Summary
Background and objectives Patterns of end-of-life care among patients with ESRD differ by race. Whether the magnitude of racial differences in end-of-life care varies across regions is not known.
Design, setting, participants, & measurements This observational cohort study used data from the US Renal Data System and regional health care spending patterns from the Dartmouth Atlas of Healthcare. The cohort included 101,331 black and white patients 18 years and older who initiated chronic dialysis or received a kidney transplant between June 1, 2005, and September 31, 2008, and died before October 1, 2009. Black–white differences in the odds of in-hospital death, dialysis discontinuation, and hospice referral by quintile of end-of-life expenditure index (EOL-EI) were examined.
Results In adjusted analyses, the odds ratios for dialysis discontinuation for black versus white patients ranged from 0.47 (95% confidence interval=0.43 to 0.51) in the highest quintile of EOL-EI to 0.63 (95% confidence interval=0.54 to 0.74) in the lowest quintile (P for interaction<0.001). Hospice referral ranged from 0.55 (95% confidence interval=0.50 to 0.60) in the highest quintile of EOL-EI to 0.82 (95% confidence interval=0.69 to 0.96) in the lowest quintile (P for interaction<0.001). The association of race with in-hospital death also differed in magnitude across quintiles of EOL-EI, ranging from 1.21 (95% confidence interval=1.08 to 1.35) in the highest quintile of EOL-EI to 1.47 (95% confidence interval=1.27 to 1.71) in the second quintile (P for interaction<0.001).
Conclusions There are pronounced black–white differences in patterns of hospice referral and dialysis discontinuation among patients with ESRD that vary substantially across regions of the United States.
Introduction
Black patients experience disproportionately high rates of ESRD (1–6). Racial differences in a range of practices and outcomes, such as predialysis care, choice of dialysis modality, access to kidney transplant, and survival, have been well described among patients with ESRD (1,6–20). Compared with these other outcomes and care practices, racial differences in patterns of end-of-life care among patients with ESRD have received much less attention (21–24).
Prior studies have found that, compared with white patients, black patients receiving chronic dialysis are less likely to complete advance directives (23), less likely to discontinue dialysis (21,22,25,26), less likely to be referred to hospice (21,25), and more likely to receive intensive interventions such as intubation, feeding tube placement, and cardiopulmonary resuscitation during the final month of life (27). Although these findings do not necessarily imply differences in the quality of end-of-life care for black and white patients—particularly if they reflect racial variation in goals and preferences at the end of life (28)—they do highlight the importance of efforts to better understand the scope of, and rationale for, racial differences in patterns of end-of-life care in this population.
There is likely substantial complexity to the relationship between race and treatment practices at the end of life. Prior studies among older Medicare beneficiaries have described considerable variation in the magnitude of racial differences in care across regions with differing patterns of health care spending (29). Prior studies have shown that systematic differences in where black and white patients with ESRD live seem to contribute to racial differences in outcomes and care practices, such as time to transplant and nephrology referral (8,30–32). To our knowledge, the extent to which racial differences in end-of-life care among patients with ESRD vary across regions has not been described. Understanding how racial differences in end-of-life treatment practices vary geographically may be particularly important, because a disproportionate number of black patients with ESRD are concentrated in a relatively small number of predominantly black zip codes (30,31). These zip codes tend to be urban, resource-poor, and located in hospital referral regions with the highest levels of Medicare spending among patients approaching the end of life (25). We evaluated the hypothesis that the magnitude of racial differences in patterns of end-of-life care among patients with ESRD would vary across hospital referral regions with differing levels of Medicare spending.
Materials and Methods
Patients and Data Sources
We used data from the US Renal Data System (USRDS), a national registry for ESRD, to identify all 366,022 patients between the ages of 18 and 100 years who initiated chronic dialysis or received a kidney transplant for the first time between June 1, 2005, and September 31, 2008. Among these patients, 142,856 died before October 1, 2009. Of these, we excluded 10,207 who were missing information on zip code, estimated GFR (eGFR), body mass index (BMI), comorbid conditions, or functional status at onset of ESRD. We excluded another 31,318 patients who were missing information on study outcomes (place of death, whether dialysis was discontinued before death, and whether they were referred to hospice), yielding an analytic cohort of 101,311 patients. Compared with decedents' excluded missing information on covariates or study outcomes, members of the analytic cohort were less likely to be black but had a similar age and sex distribution (Supplemental Table 1).
The USRDS Patients file provided information on demographic characteristics, date of first dialysis, and date of death. The USRDS Medical Evidence file provided information on eGFR, BMI, comorbid conditions (coronary artery disease, peripheral arterial disease, diabetes mellitus, stroke, congestive heart failure, chronic obstructive pulmonary disease, and cancer), functional status (whether the patient could transfer or ambulate) at onset of ESRD, whether the patient was under the care of a nephrologist before dialysis initiation, hemodialysis access type (fistula versus other), and initial treatment modality (hemodialysis, peritoneal dialysis, or transplantation). The USRDS Death file provided information on cause of death, site of death (in versus outside the hospital), whether dialysis was discontinued, and whether the patient was referred to hospice before death as reported by the nephrology provider on the Centers for Medicare and Medicaid Services (CMS) ESRD death notification form.
Predictor Variable
The primary predictor variable for most analyses was black versus white race. The association of black versus white race with each study outcome was assessed in both the overall cohort and analyses stratified by region. Patients were assigned to a hospital referral region (HRR) based on their zip code at onset of ESRD using a zip code–HRR crosswalk from the Dartmouth Atlas of Healthcare (http://www.dartmouthatlas.org/). Patterns of health care spending in each HRR were defined using the end-of-life expenditure index (EOI-EI) from the Dartmouth Atlas. The index reflects both physician spending (from the Medicare Carrier File) and acute inpatient hospital spending (from the Medicare Provider Analysis and Review File) during the last 6 months of life among Medicare beneficiaries who were between the ages of 65 and 100 years at the time of death (33,34). Only those patients who were eligible for Medicare during the 6-month period before death and not enrolled in a health maintenance organization during this time frame were used to develop the index. The index is calculated based on standardized national prices and adjusted for the demographic characteristics of Medicare beneficiaries in each hospital referral region. As such, it is intended to reflect that component of regional Medicare spending attributable to the overall quantity of medical services provided rather than local differences in pricing and demographic structure. Patients were categorized by HRR quintile of EOL-EI to be consistent with prior publications using this index (25,33,34).
Outcome Variables
Measures of end-of-life care included (1) death in versus outside the hospital, (2) whether dialysis was discontinued before death (among patients who had not received a kidney transplant), and (3) whether the patient was referred to hospice before death.
Statistical Analyses
We described patient characteristics at onset of ESRD, causes of death, and study outcomes among black and white patients using point estimates and 95% confidence intervals (95% CIs). We also described the characteristics of black and white patients living in regions in different quintiles of EOL-EI. Statistical significance was assessed using tests for trend or interaction as appropriate. We used logistic regression analysis to measure the adjusted association of black race with each outcome. Multivariate analyses were conducted both in the overall cohort and after stratification by quintile of EOL-EI, and they were adjusted for age, sex, BMI, eGFR, comorbid conditions, functional status, predialysis nephrology referral, cause of death, dialysis modality, and type of vascular access at the time of ESRD onset. We also used logistic regression analysis to compare the frequency of each outcome among patients living in the highest versus lowest quintile of EOL-EI. These analyses were adjusted for the covariates described above and stratified by race. Because patterns of end-of-life care may vary depending on how soon patients died after onset of ESRD, we conducted a subgroup analysis among patients who died within 6 months of ESRD onset.
To determine whether the magnitude of racial differences in each outcome varied by quintile of EOL-EI, we tested for interactions between race and EOL-EI quintile using the likelihood ratio test for interaction. These analyses were conducted among both the overall cohort and the subgroup that died within 6 months of ESRD onset. In all multivariate analyses, we assessed for evidence of colinearity using the variance inflation factor. All statistical analyses were conducted using Stata SE version 11.0 (StataCorp, College Station, TX). The study was approved by the Institutional Review Board at the University of Washington.
Results
Patient Characteristics
Compared with white patients, black patients were younger, included a higher percentage of women, and had a higher prevalence of diabetes and stroke and a lower prevalence of congestive heart failure, coronary artery disease, chronic obstructive pulmonary disease, and cancer. Black patients were less likely to have seen a nephrologist before ESRD onset, to have initiated hemodialysis with a fistula, and to have peritoneal dialysis or kidney transplant as their initial modality. Cardiovascular and infectious causes of death were slightly more common among black than white patients. A higher percentage of white patients died within 6 months of ESRD onset (37.1% versus 33.8%) (Table 1). Cause of death was more likely to be infection-related (11.2% versus 12.9%) and cardiovascular (46.2% versus 47.9%) in black compared with white patients.
Patient characteristics by race
The percentage of black patients ranged from 7.6% in the lowest quintile of EOL-EI to 30.6% in the highest quintile. Overall, 34.1% of all black patients and 24.4% of all white patients lived in regions in the highest quintile of EOL-EI (Table 1). Among both black and white patients, those patients living in the highest versus lowest quintile of EOL-EI were older, had a lower prevalence of most comorbid conditions, had a higher prevalence of functional impairment, and were less likely to have been referred to a nephrologist, to have peritoneal dialysis as the initial modality, and to initiate hemodialysis with a fistula (Supplemental Tables 2 and 3).
Patterns of End-of-Life Care among Black Versus White Patients
Black patients were more likely than white patients to have died in the hospital (68.4% versus 58.2%), less likely to have discontinued dialysis (16.4% versus 32.1%), and less likely to have been referred to hospice before death (15.5% versus 28.3%) (Table 2). These differences persisted in adjusted analyses and among patients who died within 6 months of ESRD onset (Table 3).
Unadjusted patterns of end-of-life care by race and quintile of end-of-life expenditure index
Adjusted association of black versus white race with patterns of end-of-life care by quintile of end-of-life expenditure index
Patterns of End-of-Life Care among Black Versus White Patients across Quintiles of EOL-EI
In all quintiles of EOL-EI, black patients were more likely than white patients to have died in the hospital and less likely to have discontinued dialysis and to have been referred to hospice in both unadjusted (Table 2) and adjusted analyses (Table 3). For both races, there were significant linear trends in site of death, discontinuation of dialysis, and referral to hospice (Table 2). In adjusted analyses, the odds of dialysis discontinuation and hospice referral for black versus white patients varied significantly by quintile of EOL-EI (P<0.001 for each interaction). Odds ratios for dialysis discontinuation and hospice referral among black versus white patients were of greatest magnitude among patients living in the highest quintile of EOL-EI and were progressively more attenuated in HRRs with a lower EOL-EI. The odds of in-hospital death by race also varied significantly by quintile of EOL-EI (interaction P value<0.001), but this relationship was nonlinear, with odds ratios for black versus white patients ranging from 1.21 (95% CI=1.08 to 1.35) in the highest quintile of EOL-EI to 1.47 (95% CI=1.27 to 1.71) in the second quintile.
Patterns of End-of-Life Care in the Highest Versus Lowest Quintile of EOL-EI by Race
In adjusted race-stratified analyses examining the extent to which study outcomes differed between patients living in the highest versus lowest quintile of EOL-EI, both black and white patients living in the highest versus lowest quintile of EOL-EI were more likely to have died in the hospital (odds ratio [OR] for black patients=1.82 [95% CI=1.58 to 2.10]; OR for white patients=1.75 [95% CI=1.66 to 1.84]), less likely to have discontinued dialysis (OR for black patients=0.33 [95% CI=0.28 to 0.39]; OR for white patients=0.45 [95% CI=0.42 to 0.47]), and less likely to have been referred to hospice (OR for black patients=0.41 [95% CI=0.34 to 0.48]; OR for white patients=0.62 [95% CI=0.58 to 0.65]) (Table 4). Results were similar in subgroup analyses among patients who died within 6 months of ESRD onset (Table 4).
Adjusted patterns of end-of-life care for patients living in the highest versus lowest quintile of end-of-life expenditure index stratified by race
Discussion
Among a national cohort of adults with ESRD, the magnitude of racial differences in patterns of end-of-life care varied substantially across regions. Among both black and white patients, differences across quintiles of EOL-EI in the proportion of patients who discontinued dialysis and were referred to hospice were as pronounced as those between black and white patients living in regions in the same quintile of EOL-EI. The magnitude of black–white differences in dialysis discontinuation and hospice referral was greatest for patients living in regions in the highest quintile of EOL-EI.
Several prior studies have reported rates of dialysis discontinuation and hospice referral for the overall ESRD population and among black compared with white dialysis patients (21–24). Murray et al. (21) examined patterns of hospice referral among US dialysis patients who died between January 1, 2000, and December 31, 2002, and had Medicare as the primary payer for dialysis. Using information from Medicare hospice claims and the CMS death notification form, Murray et al. (21) found that 13.5% of patients used hospice, 21.8% withdrew from dialysis, and 63% died in the hospital. Even among those patients who discontinued dialysis, only about one half of patients were referred to hospice (21). Black race was associated with a lower likelihood of hospice use, and there were large differences in hospice use and dialysis discontinuation across states. However, these authors (21) did not evaluate regional variation in the magnitude of black–white differences in patterns of end-of-life care. Thompson et al. (24) conducted a survey among dialysis facility professionals involved in the care of 448 patients from three renal networks who died during a 6-month period in 2005–2006 and discontinued dialysis before death. Less than one half of these patients were referred to hospice after dialysis discontinuation. The goals of the survey were to assess provider knowledge about hospice and ascertain whether hospice had been discussed before death. These authors (24) identified striking knowledge deficits among dialysis facility staff about hospice benefits for dialysis patients but found no differences by race in the proportion of decedents with whom hospice had been discussed or the proportion who chose hospice. Although large differences in study populations and design preclude direct comparison with our results, these results do highlight the potential importance of provider-related factors in shaping patterns of end-of-life care among patients with ESRD.
Reasons for the observed racial differences in patterns of end-of-life care are likely complex and reflect a variety of different factors, including differences in preferences (35–38), differences in the hospitals where patients are treated (39,40), differential access to care (41–43), differences in cultural, spiritual, and religious beliefs (44–47), differences in education and health literacy (48), and mistrust of the medical system (49). Prior studies evaluating racial differences in patterns of end-of-life care among patients with ESRD have not described the extent to which these factors may vary geographically. Our results seem to suggest that black–white differences in patterns of end-of-life care among patients with ESRD may be sensitive to regional differences in practice, with the most pronounced differences in rates of dialysis discontinuation and hospice referral occurring in those regions where Medicare beneficiaries tend to receive the highest intensity of inpatient care at the end of life.
The very limited life expectancy of patients with ESRD underlines the special importance of efforts to optimize end-of-life care in this population. Although we describe large racial and geographic differences in patterns of end-of-life care, it is unclear whether these differences translate into differences in quality of care without knowing more about the extent to which existing practices are aligned with patient preferences and values (28,50). Nevertheless, the disproportionate number African-American patients with ESRD who live in regions in the highest quintile of EOL-EI underlines the importance of understanding why there are such large differences in patterns of end-of-life care among black patients living in these areas.
Strengths of this study include that it was conducted among a large representative sample of black and white US adults with ESRD and relied on data systematically collected at the onset of ESRD and time of death. Limitations include, first, that the EOL-EI characterizes regional expenditure based on Medicare spending during the last 6 months of life and thus, may not fully capture spending for patients with ESRD specifically. Second, site of death, dialysis discontinuation, and hospice referral as reported on the CMS death notification form have not been validated. Third, we were limited in our ability to fully account for variables, such as socioeconomic status, cultural beliefs, and family structure, that may contribute to racial variation in patterns of end-of-life care.
In conclusion, there is substantial regional variation in the magnitude of racial differences in patterns of end-of-life care among adults with ESRD. Black–white differences in dialysis discontinuation and hospice referral were most pronounced in those regions with the highest levels of end-of-life Medicare spending. Efforts are needed to understand the underlying reasons for these differences and the extent to which these differences reflect differences in patient values, goals, and preferences.
Disclosures.
None.
Acknowledgments.
We acknowledge the following grant support: Ruth L. Kirschstein National Research Service Award F32 DK093167-01 from the National Institutes of Health (to B.A.T.) and Beeson Career Development Award K 1K23AG28980 from the National Institute on Aging (to A.M.O.).
This work was conducted at the University of Washington and does not represent the opinion of the US Renal Data System or the Department of Veteran Affairs.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
This article contains supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.06780712/-/DCSupplemental.
- Received July 5, 2012.
- Accepted February 1, 2013.
- Copyright © 2013 by the American Society of Nephrology