Minimizing racial disparity regarding receipt of a cadaver kidney transplant

Minimizing Racial Disparity Regarding Receipt of a Cadaver Kidney Transplant Ronald J. Ozminkowski,

PhD, Alan J. White, PhD, Andrea Hassol, MSPH, and Michael Murphy, BS

0 This report describes the impact of race on waiting list entry and receipt of a cadaver kidney transplant, after accounting for self-reported income, health and functional status, and patients’ attitudes about dialysis and transplantation as treatment alternatives. Previous studies did not account for these race-related factors and therefore produced biased estimates of the impact of race on waiting list entry and receipt of a transplant. Data for this investigation came from a telephone survey of a national sample of 456 end-stage renal disease patients and from files maintained by the United Network for Organ Sharing and the Health Care Financing Administration. Proportional hazard models were estimated with these data. The results indicated that approximately 60% of the differences between black and white waiting list entry rates and 52% of the black-white differences in transplantation rates were due to race-related differences in socioeconomic status, health and functional status, severity of illness, biological factors, the existence of contraindications to transplantation, transplant center characteristics, and patients’ attitudes about dialysis and transplantation. Potential ways to narrow racial differences further include better education about treatment alternatives for black patients, more vigorous efforts to obtain donor organs from minorities, continued research and thoughtful policy on the access-related impacts of United Network for Organ Sharing point system variances, and consolidation of some smaller waiting lists into larger regional lists. 0 1997 by the National Kidney Foundation, Inc. INDEX

WORDS:

ESRD;

dialysis;

kidney

transplantation;

access

I

N THE LAST few years, several studies have shown that kidney transplants were less likely to be provided to nonwhite patients.lm7 In addition, Kallich et al6 and Eggers7 found lower rates of waiting list entry among nonwhite patients. The magnitude of the racial discrepancies noted in these studies may have been misstated, however, for two reasons. First, with some exceptions, previous studies excluded information on suitable transplant candidates who never registered on established waiting lists.6,7 It is possible that many suitable candidates were “sequestered” in dialysis units, unable to obtain access to kidney transplant waiting lists.‘,’ Second, none of the previous studies accounted directly for patients’ own assessments of their health status, their incomes, or their attitudes about dialysis versus transplantation as treatment options. Since these factors may be correlated with race, failure to account for them may have resulted in biased estimates of the effects of race on access to transplantation. This report describes how black and white patients vary in terms of the time spent between the receipt of their first end-stage renal disease (ESRD) service and entry onto a waiting list for a cadaver kidney transplant. We also describe how much of the observed black-white differential can be accounted for by differences in socioeconomic status, attitudes about dialysis and American

JournalofKidney

Diseases,

Vol30,

No 6 (December),

to treatment.

transplantation, and self-reported health and functional status, variables that have not been taken into account in previous studies. We next describe how much of the observed black-white differential in the time between waiting list entry and transplantation is due to these factors. As in other studies, the race comparisons made are between non-Hispanic black and white patients who participated in the Medicare ESRD Program.6X7The Medicare ESRD Program covers approximately 93% of the ESRD population in the United States.” DATA SOURCES Data files from three sources were merged and used for this research. These included (1) a telephone survey of 456 ESRD patients from the continental United States whose first ESRD ser-

From The MEDSTAT Group, Inc, Ann Arbor, MI; and Abt Health Care Research Foundation, Cambridge, MA. Received February 17, 1997; accepted in revised form September 9, 1997. Supported by Grant No. HSO-7.538-01 from the Agency for Health Care Policy und Research. Address reprint requests to Ronald J. Ozminkowski, PhD, The MEDSTAT Group, Inc, 777 E Eisenhower Pkwy, Suite 500, Ann Arbor, MI 48108. E-mail: ran-ozminkowski@ medstat.com 0 1997 by the National Kidney Foundation, Inc. 0272.6386/97/3006-0004$3.00/O 1997:

pp 749-759

749

OZMINKOWSKI

750

vice was in the first calendar quarter of 1992; (2) waiting list and other files from the United Network for Organ Sharing (UNOS) from 1992 until October 3 1, 1994 (the end of the study period), and (3) the Medicare ESRD Program Management and Medical Information System data files for the study period. The telephone survey was conducted by Abt Associates Inc, one of the largest survey research firms in the country. The survey excluded patients younger than 18 years and older than 69 years because very few Medicare ESRD patients could be found who were younger than 18 years and transplantation is rare in patients older than 69 years. The survey also over-sampled blacks and those aged 60 to 69 years to allow sufficient power for statistical analyses. The survey achieved a 61.7% response rate; details of the survey sampling and other processes are available on request. The telephone survey provided information on demographics, socioeconomic status, and atti-’ tudes about dialysis and transplantation. The survey also included the SF-36 general health and functional status instrument. The SF-36 instrument has been used reliably with ESRD patient samples in other studies.‘1”2 The SF-36 was used here to create measures of health and functional status for the statistical analyses. The UNOS files provided information on biological factors, access to waiting lists, and transplant center productivity. The Program Management and Medical Information System files provided data that allowed correlates of ESRD severity and general suitability for transplantation to be included in the statistical analyses. MATERIALS

AND

METHODS

The factors that might influence waiting list entry and cadaver kidney transplantation may be categorized broadly as (1) demographic; (2) socioeconomic; (3) related to severity of illness, general health status, and biological and medical suitability for transplantation; (4) characteristics of the patients’ likely transplant centers; and (5) patients’ attitudes about dialysis versus transplantation.z.5 Since these factors also may be correlated with race, the specific aim of the research conducted here was to determine how adjusting for the influence of these factors would influence the effect of race on waiting list entry and on the odds of receiving a cadaver kidney transplant. To address this aim, five proportional hazard models were used to investigate the impact of race on waiting list entry, and five hazard models investigated the impact of race on

ET AL

receipt of a cadaver kidney transplant. The first five hazard models addressed the amount of time between the patient’s first ESRD service and waiting list entry; these models were conducted using the entire sample (n = 456). The second set of hazard models addressed the amount of time between wait listing and receipt of a cadaver kidney transplant; these models were estimated using the subset of observations who appeared on at least one waiting list for a cadaver kidney transplant (n = 270). The proportional hazard model approach was useful for addressing race-related differences in waiting list entry and transplantation because the hazard models adjusted for censoring events, such as death and other factors that may cause observations to be lost before waiting list entry or transplantation occur.

Model SpeciJications and the Impact of Race The independent variables included in the first hazard model (model 1) in each set were used to measure patient demographics (ie, age, gender, and race). Model 2 in each set augmented the hazard specification to include information on other socioeconomic factors that have not been measured in other studies (ie, self-reported income and education levels). Model 3 added information about self-reported health and functional status, biological factors (when available), the existence of contraindications to transplantation, and severity of ESRD. Model 4 added information about the productivity of the nearest transplant center and the distance from that center to the patient’s home. Finally, model 5 (referred to later as the “full transplant model”) added information concerning patients’ attitudes about dialysis and transplantation as treatment alternatives. The results obtained from each hazard model showed how black race was related to the odds of waiting list entry or to the odds of receiving a transplant, controlling for the other variables in the model. If black and white patients were equally likely to be included on a waiting list or to receive a transplant, the odds ratio for black versus white patients in each model would be very close to 1.0, with any differences from 1.0 due to random variation. This was not the case, however, so we paid particular attention to the change in the odds ratios for black versus white patients as additional categories of variables were added in the five hazard models. To do this, we first noted the odds ratio of being included on a waiting list or being transplanted for black patients vis a vis white patients in model 1, which included only the demographic factors. We then subtracted that odds ratio from the odds ratio for black patients that was obtained in model 5, the full model. The result was then divided by the difference between 1 .O and the odds ratio for black patients in model 1. This ratio estimated the percentage of the difference between black and white patients in model 1 that could be accounted for subsequently by adjusting for their different socioeconomic factors, severity of illness, medical and biological suitability for transplantation, transplant center-related factors, and attitudes about dialysis and transplant centers as treatment alternatives.

Independent

Variables

The list of independent variables used in each hazard model are noted in Table 1. Most of the variables defined are self-

RACE

AND

KIDNEY

751

TRANSPLANTATION

explanatory; however, a few are not. For example, the health and functional status indicators included in model 3 included two indicators to differentiate between patients who had, respectively, either no marked limitations in daily activities or five or more daily activity limitations. Such patients were compared in the hazard analyses with those with a moderate number of activity limitations. Model 3 also included an indicator to differentiate between patients who expected their health status to decline significantly in the coming year versus those who did not. The ESRD severity indicators used in model 3 were similar to those developed by Diamond et all3 and validated by Held et a1.14 These indicators included variables to differentiate between those who had high-risk ESRD (ie, as developed from diabetes), medium-risk ESRD (as developed from hypertension), or low-risk ESRD with one or more high-risk complicating conditions. Those patients were compared in the hazard models to those with low-risk ESRD and no complications. In addition, the severity proxies included an indicator of dialysis type (ie, hemodialysis v peritoneal dialysis), because early research noted health status differences that were associated with dialysis type.” Next, based on work conducted by Gaylin et al: model 3 included an indicator to differentiate between other patients and those with any of the following contraindications to kidney transplantation: heart or peripheral vascular diseases, transient ischemic attack, hepatitis, obesity, cirrhosis of the liver, pulmonary edema, chronic obstructive pulmonary disease, or neoplasms. These problems were identified by ICD9-CM diagnosis codes included in the Medicare Program Management and Medical Information System files. Finally, in the analyses that pertained to the amount of time between being included on a waiting list and receipt of a transplant, model 3 included several biological indicators that influenced the likelihood of finding a suitable donor organ. The first of these was an indicator to denote whether patients had blood type B, which is more prevalent nationally among transplant candidates than among kidney donors.‘6 Model 3 also included three indicators to denote whether the patient had HLA-A, -B, or -DR antigens that were more prevalent among transplant candidates than among kidney donors. These antigens were identified previously by Lefell et al.” Model 3 also included an indicator to denote whether patients had maximum panel-reactive antibody values above 80%. All other factors being equal, patients with blood type B, panel-reactive antibody values greater than 80%, and HLA antigens that were more prevalent in transplant candidates than among kidney donors should be expected to wait longer for a transplant. Two of the transplant center characteristics that were added in model 4 addressed the productivity of the patients’ nearest transplant center. These variables were based on the ratio of the number of transplants performed to the size of the waiting list at that center. Indicators were included in the hazard model to differentiate between patients whose nearest transplant center was in the upper or lower quartile of this ratio versus those whose transplant centers were moderately productive. Model 4 also included a proxy measure of the distance from the patients’ home to that center, because distance is often considered in point systems used to allocate cadaver

kidneys and because other research noted that distance may influence receipt of a kidney transplant.’ The distance measure was based on the distance between the centroid of the patient’s zip code and the centroid cf the zip code of the transplant center. An indicator was then added to the hazard model to denote patients with distances of at least 50 miles. The variables used to adjust for attitudes about dialysis and transplantation in model 5 are described in Table 1. The survey statements used to develop these variables are presented in Table 2.

Estimation Techniques All the hazard models were estimated with the SUDAAN statistical program.r8 Analyses were weighted to adjust for the oversampling of black and elderly patients used in the survey process and for nonresponse to the survey. The SUDAAN program produced standard errors that accounted for the use of the survey weights. Finally, several z-tests for differences in proportions were conducted to show correlations between race and other factors, thereby facilitating the interpretation of the results obtained from the hazard analyses.

RESULTS

Table 3 presents unadjusted descriptive statistics that compare cumulative waiting list entry and transplantation rates for blacks and whites in the patient sample. Consistent with previous studies, black patients in this study were entered onto waiting lists at a significantly slower rate than white patients. Within 1 year of first ESRD treatment, 13.5% of black respondents were included on a waiting list, which is less than half the rate for white patients. Approximately 25.2% of black patients were added to a waiting list within 2 years of first listing, compared with 41.9% of white patients. In addition, white patients were approximately three times as likely as black patients to receive a transplant within 1 year of waiting list entry (45.8% v 14.4%). Roughly the same difference was found within 2 years of waiting list entry (58.9% for white patients v 20.5% for black patients). These differences in waiting list entry and transplantation may reflect, in part, the impact of other factors that were correlated with race. Table 1 shows that nonwhite and white patients differed with regard to income, dialysis type, severity of illness, the characteristics of their likely transplant centers, and their attitudes about dialysis and transplantation as treatment alternatives. Impact of Race on Waiting List Entry Table 4 shows how adjusting for the factors included in the five hazard models influenced the

752 Table

OZMINKOWSKI 1. Proportion

of Black

and

Non-Black

Patients

With

Entire Sample

Variables Added in model 1 Patient was male Age 18-34 Age 35-39 Age 40-44 Age 45-49 Age 50-54 Age 55-59 Age 60-64 Race is Hispanic Added in model 2 Had at least high school education Had household income < $10,000 Had household income > $40,000 Income information was missing Added in model 3 The patient expected health status to decline in coming year The patient had any contraindication to transplantation The patient used hemodialysis The patient had marked limitations in five or more daily activities The patient had no marked limitations in daily activities The patient had high-risk ESRD (ie, from diabetes) The patient had intermediate-risk ESRD (ie, from hypertension) The patient had low-risk ESRD with high-risk complications The patient had type f3 blood The patient’s peak PRA value > 80 The patient had HLA-A antigens that were more prevalent in transplant candidates than in kidney donors The patient had HLA-6 antigens that were more prevalent in transplant candidates than in kidney donors The patient had HLA-DR antigens that were more prevalent in transplant candidates than in kidney donors Added in model 4 Distance to nearest transplant center was at least 50 miles Ratio of transplants performed to waiting list size at nearest transplant center was in top quartile

Characteristics

Measured

(n = 456)

in the

Five

Time to Transplant

ET AL

Hazard

Models

Sample

(n = 270)

Nonblack Proportion

Black Proportion

P Value

Nonblack Proportion

Black Proportion

0.569 0.101 0.129 0.140 0.112 0.060 0.102 0.146 0.059

0.487 0.120 0.099 0.090 0.100 0.119 0.127 0.135 0.000

0.080 0.523 0.332 0.104 0.691 0.024 0.407 0.737 Not tested

0.594 0.169 0.174 0.117 0.114 0.113 0.116 0.097 0.082

0.583 0.186 0.136 0.123 0.141 0.126 0.109 0.100 0.000

0.853 0.711 0.399 0.880 0.513 0.758 0.850 0.944 Not tested

0.804 0.265 0.208 0.112

0.590 0.579 0.055 0.090

0.000 0.000 0.000 0.442

0.849 0.214 0.274 0.079

0.740 0.548 0.078 0.033

0.025 0.000 0.000 0.118

0.504

0.480

0.602

0.617

0.518

0.104

0.564 0.561

0.589 0.807

0.587 0.000

0.511 0.483

0.486 0.743

0.682 0.000

0.276

0.286

0.818

0.181

0.179

0.968

0.305

0.309

0.930

0.415

0.366

0.417

0.281

0.256

0.547

0.328

0.187

0.010

0.155

0.323

0.000

0.122

0.302

0.000

0.347 NA NA

0.225 NA NA

0.005 NA NA

0.341 0.128 0.035

0.271 0.207 0.053

0.220 0.111 0.526

NA

NA

NA

0.399

0.843

0.000

NA

NA

NA

0.584

0.905

0.000

NA

NA

NA

0.888

0.924

0.381

0.389

0.314

0.096

0.365

0.273

0.111

0.253

0.051

0.191

0.246

0.279

0.177 (Continued

-

on following

Z-Test

page)

Z-Test

P Value

RACE

AND

KIDNEY

Table

1. Proportion

TRANSPLANTATION of Black

and

753 Non-Black

Patients

With (Cont’d)

Entire Sample

Variables Ratio of transplants performed to waiting list size at nearest transplant center was in bottom quartile Added in model 5 (full model) The patient strongly agreed with all three of the financial attitude statements in Table 2 The patient strongly disagreed with all three of the financial attitude statements in Table 2 The patient strongly agreed with both of the medical: transplant attitude statements in Table 2 The patient strongly disagreed with both of the medical: transplant statements in Table 2 The patient strongly agreed with one or more of the personal attitude statements in Table 2 The patient strongly disagreed with one or more of the personal attitude statements in Table 2 The patient strongly agreed that transplant patients are healthier than dialysis patients The patient strongly agreed with both of the dialysis time statements in Table 2 The patient strongly disagreed with both of the dialysis time statements in Table 2 The patient strongly agreed with two or more of the medical: dialysis statements in Table 2 The patient strongly disagreed with two or more of the medical: dialysis statements in Table 2 The patient strongly agreed with the daily activity attitude statements in Table 2 The patient strongly disagreed with the daily activity statements in Table 2 Abbreviations:

PRA,

panel-reactive

antibody;

Nonblack Proportion

Black Proportion

0.276

0.185

0.169

Characteristics

(n = 456)

Measured

in the

Five

Time to Transplant

Models

Sample

(n = 270)

Nonblack Proportion

Black Proportion

0.024

0.287

0.177

0.037

0.140

0.408

0.254

0.163

0.074

0.149

0.153

0.908

0.139

0.190

0.260

0.242

0.271

0.484

0.204

0.317

0.035

0.157

0.148

0.805

0.224

0.122

0.033

0.119

0.272

0.000

0.092

0.187

0.023

0.527

0.266

0.000

0.523

0.370

0.012

0.918

0.817

0.001

0.949

0.884

0.048

0.192

0.171

0.566

0.173

0.140

0.467

0.283

0.276

0.862

0.244

0.280

0.504

0.174

0.161

0.733

0.232

0.154

0.116

0.421

0.452

0.516

0.429

0.494

0.289

0.467

0.365

0.030

0.518

0.423

0.123

0.100

0.144

0.149

0.081

0.075

0.853

NA, data

estimated impact of black race on waiting list entry and transplantation. A complete list of hazard model results is available on request. The first hazard model included as independent variables only indicators for age category,

Z-Test

Hazard

P Value

not available

for entire

Z-Test -

P Value

sample.

gender, and race. This model implied that the relative waiting list entry rate for black patients was only 58% as high as that for white patients (odds ratio, 0.578; P < 0.01). Model 2 added measures of socioeconomic

754

OZMINKOWSKI Table

2. Attitudes Toward and Transplantation Description

Category Financial

Medical: transplant

Personal

Health Daily

activities

Dialysis

Medical: dialysis

time

Dialysis

of Questions Category

Included

in

1. Paying for the cyclosporine or other drugs needed after a transplant can be very difficult. 2. You are concerned about maintaining health insurance after a transplant. 3. You are concerned about losing Social Security or SSI payments after a transplant. 1. The risk of rejecting a transplanted kidney is very high. 2. The side effects from transplant drugs can be very serious. 1. The thought of having a dead person’s kidney inside your body seems very strange. 2. You have religious objections to having a transplant from a dead person. 1. In general, transplant patients are healthier than dialysis patients. 1. It’s easier to go to work or school with a transplant than it is when on dialysis. 2. One must live with many restrictions when on dialysis. 1. It is difficult to manage your personal schedule around the dialysis sessions. 2. It is inconvenient to spend so much time at the dialysis clinic or at home doing dialysis. 1. You have had problems with infections at the dialysis access site. 2. Being on dialysis causes other medical problems. dialysis only: 3. For those on peritoneal You get tired of doing the CAPD exchanges every 4 hours. 4. For those on hemodialysis only: You’ve had problems with vascular access.

NOTE. Respondents were asked whether they agree/ disagree with the statements, using a five-point scale, ranging from 1 = strongly disagree to 5 = strongly agree. Data from telephone survey of 456 ESRD patients conducted from April to June 1995.

status (based on income and educational levels) to the hazard model. The results showed that those with incomes of $40,000 or more were approximately twice as likely to enter waiting

ET AL

lists within 2 years of their first ESRD service as respondents reporting annual incomes less than $10,000. Z-tests conducted separately showed that nonwhite patients were approximately 2.2 times as likely as white patients to have incomes less than $10,000, and that white patients were approximately 3.8 times as likely as nonwhite patients to have an annual income of at least $40,000 (Table 1; z-test, P < 0.05 in both cases). Whereas black patients were still significantly less likely than white patients to enter waiting lists when the socioeconomic measures were included in the hazard model (Table 3; P < 0.05), adding those measures in model 2 increased the relative odds of waiting list entry for black patients from 0.578 to 0.716 times the rate of white patients. Thus, almost one third of the blackwhite differential found in model 1 can be attributed to differences in socioeconomic factors between the races. Model 3 added measures to control for dialysis type, ESRD severity, contraindications to transplantation, and self-reported health and functional status to the hazard model. None of these measures were significantly associated with waiting list entry, and adding these variables to the hazard model had virtually no impact on the estimated race effect. In model 4, measures related to transplant center characteristics were added to account for location-related differences in the supply of donor organs and for geographic proximity to transplant care. Neither the productivity of the nearest transplant center nor its estimated distance from the patient’s home were significantly associated with waiting list entry. Thus, including transplant center characteristics in the model had a negligible effect on the estimates of the effects of race in the waiting list entry models. Adjusting for the influence of the other factors included in model 4, black patients had a rate of waiting list entry that was approximately 0.707 times that of whites and the black-white difference continued to be statistically significant (Table 3; P < 0.05). The full model (model 5) added information on attitudes about dialysis and transplantation as treatment alternatives. The details of the relationships between these variables and waiting list entry are available on request. Bivariate analyses provided evidence of significant racial differences in some of the attitude measures (see Table

RACE

AND

KIDNEY

755

TRANSPLANTATION Table

3. Cumulative

Waiting

List

Entry

and

Transplantation

Rates Percent

Percent Characteristic

Entering

a Waiting

List (SE)

2 yr

1 yr

Total sample Whites Blacks

22.5% 30.2% 13.5%

34.4% 41.9% 25.2%

(0.014) (0.023) (0.019)

NOTE. Differences between black and white patients (P < 0.05). * Based on the entire sample (n = 456). t Based on a subsample of the patients on a waiting Data from UNOS waiting list files.

statistically

4. Odds

Ratios

for Five Hazard Black Patients

Models Entering

Baseline* Model 1: demographics only (age, gender, race) Model 2: add socioeconomic status indicators (income, education) Model 3: add indicators of health and functional status, dialysis type, ESRD severity, transplant suitability, and biological factors Model 4: add transplant center characteristics (ratio of transplantations performed to waiting list size, distance to nearest transplant center) Model 5: add indicators of attitudes toward dialysis and transplantation *The baseline waiting T Black race coefficient $ Black race coefficient

1 .oo

43.1% 58.9% 20.5%

in chi-squared

and White Patients, List and Receiving

tests

and the Cumulative a Transplant

List Entry

Odds Ratio for Black Y White Patients (95% Confidence Interval)

Model

significant

(0.083) (0.115) (0.090)

of independence

Receipt

Estimated Probability of Entering a Waiting List Within 2 yr of First ESRD Service

of a Cadaver

Odds Ratio for Black v White Patients (95% Confidence Interval)

0.344

1 .oo

Probability

Kidney

Estimated Probability of Transplantation Within 2 yr of Entering the Waiting List

0.616

(0.450-0.742)

0.216

0.405t

(0.310-0.530)

0.320

0.716$

(0.530-0.966)

0.261

0.4327

(0.321-0.581)

0.338

0.717*

(0.520-0.990)

0.261

0.670$

(0.450-0.995)

0.472

0.707$

(0.507-0.985)

0.258

0.676$

(0.458-0.996)

0.476

0.295

0.716

(0,580-l

,183)

list entry and transplantation was statistically significant was statistically significant

rates represent the mean at the 1% level. at the 5% level.

(0,462-l

values

,106)

observed

of

Transplant

0.578-t

0.828

(0.087) (0.114) (0.110)

were approximately 2.5 times as likely as white patients to express both religious objections to transplantation and uneasiness about having a dead person’s organ inside them (z-test, P < 0.05). Given the relationships between these and other attitude measures and race, adding the atti-

for Black a Waiting

Waiting

32.9% 45.8% 14.4%

list (n = 270).

l), and these differences explained a portion of the observed difference between the two groups in waiting list entry. For example, approximately 92% of white patients believed that transplant recipients were healthier than dialysis patients, compared with only 82% of nonwhite patients in our sample (z-test, P < 0.05). Nonwhite patients Table

2 yr

1 yr

(0.011) (0.020) (0.017)

were

Transplanted After Entering a Waiting List (SE)t

0.495 in the data.

756

tude measures to the hazard model reduced the black-white differential in waiting list entry. Thus, in the full model, black patients were approximately 0.828 times as likely as white patients to enter a waiting list. This black-white difference was no longer statistically significant (P > 0.10). One should keep in mind, however, that the statistical power to find a significant difference this small between black and white patients was rather low (power = 0.68, alpha = 0.10). Finally, based on the difference in the odds ratios for blacks versus whites in models 1 and 5, we calculated that nearly 60% of the blackwhite differential in waiting list entry that was found in model 1 could be attributed to differences in other factors that were correlated with race. Impact of Race on Transplantation As with the waiting list models, race had a large and statistically significant impact on transplantation rates in four of the five hazard models. The results from model 1, which controlled only for age group, gender, and race, showed that black patients on waiting lists were only approximately 41% as likely as white patients on waiting lists to obtain a kidney transplant (Table 4; odds ratio, 0.405). When income and education were added to the hazard model in model 2, the results showed that income was strongly associated with transplantation. The odds of transplantation were approximately 1.68 times as high for those with annual incomes greater than $40,000 than for those with incomes less than $10,000. However, including measures of socioeconomic status in the hazard model raised the odds that black patients would obtain a kidney transplant only slightly; model 2 implied that the transplantation rate for black patients was 0.432 times that of white patients. As in the waiting list models, the addition in model 3 of measures related to self-reported health and functional status, ESRD severity, contraindications to transplantation, and dialysis type had little impact on the estimated impact of race on transplantation. However, adding information on blood type, panel-reactive antibody values, and HLA matching did have an impact. After adding these variables to the hazard model, the odds ratio for black patients increased to

OZMINKOWSKI

ET AL

0.670, meaning that black patients were approximately two thirds as likely as white patients to obtain a transplant in that model. As in the waiting list models, the addition of the transplant center characteristics in model 4 offered little help in explaining the remaining black-white differential. In model 4, black patients were still only approximately two thirds as likely as white patients to obtain a transplant (Table 4; odds ratio, 0.676, P < 0.05). Finally, as in the waiting list model, the impact of race was no longer statistically significant in model 5, which added the attitude measures to the hazard model. In that model the relative transplantation rate for black patients was 0.716 times that of otherwise identical white respondents. Table 1 suggests that accounting for attitudes was useful because black and white patients differed significantly in terms of their thoughts about the medical consequences of transplantation, the relative health of transplant versus dialysis patients, and personal attitudes about having a dead person’s organ inside them and religious objections to transplantation. Unlike the waiting list models, the lack of statistical significance of the black race variable in model 5 was not due to low power (power = 0.91, alpha = 0.10). Overall, the addition of the measures included in models 2 to 5 accounted for 52% of the blackwhite differential observed in model 1. DISCUSSION

Earlier studies found lower rates of entry onto transplant waiting lists and lower transplantation rates among nonwhite patients. 1-3Z6,7 The reasons cited most often for these findings are as follows. First, race-related differences in HLA, blood type, and organ donation rates exist, making it relatively more difficult to find suitable matching organs for nonwhite patients.‘6,‘9-23 Second, during 15 consecutive years before UNOS began managing the organ allocation process, white kidney recipients had a higher graft survival rate than black recipients.24 Third, nonwhite patients tended to survive longer on dialysis than white patients. 14,25Finally, one study found that compliance with the posttransplant treatment regimen was significantly lower among nonwhite patients, so some might argue that nonwhite patients were less suitable candidates for transplantation in that regard.26 It is interesting, however, that the rela-

RACE

AND

KIDNEY

TRANSPLANTATION

tionship between race and compliance rates disappeared in that study when ability to pay was controlled, suggesting that compliance was related to income, not race. Nevertheless, transplant providers may still be responding to these factors by limiting access among non-white patients. Thus, it was not surprising to see lower rates of waiting list entry and transplantation among non-white ESRD patients. The access literature has progressed over time to include better controls for severity of illness, contraindications to the procedure, and correlates of socioeconomic status. These studies continued to find race-related access differences. This study was the first to use additional information on self-reported health and functional status and patients’ attitudes about dialysis and transplantation as treatment alternatives. Attitudes may be particularly important; an unpublished, internal study conducted by UNOS showed that organ refusal rates may be higher among black patients.27 Table 1 showed that nonwhite and white patients often differed with regard to their attitudes about dialysis and transplantation. These differences, plus those related to socioeconomic factors, severity of illness, contraindications to transplantation, and the other factors we measured, accounted for approximately 60% of the observed black-white difference in waiting list entry. These factors also accounted for approximately 52% of the racial difference in the odds of receiving a transplant. The fact that we found no significant differences between black and white patients in waiting list entry and in transplantation rates in our full hazard models may lead some to conclude, incorrectly, that racial barriers to waiting list entry and transplantation no longer exist. Indeed, the statistically best estimates of the black-white differences are the odds ratios for black versus white patients in the full hazard models, which were not equal to 1.0. With regard to waiting list entry, the lack of statistical significance could be an artifact of low power in the full hazard model. In general, adding information about socioeconomic factors, attitudes about treatment options, and other factors helped to close the gap between black and white patients, but the gap was not eliminated. Thus, the question remains of how the odds of obtaining a transplant can be improved for black ESRD patients.

757

There are at least four ways to address this question. These include (1) better education about the advantages and disadvantages of kidney transplantation, which should be directed at minority patients and their care providers; (2) continued efforts to increase the supply of donor organs by minorities; (3) continued research on the race-related effects of variances in point systems used to allocate cadaver kidneys; and (4) consolidating some waiting lists into larger regional lists, thereby giving waiting list patients access to a wider range of donor organs. These solutions are described briefly in turn. The first two options are related. Callender21 notes that organ donation rates among minorities are lower for several reasons, including a lack of education about transplantation and poor communication between minority patients and their providers. Better education of the potential advantages and disadvantages of transplantation may help minority ESRD patients make better treatment choices, and better education of other patients and their families may help increase organ donation rates among minorities. Successful efforts to increase organ donation among minorities may improve donor-patient differences in HLA and blood type, and thereby lessen the impact of these differences on receipt of a transplant by black patients. The third option, continued research on the impacts of variances from the UNOS kidney transplant point system, is motivated partly by a study by Lazda.22 UNOS modified its kidney transplant point system in 1989 by putting less emphasis on HLA matching and more emphasis on waiting time and other factors4 Lazda designed a point system that varied from the UNOS point system by putting even less emphasis on HLA matching and more emphasis on panelreactive antibody sensitivity and waiting time. Lazda found that the use of this point system variance in Illinois increased the odds of receiving a transplant among black patients. Work like that reported by Lazda22 should continue. In addition, the use of every point system variance should be evaluated in terms of the effects on patient and graft survival, quality of life after the transplant, and access to transplantation by minorities, low-income patients, and others for whom it is usually hard to find a matching organ. Patient and graft survival and quality of

OZMINKOWSKI

758

life outcomes were not addressed in Lazda’s study. Similar in spirit to the Lazda study,** it is interesting to note that the UNOS Board of Directors recently agreed to test yet another organ allocation method that is expected to result in higher transplantation rates among minorities. A June 2.5, 1997, press release from UNOS noted that a 3-year cross-reactive antigen groupmatching pilot study is about to begin that will modify the way that organs are distributed to minorities and highly sensitized study participants.28 The press release suggested that access barriers will be overcome for study participants, but details of the study were not provided. We anticipate that access to transplantation will be studied, in addition to the medical and quality of life outcomes that will be tested. Information about the full range of relevant outcomes will help UNOS, transplant providers, and kidney transplant candidates make more informed treatment decisions. Finally, race-related differences in the odds of receiving a transplant may be ameliorated somewhat if some of the smaller kidney transplant waiting lists were consolidated into larger regional lists. This has already happened to some extent, as the number of organ procurement organizations (each with its own waiting list) has changed over time from a high of 72 to 66. However, further consolidation is likely to be met with significant challenges, which stem in part from logistic difficulties and from the different values that may underlie the use of different point systems in contiguous areas.29 There may be strong objections to consolidating waiting lists in organ procurement organization areas that use different point systems. These objections may arise from the different beliefs about the importance of access-related versus medical outcomes that should be maximized via the particular point systems being used. Because of these challenges, consolidating waiting lists may be quite controversial. Nevertheless, we believe that these challenges should be addressed and that empirical evaluations should be conducted to estimate the impact that consolidating waiting lists will have on equity in the organ allocation process. ACKNOWLEDGMENT The authors thank and Robert Kirkman,

Drs Louis Diamond, Charles Shield III, and the Project Officer, Melford J. Hen-

ET AL

derson, for helpful comments during the study period. They also thank the two anonymous reviewers and the Journal’s editorial committee for their contributions during the peer review process. The opinions expressed in this paper are the authors’ and do not necessarily represent the opinions of their consultants, employers, or Agency for Health Care Policy and Research.

REFERENCES 1. Kjellstrand CM: Age, sex, and race inequality in renal transplantation. Arch Intern Med 14813051309, 1988 2. Held PJ, Pauly M, Bovbjberg R, Newmann J, Salvatierra 0: Access to kidney transplantation: Has the United States eliminated income and racial differences? Arch Intern Med 1482594-2600, 1988 3. Office of the Inspector General (OIG): The Distribution of Organs for Transplantation: Expectations and Practices. US Department of Health and Human Services publication OEl-01-89-00550. Washington, DC, Office of Analysis and Inspections, 1991 4. Sanfilippo FP, Vaughn WK, Peters TG, Shield CF, Adams PL, Lorber MI, Williams GM: Factors affecting the waiting time of cadaveric kidney transplant candidates in the United States. JAMA 267:247-252, 1992 5. Gaylin DS, Held PJ, Port FK, Hunsicker LG, Wolfe RA, Kahan BD, Jones CA, Agodoa LY: The impact of comorbid and sociodemographic factors on access to renal transplantation. JAMA 269:603-608, 1993 6. Kallich JD, Adams JL, Lindsay Barton P, Spritzer KL: Access to Cadaveric Kidney Transplantation. Santa Monica, CA, RAND Carp, 1993 7. Eggers PW: Racial differences in access to kidney transplantation. Health Care Fin Rev 17:89-104, 1995 8. Warren E, Hull AR, Prati RC: Patient status in dialysis units. Semin Nephrol 2:186-188, 1982 9. Evans R: Organ donation: New ways needed to keep up with the demand. Nephrol News Issues 1990, pp 16-17 10. HCFA: Health Care Financing Research Report: End Stage Renal Disease, 1991. HCFA Publication Number 03338. Baltimore, MD, Health Care Financing Administration, 1993 11. Kurtin PS, Davies AR, Meyer KB, DeGiacome JM, Kantz ME: Patient-based health status measurements in outpatient dialysis: Early experiences in developing an outcomes assessment program. Med Care 30:MS136-MS149, 1992 12. Meyer KB, Espindle DM, DeGiacomo JM, Jenuleson CS, Kurtin PS, Ross Davies A: Monitoring dialysis patients health status. Am J Kidney Dis 24:267-279, 1994 13. Diamond L, Held PJ, Palumbo MJ: Developing a workable casemix index and relates cost: Experience for patients with chronic renal failure. Urban Institute Working Paper 3064-l 1. Washington, DC, The Urban Institute, 1984 14. Held PJ, Pauly M, Diamond L: Survival analysis of patients under-going dialysis. JAMA 257:645-650, 1987 15. Health Care Financing Administration (HCFA): Special Report: Findings from the National Kidney Dialysis and Kidney Transplantation Study. Baltimore, MD, US Depatiment of Health and Human Services, Health Care Financing Administration, Office of Research and Demonstration, 1987 16. Ellison MD, Breen TJ, Guo TG, Cunningham PRG, Daily OP: Blacks and whites on the UNOS renal waiting list:

RACE

AND

KIDNEY

TRANSPLANTATION

Waiting times and patient demographics compared. Transplant Proc 25:2462-2466, 1993 17. Lefell MS, Steinberg AG, Bias WB, Machan CH, Zachary AA: The distribution of HLA antigens and phenotypes among donors and patients in the UNOS registry. Transplantation 58:1119-1130, 1994 18. Software for Survey Data Analysis (SUDAAN), Version 5.30. Research Triangle Park, NC, Research Triangle Institute, 1991 19. Bayton JA, Jennings PS, Callender CO: The role of blacks in donation and the organ and tissue transplantation process. Transplant Proc 21:3971-3972, 1989 20. Burdick JF, Diethelm A, Thompson JS, Van Buren CT, Williams GM: Organ sharing: Present realities and future possibilities. Transplantation 51:287-292, 1991 21. Callender CO: Organ donation in the black population: Where do we go from here? Transplant Proc 19:36-40, 1987 (SUPPl 2) 22. Lazda VA: An evaluation of a local variance of the United Network for Organ Sharing (UNOS) point system on the distribution of cadaver kidneys to waiting minority recipients. Transplant Proc 23:901-902, 1991 23. Hauptman PJ, O’Connor KJ: Procurement and allocation of solid organs for transplantation. N Engl J Med 336:422-431, 1997

759 24. Kondo K, Shibue T, Iwaki Y, Terasaki P: Racial effect on kidney transplants, in Terasaki P (ed): Clinical Transplant 1987. Los Angeles, CA, UCLA Tissue Typing Laboratory, 1988, pp 339-349 25. Held PJ, Friedrich KP, Blagg CR, Agodoa LYC: The United States Renal Data System’s 1990 annual report: Executive summary. Am J Kidney Dis 16:5-12, 1990 (suppl 2) 26. Rovelli M, Palmeri D, Vossler E, Bartus S, Hull D, Schweizer R: Noncompliance in renal transplant recipients: Evaluation by sociodemographic group. Transplant Proc 21:3979-3981, 1989 27. United Network for Organ Sharing: Background: Problems and Concerns in Equitable Organ Allocation. Statement of Principles and Objectives-Appendix D. Richmond, VA, United Network for Organ Sharing (web page address: http://204.127.237.1l/eg-bkgnd.htm) 28. United Network for Organ Sharing: Press Release: New Method of Matching Donated Kidneys May Mean More Transplants for Minorities. Richmond, VA, United Network for Organ Sharing, June 25, 1997 29. Ozminkowski RJ, Hassol A, White AJ, Murphy M, Dennis JM, Shield CF: Socioeconomic factors and multiple listing for cadaveric kidney transplantation among Medicare End-Stage Renal Disease Program beneficiaries. Transplant Rev 11:70-75, 1997