Sex and Racial/Ethnic Differences in Home Hemodialysis Mortality

Abstract

Key Points

• Women on home hemodialysis have higher 1-year mortality than men, and women and men have comparable survival on long-term follow-up.

• Compared with White patients on home hemodialysis, there was no difference in all-cause mortality for Black patients, Hispanics, or Native Americans.

• Among patients undergoing home hemodialysis, Asians had a lower risk of all-cause mortality than White patients.

Background

Women and minorities constitute substantial portions of the prevalent population of patients with kidney failure. Little is known about sex and racial/ethnic differences in mortality among patients with kidney failure on home hemodialysis in the United States.

Methods

Using the United States Renal Data System, we retrospectively evaluated a cohort of 42,849 patients who started home hemodialysis between January 1, 2005, and December 31, 2015. We examined the association of sex and race/ethnicity with the outcome of all-cause mortality using adjusted Cox proportional hazard models and logistic regression models.

Results

In the study cohort, 40.4% were women, and 57.4% were White. Women on home hemodialysis had higher unadjusted death rates (26.9 versus 22.4 per 100 person-years) compared with men. There was no difference in adjusted all-cause mortality between men and women, but women had an 8% higher adjusted risk of all-cause mortality at 1 year after initiating home hemodialysis (odds ratio 1.08, 95% confidence interval [CI], 1.01 to 1.15). Regarding race/ethnicity, Hispanic, White, and Black patients had higher unadjusted death rates compared with Asians and Native Americans (25.1 versus 24.8 versus 23.2 versus 17.4 versus 16.6 per 100 person-years). There was no difference in adjusted all-cause mortality in Black, Hispanic, and Native Americans compared with White patients, while Asians had a lower risk of all-cause mortality than did White patients (hazard ratio, 0.81; 95% CI, 0.72 to 0.92). There was no difference in adjusted 1-year mortality for Asian, Black, Hispanic, and Native American patients compared with White patients.

Conclusions

Among patients undergoing home hemodialysis, women have higher 1-year mortality than men, and women and men have comparable survival on long-term follow-up after adjusting for other covariates. Compared with White patients, there was no difference in adjusted survival on long-term follow-up for Black patients, Hispanics, or Native Americans, while Asians had better survival. Our results suggest the need for population-wide strategies to overcome differences in home hemodialysis care.

Introduction

Kidney failure is a global health problem, and patients with kidney failure have a high annual mortality of 20%–25%.¹ Home hemodialysis, synonymous to more frequent dialysis in the United States, provides better patient outcomes, particularly superior quality of life, reductions in left ventricular index, better control of hypertension, improved phosphate control, and opportunities for employment and rehabilitation compared with in-center hemodialysis.^2–4

Home hemodialysis leads to improved survival in patients with kidney failure compared with conventional in-center hemodialysis.^5,6 Sex and racial/ethnic differences in mortality for patients receiving in-center hemodialysis are well known.^7–10 For instance, although women in the general population have longer life expectancy than do men, women with kidney failure lose this survival advantage once they initiate in-center hemodialysis.¹¹ Black patients have a 13%–45% lower mortality than do White patients receiving in-center hemodialysis.^7,9–12 However, little is known concerning survival differences by sex and race/ethnicity in the home hemodialysis population.

Socioeconomic factors differ between patients receiving home dialysis and those receiving in-center hemodialysis. Notably, home hemodialysis patients tend to be wealthier, more educated, and more likely to have received predialysis nephrology care than those receiving in-center hemodialysis.^13,14 Many of these characteristics are less prevalent among women and minorities with kidney failure, suggesting the need to study sex and racial/ethnic mortality differences after accounting for them in the home hemodialysis population.¹⁵ We used the United States Renal Data System (USRDS) database, a national registry of patients receiving chronic dialysis, to compare the mortality risk by sex and race/ethnicity after adjusting for several covariates including socioeconomic factors in patients who initiated home hemodialysis between January 1, 2005, and December 31, 2015.

Methods

Data Sources and Study Population

We used the USRDS database, a national registry of patients receiving chronic dialysis, to perform an observational retrospective cohort study; the registry contains information from the Medical Evidence Form of the Centers for Medicare and Medicaid Services (CMS; form CMS-2728).¹⁶ Forty-two thousand eight hundred and forty-nine patients who initiated home hemodialysis at any time between January 1, 2005, and December 31, 2014, were included. The sample excluded patients with no baseline CMS-2728 form, those lacking any records in the treatment history file, those dying on the same day as their first service, and those with missing data on sex. Figure 1 illustrates the study cohort derivation. Because the data were deidentified, the University of Cincinnati Institutional Review Board committee classified the study into the exempt category.

Figure 1.

Cohort selection flow diagram.

The CMS-2728 form was used to obtain information on demographics, body mass index, ethnicity, comorbidities (congestive heart failure, hypertension, and diabetes mellitus), predialysis nephrology care, history of unemployment, and poor functional status defined by any of three comorbid conditions as specified in the form CMS-2728—inability to ambulate, inability to transfer, or need of assistance with daily activities.¹³ The patients' file was used to ascertain information on the year of dialysis initiation, race, sex, age, cause of ESKD, and date of death. Race and ethnicity were combined into one variable, categorized as Hispanic, Non-Hispanic Asian, non-Hispanic Black, non-Hispanic Native American, and non-Hispanic White. Predialysis nephrology care was grouped into none, ≤12 months, >12 months, and unknown. The treatment history file was used to obtain information on dialysis modalities during the study period and history of kidney transplants. The residence file was used to obtain information on patients' zip codes of residence. These zip codes were combined with zip code level data from the US Census Bureau American Community Survey 5-year estimates from 2007 to 2011 to determine neighborhood socioeconomic status, which were defined as the percent of zip code residents living below the federal poverty level and grouped into approximate quartiles: I (≤5.1%), II (5.2%–9.6%), III (9.7%–16.3%), IV (16.4% or more), and unknown.^17,18 Rurality of the neighborhood was determined using the rural-urban commuting area (RUCA) code version 2.0 and grouped into four categories: metropolitan (RUCA 1.0–3.9), micropolitan (RUCA 4.0–6.0), rural (RUCA 7.0–10.6), and unknown.^18,19 Groups were created on the basis of clinical relevance with patients with unavailable information on covariates categorized into a “missing” group for that covariate, as shown in Table 1. The covariates for adjusted analyses were chosen on the basis of their known clinical relevance (Table 2).

Table 1.

Characteristics of incident home hemodialysis patients from 2005 to 2015, overall and by sex

Patient Characteristics	All, N=42,849	Men, N=25,547, 59.6%	Women, N=17,302, 40.4%	P Value
Age (yr)	60.1±16.0	59.5±16.6	60.9±16.5	<0.001
Age group (yr)				<0.001
Younger than 18	0.2	0.2	0.3
18–29	3.2	3.0	3.5
30–39	8.3	8.4	8.2
40–49	14.2	15.0	13.1
50–59	20.4	21.5	18.8
60–69	23.3	23.7	22.7
70–79	18.6	17.8	19.8
80 and older	11.7	10.4	13.7
Male patients	59.6
Race/ethnicity				<0.001
Asian	2.9	2.9	2.9
Black	29.5	27.3	32.9
Hispanic	9.5	9.3	9.8
Native American	0.4	0.4	0.4
White	57.4	59.9	53.8
Unknown/Other	0.2	0.2	0.2
Region				<0.001
Midwest	44.5	42.1	47.9
Northeast	10.5	11.1	9.7
South	33.3	34.3	31.9
West	11.7	12.5	10.5
Unknown	0.0	0.0	0.0
Poverty (nonmissing)	2.5±1.1	2.5±1.1	2.6±1.1	<0.001
Poverty quartiles				<0.001
≤5.1%	24.1	25.1	22.6
5.2%–9.6%	25.3	25.7	24.6
9.7%–16.3%	24.6	25.1	24.0
≥16.4%	25.0	22.9	28.0
Missing	1.1	1.2	0.9
Rurality				<0.001
Metropolitan	83.6	82.5	85.3
Micropolitan	7.0	7.3	6.6
Rural	8.3	9.2	7.0
Missing	1.1	1.1	1.1
Body mass index (kg/m2)	30.5±8.8	30.0±8.2	31.3±9.6	<0.001
Body mass index (kg/m 2)				<0.001
<18.5	3.2	2.8	3.8
18.5 to <25	24.7	25.3	23.9
25 to <30	24.7	27.1	21.2
≥30	42.2	39.9	45.7
Missing	5.2	5.0	5.4
Congestive heart failure	25.2	24.1	26.8	<0.001
Hypertension	81.0	81.4	80.4	0.006
Diabetes mellitus	45.6	44.9	46.6	<0.001
Poor functional status	14.9	12.7	18.1	<0.001
Unemployed	19.4	18.3	20.9	<0.001
Nursing home	10.3	8.6	12.7	<0.001
ESKD cause				<0.001
Diabetes mellitus	38.8	38.3	39.5
Hypertension/large vessel disease	28.0	28.5	27.3
Malignancy	2.3	2.6	1.8
Cystic/hereditary	3.4	3.4	3.4
Secondary glomerulonephritis/vasculitis	2.0	1.2	3.2
Glomerulonephritis	8.2	9.0	7.0
Interstitial nephritis/pyelonephritis	2.2	2.1	2.2
Other/missing	15.2	14.9	15.6
Prior nephrology care				<0.001
None	17.7	17.8	17.4
≤12 mo	25.0	25.7	24.1
>12 mo	20.0	20.9	18.7
Unknown	37.3	35.6	39.9

Values are percentage or mean ± SD.

Table 2.

Characteristics of incident home hemodialysis patients from 2005 to 2015 by race/ethnicity

Patient Characteristics	Asian ,N=1233, 2.9%	Black ,N=12,658, 29.5%	Hispanic ,N=4081, 9.5%	Native American ,N=181, 0.4%	White ,N=24,611, 57.4%	Unknown/Other, N=85, 0.2%	P Value
Age (yr)	59.4±7.1	57.0±15.9	58.3±16.8	53.2±13.6	62.0±15.6	54.8±16.9	<0.001
Age group (yr)							<0.001
Younger than 18	0.2	0.2	0.2	0.0	0.2	1.2
18–29	4.9	3.6	5.2	2.8	2.6	7.1
30–39	9.2	12.0	10.6	13.8	6.0	7.1
40–49	14.9	18.2	14.5	22.1	12.1	21.2
50–59	18.5	21.5	19.6	30.9	20.0	22.4
60–69	20.3	20.5	21.1	18.2	25.3	20.0
70–79	19.1	15.8	18.5	9.9	20.1	16.5
80 and older	13.0	8.4	10.3	2.2	13.7	4.7
Male patients	59.5	55.1	58.4	58.0	62.2	54.1	<0.001
Region							<0.001
Midwest	34.7	45.5	34.7	27.1	46.2	42.4
Northeast	8.4	8.3	8.3	6.6	12.1	11.8
South	18.3	42.2	32.8	36.5	29.6	25.9
West	38.5	4.1	24.3	29.8	12.1	20.0
Unknown	0	0	0	0	0	0
Poverty (nonmissing)	2.2±1.1	3.1±1.0	2.9±1.1	3.1±1.0	2.2±1.0	2.4±1.1	<0.001
Poverty quartiles							<0.001
≤5.1%	35.0	10.3	14.0	9.4	32.4	27.1
5.2%–9.6%	26.4	17.3	20.7	16.6	30.1	27.1
9.7%–16.3%	24.2	23.9	27.3	32.0	24.5	24.7
≥16.4%	13.5	47.3	37.0	40.3	11.9	18.8
Missing	1.0	1.2	1.0	1.7	1.0	2.4
Rurality							<0.001
Metropolitan	93.9	91.1	90.8	53.0	78.3	81.2
Micropolitan	2.5	4.3	4.0	12.7	9.1	9.4
Rural	1.8	3.9	3.3	33.2	11.5	7.1
Missing	1.8	0.7	1.9	1.1	1.1	2.4
Body mass index (kg/m2)	26.7±6.8	30.8±9.1	29.8±8.2	32.0±8.0	30.7±8.8	29.6±10.1	<0.001
Body mass index (kg/m 2)							<0.001
<18.5	5.5	3.5	2.5	2.8	3.1	8.2
18.5 to <25	41.3	23.8	27.4	15.5	24.0	23.5
25 to <30	25.8	24.0	26.5	21.6	24.7	31.8
≥30	23.6	43.4	39.1	54.7	43.0	30.6
Missing	3.8	5.3	4.6	5.5	5.2	5.9
Congestive heart failure	19.8	23.8	24.5	20.4	26.4	25.9	<0.001
Hypertension	81.9	84.2	82.3	84.5	79.1	70.6	<0.001
Diabetes mellitus	47.3	43.1	54.9	59.1	45.2	42.4	<0.001
Poor functional status	11.0	15.5	15.8	5.5	14.7	23.5	<0.001
Unemployed	19.1	26.7	26.2	31.5	14.4	15.3	<0.001
Nursing home	6.7	12.3	10.2	1.7	9.5	12.9	<0.001
ESKD cause							<0.001
Diabetes mellitus	44.0	36.8	50.6	59.1	37.4	36.5
Hypertension/large vessel disease	24.2	37.2	22.3	12.2	24.5	27.1
Malignancy	1.3	1.0	1.4	0.0	3.2	0.0
Cystic/hereditary	2.7	1.4	2.3	3.9	4.7	4.7
Secondary glomerulonephritis/vasculitis	1.6	2.3	2.5	2.2	1.8	1.2
Glomerulonephritis	11.1	7.1	7.7	11.1	8.7	5.9
Interstitial nephritis/pyelonephritis	1.8	0.6	1.4	1.1	3.1	0.0
Other/missing	13.3	13.6	11.8	10.5	16.7	24.7
Prior nephrology care							<0.001
None	17.8	19.7	23.1	18.2	15.7	15.3
≤12 mo	25.7	22.4	24.7	27.1	26.4	20.0
>12 mo	18.8	14.8	15.3	21.6	23.5	24.7
Unknown	37.7	43.0	37.0	33.2	34.4	40.0

Values are percentage or mean ± SD.

The outcome of interest was all-cause mortality while patients were on home hemodialysis. The major predictors were sex and race/ethnicity.

Statistical Analyses

Summary statistics are presented as percentages for categorical variables and mean±SD for continuous variables. Chi-squared tests and Wilcoxon tests were used to assess simple bivariate relationships between variables. Changes in our sample across years of incident home hemodialysis were assessed linearly across years for patients' age, sex, race, region, local poverty level, and rurality, each individually. The outcome of all-cause mortality, and each patient's time under observation, included time during home hemodialysis only, excluding time during interruptions to home hemodialysis and was censored for transplant, recovered function, last known home hemodialysis, and January 31, 2016. Cox proportional hazards models were nonparsimonious and included covariates, in addition to those for sex and race/ethnicity, age, poverty quartile, rurality, region, year of first home hemodialysis, number of years between first dialysis and first home hemodialysis, body mass index, comorbidities, predialysis nephrology care, unemployment history, and functional status. These models also included a time-varying covariate representing the accumulated number of months during which home hemodialysis was interrupted by the use of a different modality. In addition, covariates associated with 1-year mortality were assessed using multivariable logistic regression models applied to patients who died within 1 year of observation time versus those who survived at least 1 year, excluding those who lacked 1 year of follow-up but who did not die during their follow-up. These models also excluded patients starting home hemodialysis during 2015 or later, to ensure the availability of 1-year follow-up time.

We also estimated Kaplan-Meier curves of mortality during observation time as defined above, due to sex and race, unadjusted for other covariates, and compared curves using a log-rank test. All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC).

Results

Baseline Demographics and Clinical Characteristics

The mean age of cohort members was 60±16 years (SD), with 40.4% of them being women. Regarding racial/ethnic categories, 57.4% were White, 29.5% were Black, 9.5% were Hispanic, 2.9% were Asian, and 0.4% were Native American. The mean time between initiating dialysis and home hemodialysis was 39 months. Only 46% of patients initiated home dialysis within 12 months of their first dialysis date. The mean of total time on home hemodialysis was 16.2 months. The mean body mass index was 30.5±8.8 kg/m². Most of the patients resided in metropolitan areas (83.6%) and most often received home hemodialysis in the midwestern region (44.5%). Diabetes (38.8%) was the most common cause of kidney failure among home hemodialysis patients, followed by hypertension (28.0%) (Table 1).

Table 1 shows the characteristics of home hemodialysis patients by sex. Women were older than men (mean age 60.9±16.5 years versus 59.5±16.6 years). Women compared with men were more likely to be Black (32.9% versus 27.3%) and less likely to be White (53.8% versus 59.9%). Women compared with men were more likely to have comorbidities of diabetes (46.6% versus 44.9%), congestive heart failure (26.8% versus 24.1%), and poor functional status (18.1% versus 12.7%). Women less frequently received predialysis nephrology care (42.7% versus 46.6%). A higher proportion of women on home hemodialysis resided in midwestern region (47.9% versus 42.1%) and in metropolitan areas (85.3% versus 82.5%) than did men. A lower proportion of women than men resided in zip codes in the highest socioeconomic status quartile (22.6% versus 25.1%). Women were more likely to be unemployed (20.9% versus 18.3%).

In the home hemodialysis population, Native American patients were youngest (mean age 53.2±13.6 years), and White patients were oldest (mean age 62.0±15.6 years). The male proportion was highest in White patients (62%) and lowest in Black patients (55%). The most common cause of kidney failure for Native Americans was diabetes (59%), while for Black patients, it was diabetes and hypertension/large vessel disease (37% each). Asians and Native Americans were less likely to have the comorbidities of congestive heart failure (20%, 20%) than were members of other racial/ethnic groups and the least likely to have poor functional status (11%, 6%), respectively. The most common region in which Asians lived was the western region (39%), while Native Americans lived most in the southern region (37%). Black patients, Hispanics, and White patients lived most often in the midwestern region (46%, 35%, and 46% respectively). Asians and White patients were more likely to reside in zip codes in the lowest quartile of poverty (<5.1%) than were members of other racial groups. Asians were more likely to live in metropolitan areas than were White patients (94% versus 78%) (Table 2).

The mean age at home hemodialysis initiation decreased slightly over the years from 2005 to 2015 (P = 0.003). Across all years, 59.6% of patients were male, and this proportion did not change significantly between 2005 and 2015. The patient racial mix changed across years (P<0.001), with the proportion of patients who were Hispanic increasing from 2005 to 2015, the White proportion decreasing, and the Black proportion remaining fairly flat (Supplemental Figure 1). The greatest number of patients on home hemodialysis was in the midwestern region (44.5%) and the fewest in the northeastern region (10.5%). Across the years the mix of regions changed (P<0.001), with the proportion of patients residing in the northeastern region remaining fairly constant, while increasing in the southern and western regions, and decreasing in the midwestern region. There was an increase in the proportion of home hemodialysis patients residing in the higher poverty quartile from 2005 to 2015. Although most home hemodialysis patients resided in metropolitan areas (83.6% overall), there was an increase in proportion of patients from rural and micropolitan areas from 2005 to 2015 (Supplemental Figure 2).

Sex and Racial/Ethnic Differences in Mortality in Home Hemodialysis

On home hemodialysis, women compared with men had a higher death rate (26.9 versus 22.3 deaths per 100 person-years). In the Cox model adjusted for socioeconomic factors and other covariates, there was no difference in all-cause mortality between men and women (hazard ratio [HR] 0.986, 95% confidence interval [CI], 0.950 to 1.024) (Figure 2, Table 3). The adjusted risk of dying before 1 year of home hemodialysis was slightly higher in women than in men (odds ratio [OR] 1.08, 95% CI, 1.01 to 1.15) (Figure 3, Table 4). Hispanic, Black, and White patients had higher death rates per 100 person-years than did Asian and Native American patients (Hispanics, 25.1; White patients, 24.8; Black patients, 23.2; Asians, 17.4; and Native Americans, 16.6). In the adjusted Cox model, Asians had a lower likelihood of death than did White patients (HR, 0.82; 95% CI, 0.72 to 0.92), while Black, Hispanic, and Native American had risks of death not significantly different from that of White patients (Figure 2, Table 3). In the adjusted logistic multivariate regression model, there was no difference in 1-year mortality among different races and ethnicities (Figure 3, Table 4).

Figure 2.

Main effects time-to-event models showing sex and race/ethnicity associated with mortality, censoring for transplants or recovered function, in home hemodialysis patients. *The model predicts time to death while on home hemodialysis only, censored for transplant and recovered function, and includes a time-varying covariate for cumulative months of prior interruptions to home hemodialysis for those patients who had interruptions. HRs are adjusted covariates of year of first home hemodialysis, number of years between first dialysis and first home hemodialysis, age, sex and race/ethnicity, region, poverty quartile, rurality, body mass index, congestive heart failure, hypertension, unemployment history, functional status, cause of ESKD, and predialysis nephrology care.

Table 3.

Hazard ratios from a multivariable Cox model predicting death while on home hemodialysis

Variable	Hazard Ratio (95% CI)
Year of first home hemodialysis	1.00 (0.99 to 1.00)
Years between first dialysis and home hemodialysis	1.00 (0.99 to 1.01)
Age group (yr) (P<0.001)
Younger than 18	0.30 (0.14 to 0.65)
18–29	0.27 (0.22 to 0.34)
30–39	0.36 (0.32 to 0.40)
40–49	0.59 (0.54 to 0.63)
50–59	Reference
60–69	1.56 (1.48 to 1.65)
70–79	2.43 (2.29 to 2.58)
80 and older	3.54 (3.31 to 3.78)
Cumulative months of prior interruption	1.03 (1.02 to 1.03)
Female	0.99 (0.95 to 1.02)
Race/ethnicity (P=0.01)
Asian	0.81 (0.72 to 0.92)
Black	0.96 (0.92 to 1.01)
Hispanic	1.01 (0.95 to 1.08)
Native American	1.17 (0.87 to 1.57)
White	Reference
Unknown/other	0.95 (0.68 to 1.34)
Region (P<0.001)
Midwest	1.36 (1.31 to 1.42)
Northeast	0.80 (0.75 to 0.86)
South	Reference
West	0.68 (0.63 to 0.73)
Unknown	0.00 (0.00 to 0.00)
Poverty quartiles (P<0.001)
≤5.1%	Reference
5.2%–9.6%	1.02 (0.97 to 1.07)
9.7%–16.3%	1.07 (1.01 to 1.13)
≥16.4%	1.15 (1.09 to 1.22)
Missing	1.14 (0.94 to 1.38)
Rurality (P<0.001)
Metropolitan	Reference
Micropolitan	0.76 (0.70 to 0.82)
Rural	0.74 (0.69 to 0.79)
Missing	1.11 (0.92 to 1.33)
Body mass index (kg/m 2) (P<0.001)
<18.5	1.18 (1.07 to 1.30)
18.5 to <25	Reference
25 to <30	0.88 (0.83 to 0.92)
≥30	0.83 (0.79 to 0.87)
Missing	0.89 (0.79 to 1.00)
Congestive heart failure	1.27 (1.22 to 1.33)
Hypertension	0.90 (0.86 to 0.95)
Poor functional status	1.47 (1.40 to 1.54)
Unemployed	1.05 (0.99 to 1.10)
ESKD cause (P<0.001)
Diabetes mellitus	Reference
Hypertension/large vessel disease	0.81 (0.78 to 0.85)
Malignancy	0.95 (0.84 to 1.06)
Cystic/hereditary/congenital	0.45 (0.38 to 0.52)
Secondary glomerulonephritis/vasculitis	0.68 (0.57 to 0.81)
Glomerulonephritis	0.56 (0.51 to 0.61)
Interstitial nephritis/pyelonephritis	0.70 (0.61 to 0.81)
Other/missing	0.79 (0.75 to 0.84)
Prior nephrology care (P<0.001)
None	Reference
≤12 mo	0.83 (0.78 to 0.88)
>12 mo	0.67 (0.63 to 0.71)
Unknown	1.06 (1.01 to 1.12)

The model predicts time to death while on home hemodialysis only, censored for transplant and recovered function, and includes a time-varying covariate for cumulative months of prior interruptions to home hemodialysis for those patients who had interruptions. CI, confidence interval.

Figure 3.

Logistic regression models showing unadjusted and adjusted 1-year mortality by sex and by race/ethnicity. *The models predict death before 1 year versus survival for a contiguous year on home hemodialysis, excluding those who lacked 1 year of follow-up but who did not die during their follow-up. Incident year 2015 is not included. ORs are adjusted covariates of year of first home hemodialysis, number of years between first dialysis and first home hemodialysis, age, sex and race/ethnicity, region, poverty quartile, rurality, body mass index, congestive heart failure, hypertension, unemployment history, functional status, cause of ESKD, and predialysis nephrology care.

Table 4.

Results from a multivariable logistic model predicting death while on home hemodialysis within 1 year

Variable	Odds Ratio (95% CI)
Year of first home hemodialysis	1.01 (0.99 to 1.02)
Years between first dialysis and home hemodialysis	0.99 (0.98 to 1.00)
Age group (yr) (P<0.001)
Younger than 18	0.31 (0.11 to 0.88)
18–29	0.26 (0.19 to 0.35)
30–39	0.29 (0.24 to 0.35)
40–49	0.52 (0.46 to 0.59)
50–59	Reference
60–69	1.85 (1.68 to 2.03)
70–79	3.54 (3.21 to 3.91)
80 and older	6.30 (5.60 to 7.10)
Female patients	1.08 (1.01 to 1.15)
Race/ethnicity (P=0.04)
Asian	0.86 (0.70 to 1.06)
Black	1.05 (0.96 to 1.14)
Hispanic	1.10 (0.98 to 1.24)
Native American	1.63 (1.00 to 2.66)
White	Reference
Unknown/other	0.55 (0.25 to 1.20)
Region (P<0.001)
Midwest	1.86 (1.73 to 2.00)
Northeast	0.73 (0.64 to 0.83)
South	Reference
West	0.58 (0.52 to 0.66)
Unknown	0.00 (<0.01 to >999.99)
Poverty quartiles (P<0.001)
≤5.1%	Reference
5.2%–9.6%	1.07 (0.98 to 1.17)
9.7%–16.3%	1.15 (1.05 to 1.27)
≥16.4%	1.34 (1.21 to 1.48)
Missing	1.26 (0.91 to 1.75)
Rurality (P<0.001)
Metropolitan	Reference
Micropolitan	0.58 (0.51 to 0.67)
Rural	0.57 (0.51 to 0.65)
Missing	1.36 (0.99 to 1.87)
Body mass index (kg/m2) (P<0.001)
<18.5	1.29 (1.07 to 1.55)
18.5 to <25	Reference
25 to <30	0.86 (0.79 to 0.94)
≥30	0.74 (0.68 to 0.81)
Missing	0.82 (0.68 to 1.00)
Congestive heart failure	1.42 (1.32 to 1.53)
Hypertension	0.82 (0.75 to 0.89)
Poor functional status	2.09 (1.91 to 2.28)
Unemployed	1.13 (1.04 to 1.24)
ESKD cause (P<0.001)
Diabetes mellitus	Reference
Hypertension/large vessel disease	0.73 (0.67 to 0.79)
Malignancy	0.88 (0.72 to 1.09)
Cystic/hereditary/congenital	0.31 (0.23 to 0.41)
Secondary glomerulonephritis/vasculitis	0.62 (0.46 to 0.83)
Glomerulonephritis	0.45 (0.38 to 0.53)
Interstitial nephritis/pyelonephritis	0.63 (0.50 to 0.80)
Other/missing	0.77 (0.70 to 0.86)
Prior nephrology care (P<0.001)
None	Reference
≤12 mo	0.68 (0.62 to 0.75)
>12 mo	0.50 (0.45 to 0.56)
Unknown	1.06 (0.97 to 1.17)

The models predict death before 1 year versus survival for a contiguous year on home hemodialysis, excluding those who lacked 1 year of follow-up but who did not die during their follow-up. Incident year 2015 is not included. CI, confidence interval.

Table 3 shows the adjusted multivariable Cox model–based HRs for death while on home hemodialysis. Compared with patients aged 50–60 years, the likelihood of death for adults increased with increasing age at time of home hemodialysis initiation: age younger than 18 years (HR, 0.30; CI, 0.14 to 0.650), age 18–29 (HR, 0.27; 95% CI, 0.22 to 0.34), age 30–39 years (HR, 0.36; 95% CI, 0.32 to 0.40), 40–49 years (HR, 0.59; 95% CI, 0.54 to 0.63), 60–69 years (HR, 1.56; 95% CI, 1.48 to 1.65), 70–79 years (HR, 2.43; 95% CI, 2.29 to 2.58), and older than 80 years (HR, 3.54; 95% CI, 3.31 to 3.78). Compared with patients residing in metropolitan areas, patients living in rural areas (HR, 0.74; 95% CI, 0.69 to 0.79) and micropolitan areas (HR, 0.76; 95% CI, 0.70 to 0.82) had a lower likelihood of death. Compared with patients in southern areas, patients in midwestern areas had a higher likelihood of death (HR, 1.36; 95% CI, 1.31 to 1.42), and patients in northeastern (HR, 0.80; 95% CI, 0.75 to 0.86) and western areas (HR, 0.68; 95% CI, 0.63 to 0.73) had a lower likelihood of death. Compared with patients residing in the least poor poverty quartile (<5.1%), patients residing in the highest poverty quartile (>16.4%) had a higher hazard of dying (HR, 1.15; 95% CI, 1.09 to 1.22). Compared with patients with kidney failure due to diabetes, patients with kidney failure due to cystic disease (HR, 0.45; 95% CI, 0.38 to 0.52), glomerulonephritis (HR, 0.56; 95% CI, 0.51 to 0.61), hypertension (HR, 0.81; 95% CI, 0.78 to 0.85), and interstitial nephritis (HR, 0.70; 95% CI, 0.61 to 0.81) all had lower hazards of dying. Compared with patients with no predialysis care, patients receiving predialysis care ≤12 months (HR, 0.83; 95% CI, 0.78 to 0.88) and >12 months (HR, 0.67; 95% CI, 0.63 to 0.71) had lower likelihoods of death on home dialysis. There was no difference in mortality regarding employment status. Table 4 shows the ORs from the adjusted logistic regression model predicting 1-year mortality for home hemodialysis patients.

Figure 4 shows Kaplan-Meier curves of survival until death while on hemodialysis, censoring for transplant or recovered function, separated by sex and by race. Women had lower survival than did men (log rank P<0.001). Races overall differed significantly (P<0.001) with the highest death rates for White, Black, and Hispanic patients and the lowest rates for Asian and Native American patients.

Figure 4.

Kaplan-Meier curves of survival while on home hemodialysis. (A) by sex and (B) by race. *Survival to death while on home hemodialysis only, censored for transplant and recovered function.

Discussion

Our study reports sex and racial/ethnic differences in mortality in patients who began home hemodialysis from 2005 through 2015 in the United States. Women had higher adjusted 1-year mortality than did men; however, they had comparable survival on long-term follow-up after adjusting for socioeconomic status and other covariates in the home dialysis population. There were no racial/ethnic differences in adjusted mortality in the home dialysis population in the long-term follow up, except for Asians who had lower mortality than did White patients. Finally, residing in midwestern geographical region was associated with a higher adjusted risk of mortality in the home hemodialysis population.

Women were more likely than were men to die within 1 year on home hemodialysis. Several potential mechanisms might explain this survival disadvantage of women over men at 1 year. First, women are older than men in our home dialysis population, and mortality increases with advancing age in patients with kidney failure; however, the sex difference in 1-year survival persisted even after adjusting for age.¹¹ Second, although use of arteriovenous access is associated with lower mortality, fewer women than men initiate home dialysis with arteriovenous access.^15,20 Third, duration and frequency of home dialysis may contribute to this loss of survival advantage in women, especially in the first year. In the Frequent Hemodialysis Network Nocturnal trial, the more frequent dialysis was associated with higher small solute clearance, improved phosphate control, reduced extracellular fluid volume, and regression of left ventricular hypertrophy compared with conventional hemodialysis.²¹ The differences between duration and number of home sessions by sex and its effect on mortality need further exploration.

The median survival time did not differ between men and women at long-term follow-up. This is in contrast with the general population where women have a survival advantage over men but is similar to what has been reported in the in-center hemodialysis population, where this survival advantage is not sustained.^22–24 So why do women not live longer than do men once they initiate home dialysis? While the reasons are not clear, we speculate predialysis health differences may be a factor. For example, the comorbidity of diabetes mellitus results in excess mortality for women among patients on dialysis.^22,25 In this study, women compared with men received less predialysis nephrology care and had higher rates of diabetes, congestive heart failure, and poor functional status. In addition, lower dialysis doses and inferior treatment of anemia and mineral bone disease that women receive compared with men may play a role.^26–28 Finally, dialysis initiation is associated with premature menopause in women, which may result in loss of the cardiac-related protective effects of endogenous estrogen and possibly higher cardiovascular mortality.^29,30

In our home hemodialysis cohort, there was no difference in the risk of death between Black, Hispanic, and Native American patients, compared with White patients, after adjusting for demographics, socioeconomic, and other clinical factors. This contrasts with previously reported outcomes in the in-center hemodialysis population, where Black, Hispanic, and Native American patients have a lower mortality than White patients.¹¹ It is speculated that older age, higher cardiovascular comorbidities, and fewer coexisting illnesses in White patients contribute to their lower survival in in-center hemodialysis population. Another study reported that, contrary to our observation, among home hemodialysis patients, only Black patients had a lower risk of death than did White patients; however, the report was limited by a small cohort and very small number of deaths among other races.³¹ In the absence of demographic and clinical covariates explaining the comparable mortality, we speculate that differences in prescription patterns of home dialysis treatment, differences in social support, and inadequate access to health care may explain comparable survival for White patients and minorities (except Asians). We propose a built-in additional support infrastructure to overcome these potential barriers and deliver home hemodialysis therapies to these vulnerable groups, especially to achieve the targets of the Advancing American Kidney Health Initiative. For example, structured modality education, staff-assisted, and self-care dialysis could bridge the racial and ethnic gap of home hemodialysis utilization and possibly contribute to improvement in survival.

Our study shows geographical differences in mortality among patients on home hemodialysis, with the highest mortality rate in the Midwest, an important finding given that the number of patients on home hemodialysis was largest in the midwestern region. Our potential explanation for the findings is reduced access to health care in midwestern regions. Previous studies have reported increased risk of death with longer travel times in both hemodialysis and peritoneal dialysis patients,^32,33 although causality for these findings remains unclear in these associations. However, to our knowledge, the association of geographical region with outcomes in home hemodialysis patients in the United States has not been addressed. Future studies are needed to examine practice patterns, access to care, and need for remote monitoring and may lead to possible explanations for these poorer outcomes.

A significant strength of our study is that it includes all patients with kidney failure in the United States who began home hemodialysis 2005–2015, thus providing us with information about differences in mortality rates by sex and race/ethnicity. In addition, we considered various socioeconomic determinants including socioeconomic status, geographical region, rurality, and predialysis care, which make our study unique. The limitations of our study include the observational design, which precludes the determination of causality. In addition, variability in the quality and completeness of the data recorded on form 2728, including the high rate of missing data for predialysis nephrology care, is an inherent limitation when using USRDS data. Information of practice patterns, and patient-level variables of health literacy and duration and frequency of home dialysis treatment sessions, which may affect home hemodialysis, are not captured in the database and were not available for our analysis.

In conclusion, our study demonstrates that women were less likely to survive for 1 year on home hemodialysis than were men, but men and women had comparable survival on long-term follow-up. There were no significant differences in mortality across race/ethnicity except that Asians had a better survival than did White patients on long-term follow-up. This study provides an opportunity to develop interventions that may mitigate differences in mortality in women and minorities in the home hemodialysis population.

Disclosures

N. Gupta reports the following: research funding: Quanta; honoraria: UptoDate. S. Shah reports the following: honoraria: ACKD Journal, Otsuka Pharma, Vifor Pharma and speakers bureau: AstraZeneca. C.V. Thakar reports the following: research funding: Natera Inc; honoraria: LEK, NKF, NxStage, Teladoc, Vifor; and speakers bureau: NxStage. All the remaining authors have nothing to disclose.

Funding

S.S. is supported by National Institutes of Health K23 career development award 1K23HL151816-01A1.

Acknowledgments

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funders of the study had no role in study design; collection, analysis, and interpretation of data; writing the report; and the decision to submit the report for publication. The results presented in this paper have not been published previously in whole or part, except in abstract format. The data reported here have been supplied by the United States Renal Data System (USRDS). The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or interpretation of the US Government.

Author Contributions

S. Shah conceptualized the study; K. Meganathan and S. Shah were responsible for data curation; S. Shah was responsible for funding acquisition; A.L. Christianson, A.C. Leonard, and S. Shah were responsible for investigation; A.L. Christianson, N. Gupta, A.C. Leonard, K. Meganathan, S. Shah, and C.V. Thakar were responsible for methodology; N. Gupta and S. Shah were responsible for project administration; S. Shah and C.V. Thakar were responsible for resources and provided supervision; A.C. Leonard, K. Meganathan, S. Shah, and C.V. Thakar were responsible for visualization; S. Shah wrote the original draft; A.L. Christianson, N. Gupta, A.C. Leonard, K. Meganathan, S. Shah, and C.V. Thakar reviewed and edited the manuscript; A.L. Christianson was responsible for formal analysis; A.L. Christianson, A.C. Leonard, and K. Meganathan were responsible for software; and A.C. Leonard was responsible for validation.

Data Sharing Statement

All data are included in the manuscript and/or supplemental material.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/KN9/A251.

Supplemental Figure 1. The proportion of home hemodialysis patients over time in groups defined by (A) age, (B) sex, and (C) race/ethnicity.

Supplemental Figure 2. The proportion of home hemodialysis patients over time in groups defined by (A) geographical region, (B) neighborhood poverty, and (C) neighborhood rurality.