Racial Differences in Determinants of Live Donor Kidney Transplantation in the United States

Abstract

Few studies have compared determinants of live donor kidney transplantation (LDKT) across all major US racial-ethnic groups. We compared determinants of racial-ethnic differences in LDKT among 208 736 patients who initiated treatment for end-stage kidney disease during 2005–2008. We performed proportional hazards and bootstrap analyses to estimate differences in LDKT attributable to sociodemographic and clinical factors. Mean LDKT rates were lowest among blacks (1.19 per 100 person-years [95% CI: 1.12–1.26]), American Indians/Alaska Natives-AI/ANs (1.40 [1.06–1.84]) and Pacific Islanders (1.10 [0.78–1.84]), intermediate among Hispanics (2.53 [2.39–2.67]) and Asians (3.89 [3.51–4.32]), and highest among whites (6.46 [6.31– 6.61]). Compared with whites, the largest proportion of the disparity among blacks (20%) and AI/ANs (29%) was attributed to measures of predialysis care, while the largest proportion among Hispanics (14%) was attributed to health insurance coverage. Contextual poverty accounted for 16%, 4%, 18%, and 6% of the disparity among blacks, Hispanics, AI/ANs and Pacific Islanders but none of the disparity among Asians. In the United States, significant disparities in rates of LDKT persist, but determinants of these disparities vary by race-ethnicity. Efforts to expand preESKD insurance coverage, to improve access to high-quality predialysis care and to overcome socioeconomic barriers are important targets for addressing disparities in LDKT.

Introduction

Despite the success of live donor kidney transplantation (LDKT) as an optimal treatment for patients with end-stage kidney disease (ESKD) (1,2), significant racial-ethnic disparities in attainment of LDKT persist (3–5). In the United States, racial-ethnic minorities experience disproportionately high rates of ESKD; yet for over two decades, minority patients have been consistently less likely than their white counterparts to undergo LDKT (3–6).

Determinants of disparities in LDKT appear to operate on multiple levels and may include racial-ethnic differences in individual- and area-level factors such as differences in socioeconomic status, chronic disease burden, healthcare access, and in attitudes and knowledge regarding LDKT (7–14). As evidenced by geographic differences in transplant rates, population factors such as the disproportionately high burden of clinical comorbidities within minority communities and disparities in area resource allocation (i.e. contextual poverty) may also contribute to these differences (15–19).

To date, prior studies have not quantified the extent to which these individual and area-level factors might contribute to disparities in LDKT across all major US racial-ethnic groups. Furthermore, the extent to which these measures might differentially impact LDKT access for specific racial-ethnic groups in the United States is also poorly understood. To address these issues, we examined data on a nationally inclusive population of 208 736 adults who initiated treatment for ESKD during 2005–2008 in order to better understand the relative contributions of individual- and area-level determinants of LDKT across major racial-ethnic groups in the United States. We hypothesized that LDKT rates would be lower among racial-ethnic minorities compared with whites but that determinants of delays in LDKT would vary according to race-ethnicity.

Methods

Data sources

We obtained individual patient-level data from the US Renal Data System (USRDS) registry (3) and the United Network for Organ Sharing (UNOS) Kidney Transplant file, and area-level socioeconomic data from the 2000 US Census at the level of the five-digit zip code tabulation area.

Study sample

We identified all persons aged 18–70 years who initiated treatment for ESKD in the United States between June 30, 2005 and September 23, 2008 (N = 220 811). We selected the initial time period in order to leverage data on predialysis care measures from the Centers for Medicare and Medicaid Services (CMS) 2005 Medical Evidence Form (Form-2728-U3) (3). To reduce the potential for race-ethnicity misclassification bias, we excluded 691 (0.3%) subjects who did not have a race or ethnicity assignment in the USRDS Patients file or Medical Evidence file, or for whom race was reported differently in these two files. Because kidney transplantation is rarely performed in subjects with known malignancies, we excluded 11 342 subjects who were identified as having cancer at the time of ESKD. To further limit our study population to those individuals who might be “medically eligible” for LDKT, we restricted the cohort to patients aged 18–70 who receive the vast majority of LDKT (92% of LDKT have been performed in patients younger than 65) (26). Finally, because area-based proxies of socioeconomic status were important covariates of interest, we further excluded 42 subjects who did not have a residential zip code. The analytic sample consisted of the remaining 208 736 subjects.

Outcome variables

The primary outcome was time from ESKD onset to receipt of a first LDKT. For all analyses, we censored patients at the time of death (N = 37 190), deceased-donor kidney transplantation (N = 8 214) or the end of the study observation period on September 30, 2008. Patients who underwent preemptive LDKT (N = 4 157) were assigned a survival time of 1 day.

Primary explanatory variable

The primary explanatory variable for all analyses was patient race-ethnicity based on information collected at the time of ESKD onset. We defined race-ethnicity as non-Hispanic white, non-Hispanic black, Asian, Hispanic, Pacific Islander or American Indian/Alaska Native (AI/AN) based on the USRDS designation (3). USRDS defines patient race and ethnicity based upon data obtained from the Medical Evidence Form (CMS 2728), the CMS Medicare Enrollment Database and the CMS Renal Management Information System/ Renal Beneficiary and Utilization System (REMIS/REBUS) identification file (3).

Patient-level covariates

We extracted patient-level covariates that might confound or potentially mediate the association of race-ethnicity and time to transplantation based on prior studies (6–24). Sociodemographic covariates included age, sex and health insurance coverage (Medicare, Medicaid, employer group insurance, other insurance or no insurance). We examined the following comorbid conditions from the 2005 Medical Evidence form: cardiovascular disease, chronic lung disease, diabetes, poor functional status (requiring assistance with daily activities, inability to ambulate or transfer or institutionalized at an assisted living or nursing home facility), and active drug, alcohol or tobacco use at the time of initiating treatment for ESKD. We further identified patients who received nephrology care, who had moderate to severe anemia (hemoglobin <10 g/dL), who had hypoalbuminemia (serum albumin concentration <3.5 g/dL) and who were prescribed an erythropoiesis-stimulating agent prior to ESKD because prior studies have reported racial-ethnic differences in these process measures and because these measures are associated with time to transplantation (14). We divided patients into World Health Organization-designated categories of body mass (Quetelet's) index (25). Finally, because regional “competing” rates of deceased donor kidney transplantation might influence a subject's decision to pursue LDKT, we assigned each patient to the Organ Procurement Organization (OPO) region in which she or he resided at ESKD onset (26).

Zip code-level covariates

We included a variable for poverty based on US Census estimates of the percentage of residents living in poverty within the zip code where each patient resided at ESKD onset (<5%, 5–9%, 10–14%, 15–19% or ≥ 20% of the population) (18,19,27). The US Census defines a “poverty area” as an area where at least 20% of residents are poor (28). We further included a variable corresponding to the percentage of linguistically isolated households in each patient's zip code of residence (<1%, 1–4%, 5–9%, 10–19%, or 20% of the population). The US Census defines a linguistically isolated household as one in which all members aged 14 years and older speak a non-English language and also speak English less than “very well” (29).

Statistical analysis

We calculated mean crude rates of LDKT by race-ethnicity. We analyzed the associations of race-ethnicity and time to LDKT from ESKD onset using proportional hazards (“Cox”) regression. The referent group for all analyses was non-Hispanic whites. To a base model comprising fixed, demographic factors (age, race-ethnicity and sex), we incorporated individual-level comorbid conditions, unhealthy behaviors (active tobacco smoking, alcohol dependence and drug abuse), body mass index, direct and indirect measures of access to nephrology (predialysis) care prior to ESKD (anemia, hypoalbuminemia, administration of predialysis erythropoiesis stimulating agent and receipt of predialysis nephrology care) and zip code-level proxies for contextual poverty and household linguistic isolation to examine their effects on the coefficient estimates for race-ethnicity. We used scaled Schoenfeld residual plots against time and estimated log (−log [survivor function]) versus time survival curves to assess the proportionality assumption and found no violations. We found no evidence of collinearity among the vector of explanatory variables using the variance inflation factor and Eigen values. To account for potential correlations due to practice patterns within OPO, we obtained robust sandwich estimates for the Cox model using the OPO as a cluster variable (30,31). We confirmed model fit using Cox-Snell residuals.

To assess the degree of reduced or delayed LDKT attributable to measured demographic, clinical, socioeconomic and linguistic factors as well as regional deceased donor organ availability, we calculated the proportion of the reduced rate of transplantation for each racial-ethnic group (32). We performed bootstrap analyses (with 100 replications) to calculate the bias-corrected confidence intervals of the hazard ratio estimates for LDKT by race-ethnicity. With bootstrapping, the data were randomly resampled to estimate the distribution of the variance (i.e. confidence intervals) in the hazard ratio estimate that occurred with the addition of a variable (or group of variables) to a multivariable model incorporating the covariates indicated above (33).

To evaluate whether differences in LDKT rates were affected by differences in death rates (i.e. informative censoring), we repeated the main analyses using competing risks regression according to the method of Fine and Gray (34). These analyses, in which death was treated as a competing risk and end of study as administrative censoring, were adjusted for the same variables described above. The competing risk model provided an estimate of the rate of LDKT for patients while adjusting for differences in death rates by race-ethnicity. To estimate the influence of differential rates of deceased donor kidney transplantation (DDKT) on racial differences in LDKT, we performed additional analyses examining racial-ethnic differences in parameter estimates for any (i.e. either living or deceased) transplant. Finally, to examine the influence of preemptive transplantation and older age on the parameter estimates, we repeated all analyses in a reduced cohort of patients aged 18–64 years that excluded patients who received preemptive transplants. We considered a two-tailed P-value <0.05 as statistically significant. All statistical analyses were conducted using Stata Statistical Software (Stata MP version 11.0, Stata Corp, College Station, TX).

Results

A total of 208 736 adults aged 18–70 years initiated treatment for ESKD in the United States during the study period. As expected, the distribution of health insurance coverage differed substantially according to race-ethnicity. The prevalence of employer group health coverage was highest among whites and Asians, and the prevalence of Medicaid enrollees and uninsured patients were highest among blacks, Hispanics and AI/ANs. In this cohort, the prevalence of diabetes was particularly high among Hispanics, Pacific Islanders and AI/ANs. Measures of high-quality predialysis care were invariably worse among most racial-ethnic minority groups as compared with whites (Table 1).

Table 1.

Characteristics of all US adult patients aged 18–70 years who initiated treatment for ESKD during June 2005 to September 2008 by race-ethnicity

	Race or ethnicity
Characteristic1	White (N = 94 316)	Black (N = 69 761)	Asian (N = 6 674)	Hispanic (N = 33 456)	Pacific Islander (N = 1 953)	American Indian/Alaska Native (N = 2 576)
Age, mean (SD), years	55.7 (11.3)	52.1 (11.8)	54.0 (12.2)	52.6 (12.4)	53.6 (11.5)	53.1 (11.3)
Age category, %, years
18–29	3	5	5	7	4	4
30–39	7	11	9	9	10	10
40–49	15	21	16	18	18	20
50–59	30	32	29	31	31	33
60–70	45	31	40	35	37	33
Female, %	41	46	42	42	46	48
Health insurance coverage2,%
Employer group	38	28	37	21	30	17
Medicare	43	34	24	30	25	32
Medicaid	21	34	31	38	30	39
No coverage	7	15	10	17	12	8
Other coverage	19	10	17	14	20	31
Clinical measures, %
Diabetes	55	52	53	64	68	77
CVD3	45	40	32	39	41	46
Poor functional status	13	11	6	11	9	11
Tobacco use	10	9	3	3	5	8
BMI ≥ 30 kg/m2	45	45	18	36	43	44
Predialysis access measures, %
Nephrology care	63	53	60	50	56	59
Prescribed ESA	31	24	33	24	30	29
Hemoglobin <10 g/dL	40	52	44	48	46	46
Serum albumin <3.5 g/dL	47	52	48	55	58	67
Zip code-level: percentage of residents living in poverty, %
<5%	19	6	19	5	9	3
5–9%	32	16	31	14	28	12
10–14%	24	18	19	16	18	13
15–19%	13	16	13	17	11	15
≥20%	9	40	14	43	13	53
Zip code-level: percentage of linguistically isolated households, %
<1%	16	8	1	1	1	9
1–4%	50	50	16	8	14	37
5–9%	15	16	17	12	17	19
10–19%	9	13	24	20	20	12
≥20%	6	10	38	54	27	19

¹Due to rounding, percentages may not total 100% for some categories.

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

³Cardiovascular disease.

The study population was distributed across 24 414 residential zip codes in the United States with a high proportion of blacks, Hispanics and AI/ANs residing in areas of high poverty. Additionally, over half of Hispanics, over one-third of Asians and over one-quarter of Pacific Islanders lived in areas where 20% or more households were linguistically isolated compared with 6% of whites and 10% of blacks (Table 1).

Time from ESKD onset to first LDKT

Overall, 10 201 subjects received a first LDKT during 275 325 person-years. Mean rates of LDKT were lowest among blacks, AI/ANs and Pacific Islanders, intermediate among Hispanics and Asians, and highest among whites (Figure 1). In analyses adjusted for individual-level clinical factors and individual- and zip code-level sociodemographic factors, the lower transplant rates among racial-ethnic minority groups compared with whites were attenuated in all groups except among Asians (Tables 2 and 3).

Figure 1. Mean unadjusted rates (per 100 person-years) of living (LDKT) and deceased (DDKT) donor kidney transplant by race-ethnicity.

Error bars represent the 95% confidence interval for the point estimate.

Table 2.

Hazard ratios for race-ethnicity and time from ESKD onset to live-donor kidney transplantation among adults aged 18–70 years with incident ESKD during June 2005 to September 2008

Race or ethnicity	At-risk (N = 208 736)	LDKT events (n = 10 201)	Rate per 100 person-years (95% CI)	LDKT events biological donors, N (%)	Model 1 HR (95% CI)	Model 2 HR (95% CI)
White	94 316	7411	6.46 (6.31–6.61)	3974 (54)	1.00 (Ref.)	1.00 (Ref.)
Black	69 761	1168	1.19 (1.12–1.26)	735 (63)	0.16 (0.15, 0.17)	0.27 (0.24, 0.31)
Asian	6674	360	3.89 (3.51–4.32)	206 (57)	0.56 (0.51, 0.63)	0.48 (0.40, 0.57)
Hispanic	33 456	1179	2.53 (2.39–2.67)	791 (67)	0.33 (0.31, 0.35)	0.58 (0.54, 0.62)
Pacific Islander	1953	32	1.10 (0.78–1.84)	20 (63)	0.19 (0.15, 0.26)	0.23 (0.15, 0.33)
AI/AN	2576	51	1.40 (1.06–1.84)	35 (68)	0.19 (0.14, 0.28)	0.42 (0.29, 0.60)

Model 1: Adjusted for age, sex and race-ethnicity.

Model 2: Adjusted for age, sex, race-ethnicity, health insurance coverage (employer group insurance, Medicare, Medicaid, other insurance or no insurance), clinical factors [cardiovascular disease, chronic lung disease, diabetes, poor functional status (inability to walk, inability to transfer), body mass (Quetelet's) index, drug abuse, alcohol dependence or tobacco use], predialysis access measures [serum albumin concentration, hemoglobin level, nephrology care and prescribed erythropoiesis stimulating agent], zipcode poverty, zipcode linguistic isolation and OPTN/UNOS region.

Table 3.

Hazard ratio estimates for time from ESKD onset to live donor kidney transplantation among adults aged 18–70 years with incident ESKD during 2005–2008

Variable	Hazard ratio	95% CI
Age	0.98	0.97–0.99
Age squared	0.99	0.99–0.99
Sex
Male	Referent	Referent
Female	1.02	0.97–1.07
Health insurance coverage
Employer group	Referent	Referent
Medicare	0.63	0.56–0.69
Medicaid	0.35	0.30–0.40
No coverage	0.25	0.22–0.28
Other coverage	0.86	0.80–0.91
Race-ethnicity
White	Referent	Referent
Black	0.27	0.24–0.31
Asian	0.48	0.40–0.57
Hispanic	0.58	0.54–0.62
Pacific Islander	0.23	0.15–0.33
American Indian/Alaska Native	0.42	0.29–0.60
Clinical measures
Diabetes	0.60	0.55–0.68
CVD	0.47	0.42–0.49
Poor functional status	0.26	0.20–0.33
Chronic lung disease	0.46	0.29–0.50
Tobacco use	0.42	0.36–0.49
Drug use	0.28	0.17–0.33
Alcohol dependence	0.26	0.19–0.37
BMI	1.17	1.13–1.20
BMI squared	1.00	1.00–1.00
Predialysis access measures
Predialysis nephrology care	2.08	1.84–2.38
Albumin	1.65	1.57–1.73
Hemoglobin	1.12	1.11–1.14
Prescribed ESA	0.88	0.85–0.91
Zip-code poverty
<5%	Referent	Referent
5–9%	0.75	0.68–0.82
10–14%	0.58	0.52–0.63
15–19%	0.58	0.52–0.64
≥20%	0.49	0.42–0.58
Household linguistic isolation
<1%	Referent	Referent
1–4%	0.97	0.89–1.06
5–9%	0.93	0.82–1.05
10–19%	0.90	0.73–1.12
≥20%	0.87	0.57–1.32

Estimates are from multiply imputed hierarchical models that are clustered on the Organ Procurement Organization.

Race- or ethnicity-specific determinants of LDKT

Among racial-ethnic minority groups, the degree to which reduced rates of LKDT were attributable to adjustment for measured factors varied significantly by race-ethnicity (Table 3). Among blacks and AI/ANs, the largest proportions of the disparity in LDKT compared with whites were attributed to adjustment for measures of predialysis care, while the largest proportion of the disparity among Hispanics compared with whites was attributed to adjustment for health insurance coverage. Substantial proportions of the disparity in transplant rates among blacks and AI/ANs compared with whites were attributed to adjustment for contextual poverty. Among all groups, there was little evidence that the disparity in LDKT compared with whites was attributable to household linguistic isolation. Further, differences in underlying comorbid conditions including diabetes and unhealthy behaviors accounted for little to none of the reduced rate of LDKT among all minority groups (Table 3).

Differential rates of LDKT among racial-ethnic minority groups did not appear to result from racial-ethnic differences in rates of DDKT (Figure 1) or in competing rates of death (data not shown). In additional analyses that examined time to any transplant, disparities in transplant rates were slightly attenuated among black and Asian patients, and slightly worse among Hispanic, AI/AN and Pacific Islander patients (Table S1). Similar differences in LDKT by race-ethnicity were observed when patients who received a preemptive transplant and those older than 64 years were excluded (Table S2). In this reduced cohort of patients, the largest proportion of the disparity in LDKT among blacks and AI/ANs compared with whites was due to contextual poverty (11% and 17%, respectively), while the largest proportion (17%) among Hispanics compared with whites was attributed to health insurance coverage. As expected, the influence of predialysis care diminished when excluding patients who were preemptively transplanted.

Discussion

In this national study of US adults who initiated treatment for ESKD in recent years, we observed lower relative rates of LDKT among racial-ethnic minority as compared to white patients. In particular, black, AI/AN and Pacific Islander patients experienced age- and sex-adjusted rates of LDKT that were over 80% lower than white patients. Additionally, determinants (i.e. demographic, clinical, socioeconomic and linguistic) of reduced or delayed LDKT differed among racial-ethnic minority patients, as did the relative importance of these determinants in explaining variation in LDKT.

We found that a large proportion of the reduced rate of LDKT among all racial-ethnic minority groups was attributable to differences in predialysis care measures. Suboptimal predialysis care may contribute to poor provider-patient and provider-family communication regarding the full range of treatment options. Limited communication during this critical phase prior to ESKD onset appears to associate with less knowledge about the benefits of LDKT and less willingness to accept nephrologists’ recommendations for transplantation (11,14). Improved predialysis care with timely LDKT communication and decision support may improve patients’ knowledge and consideration of LDKT (10,13).

In addition to predialysis care measures, we found that contextual poverty was an important determinant of differences in LDKT rates for most minority groups, particularly black and AI/AN patients. This observation is consistent with previous findings, which suggest that patients living in neighborhoods with higher degrees of poverty may be less likely to receive transplants than patients living in more affluent areas (19,22). Patients living in areas with a higher degree of poverty may be more likely to encounter geographic and socioeconomic barriers including resource deprivation and financial disincentives to completing the required steps needed to facilitate successful LDKT (7,35). Clustering of risk factors for chronic kidney disease (CKD) and other clinical comorbidities within impoverished areas may also reduce the potential living donor pool for patients residing in poor as compared with more affluent areas (15,18).

A significant proportion of the reduced rate of LDKT among black, Hispanic and AI/AN patients was also attributable to differences in health insurance coverage. While the US ESKD Medicare program provides health insurance coverage for the vast majority of patients after initiation of renal replacement therapy, these benefits do not extend to many patients in earlier stages of CKD (3,36). Reduced health care access due to suboptimal health insurance coverage and poor availability or utilization of routine medical care in earlier stages of CKD may result in delayed nephrology care and fewer preemptive referrals for transplantation. These delays in turn could result in lower rates of transplant evaluations, higher rates of incomplete workups and reduced access to transplantation (23).

In contrast with the other minority groups, we found relatively few determinants of the reduced rate of LDKT among Asian patients. Asians tend to reside in distinct regions of the United States where local community (e.g. access to transplant programs) and cultural factors may limit pursuit of LDKT even in the presence of adequate health insurance coverage and community resources. Somewhat paradoxically, we found little evidence that delayed LDKT among Asians, Pacific Islanders and Hispanics relative to whites was attributable to household linguistic isolation. While patients residing in linguistically isolated households might encounter difficulty understanding and navigating the transplant process, it is plausible that the increased availability of interpreters and educational materials on LDKT in diverse languages has reduced the relative influence of these factors (37). It is also possible that area-level measures of household linguistic isolation are relatively insensitive proxies for patients who encounter linguistic challenges during the LDKT process. Furthermore, the substantial “residual” disparity in LDKT by race-ethnicity after extensive adjustment for measured factors suggests a potential role of cultural factors. Previous work among black and Hispanic patients reported that distinct cultural barriers may limit the ability of medical providers to understand their patients’ values and cultural beliefs, which influence preferences for LDKT (10,38,39). Future efforts are needed to more comprehensively address these barriers and to identify impediments to LDKT among less well-studied groups, such as Asians and Pacific Islanders.

While prior research has reported that blacks and “other” racial-ethnic minorities are less likely to undergo LDKT (6,8), our study uniquely and comprehensively examines the relative influence of race-ethnicity-specific, patient-, and area-level factors in explaining variation in LDKT among all major US racial-ethnic groups of patients initiating ESKD treatment. Previously, Weng et al. observed that black kidney transplant candidates at a single transplant center were less likely to receive LDKTs due to lower rates of donor recruitment and conversion (8). Additionally, Gore et al. found that older age, less education, zip code-level income and Medicare insurance coverage were associated with overall lower odds of receiving LDKT among US adults undergoing primary kidney transplantation during 2004– 2006 (6). Our study findings extend those of prior work by identifying race-ethnicity-specific factors that significantly contributed to delays in LDKT among all major US racial-ethnic groups (including less well-studied groups, such as Asians, Pacific Islanders and AI/ANs) and by quantifying the contribution of these factors to delayed LDKT.

Our findings may guide the development of interventions designed to more specifically address racial-ethnic disparities in LDKT. Based on our results, efforts to expand health insurance coverage in earlier stages of CKD, to improve access to high-quality predialysis care, and to overcome socioeconomic barriers to LDKT are particularly important to address racial-ethnic disparities in LDKT. For example, area-level studies that examine the distribution and quality of care within CKD clinics, particularly those that serve racial-ethnic minority communities, may provide valuable insight on how (and whether) expansion of health insurance coverage might change access to high-quality predialysis care for racial-ethnic minority groups. Similarly, studies that incorporate social workers and financial counselors into LDKT educational efforts have reported early success in reducing patients’ financial concerns regarding the LDKT process (40). However, the impact of such interventions on subsequent progress through the LDKT process has yet to be established. Lastly, the large residual disparity after extensive covariate adjustment warrants continued development of culturally-sensitive, literacy-appropriate LDKT education interventions and formal testing of their effects on patient-centered decision-making. Such decision-making would ideally incorporate patient and family engagement in ESKD treatment options while respecting cultural factors, such as cultural norms, preferences and values (41,42).

Our study has a number of strengths including analysis of a well-characterized national population of adult patients who initiated ESKD treatment and comprehensive follow-up for kidney transplantation and death. We also included comparative data for racial-ethnic groups that have been historically understudied with respect to LDKT (i.e. Asians, Hispanics, Pacific Islanders, AI/ANs). Our study also has several potential limitations. First, we based our assessment of race-ethnicity on provider-reported data obtained from the CMS medical evidence records, which has the potential for misclassification. However, the direction and magnitude of our results align with those from other studies in kidney transplantation, some of which were based on self-reported ethnicity, and thus we suspect that the impact of this bias, if present, was likely small. Second, measures of poverty and linguistic isolation were only available at the zip-code level. Therefore, we were unable to control for individual-level income and linguistic isolation, which may differ from zip-code level measures. While contextual poverty may not comprehensively reflect an individual patient's financial resources for undertaking the LDKT process, it serves as an important community-level proxy for geographic and socioeconomic barriers to care (16). Third, our results are potentially limited by under-reporting of several measures, such as patient comorbid medical conditions and receipt of predialysis nephrology care (43). However, given the magnitude of racial-ethnic differences in relative LDKT attainment, it seems unlikely that observed differences in LDKT were caused by residual confounding from differential ascertainment of comorbidities or predialysis nephrology care alone. Fourth, the shortened follow-up period for patients who initiated ESKD treatment in 2008 may have precluded identification of a living donor. Although we observed similar estimates when restricting the analyses to patients who initiated treatment during 2005–2006, extending the follow-up period for these patients may have allowed us to further refine our estimates. Finally, we were unable to account for other individual- or center-level characteristics (e.g. cultural attitudes, disparities in provision of transplant education) that might contribute to disparities in LDKT (44), as many patients had yet to list or select a specific transplant center. A recent report observed that black patients were less likely to receive LDKT at every transplant center in the United States, but that this disparity was attenuated in centers with higher rates of kidney transplants (45).

Despite two decades of recognition, we found that relative rates of LDKT remain substantially lower among racial-ethnic minority patients as compared with white counterparts. However, determinants of delayed LDKT differed according to race-ethnicity. Black, Hispanic and AI/AN patients face persistent difficulty in accessing LDKT due to socioeconomic factors and limited health insurance coverage. In contrast with other minority groups, few measured determinants including household linguistic isolation appeared to explain low rates of LDKT among Asian patients. Differences in measures of predialysis care contributed substantially to delayed LDKT among all racial–ethnic minority groups. Geographically targeted efforts to address these important determinants of delayed LDKT may help to reduce pervasive racial-ethnic differences in kidney transplantation in the United States.

Table 4.

Percentage (95% CI) of race-ethnicity effect attributed to adjustment for patient and area-level determinants of live-donor kidney transplantation among adults aged 18–70 years who initiated treatment for ESKD during 2005–2008

	Race-ethnicity
	White (N = 94 316)	Black (N = 69 761)	Asian (N = 6674)	Hispanic (N = 33 456)	Pacific Islander (N = 1953)	American Indian/Alaska Native (N = 2576)
Race-ethnicity effect attributed to adjustment for Zip code/contextual poverty
Percent (95% CI)	—	15.7 (13.8, 17.5)	–7.0 (–8.2, –5.8)	3.7 (2.6, 4.7)	6.2 (4.1, 8.4)	18.1 (17.2, 19.1)
Health insurance coverage
Percent (95% CI)	—	10.2 (9.1, 11.6)	1.3 (0.4, 3.5)	17.1 (15.7, 19.9)	–0.2 (–4.0, 2.9)	8.3 (3.5, 14.5)
Predialysis care measures (receipt of nephrology care and/or ESA, hypoalbuminemia, anemia)
Percent (95% CI)	—	20.4 (19.5, 22.9)	7.6 (4.5, 9.7)	14.4 (12.2, 16.4)	12.2 (5.3, 18.6)	28.7 (21.1, 35.2)
Differences in body size and diabetes status
Percent (95% CI)	—	0.3 (–0.3, 1.0)	–9.3 (-–0.7, –8.3)	–1.9 (–3.2, –1.1)	–2.1 (–4.4, 1.6)	1.4 (–0.7, 2.4)
Household linguistic isolation
Percent (95% CI)	—	–1.0 (–1.8, 0.1)	–5.5 (–7.7, –2.4)	3.6 (0.8, 6.5)	1.9 (–1.8, 6.1)	2.6 (0.7, 3.7)
Underlying comorbid conditions (cardiovascular disease, chronic lung disease, functional status)
Percent (95% CI)	—	–2.2 (–2.8, –1.6)	–4.5 (–5.4, –3.5)	–3.4 (–4.1, –2.7)	–1.6 (–3.3, 1.1)	–3.4 (–5.1, –1.3)
Unhealthy behaviors (alcohol dependence, tobacco or drug use)
Percent (95% CI)	—	–1.4 (–1.8, –0.9)	–3.8 (–4.4, –3.1)	–3.7 (–4.2, –3.2)	–2.9 (–3.9, –1.4)	0.7 (–0.4, 2.5)

Negative values indicate that, after adjustment for the specific factors, differences in time to LDKT compared with whites were larger than observed in the age-, sex- and race-adjusted models.

Abbreviations

AI/AN

American Indian/Alaska Native

CKD

chronic kidney disease

ESA

erythropoieis-stimulating agent

ESKD

end-stage kidney disease

LDKT

live donor kidney transplantation

OPO

Organ Procurement Organization

UNOS

United Network for Organ Sharing

US

United States

USRDS

United States Renal Data System