Racial and Ethnic Variations in Mortality Rates for Patients Undergoing Maintenance Dialysis Treated in US Territories Compared with the US 50 States

Visual Abstract

Abstract

Background and objectives

In the United States mortality rates for patients treated with dialysis differ by racial and/or ethnic (racial/ethnic) group. Mortality outcomes for patients undergoing maintenance dialysis in the United States territories may differ from patients in the United States 50 states.

Design, setting, participants, & measurements

This retrospective cohort study of using US Renal Data System data included 1,547,438 adults with no prior transplantation and first dialysis treatment between April 1, 1995 and September 28, 2012. Cox proportional hazards regression was used to calculate hazard ratios (HRs) of death for the territories versus 50 states for each racial/ethnic group using the whole cohort and covariate-matched samples. Covariates included demographics, year of dialysis initiation, cause of kidney failure, comorbid conditions, dialysis modality, and many others.

Results

Of 22,828 patients treated in the territories (American Samoa, Guam, Puerto Rico, Virgin Islands), 321 were white, 666 were black, 20,299 were Hispanic, and 1542 were Asian. Of 1,524,610 patients in the 50 states, 838,736 were white, 444,066 were black, 182,994 were Hispanic, and 58,814 were Asian. The crude mortality rate (deaths per 100 patient-years) was lower for whites in the territories than the 50 states (14 and 29, respectively), similar for blacks (18 and 17, respectively), higher for Hispanics (27 and 16, respectively), and higher for Asians (22 and 15). In matched analyses, greater risks of death remained for Hispanics (HR, 1.65; 95% confidence interval, 1.60 to 1.70; P<0.001) and Asians (HR, 2.01; 95% confidence interval, 1.78 to 2.27; P<0.001) living in the territories versus their matched 50 states counterparts. There were no significant differences in mortality among white or black patients in the territories versus the 50 states.

Conclusions

Mortality rates for patients undergoing dialysis in the United States territories differ substantially by race/ethnicity compared with the 50 states. After matched analyses for comparable age and risk factors, mortality risk no longer differed for whites or blacks, but remained much greater for territory-dwelling Hispanics and Asians.

Introduction

CKD is a noncommunicable disease that is now well recognized as a major source of premature morbidity and mortality (1). The estimated prevalence of CKD varies by racial and/or ethnic (racial/ethnic) group (2), as well as geographic location (3). Ethnic subgroup differences for Hispanics have been noted in CKD prevalence with a nearly two-fold difference across persons of Cuban (12%), Mexican (13%), Puerto Rican (17%), and South American (8%) backgrounds, with the highest rate noted among Puerto Ricans (4). Racial/ethnic minorities in the United States are more likely to develop ESKD than nonminority groups, and are often treated with maintenance dialysis (5).

Although there have been extensive analyses of CKD and ESKD in the United States 50 states, there are few studies about kidney disease in the United States territories. The US Renal Data System (USRDS) Annual Data Report (including the most recent version) mainly focuses on individuals residing in the 50 states and excludes patients from the territories (6). The United States has five territories that are permanently inhabited and comprise Puerto Rico, the US Virgin Islands, Guam, the Northern Mariana Islands, and American Samoa, with a disproportionately high rate of diabetes-related ESKD (7,8). Similar to the 50 states, ESKD treatment in the territories is covered by Medicare, although the reimbursement rate is much lower (9). Further, in Puerto Rico, Medicare assumes primary coverage in lieu of commercial insurers after only 3 months rather than 30 months of ESKD in the 50 states (9). Most analyses in the 50 states found racial/ethnic minorities on maintenance dialysis have a 15%–20% lower adjusted mortality rate than their majority peers, with Hispanic patients consistently having the lowest mortality rates across all age groups (10–14).

Mortality outcomes for patients undergoing dialysis in the territories are not well known. We hypothesized that mortality outcomes for patients in the territories may differ from patients in the 50 states. We therefore undertook a national population study to examine all-cause mortality by racial/ethnic group among patients with ESKD treated with dialysis in the United States territories compared with patients treated in the United States 50 states.

Materials and Methods

This project was approved by the Institutional Review Board at University of Virginia. Using the USRDS Core Standard Analysis Files (15), we identified all patients aged 18 years or older, with no prior kidney transplantation who initiated the first maintenance dialysis between April 1, 1995 and September 28, 2012. We excluded 3% of patients with missing body mass index (BMI) and 2% with missing eGFR. The final cohort consisted of 1,547,438 patients, including 22,828 patients in the four territories (295 in American Samoa, 1507 in Guam [including the Northern Mariana Islands], 20,289 in Puerto Rico, 737 in the Virgin Islands), and 1,524,610 nonterritory patients.

Study Variables

Outcomes.

The outcomes of interest were mortality differences between patients undergoing dialysis in the territories and those in the 50 states in the same racial/ethnic group. Survival time was calculated from the date of dialysis initiation to the date of death, kidney transplant, or administrative end of study (September 30, 2012), whichever occurred first, with censoring at transplantation. Death and transplant dates were obtained from the USRDS Patient File. Causes of deaths due to cardiovascular disease and infection were ascertained on the basis of the primary cause of death in the Centers for Medicare and Medicaid Services (CMS) ESKD Death Notification Form (CMS-2746) (16).

Primary and Grouping Variables.

The primary explanatory variable is an indicator for territory (versus nonterritory). The grouping variable is race/ethnicity. We obtained patient race from the USRDS Patient File containing six categories (white, black, Asian, Native American, other, and unknown). This single Asian category, combining Pacific Islanders and other Asians (such as Chinese and Koreans), was used in primary analyses. Analyses separating these two subgroups were also performed with the subgroup ascertained from the CMS Medical Evidence (ME) form (CMS-2728). Patient ethnicity (Hispanic/non-Hispanic) was obtained from the ME form. We present our analyses for non-Hispanic whites, non-Hispanic blacks, non-Hispanic Asians, and Hispanics. Other groups were excluded owing to small sample size.

Covariates.

Two sets of covariates were included in adjusted analyses. The first set were demographics (age at ESKD onset, sex), BMI, year of dialysis initiation (2000 or before, 2001–2005, and 2006–2012), cause of kidney failure (diabetes, hypertension, GN, and other/unknown), insurance at ESKD onset (private, Medicare, and Medicaid/none), and eGFR. The second set of covariates included comorbid conditions (hypertension, diabetes, cardiac failure, atherosclerotic heart disease, other cardiac disease, cerebrovascular disease, peripheral vascular disease, chronic obstructive pulmonary disease, and cancer); any of smoking, alcohol abuse, or drug use; and dialysis modality at dialysis initiation (hemodialysis or peritoneal dialysis). We also provided descriptive summaries for nephrology visit before dialysis and vascular access type used at dialysis initiation, but did not include them in modeling analyses because data were only available after 2005. Data on dialysis modality are on the basis of the Modality Sequence File, whereas all other covariate values are from the ME files.

Statistical Analyses

We calculated crude mortality rates, the number of deaths per 100 patient-years, of each race/ethnicity in the group of four territories combined and the group of the 50 states combined. For each race/ethnicity, we obtained unadjusted hazard ratio (HR) for the territory group versus the 50 states with Cox proportional hazards regression and adjusted HR adjusting for all the covariates described above. We also examined the Fine–Gray subdistribution HRs that account for the competing risk of kidney transplantation, with adjustment for the same covariates (17).

Because of great differences in the number of subjects between the two groups, we performed another set of matched analyses. Each territory patient was matched 1:1 to a patient residing in the 50 states with the same race/ethnicity. We constructed two sets of matched samples. The first sample was matched on the basis of the first set of covariates described above: age at ESKD onset (±5 years), sex, BMI (±5 kg/m²), year of dialysis initiation, cause of kidney failure, insurance at ESKD onset, and eGFR (±3 ml/min per 1.73 m²). Almost every territory patient has a match of the same race/ethnicity from the 50 states. The second sample was on the basis of the first covariate set with the addition of the second set of covariates: comorbid conditions; any of smoking, alcohol abuse, or drug use; and dialysis modality. This second matched sample contains most, but not all, of the territory patients from the original territory group; those for whom a nonterritory match was not found were not included. We then repeated the survival analyses using these matched samples.

Results

Of 1,524,610 patients in the 50 states, 838,736 (55%) were white, 444,066 (29%) were black, 182,994 (12%) were Hispanic, and 58,814 (4%) were Asian. Of 22,828 patients treated in the four territories, 321 (1%) were white, 666 (3%) were black, 20,299 (89%) were Hispanic, and 1542 (7%) were Asian. Whites and Asians in the territories were, on average, 7 years and 4 years younger, respectively, than their counterparts in the 50 states, in contrast to an average age of 2 years older for both black and Hispanic patients in the territories than in the 50 states (Table 1). Patients in the territories had a higher prevalence of diabetes-related ESKD (64% versus 45%) and were less likely to have pre-ESKD erythropoietin use (20% versus 27%), a nephrology visit before dialysis (50% versus 58%), or use of an arteriovenous fistula (10% versus 14%). White, black, and Hispanic groups had similar average BMIs, but Asians in the territories had higher BMIs than the Asians in the 50 states (28 versus 25 kg/m²). Several patient characteristics also differed across the four territories (Supplemental Table 1). For instance, Samoans tended to have higher BMIs and Puerto Ricans were more likely to have received predialysis erythropoietin than those residing in the other territories.

Table 1.

Characteristics of patients undergoing dialysis by race/ethnicity according to territories and 50 states

Characteristic	Territories					50 States
	Overall	White	Black	Hispanic	Asian	Overall	White	Black	Hispanic	Asian
Patients, n	22,828	321	666	20,299	1542	1,524,610	838,736	444,066	182,994	58,814
Age, yr, mean±SD	61±14	59±15	60±14	61±14	58±13	63±15	66±14	58±15	59±15	62±16
Male, %	59	63	55	60	53	55	57	50	55	54
BMI, kg/m2, mean±SD	27±6	28±7	27±7	27±6	28±8	28±7	28±7	28±8	28±7	25±6
Employed, %	80	78	63	81	68	83	89	76	75	79
eGFR, ml/min per 1.73 m2,at ESKD initiation, mean±SD	9±4	8±4	9±5	9±4	8±5	10±5	10±5	10±5	9±5	9±4
Insurance status at ESKD initiation, %
Private	46	51	43	45	56	50	61	37	32	45
Medicare	29	27	32	30	20	29	26	33	30	29
Medicaid/none	25	22	26	25	25	21	13	29	38	26
Cause of kidney failure, %
Diabetes	64	58	54	64	71	45	42	44	61	50
Hypertension	16	19	29	15	18	28	27	35	19	25
GN	8	7	4	8	4	8	8	7	7	12
Urologic	2	w	w	2	2	2	3	1	1	1
Polycystic kidneys	2	w	2	2	1	2	3	1	1	2
Other	7	9	8	7	3	11	13	9	7	6
Unknown	2	w	w	2	1	4	4	3	3	4
Dialysis modality at dialysis initiation, %
Hemodialysis	91	91	97	91	97	93	91	95	94	91
Peritoneal dialysis	9	9	3	9	3	7	9	5	6	9
Presence of comorbidities, %
Hypertension	82	65	84	83	82	80	78	83	82	81
Diabetes	58	47	60	57	71	51	48	50	63	52
Congestive heart failure	34	31	43	34	30	33	36	29	28	26
Atherosclerotic heart disease	30	19	15	32	18	24	31	15	18	18
Other cardiac disease	7	25	10	6	15	13	16	11	8	11
Cerebrovascular accident/transient ischemic attack	8	6	7	8	5	9	10	10	7	8
Peripheral vascular disease	26	20	24	28	6	14	17	10	12	8
Chronic obstructive pulmonary disease	3	w	3	3	3	8	12		4	3
Cancer	4	w	4	4	2	7	9	5	3	3
Immobility	8	6	5	8	10	6	6	5	5	4
Smoking, alcohol abuse or illicit drug use	3	6	4	3	7	7	8	9	4	2
Erythropoietin pre-ESKD (yes), %	20	22	10	22	10	27	30	24	23	32
Pre-ESKD care at dialysis initiation (limited to 2005 or later)
Nephrologist visit before dialysis (yes), %	50	42	39	52	29	58	62	54	50	60
Access type at dialysis initiation, %
Arteriovenous fistula	10	11	12	10	6	14	14	12	12	15
Arteriovenous graft	2	w	4	2	3	3	3	5	2	3
Central venous catheter	82	81	82	81	88	76	74	77	80	72
Other vascular access	6	7	2	7	3	7	8	6	6	10

Cells with numbers that are too small to be reported are replaced with the symbol w according to the US Renal Data System privacy and confidentiality. Summing multiple categories in the same variable may not be 100% because of rounding. BMI, body mass index.

Mortality Analysis of the Overall Population by Race/Ethnicity

The median follow-up time was 23 months (interquartile range, 8–46 months) for the territory group and 25 months (interquartile range, 9–49 months) for the 50 states. White patients undergoing dialysis in the territories had a much lower crude mortality rate than their counterparts in the 50 states (14 versus 29 deaths per 100 patient-years) (Table 2). Blacks in the territories and 50 states had similar mortality rates (18 and 17 deaths per 100 patient-years). By contrast, Hispanics in the territories had a much higher mortality rate than their counterparts in the 50 states (27 versus 16), with 27 deaths per 100 patient-years for Puerto Rican Hispanics alone and 24 for Virgin Islander Hispanics alone. Asians in the territories also had a much higher mortality rate than their 50 states counterparts (22 versus 15), with 19 deaths per 100 patient-years for Samoan Asians alone and 23 for Guamanian Asians. There were similar differences in mortality rates between territories and 50 states for each racial/ethnic group when hemodialysis and peritoneal dialysis were separately analyzed (Supplemental Table 2).

Table 2.

Race/ethnicity-specific mortality rates and hazard ratios of death for territories versus 50 states in the whole cohort

Race	Deaths per 100 Patient-Year (Percentage of Deaths)		Unadjusted (Territories versus 50 States)		Adjusted (Territories versus 50 States)
	Territories	50 States	HR (95% CI)	P Value	HR (95% CI)	P Value
White	14 (59%)	29 (70%)	0.53 (0.46 to 0.61)	<0.001	0.75 (0.65 to 0.86)	<0.001
Black	18 (60%)	17 (59%)	1.08 (0.98 to 1.19)	0.13	1.04 (0.94 to 1.15)	0.45
Hispanic	27 (70%)	16 (53%)	1.62 (1.59 to 1.65)	<0.001	1.61 (1.58 to 1.63)	<0.001
Asian	22 (64%)	15 (50%)	1.47 (1.38 to 1.56)	<0.001	1.95 (1.82 to 2.08)	<0.001

Adjusted for demographics (age at ESKD onset, sex); body mass index; year of dialysis initiation (2000 or before, 2001–2005, and 2006–2012); cause of kidney failure (diabetes, hypertension, GN, and other/unknown); insurance at ESKD onset (private, Medicare, and Medicaid/none); eGFR; presence/absence of each of the comorbid conditions (hypertension, diabetes, cardiac failure, atherosclerotic heart disease, other cardiac disease, cerebrovascular disease, peripheral vascular disease, chronic obstructive pulmonary disease, and cancer); any of smoking, alcohol abuse, or drug use; and dialysis modality (hemodialysis or peritoneal dialysis). HR, hazard ratio (territories versus 50 states), 95% CI, 95% confidence interval.

For all racial/ethnic groups, cardiovascular disease was the major cause of death in both territories and 50 states, accounting for more than 40% of the total deaths (Figure 1). The percentages of deaths due to infection were, however, consistently greater in territories than in the 50 states, regardless of race and ethnicity.

Figure 1.

Percentages of deaths due to cardiovascular disease and infection by racial/ethnic group in the territories and 50 states. (A) Cardiovascular disease; (B) infection. Cardiovascular disease was the major cause of death in both territories and the 50 states. The percentages of deaths due to infection were greater in the territories than in the 50 states, regardless of race/ethnicity.

After multivariable adjustment for demographics and medical conditions, the mortality risk for white patients in the territories remained lower than that of white patients in the 50 states (HR, 0.75; 95% confidence interval [95% CI], 0.65 to 0.86; P<0.001), but did not differ for black patients (HR,1.04; 95% CI, 0.94 to 1.15; P=0.45) (Table 2). Conversely, Hispanic and Asian patients in the territories still had higher risks of death (HR, 1.61; 95% CI, 1.58 to 1.63; P<0.001 and HR, 1.95; 95% CI, 1.82 to 2.08; P<0.001, respectively).

Mortality Analysis of Matched Samples by Race/Ethnicity

The characteristics of the matched samples are presented in Supplemental Table 3. As expected, the first samples achieved covariate balance on the first set of covariates whereas the second samples achieved balance in both sets of covariates. Although not all territory patients were matched and included in the second territory samples (matched rates 67%–89%), the characteristics of the second and first territory samples within the same racial/ethnic group were largely similar. However, for whites, the nonterritory matched samples were much younger (mean age 59–60 years) than the original (unmatched) nonterritory white population (mean age 66 years; Table 1), and also had a lower prevalence of most comorbid conditions.

For Hispanics, both territory and nonterritory groups in the first and second matched samples had similar crude mortality rates as their original unmatched populations (Table 3), indicating that the greater adjusted mortality risk for Hispanic territory patients obtained in the previous regression model is robust (HR, 1.61; 95% CI, 1.58 to 1.63; HR, 1.62; 95% CI, 1.58 to 1.66; and HR, 1.65; 95% CI, 1.60 to 1.70; Table 3). For all Asians, the HR of 2.01 (95% CI, 1.78 to 2.27) in the second matched sample was also consistent with the greater adjusted mortality risk experienced by territory Asians. This greater mortality risk was consistent when Pacific Islanders and other Asians were analyzed separately (HR, 1.95; 95% CI, 1.60 to 2.37; and HR, 1.45; 95% CI, 1.10 to 1.91, respectively; Supplemental Table 4). For blacks, there were comparable mortality rates of the territory and nonterritory groups in the matched samples, similar as before.

Table 3.

Race/ethnicity-specific mortality rates and hazard ratios of death for territories versus 50 states in the original cohort and two matched samples

Race	Cohort/Sample	Death Rate (Deaths per 100 Patient-Years)		Territories versus 50 States
		Territories	50 States	Adjusted HR (95% CI)	P Value
White	Original white cohort (territories n=321, nonterritories n=838,736)	14	29	0.75 (0.65 to 0.86)	<0.001
	First matched sample (n=321 each group)	14	22	0.70 (0.57 to 0.86)	0.001
	Second matched sample (n=286 each group)	15	16	0.96 (0.77 to 1.20)	0.72
Black	Original black cohort (territories n=666, nonterritories n=444,066)	18	17	1.04 (0.94 to 1.15)	0.45
	First matched sample (n=664 each group)	18	17	1.11 (0.97 to 1.28)	0.13
	Second matched sample (n=578 each group)	18	19	0.93 (0.81 to 1.08)	0.37
Hispanic	Original Hispanic cohort (territories n=20,299, nonterritories n=182,994)	27	16	1.61 (1.58 to 1.63)	<0.001
	First matched sample (n=20,167 each group)	27	17	1.62 (1.58 to 1.66)	<0.001
	Second matched sample (n=14,769 each group)	25	16	1.65 (1.60 to 1.70)	<0.001
Asian	Original Asian cohort (territories n=1542, nonterritories n=58,814)	22	15	1.95 (1.82 to 2.08)	<0.001
	First matched sample (n=1508 each group)	22	13	1.83 (1.66 to 2.02)	<0.001
	Second matched sample (n=1032 each group)	22	12	2.01 (1.78 to 2.27)	<0.001

The first matched sample contains territory patients matched 1:1 to nonterritory patients of the same race/ethnicity on the following covariates: age at ESKD onset (±5 years), sex, body mass index (±5 kg/m²), year of dialysis initiation (2000 or before, 2001–2005, or 2006–2012), cause of kidney failure (diabetes, hypertension, GN, or other/unknown), insurance at ESKD onset (private, Medicare, or Medicaid/none), and eGFR (±3 ml/min per 1.73 m²). The presented HRs were further adjusted for these covariates to account for residual covariate imbalance in the matched data. The second matched sample contains territory patients matched 1:1 to nonterritory patients of the same race/ethnicity on the above covariates as well as the following covariates: presence/absence of each comorbid condition (hypertension, diabetes, cardiac failure, atherosclerotic heart disease, other cardiac disease, cerebrovascular disease, peripheral vascular disease, chronic obstructive pulmonary disease, and cancer); any of smoking, alcohol abuse, or drug use; and dialysis modality (hemodialysis or peritoneal dialysis). The presented HRs were further adjusted for all these covariates. HR, hazard ratio (territories versus 50 states); 95% CI, 95% confidence interval.

For whites, the crude mortality rate of the territory group in the matched samples remained almost unchanged (around 14–15 deaths per 100 patient-years; Table 3), but the mortality rate of the 50 states decreased greatly in the first matched sample and further decreased in the second sample (22 and 16 deaths per 100 patient-years from the original 29). This is expected because the matched nonterritory samples for whites contained younger and healthier whites selected from their original nonterritory white population. As a result, with comparable covariate conditions, white patients in the territories and the 50 states had similar mortality risks (HR, 0.96; 95% CI, 0.77 to 1.20). These results, including previous results for the other three racial/ethnic groups, were consistent with the Fine–Gray HRs that account for the competing risk of kidney transplantations (Supplemental Table 5).

Discussion

This is the first report to our knowledge to assess mortality rates for patients undergoing maintenance dialysis in all the United States territories. We found that mortality differences between the territories and the 50 states differed greatly by race/ethnicity. Our initial finding of greater mortality risk of white patients in the 50 states appeared to be explained by their older age and greater number of comorbidities. When the two groups were comparable on age, comorbidities, and other risk factors, their adjusted mortality rates reached parity. Black patients living in the territories had a similar mortality risk to those in the 50 states. By contrast, Hispanic and Pacific Islander and other Asian patients in the territories had greater adjusted mortality risks than their racial/ethnic counterparts living in the 50 states.

Most dialysis-related mortality studies comparing minority groups with white patients have reported much lower adjusted rates of mortality among Hispanic patients (4,5) and Asian patients receiving dialysis treatments (18). Frankenfield et al. (12) examined 2001–2005 ESKD Continuous Performance Measures data and found that compared with non-Hispanic white patients undergoing dialysis, patients of varying Hispanic backgrounds (Mexican American, Puerto Rican, Cuban American, and Hispanic other) had lower mortality rates with the exception of Puerto Ricans who did not differ from whites. Wong et al. (18) reported Asian patients had a 0.75 adjusted relative risk for mortality compared with white patients in a USRDS analysis. Similarly, Hall et al. found lower rates of mortality among Asian/Pacific Island patients undergoing dialysis compared with their white peers, with the exception of Chamorros, an indigenous group in Guam, whose mortality rates did not differ from whites (19). Indeed, a national analysis of mortality risk by race/ethnicity and age by Yan et al. (10) reported that Hispanic patients undergoing dialysis had the lowest mortality rates, followed by black and finally non-Hispanic white patients. The only exception was black patients aged 18–30 years, who had a higher mortality risk than non-Hispanic white patients (10). Mortality differences can be due to younger age of onset of kidney disease and less comorbid for many racial/ethnic minorities, but these differences usually persist when adjusted for these factors, although it attenuates and occasionally disappears in younger ESKD populations (10–14). In this study, the greater mortality risks for Hispanic and Asian and Pacific Islander patients in the territories persisted in the matched analyses, where territory and nonterritory patients of the same racial/ethnic subgroup were comparable on age, comorbidities, and other risk factors, whereas mortality risks for black and white patients in the territories and 50 states did not differ.

The reasons for persistently higher mortality rates in the territories by Hispanics and Asians are complex and may be driven by dietary patterns, health beliefs and health behaviors, access to care, quality of care, insurance status, health care system factors, underlying illness at time of dialysis initiation, and biologic and genetic factors, among other potentially important factors we were unable to capture. For instance, in Puerto Rico, with a median annual household income of only $19,686 (20), the monthly dialysis copay could force many patients to choose between routine dialysis and other life-sustaining decisions. The territories suffer from high poverty rates, ranging from 22% to 57% (42% in Puerto Rico) compared with 12% for the 50 states (21,22), simulating a state of chronic recession, which is associated with adverse health outcomes for chronic diseases (23–26) such as hypertension and stroke, important comorbidities for patients on dialysis. Medicare provider reimbursement is low in the territories, with the lowest in Puerto Rico with a wage index floor at 0.40 (27,28). The wage index is an adjustment factor for Medicare ESKD bundled payment to account for geographic differences in area labor costs in which the dialysis facility is located (29). The territories have the lowest overall dialysis wage indices and the highest projected 2021 dialysis facility payment reductions—this is disconcerting as a lower level of reimbursement per treatment can have an adverse effect on a facility, including higher risk of closure and the resulting decrease in access to care. Thus, the impact of a poor economic state on the overall health infrastructure in Puerto Rico and the other territories may contribute to the higher mortality rate in a highly vulnerable dialysis patient population living in the territories. Another uncertainty is that the Hispanic ethnic and Asian subgroups living in the territories (e.g., Puerto Rican ethnic subgroup [30] and Samoan Asian subgroup [31]) may have specific biologic and genetic profiles that increase their risk for adverse outcomes in those settings. Poor dietary practices are highly prevalent in the territories (32,33). The influence of such factors on CKD progression and CKD-related mortality is not known but could be highly relevant for greater mortality rates among Hispanic and Asian subgroups in the territories.

There are several limitations to our current findings. First, there are differences in payer mix and structure in the territories and the 50 states. For instance, Medicare intermediaries have different interpretations of the 30-month rule for private insurance to be a primary source of ESKD payment for the first 30 months in the 50 states, but in Puerto Rico, that happens for only the first 3 months (9), which may influence facility resources and patient outcomes. Second, matched analyses on the basis of smaller racial/ethnic subgroups (e.g., Puerto Ricans, Mexicans, Chinese, Koreans, etc.) can greatly help us understand the mechanisms (such as genetic or sociocultural factors versus other factors) behind the greater risk of death among Hispanics and Asians in the territories. Lack of data to identify specific Hispanic and Asian subgroups in the USRDS limits our ability to do so. Third, the number of dialysis patients in most of the United States territories outside of Puerto Rico is relatively low, which may bias the relative contribution of Puerto Rico to our analysis. Finally, it is unclear whether the study findings would have been much different if the recent 4 years data (2013–2016) were included.

To our knowledge, this study is the first to document important differences in dialysis mortality for various racial/ethnic groups in the territories versus the 50 states. We found notably higher mortality rates for Hispanic and Asian patients undergoing dialysis in the territories than their counterparts in the 50 states. Mortality risk did not appear to differ between the territories and the 50 states for whites or blacks. Further studies are needed to better understand the influence of issues such as genetic factors, insurance coverage, health infrastructure, health beliefs and behaviors, social networks, and other subtleties in the United States territories that may add critical insights to our observations.

Disclosures

Dr. Agodoa is an employee of the National Institutes of Health. Dr. Harford is the CEO of Atlantis Healthcare (Puerto Rico). Dr. Clark, Dr. Colon, Dr. Flaque, Dr. Georges, Dr. Nee, Dr. Norris, Dr. Rodriguez, Dr. Shen, and Dr. Torre, Dr. Yan, and Mr. Yu have nothing to disclose.

Funding

Dr. Norris is supported by grants from National Institute on Aging (P30AG021684), National Heart, Lung, and Blood Institute (R25HL126145), and National Institutes of Health National Center for Advancing Translation Sciences (NCATS; UL1TR000124). Dr. Shen is supported by grants from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK; K23DK103972) and NCATS (KL2TR000122). Dr. Yan and Mr. Yu are supported by a grant from the NIDDK (1R01DK112008-01A1).

Supplemental Material

This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.03920319/-/DCSupplemental.

Supplemental Table 1. Characteristics of patients undergoing dialysis by the four territories.

Supplemental Table 2. Mortality rates according to dialysis modality at dialysis initiation.

Supplemental Table 3. Demographic and clinical characteristics of the two matched samples for each race/ethnicity.

Supplemental Table 4. Hazard ratios for death separately by Pacific Islanders and other Asians in the original Asian cohort and the second matched sample.

Supplemental Table 5. Hazard ratios for territories versus 50 states in the Cox hazards and Fine–Gray hazards regression.