomplete evaluation, as reported by the Director of the ETC Kidney Transplant Program, were abstracted from patient records and included incomplete evaluation requirements, medical contraindication, financial or insurance reasons, patient choice, psychosocial, referred elsewhere, died, or unknown. For waitlisted patients, we examined blood type (A, B, AB or O), peak panel reactive antibody (PPRA) (0, 1–19.9% and ≥ 20%) and inactive listing (UNOS status 7) at the start of waitlisting (yes/no).

We defined individual SES using several variables: health insurance, highest education and employment status. We also examined the role of distance to transplant center as a proxy for access to care. Health insurance was categorized as private (employer), public (Medicare or Medicaid), other coverage, or no coverage. Patients with more than one type of insurance were categorized as employer if employer was listed anywhere in coverage type, and patients with Medicaid and other insurance were categorized as Medicaid. Additional SES data on education (less than high school degree, completed high school, some college, completed college, or unknown) and employment status (employed or full-time student, unemployed or disabled, or retired), were documented by two to three transplant coordinators over the period of the study and were abstracted from the psychosocial evaluation and history and physical notes from patient EMRs; these data were only available for patients who started the transplant evaluation process.

Neighborhood-level covariates

Using patient residential address at the time of transplant referral, we estimated neighborhood SES with 2005–2009 American Community Survey data on the proportion of individuals residing below the federal poverty line within a patient's residential census tract. We categorized neighborhood poverty a priori as (0–4.9%, 5–9.9%, 10–14.9%, 15–19.9% and ≥20%). Rural Urban Commuting Area (RUCA) codes were obtained from the Community Health Status Indicators Project and considered a proxy for neighborhood-level access to care.

Data analysis

Differences in the means and proportions of patient demographic and clinical characteristics by race were examined using chi-square tests and t-tests (or nonparametric equivalents of the t-test). To examine whether racial differences exist in the time to deceased donor transplant, we examined the rate of completion of each intermediate step from dialysis start to patient referral, referral to evaluation start, evaluation start to completion and waitlisting to transplant receipt among patients who completed the prior step using Kaplan–Meier estimation methods and the log-rank test for significance.

Prior to model assessment, all covariates and interaction terms were entered into an initial model to assess for covariate collinearity. To assess whether racial disparities varied across SES for each outcome, we examined interactions between race and each SES in multivariable Cox models using the likelihood ratio test to assess significance (17). To assess whether SES explained the racial differences in access to each transplant step, we examined sequential Cox models. We quantified the proportion of the effect of reduced access to each transplant step among black versus white patients explained by specific factors of interest using the equation (HRcrude– HR adjusted)/(HRcrude– 1.0).

For all multivariable-adjusted models, both patient- and census tract-level variables were considered as potential confounders. We used the robust sandwich variance estimator with census tract as the cluster variable to examine neighborhood poverty and individual level covariates simultaneously, while also accounting for potential correlation of patients within neighborhoods (18). We also controlled for time due to the introduction of an in-center education program that was implemented in September 2007 at the ETC. We evaluated confounding by comparing meaningful changes in point estimates from a full model containing all a priori covariates to all other potential models (19,20) and by examining causal graphs to examine intermediate pathways between race, SES and transplant access (21,22). In multivariable analyses, we used the Markov Chain Monte Carlo method for multiple imputation methods for missing covariate information (23).

SAS 9.2 was used for all statistical analyses. ArcGIS 9.2 was used for geocoding, spatial joining and calculating distance from patient residence to the ETC. For all analyses, two-tailed p < 0.05 was considered statistically significant. This study protocol was approved by the Emory IRB.

Sensitivity analyses

In Cox models, we assumed that censoring due to death and living donor transplant was an independent, rather than random, censoring event and calculated cause-specific hazard ratios (HR) and 95% confidence intervals (CI) for deceased donor transplant. However, we considered these events in a competing risk model by examining how the effect of race changed when we considered that all patients who were censored due to death and living donor transplant either received or did not receive a deceased donor transplant.


Study population characteristics

Among the 2291 adult (> 18 years) incident ESRD patients referred for transplant, the mean age at ESRD start was 49.4 ± 13.9 years, 64.9% were black, 57.3% were male, 33.6% lived in impoverished communities and 16.1% had no health insurance coverage (Table 1). Compared to whites, a greater proportion of black patients were younger (47.4 years vs. 53.1 years), female (45.3% vs. 38.0%) and had hypertension as the primary cause of ESRD (37.2% vs. 20.3%, p < 0.005). Black patients had reduced prevalence of cardiovascular disease (38.9% vs. 45.8%, p = 0.0013) compared to whites. In addition, blacks were more likely to have lower serum albumin (<3.5 g/dL), lower hemoglobin (<10 g/dL) and lower predialysis ESA use than whites (p<0.01 for all comparisons).

Table 1.  Characteristics of patients referred for renal transplantation
 Study population N=2291White N = 805 (35.1%)Black N=1486 (64.9%)p-Value for race difference
  1. 1ESA = erythropoiesis-stimulating agent.

  2. 2Health insurance coverage may sum to >100% in patients with multiple sources of coverage.

Patient-level characteristics at ESRD start
Age, mean (SD), years49.4 ± 13.953. 1 ± 13.947.4 ± 13.4< 0.0001
Age category, N (%), years   < 0.0001
 20–39598 (26.1%)151 (18.8%)447 (30.1%) 
 40–49506 (22.1%)152 (18.9%)354 (23.8%) 
 50–59634 (27.7%)224 (27.8%)410 (27.6%) 
 60–69400 (17.5%)188 (23.4%)212 (14.3%) 
 70–85153 (6.7%)90 (11.2%)63 (4.2%) 
Male sex, N (%)1312 (57.3%)499 (62.0%)813 (54.7%)0.0008
Cause of ESRD, N (%)   < 0.0001
 Diabetes879 (38.4%)331 (41.1%)548 (36.9%) 
 Hypertension715 (31.2%)163 (20.3%)552 (37.2%) 
 Glomerulonephritis248 (10.8%)111 (13.8%)137 (9.2%) 
 Other449 (19.6%)200 (24.8%)249 (19.6%) 
Distance to transplant center, median, IQR, miles66.6 (16.4, 98.0)54.0 (28.6, 98.5)27.6 (12.6, 96.8)< 0.0001
Clinical and laboratory measures at ESRD start
 BMI > 35 kg/m2422 (18.4%)140 (17.4%)282 (19.0%)0.3499
 Tobacco use156 (6.8%)68 (8.5%)888 (5.9%)0.0293
 Cardiovascular disease947 (41.3%)369 (45.8%)578 (38.9%)0.0013
 History of cancer37 (1.6%)22 (2.7%)15 (1.0%)0.0018
 Serum albumin < 3.5 g/dL1548 (67.6%)476 (59.1%)1072 (72.1%)< 0.0001
 Hemoglobin < 10 g/dL1234 (53.9%)331 (41.1%)903 (60.8%)< 0.0001
 Predialysis ESA 1664 (29.0%)282 (35.0%)382 (25.7%)< 0.0001
Socioeconomic characteristics at ESRD start
Health insurance coverage2    
 Medicaid459 (20.0%)153 (19.0%)306 (20.6%)0.3652
 Medicare645 (28.2%)253 (31.4%)392 (26.4%)0.0103
 Employer group1004 (43.8%)406 (50.4%)598 (40.2%)< 0.0001
 Other coverage241 (10.5%)128 (15.9%)113 (7.6%)< 0.0001
 No coverage368 (16.1%)83 (10.3%)285 (19.2%)< 0.0001
Neighborhood poverty (% census tract below poverty)   < 0.0001
 0–4.9%170 (7.4%)102 (12.7%)68 (4.6%) 
 5–9.9%462 (20.2%)222 (27.6%)240 (16.2%) 
 10–14.9%515 (22.5%)208 (25.8%)307 (20.7%) 
 15–19.9%375 (16.4%)119 (14.8%)256 (17.2%) 
 >20%769 (33.6%)154 (19.1%)615 (41.4%) 
Degree of rurality   < 0.0001
 Urban1709 (74.6%)529 (65.7%)1180 (79.4%) 
 Large rural313 (13.7%)138 (17.1%)175 (11.8%) 
 Small rural146 (6.4%)62 (7.7%)84 (5.7%) 
 Remote small rural123 (5.4%)76 (9.4%)47 (3.2%) 

Racial differences in access to renal transplant steps

Among all patients eligible to progress to the next transplant step, a greater proportion of white versus black patients proceeded in starting the evaluation, waitlisting and receiving a transplant (p < 0.05), but no significant racial differences were observed in evaluation completion. Figure 2, panel A shows the racial differences in access to each transplant step among those who completed the prior step.

Figure 2.

Racial differences in transplant step completion and duration. Panel A shows the proportion of eligible patients completing each transplant step. Among all patients referred, 60.6% of whites and 51.5% of blacks started the evaluation; among all those who started the evaluation, 92% of white and 90.8% of blacks completed the evaluation process. Among patients who completed the evaluation requirements, 71.0% of white and 59.6% of black patients were placed on the deceased donor waiting list. Among waitlisted patients, 30.0% of whites and 18.1% of black patients received a deceased donor transplant during follow-up. Racial differences were also observed in the median time to completion for each step (panel B). Differences in the time to referral and time from referral to evaluation were evident. The greatest racial differences were observed in the final step, waitlisting to transplantation. *p < 0.001.

Among the 2291 patients referred to the ETC, only 54.7% of patients came to the first evaluation appointment. A greater proportion of black versus white patients did not start the evaluation (48.5% vs. 39.4%, p < 0.0001). Most (91.3%) patients completed the standard requirements for the transplant evaluation. Reasons for not completing the evaluation were comparable among racial groups with the exception of “incomplete evaluation requirements”, where black patients were significantly more likely to have incomplete requirements than white patients (45.7% vs. 17.9%) and have psychosocial reasons reported as a reason for not completing the evaluation process (10.0% vs. 0%) (Figure 1; p < 0.0001 for both comparisons).

A total of 733 patients (64.1% of patients who completed the standard evaluation requirements) were placed on the deceased donor waiting list. Among those waitlisted, 414 (56.4%) were inactive at some point during follow-up, with a median time inactive of 260 days (interquartile range [IQR]: 72, 585). Among those inactive, nearly half (46.8%) were first listed as inactive, with an additional 28.2% of patients inactive at some time after active listing (Figure 1). A greater proportion of inactive patients were black (68.1%) than white (40.8%) (p = 0.0283). Among all listed patients, 177 (24.1%) received a deceased donor transplant during the study period.

Racial differences in the duration of time patients remained in each transplant step were also observed (Figure 2, panel B). The overall median time from transplant referral to deceased donor transplant was 743 days (IQR: 453, 977) for whites and 1,096 days (IQR: 741, 1385) for black patients. The greatest racial differences were observed from ESRD start to referral and once a patient was placed on the waiting list (Figure 2, panel B).

Role of SES in racial differences to transplant access

Socioeconomic differences by race were also observed. For example, 19.2% of blacks and 10.3% of whites had no health insurance at the time of dialysis initiation, and black ESRD patients were twofold more likely to live in poor neighborhoods compared to white patients (Table 1; p < 0.0001 for both comparisons). Of those who did not have health insurance at the start of ESRD, the mean age was 41.8 years, the majority (77.6%) were black and 42.1% lived in neighborhoods where >20% of the census tract lived below the federal poverty line (data not shown).

SES measures for insurance, education, employment, distance, neighborhood poverty and degree of rurality were independently associated with completing several transplant steps, and there was a general trend of higher SES at each progressive step of the transplant process. For example, 43.8% of referred patients had private insurance (Table 1), but this increased to 56.8% of those who started the evaluation, 57.9% of those who completed the evaluation, 69.4% of those waitlisted and 74.0% of those who received a transplant (Table 2).

Table 2.  Proportion of individuals completing each transplant step by race and SES
Started evaluation processStudy population N = 1253White n = 488 (39.0%)Black n = 765 (61.0%)p-Value for race differencep-Value for association of SES and completing transplant step*
  1. *p-Values represent the significance of the differences in the means or proportions of SES characteristics by each outcome (transplant step) using chi-square tests or t-tests.

  2. 1Columns do not add up to the study population due to missing data on education for patients who started the evaluation (n = 29); patients who completed the evaluation (n = 28), patients waitlisted (n = 12) and patients transplanted (n = 3).

Health insurance coverage at time of evaluation, N (%)< 0.0001< 0.0001
 Private712 (56.8%)323 (66.2%)389 (50.9%)  
 Public (or other)541 (42.3%)165 (33.8%)376 (49.2%)  
Education, N (%) 1   0.0101< 0.0001
 Less than high school224 (17.9%)80 (16.4%)144 (18.8%)  
 Completed high school419 (33.4%)168 (34.4%)251 (32.8%)  
 Some college339 (27.1%)113 (23.2%)226 (29.5%)  
 Completed college242 (19.3%)115 (23.6%)127 (16.6%)  
Employment, N (%)   0.0003< 0.0001
 Employed425 (34.0%)163 (33.5%)262 (34.1%)  
 Unemployed or disabled462 (36.9%)209 (42.9%)253 (33.1%)  
 Retired366 (29.2%)115 (23.6%)251 (32.8%)  
Distance to center, median (IQR), miles35.4 (16.5, 89.0)46.8 (25.5, 89.4)23.2 (12.8, 87.2) < 0.0001
Neighborhood poverty (% census tract below poverty), N (%)< 0.0001< 0.0001
 <5% (wealthiest)114 (9.1%)74 (15.2%)40 (5.2%)  
 5–9.9%278 (22.2%)141 (29.0%)137 (17.9%)  
 10–14.9%306 (24.4%)129 (26.4%)177 (23.1%)  
 15–19.9%202 (16.1%)64 (13.1%)138 (18.0%)  
 >20% (poorest)353 (28.2%)80 (16.4%)273 (35.7%)  
Degree of rurality   < 0.00010.0186
 Urban960 (76.6%)324 (66.4%)636 (83.1%)  
 Large rural152 (12.1%)78 (16.0%)37 (4.8%)  
 Small rural72 (5.8%)35 (7.2%)37 (4.8%)  
 Remote small rural69 (5.5%)51 (10.4%18 (2.4%)  
evaluationpopulationn = 449n = 695for race 
Health insurance coverage at time of evaluation, N (%)< 0.00010.0204
 Private663 (57.9%)305 (67.9%)358 (51.5%)  
 Public (or other)481 (42.1%)144 (32.1%)337 (48.5%)  
Education, N (%)1   0.01010.2746
 Less than high school197 (17.7%)68 (15.5%)129 (19.0%)  
 Completed high school382 (34.2%)156 (35.6%)226 (33.3%)  
 Some college313 (28.1%)108 (24.7%)205 (30.2%)  
 Completed college224 (20.1%)106 (24.2%)118 (17.4%)  
Employment, N (%)   0.00070.1058
 Employed398 (34.8%)156 (34.7%)242 (34.8%)  
 Unemployed or disabled417 (36.5%)189 (42.1%)228 (32.8%)  
 Retired329 (28.8%)104 (23.2%)225 (32.4%)  
Distance to center, median (IQR), miles35.6 (16.6, 89.1)46.5 (25.1, 89.4)24.0 (13.1, 88.4)< 0.00010.6022
Neighborhood poverty (% census tract below poverty), N (%)< 0.00010.0650
 <5% (wealthiest)111 (9.7%)73 (16.3%)28 (5.5%)  
 5–9.9%253 (22.1%)130 (28.9%)123 (17.7%)  
 10–14.9%276 (24.1%)115 (25.6%)161 (23.2%)  
 15–19.9%178 (15.6%)57 (12.7%)121 (17.4%)  
 >20% (poorest)326 (28.5%)74 (16.5%)252 (36.3%)  
Degree of rurality   < 0.00010.1948
 Urban875 (76.5%)300 (66.8%)575 (82.7%)  
 Large rural143 (12.5%)70 (15.6%)73 (10.5%)  
 Small rural67 (5.9%)33 (7.4%)34 (4.9%)  
 Remote small rural59 (5.2%)46 (10.2%)13 (1.9%)  
Listed forpopulationN = 319N = 414for race 
transplantN = 733(43.5%)(56.4%)difference 
Health insurance coverage at time of evaluation, N (%)< 0.0001< 0.0001
 Private509 (69.4%)246 (77.1%)263 (63.5%)  
 Public (or other)224 (30.6%)73 (22.9%)151 (36.5%)  
Education, N (%)1   0.1761< 0.0001
 Less than high school83 (11.5%)34 (10.7%)49 (12.1%)  
 Completed high school250 (34.7%)114 (36.0%)136 (33.7%)  
 Some college207 (28.7%)80 (25.2%)127 (31.4%)  
 Completed college181 (25.1%)89 (28.1%)92 (22.8%)  
Employment, N (%)   0.0342< 0.0001
 Employed312 (42.6%)132 (41.4%)180 (43.5%)  
 Unemployed230 (31.4%)115 (36.1%)115 (27.8%)  
 Retired191 (26.1%)72 (22.6%)119 (28.7%)  
Distance to center, median (IQR), miles32.5 (16.7, 89.2)43.8 (23.9, 93.2)22.2 (13.7, 79.7)< 0.00010.0804
Neighborhood poverty (% census tract below poverty), N (%)< 0.0001< 0.0001
 <5% (wealthiest)87 (11.9%)59 (18.5%)28 (6.8%)  
 5–9.9%186 (25.4%)95 (29.8%)91 (22.0%)  
 10–14.9%180 (24.6%)77 (24.1%)103 (24.9%)  
 15–19.9%101 (13.8%)36 (11.3%)65 (15.7%)  
 >20% (poorest)179 (24.4%)52 (16.3%)127 (30.7%)  
Degree of rurality   < 0.00010.0354
 Urban580 (79.1%)222 (69.6%)358 (86.5%)  
 Large rural82 (11.2%)44 (13.8%)38 (9.2%)  
 Small rural35 (4.8%)21 (6.6%)14 (3.4%)  
 Remote small rural36 (4.9%)32 (10.0%)4 (1.0%)  
Received deceasedpopulationN = 99N = 78for race 
donor transplantN = 177(55.9%)(44.1%)difference 
Health insurance coverage at time of evaluation, N (%)0.19800.2580
 Private131 (74.0%)77 (77.8%)54 (69.2%)  
 Public (or other)46 (26.0%)22 (22.2%)24 (30.8%)  
Education, N (%)1   0.01200.3375
 <HS16 (9.2%)9 (9.1%)7 (9.3%)  
 Completed HS55 (31.6%)30 (30.3%)25 (33.3%)  
 Some college54 (31.0%)24 (24.2%)30 (40.0%)  
 Completed college49 (28.2%)36 (36.4%)13 (17.3%)  
Employment, N (%)   0.88270.4738
 Employed81 (45.8%)45 (45.5%)36 (46.2%)  
 Unemployed53 (29.9%)31 (31.3%)22 (28.2%)  
 Retired43 (24.3%)23 (23.2%)20 (25.6%)  
Distance to center, median (IQR), miles43.6 (19.9, 93.8)48.5 (25.1, 98.4)58.9 (15.4, 93.0)0.02560.0172
Neighborhood poverty (% census tract below poverty), N (%)0.05490.0483
 <5% (wealthiest)31 (17.5%)24 (24.2%)7 (9.0%)  
 5–9.9%40 (22.6%)24 (24.2%)16 (20.5%)  
 10–14.9%43 (24.3%)22 (22.2%)21 (26.9%)  
 15–19.9%26 (14.7%)13 (13.2%)13 (16.7%)  
 >20% (poorest)37 (20.9%)16 (16.2%)21 (16.9%)  
Degree of rurality   0.00160.3872
 Urban132 (74.6%)65 (65.7%)67 (85.9%)  
 Large rural25 (14.1%)15 (15.2%)10 (12.8%)  
 Small rural8 (4.5%)7 (7.1%)1 (1.3%)  
 Remote small rural12 (6.8%)12 (12.1%)0 (0.0%)  

In multivariable Cox models, we found no statistically or clinically significant interactions between race and any SES measure. Table 3 shows the multivariable sequential modeling results. Overall, demographic and clinical factors explained 20.8% of the reduced transplant rate among blacks versus whites, and individual and neighborhood SES factors explained 30.6%. The degree to which demographic, clinical and SES factors explained racial disparities varied at each transplant step. SES factors accounted for 26.9% and demographic and clinical factors 16.4% of the lower rate of referral among blacks versus whites, but accounted for little to none of the disparities observed in the rate of starting the transplant evaluation once referred. Once waitlisted, clinical, demographic and clinical factors explained more than half (55.8%) of the reduced transplant rate among black versus white patients, and clinical and demographic factors (38.4%) explained a greater fraction of the disparity than SES factors (30.4%). However, large racial differences persisted even after adjustment for clinical, demographic and SES factors (Table 3).

Table 3.  Multivariable sequential Cox modeling results for effect of race on access to each transplant step
Overall outcomeBlack:white hazard ratio (95% CI)p-Value for race difference% Effect explained by adjustment of specific factors*
  1. 1Model 2 adjusts for the following covariates: age, sex, etiology of ESRD, cardiovascular disease, BMI > 35, ESA use, hypoalbuminemia (<3.5 g/dL), low hemoglobin (<10 g/dL) and blood type; Model 3 adjusts for health insurance, education, distance to center, degree of rurality and census tract poverty; Model 4 adjusts for all covariates included in both model 2 and model 3.

  2. 2Model 2 adjusts for the following covariates: age, sex, etiology of ESRD, cardiovascular disease, BMI > 35, ESA use, hypoalbuminemia (<3.5 g/dL), low hemoglobin (<10 g/dL), inactive waitlisting (yes/no), blood type and PRA. Model 3 adjusts for health insurance, education, distance to center, degree of rurality and census tract poverty; model 4 adjusts for all covariates included in both model 2 and model 3.

  3. *Because of a protective effect, we calculated the percentage of the effect explained by adjustment of factors using the inverse of the HR in the following formula: [(1/HRcrude) – (1/HRadjusted)]/[(1/HRcrude) – 1.0)]*100, where each adjusted model (model 2, 3, 4) was compared with the unadjusted model to calculate the percentage of the effect of reduced access to each transplant step by the specific factors adjusted for in each model.

Overall time from ESRD start to receipt of deceased donor transplant1
 Model 1: unadjusted0.29 (0.25–0.33)< 0.0001 
 Model 2: demographic + clinical factors0.34 (0.25–0.47)< 0.0001+ 20.8%
 Model 3: individual + neighborhood SES factors0.37 (0.26–0.51)< 0.0001+ 30.6%
 Model 4: clinical + demographic + SES factors0.41 (0.28–0.58)< 0.0001+ 41.2%
Step 1: Time from ESRD start to referral1
 Model 1: unadjusted0.60 (0.55–0.66)< 0.0001 
 Model 2: demographic + clinical factors0.64 (0.57–0.73)< 0.0001+ 16.4%
 Model 3: individual + neighborhood SES factors0.67 (0.59–0.76)< 0.0001+ 26.9%
 Model 4: clinical + demographic + SES factors0.70 (0.61–0.80)< 0.0001+ 35.8%
Step 2: Time from referral to evaluation start1
 Model 1: unadjusted0.72 (0.69–0.76)< 0.0001 
 Model 2: demographic + clinical factors0.73 (0.64–0.82)< 0.0001+ 5.1%
 Model 3: individual + neighborhood SES factors0.70 (0.62–0.80)< 0.0001−10.3%
 Model 4: clinical + demographic + SES factors0.72 (0.63–0.83)< 0.0001+ 0%
Step 3: Time from evaluation start to evaluation completion1
 Model 1: unadjusted0.95 (0.90–0.99)0.0384 
 Model 2: clinical + demographic0.96 (0.84–1.09)0.4946+ 20%
 Model 3: individual + neighborhood SES factors0.97 (0.84–1.10)0.6110+ 40%
 Model 4: clinical + demographic + SES factors1.00 (0.87–1.15)0.9748+ 100%
Step 4: Time from waitlisting to transplant (among waitlisted patients)2
 Model 1: unadjusted0.42 (0.31–0.57)< 0.0001 
 Model 2: clinical + demographic0.54 (0.39–0.75)0.0003+ 38.4%
 Model 3: individual + neighborhood SES factors0.51 (0.36–0.72)0.0002+ 30.4%
 Model 4: clinical + demographic + SES factors0.62 (0.42–0.91)0.0139+ 55.8%

Sensitivity analyses

A total of 656 patients were censored during the study period, 115 (9.2%) were censored after receiving a living donor transplant and 541 (23.6%) died. Racial differences were evident among the population of patients who received a living donor transplant (69.6% white, 30.4% black recipients, p < 0.0001), representing 16.4% of the white but only 4.6% of the black ESRD study population. Compared to the study population, a greater proportion of living donor recipients lived in wealthy neighborhoods, had private insurance, were employed and lived in rural areas (p < 0.05 for each comparison). Racial differences were also apparent in the proportion of patients who died during the follow-up, where 27.7% of whites and 21.4% of blacks died (p = 0.0007). In sensitivity analyses that examined the competing risk of living donor transplant, there were similar racial disparities in each transplant step. When all patients who received a living donor transplant were considered as having a transplant, the probability of receiving a transplant at any given time after referral was 64% lower among black versus white patients (HR = 0.36; 95% CI: 0.27–0.47). Results for other transplant steps were within 10% of the main analysis.


Black ESRD patients evaluated at a large transplant center in the Southeastern United States have reduced access to renal transplantation, where black patients are 59% less likely to receive a transplant at any given time compared to white patients. Individual- and neighborhood-level SES factors explained 30.6% of the reduced rate of transplant among black versus white patients, but substantial racial disparities persisted in several steps of the renal transplant process even after accounting for demographic, clinical and SES factors, particularly in referral for transplant and the rate of transplantation among waitlisted patients. The results of this study suggest that improving access to healthcare may reduce some, but not all of the racial disparities observed in access to kidney transplantation.

Previous studies have also documented racial disparities in access to transplant referral and evaluation completion. In a study of dialysis patients in Indiana, Kentucky and Ohio, Alexander et al. (8) reported that black patients were less likely to be interested in transplant, complete the pretransplant workup and move up a waiting list compared to white patients. Other studies have reported that blacks complete the transplant evaluation process slower (11) and are less likely to be rated as appropriate candidates for transplant even after evaluation compared to white patients (12).

The disparities observed between racial and ethnic groups are not entirely explained by clinical or biologic factors, and SES has been hypothesized to play a role (8,24,25). Schold et al. recently reported that racial disparities observed in a population of patients referred for renal transplantation in Florida are largely explained by SES, as measured by health insurance and county income (26). In contrast, Hall et al. reported that racial disparities in access to waitlisting were somewhat attenuated after adjusting for SES, where health insurance coverage and zip code poverty explained 21% of the reduced rate of waitlisting among black patients in the United States. However, once a patient was waitlisted, SES accounted for little if any of the racial disparities (1). Similarly, in our study, SES as measured by more sensitive SES measures accounted for 30.6% of the reduced rate of transplant but SES did not entirely explain the racial disparity observed in access to renal transplant. Additionally, even though black patients in our study had lower SES on average, the racial disparity in transplant access was consistent across levels of SES.

There are several potential explanations why racial disparities that are unexplained by demographic, clinical and SES factors measured in this study exist in early steps of the renal transplant process. Racial bias may partially explain why black patients have reduced access to transplant compared to white patients (3,27,28). Klassen et al. examined the role of racial discrimination among adult renal transplant eligible patients in hemodialysis centers in Baltimore, finding that patients who reported a lifetime experience of racial discrimination experienced reduced access to waitlisting (29). Variations in referral for transplant at the referring provider or dialysis facility level may also account for the observed racial differences. Prior research suggests that a small number of providers may account for the racial differences in the quality of patient care (30,31). Limited access to healthcare may disproportionately affect minority patients (3) and if not adequately measured, could explain some of the observed racial disparity. In a recent study, Prakash et al. found that as the percentage of black patients in a neighborhood increases, the likelihood of access to pre-ESRD nephrology care decreases (32). Compared to whites, black ESRD patients in our study population were more likely to lack insurance coverage, have a lower prevalence of predialysis ESA use, and more anemia and hypoalbuminemia at incident ESRD, all of which are proxies for early access to healthcare and may portend poorer health status at time of referral for evaluation. We also used neighborhood poverty, degree of rurality and distance to transplant center as proxies for access to care, but perhaps better measures of healthcare quality and accessibility could explain some of the observed racial differences in access to kidney transplantation.

Our study has several strengths. The racial distribution of our study (61% black) provides us with ample study power to examine racial differences in access to each step of the renal transplant process. Follow-up data for this analysis was validated using USRDS-linked with UNOS data to capture virtually all waitlist and transplant outcomes, thus limiting selection bias due to loss to follow-up. We assessed SES using both individual- and neighborhood-level measures, which permitted the evaluation of poverty in a multilevel framework. The use of census data to estimate neighborhood SES, as opposed to zip code data, is more sensitive (33). National studies that examine access to the deceased donor waitlist and transplant receipt typically include all dialysis patients, even though some may have medical comorbidities that preclude them from transplantation (9). Since 30% of dialysis patients may be medically ineligible for renal transplantation, an examination of racial disparities among referred patients is preferred (29). We examined the proportion of patients progressing to each step based on the number of patients completing the prior step, rather than including all dialysis patients.

Limitations to our study should be noted. Our study was conducted at a single transplant center and results may not be generalizable to all ESRD patients. For example, the racial disparities observed in the Southeast may be more pronounced. In our study, which took place over a median follow-up of nearly 4 years, black patients comprised 56% of the waiting list but only represented 44% of those transplanted. However, on a national level, blacks received 33.4% of all deceased donor transplants, nearly equivalent to their waiting list proportion (34.1%) (34). At this time, the assessment of factors which impact transplant referral and evaluation are not captured in national surveillance databases but these data are important to better understand disparities in the continuum of the transplant process in other regions throughout the United States. Regional and geographic differences between the Southeast and other regions may also limit generalizability of these findings, such as differences in care for ESRD patients (35), average transplant waiting times (36), or transplant center characteristics (37).

Although our study adjusted for demographic, clinical and SES factors, there may be unmeasured factors unaccounted for in our analyses, such as changes in health status. Information on the reasons a patient had incomplete evaluation requirements or inactive waitlisting was not available in our study and could explain part of the observed racial disparity. This study was retrospective in nature and the causal pathway between SES and racial disparities cannot be discerned. In addition, the source population of ESRD patients eligible for transplant in this study is not known. Our results may underestimate racial disparities if the rate of referral is higher among whites versus blacks. Additionally, we were unable to determine if dialysis facility factors contributed to the observed racial disparities; it would be helpful to collect national data on referral and evaluation rates by dialysis facility to identify center practice variations and target interventions to improve equitable access to transplant.

To our knowledge, this is the first study to examine race and both individual- and census tract-level SES in access to each step of the renal transplant process. We found that racial disparities are evident in several steps of the renal transplant process, and that while SES explained some of the racial differences, black patients had a 59% lower rate of transplant than white patients even after accounting for demographic, clinical and SES factors. This suggests that improving access to care for patients may help reduce some, but not all, of the racial disparities in access to transplant. Efforts to improve equity in access to renal transplantation may need to focus on earlier steps of transplant access, in addition to the waitlisted population.


We would like to acknowledge Paul Eggers, PhD, and Rebecca Zhang, MS, for assistance in matching Emory Transplant Center data with USRDS data. A portion of this work was presented at the American Transplant Congress 2011 Annual Meeting (Philadelphia, PA) Abstract 636. This study was funded in part by the Emory University Race and Difference Initiative. The interpretation and reporting of the data presented here are the responsibility of the authors and in no way should be seen as an official policy or interpretation of the US government.


The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.