Racial and Neighborhood-Level Disparities in COVID-19 Incidence among Patients on Hemodialysis in New York City

Abstract

Background 

The coronavirus disease 2019 (COVID-19) pandemic has disproportionately affected socially disadvantaged populations. Whether disparities in COVID-19 incidence related to race/ethnicity and socioeconomic factors exist in the hemodialysis population is unknown.

Methods 

Our study involved patients receiving in-center hemodialysis in New York City. We used a validated index of neighborhood social vulnerability, the Social Vulnerability Index (SVI), which comprises 15 census tract–level indicators organized into four themes: socioeconomic status, household composition and disability, minority status and language, and housing type and transportation. We examined the association of race/ethnicity and the SVI with symptomatic COVID-19 between March 1, 2020 and August 3, 2020. COVID-19 cases were ascertained using PCR testing. We performed multivariable logistic regression to adjust for demographics, individual-level social factors, dialysis-related medical history, and dialysis facility factors.

Results 

Of the 1378 patients on hemodialysis in the study, 247 (17.9%) developed symptomatic COVID-19. In adjusted analyses, non-Hispanic Black and Hispanic patients had significantly increased odds of COVID-19 compared with non-Hispanic White patients. Census tract–level overall SVI, modeled continuously or in quintiles, was not associated with COVID-19 in unadjusted or adjusted analyses. Among non-Hispanic White patients, the socioeconomic status SVI theme, the minority status and language SVI theme, and housing crowding were significantly associated with COVID-19 in unadjusted analyses.

Conclusions 

Among patients on hemodialysis in New York City, there were substantial racial/ethnic disparities in COVID-19 incidence not explained by neighborhood-level social vulnerability. Neighborhood-level socioeconomic status, minority status and language, and housing crowding were positively associated with acquiring COVID-19 among non-Hispanic Whites. Our findings suggest that socially vulnerable patients on dialysis face disparate COVID-19–related exposures, requiring targeted risk-mitigation strategies.

Introduction

The novel coronavirus disease 2019 (COVID-19) pandemic has disproportionately affected socially disadvantaged populations, including Black and Hispanic individuals,^1–3 those with limited English proficiency,⁴ and persons of low socioeconomic status.⁵ Individuals with kidney failure have also been disproportionately affected; an analysis of early Medicare claims data confirmed that individuals with kidney failure were at 3.5 times the risk of acquiring COVID-19 compared with other Medicare fee-for-service beneficiaries.⁶

Although COVID-19 hospitalization and mortality outcomes among patients on dialysis have been reported,^7–10 there have been fewer reports characterizing factors associated with COVID-19 incidence¹¹ and in particular, social factors, such as race/ethnicity and neighborhood-level characteristics. One study of a large national sample of patients on hemodialysis found that patients who were non-Hispanic Black, were Hispanic, or resided in high poverty or majority Black and Hispanic neighborhoods were more likely to have severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antibodies.¹² Greater examination of neighborhood-level characteristics may reveal drivers of racial/ethnic disparities seen in COVID-19 incidence in the dialysis population, with greater viral transmission in neighborhoods with more housing crowding and a higher essential workforce.^{13,14 (preprint)}

The Social Vulnerability Index (SVI) is a composite of census indicators used by the Centers for Disease Control and Prevention for emergency preparedness and disaster response. County-level SVI is strongly associated with county-level COVID-19 cases in the general population, but how SVI associates with individual-level COVID-19 cases in the dialysis population is unknown.^{5,15 (preprint),16,17} Furthermore, the relationship between race/ethnicity and SVI is incompletely characterized. Therefore, we examined racial/ethnic differences in COVID-19 incidence among patients on hemodialysis in New York City during the first wave of the COVID-19 pandemic and assessed if SVI explained racial/ethnic differences in COVID-19 incidence.

Methods

Study Design and Population

We performed a retrospective cohort study of prevalent patients on in-center hemodialysis at eight dialysis units within a nonprofit nephrology organization in New York City. We included patients receiving in-center hemodialysis as of March 1, 2020 and followed them until August 3, 2020. We excluded patients who had AKI requiring dialysis (six patients), those on home dialysis modalities (five patients), and those with missing SVI values (11 patients). As a supplemental data source, we obtained publicly available data on COVID-19 case rates by race and zip code tabulation area (ZCTA) from the New York City Department of Health and Mental Hygiene.¹⁸

Predictor, Outcome, and Covariates

Our primary predictor was patient race/ethnicity obtained from the electronic health record (EHR) and classified as (1) non-Hispanic White; (2) non-Hispanic Black; (3) Hispanic; (4) Asian or Pacific Islander; or (5) other, unknown, or missing. Our coprimary predictor was neighborhood social vulnerability as measured by overall SVI, the four SVI themes, and housing crowding. We used census tract–level SVIs for New York State calculated from the 2014–2018 American Community Survey 5-year estimates.¹⁹ The SVI is a composite of 15 census indicators organized into four themes: (1) socioeconomic status (percentage of persons below poverty, percentage unemployed, percentile per capita income, percentage with no high school diploma), (2) household composition and disability (percentage aged 65 or older, percentage aged 17 or younger, percentage with a disability, percentage of single-parent households), (3) minority status and language (percentage minority, percentage who speak English “less than well”), and (4) housing type and transportation (percentage in multiunit housing, percentage in mobile homes, percentage of households with more people than rooms [housing crowding], percentage of households with no vehicle, percentage in group quarters). Each census tract is given a percentile ranking for the four SVI themes and an overall ranking, compared with other census tracts in New York State. We geocoded patient addresses to census tracts and ZCTAs using the US Census Bureau website and linked them to census tract–level SVI.²⁰ Overall SVI, the four SVI themes, and housing crowding percentiles were modeled as continuous variables. Additionally, overall SVI was divided into quintiles within our study population.

Our primary outcome of interest was confirmed or presumed COVID-19. We created a COVID-19 tool in the Research Electronic Data Capture system and surveyed our EHR weekly using a direct data connection and automated scripting to identify and track patients with COVID-19. Patients were tested for COVID-19 during routine clinical care when there was clinical suspicion on the basis of signs and symptoms. Testing was performed on nasopharyngeal swab specimens sent for RT-PCR assay for SARS-CoV-2. A positive result from a SARS-CoV-2 PCR nasal swab was verified by chart review and classified as a confirmed COVID-19 case. Because New York City was limiting PCR testing early in the pandemic to primarily hospitalized patients, patients with clinical signs and symptoms of cough, fever, or respiratory symptoms were classified as presumed cases on the basis of manual chart review.

Covariates included patient demographics (age, sex), individual-level social factors (employment, marital status, transportation type to dialysis), dialysis-related medical history (vintage [length of time on dialysis], cause of kidney failure), and dialysis facility factors (unit fixed effects and dialysis unit SVI). Individual-level social factors were pulled from structured fields in the EHR. Transportation to dialysis was classified as (1) van service (including ambulette, ambulance, and Access-A-Ride), (2) private vehicle (including taxi services), or (3) public transportation (including bus or subway). A missing indicator category was used for missing covariates; multiple imputation was not used because covariates were thought to be missing not at random. Missingness was as follows: race/ethnicity (9%), employment (35%), marital status (22%), and transportation type (44%).

Statistical Analyses

We first compared the demographics, individual-level social factors, and dialysis-related medical history of patients on hemodialysis residing in neighborhoods with high SVI (higher than median SVI in our study cohort, reflecting higher social vulnerability) versus low SVI (lower than median SVI, reflecting lower social vulnerability). We then compared characteristics of patients on hemodialysis who were COVID-19 positive and not COVID-19 positive. Differences in characteristics were assessed using chi-squared tests for categorical variables and Wilcoxon rank-sum tests for continuous variables. We also examined the association of ZCTA-level COVID-19 cumulative incidence (COVID-19 cumulative PCR+ count divided by ZCTA population denominator) within the New York City general population with COVID-19 among patients on hemodialysis residing in ZCTAs using unadjusted logistic regression.

We then performed logistic regression to examine the association between patient race/ethnicity and overall SVI quintile with COVID-19 positivity, accounting for race/ethnicity and SVI interactions. We used two nested multivariable regression models. Model 1 adjusted for age, sex, race/ethnicity or SVI, individual-level social factors, and dialysis-related medical history. Model 2 additionally adjusted for dialysis facility factors.

To examine if our findings were consistent over time, as a sensitivity analysis we stratified our results by three time periods: March 1 to March 21 (before the stay-at-home order), March 22 to April 14 (premask order), and April 15 to June 7 (before reopening). We also performed a time-to-event analysis using Fine and Gray subdistribution hazard models to estimate the association of race/ethnicity and overall SVI quintile with the subhazard of COVID-19, accounting for the competing risk of death and censoring at transplant, dialysis withdrawal, and modality change. All analyses were clustered at the level of the dialysis unit. Our study was approved by the Weill Cornell Medicine Institutional Review Board. Data analyses were performed using Stata/IC, version 15.1 (StataCorp) and R version 4.0.2 statistical software.

Results

Patient Characteristics

Of the 1378 patients on hemodialysis in our study cohort, 294 (21.3%) were non-Hispanic White; 578 (41.9%) were non-Hispanic Black; 207 (15.0%) were Hispanic; 174 (12.6%) were Asian or Pacific Islander; and 125 (9.0%) had other, unknown, or missing race/ethnicity. Patients on hemodialysis lived in census tracts with higher social vulnerability (Figure 1) (mean SVI 68) compared with the general population of New York City (mean population-weighted SVI 61 in Manhattan, Queens, and Brooklyn).

Figure 1.: Patients on hemodialysis in New York City had variable census tract-level Social Vulnerability Indices (SVI). Higher SVI percentile indicates greater neighborhood social vulnerability. Patients on hemodialysis in our cohort did not reside in census tracts pictured in grey.

Patients who were non-Hispanic Black and Hispanic were younger, were more likely to be disabled, and had a longer dialysis vintage than patients in other race/ethnic categories (Supplemental Table 1). Sex, marital status, transportation to dialysis, and primary cause of kidney failure also differed by race/ethnicity. Patients living in high-SVI neighborhoods were younger, more likely to be women (46% versus 39%), more likely to be non-Hispanic Black (51% versus 33%) and Hispanic (17% versus 13%), less likely to be married (27% versus 39%), and more likely to have diabetes (44% versus 38%) as the cause of their kidney failure, compared with those living in low-SVI neighborhoods (Table 1).

Table 1. - Characteristics of patients on in-center hemodialysis by overall SVI (n=1378)

Characteristic	Low SVI, a %, n=689	High SVI, a %, n=689	P Value
Demographics
Age, yr
18–44	7	14	<0.001
45–64	37	42
65–79	39	34
≥80	17	11
Sex
Men	61	54	0.01
Women	39	46
Individual-level social factors
Race/ethnicity
Non-Hispanic White	32	11	<0.001
Non-Hispanic Black	33	51
Hispanic	13	17
Asian or Pacific Islander	15	10
Other, unknown, or missing	7	11
Employment
Full time or part time	16	8	<0.001
Retired (age)	21	19
Retired (disabled)	19	21
Unemployed	8	11
Homemaker, medical leave, or student	4	4
Missing	32	37
Marital status
Married	39	27	<0.001
Divorced or separated	9	9
Widowed	7	7
Single	25	33
Missing	20	24
Transportation type
Van service	33	35	0.76
Private vehicle	20	21
Public transportation	2	2
Other, unknown, or missing	45	42
Dialysis-related medical history
Dialysis vintage, yr b	3.0 [1.2–5.9]	3.6 [1.7–6.4]	0.01
Primary kidney failure cause
Diabetes	38	44	0.02
Hypertension	31	26
GN	11	11
Cystic kidney disease	3	2
HIV	1	3
Malignancy	2	0.4
Post-transplant	5	4
Other or unknown	9	10

Data are from noninstitutionalized adults residing in the New York City area receiving hemodialysis in one of eight dialysis units in Manhattan, Brooklyn, and Queens. Percentages may not add to 100% due to rounding. P values are presented for chi-squared tests for categorical variables and Wilcoxon rank-sum tests for continuous variables.

^aHigh (above-median) SVI indicates greater neighborhood social vulnerability. Low (below-median) SVI indicates lower neighborhood social vulnerability.

^bDialysis vintage is presented as median [interquartile range] and reported among patients classified as ESKD as of March 1, 2020 (n=1362).

COVID-19 and Patient Characteristics

A total of 247 (17.9%) patients developed symptomatic COVID-19, of whom 230 (93.1%) were laboratory-confirmed positive and 17 (6.9%) were presumed positive. The timing of COVID-19 cases among patients on hemodialysis coincided with the New York City–wide surge (Supplemental Figure 2). Patients with COVID-19 were more likely to be single (30% versus 25%), more likely to travel to dialysis using a van service (45% versus 31%), and more likely to have diabetes as their cause of kidney failure (51% versus 39%) (Table 2). ZCTA-level COVID-19 cumulative incidence in the New York City general population was strongly associated with COVID-19 among patients on hemodialysis (odds ratio [OR], 1.34; 95% confidence interval [95% CI], 1.14 to 1.58 per percentage increase in PCR+ cases divided by the ZCTA population denominator) (Supplemental Figure 2).

Table 2. - Patient characteristics by COVID-19 positivity among patients on hemodialysis (n=1378)

Characteristic	COVID-19 Positive, %, n=247	Not COVID-19 Positive, %, n=1131	P Value
Demographics
Age, yr
18–44	8	11	0.21
45–64	38	40
65–79	42	35
≥80	13	14
Sex
Men	52	58	0.07
Women	48	42
Individual-level social factors
Employment
Full time or part time	10	13	0.12
Retired (age)	23	19
Retired (disabled)	25	19
Unemployed	9	10
Homemaker, medical leave, or student	3	4
Missing	30	36
Marital status
Married	33	33	0.002
Divorced or separated	12	8
Widowed	12	6
Single	25	30
Missing	19	23
Transportation type
Van service	45	31	0.001
Private vehicle	18	21
Public transportation	2	2
Other, unknown, or missing	36	45
Dialysis-related medical history
Dialysis vintage, yr a	3.5 [1.5–6.0]	3.2 [1.4–6.1]	0.98
Primary kidney failure cause
Diabetes	51	39	0.05
Hypertension	25	29
GN	10	11
Cystic kidney disease	2	2
HIV	2	2
Malignancy	1	1
Post-transplant	2	5
Other or unknown	7	10

Percentages may not add to 100% due to rounding. P values are presented for chi-squared tests for categorical variables and Wilcoxon rank-sum tests for continuous variables.

^aDialysis vintage is presented as median [interquartile range] and reported among patients classified as ESKD as of March 1, 2020 (n=1362).

Race/Ethnicity and COVID-19

Non-Hispanic Black (OR, 1.68; 95% CI, 1.14 to 2.48) and Hispanic patients (OR, 2.66; 95% CI, 1.52 to 4.65) had higher odds of acquiring symptomatic COVID-19, compared with non-Hispanic White patients in unadjusted analyses (Table 3). Similar relative risks by race/ethnicity were seen in the New York City general population (Supplemental Table 2).¹⁸ Adjusting for ZCTA-level COVID-19 cumulative incidence somewhat attenuated racial/ethnic disparities seen among patients on dialysis (non-Hispanic Black adjusted odds ratio [aOR], 1.47; 95% CI, 1.03 to 2.09 and Hispanic aOR, 2.23; 95% CI, 1.39 to 3.58). Non-Hispanic Black and Hispanic patients were more likely to acquire COVID-19 in multivariable analyses accounting for age, sex, SVI, social factors, and dialysis-related medical history. In multivariable analyses additionally accounting for dialysis facility factors (unit fixed effects and dialysis unit SVI) (Supplemental Table 3), non-Hispanic Black (aOR, 1.76; 95% CI, 1.25 to 2.48) and Hispanic patients (aOR, 2.66; 95% CI, 1.50 to 4.75) had increased odds of COVID-19 compared with non-Hispanic White patients. Patients who were Asian or Pacific Islander were not at increased risk of incident COVID-19 compared with non-Hispanic White patients in unadjusted or adjusted models. Racial/ethnic disparities were largely consistent across time periods during the first wave of the pandemic and in time-to-event analyses accounting for the competing risk of death (Supplemental Tables 4 and 5).

Table 3. - Association of race/ethnicity and SVI with COVID-19 among patients on hemodialysis (n=1378)

Variable	Unadjusted OR (95% CI)	P Value	Model 1 aOR (95% CI)	P Value	Model 2 aOR (95% CI)	P Value
Race/ethnicity
Non-Hispanic White (reference), n=294	1	—	1	—	1	—
Non-Hispanic Black, n=578	1.68 (1.14 to 2.48)	0.01	1.77 (1.20 to 2.62)	0.004	1.76 (1.25 to 2.48)	0.001
Hispanic, n=207	2.66 (1.52 to 4.65)	0.001	2.90 (1.72 to 4.88)	<0.001	2.66 (1.50 to 4.75)	0.001
Asian or Pacific Islander, n=174	1.32 (0.69 to 2.50)	0.40	1.19 (0.56 to 2.53)	0.66	1.22 (0.63 to 2.36)	0.56
Other, unknown, or missing, n=125	1.21 (0.70 to 2.06)	0.50	1.30 (0.87 to 1.94)	0.19	1.37 (0.94 to 2.02)	0.11
Overall SVI
Quintile 1 (reference)	1	0.50	1	0.92	1	0.80
Quintile 2	1.34 (0.80 to 2.23)		1.06 (0.67 to 1.66)		1.01 (0.60 to 1.68)
Quintile 3	1.32 (0.86 to 2.02)		0.99 (0.75 to 1.32)		1.00 (0.73 to 1.38)
Quintile 4	1.55 (0.96 to 2.53)		1.16 (0.85 to 1.58)		1.19 (0.88 to 1.60)
Quintile 5	1.33 (0.76 to 2.31)		1.05 (0.63 to 1.75)		1.11 (0.66 to 1.87)

SVI is calculated at the census tract level. Quintile 1 of SVI represents the lowest level of social vulnerability. Quintile 5 of SVI represents the highest level of social vulnerability. Model 1 was adjusted for age, sex, race or SVI, individual-level social factors (employment, marital status, transportation type to dialysis), and dialysis-related medical history (dialysis vintage, primary kidney failure cause). Model 2 was additionally adjusted for dialysis facility factors (unit fixed effects and dialysis unit SVI). —, reference values. Bolded values are statistically significant at a threshold of P<0.05.

Neighborhood Social Vulnerability and COVID-19

The interaction terms of race/ethnicity and SVI themes were statistically significantly associated with COVID-19 (Supplemental Table 6). Because of statistical evidence of an interaction, we stratified analyses by race/ethnicity when examining the association between overall SVI, SVI themes, housing crowding, and COVID-19 (Table 4). Overall SVI, modeled either continuously or in quintiles, was not significantly associated with COVID-19 in unadjusted or adjusted models across racial/ethnic categories. Among non-Hispanic White patients, the socioeconomic status SVI theme (OR, 1.11; 95% CI, 1.02 to 1.20 per 10-percentile increase), minority status and language SVI theme (OR, 1.22; 95% CI, 1.07 to 1.38), and housing crowding (OR, 1.24; 95% CI, 1.11 to 1.37) were significantly associated with COVID-19 in unadjusted analyses. In analyses adjusted for demographics, social factors, dialysis-related medical history, and dialysis facility factors, overall SVI or SVI themes were not associated with COVID-19 across racial/ethnic categories, but housing crowding remained significantly associated with COVID-19 among non-Hispanic White patients (aOR, 1.14; 95% CI, 1.03 to 1.26) (Supplemental Table 7).

Table 4. - Association of SVI with COVID-19 among patients on hemodialysis (n=1378), stratified by race/ethnicity (unadjusted analyses)

Variable	Non-Hispanic White, n=294, OR (95% CI)	P Value	Non-Hispanic Black, n=578, OR (95% CI)	P Value	Hispanic, n=207, OR (95% CI)	P Value	Asian or Pacific Islander, n=174, OR (95% CI)	P Value	Other, Unknown, or Missing, n=125, OR (95% CI)	P Value
Overall SVI a
SVI	1.08 (0.96 to 1.21)	0.20	0.99 (0.92 to 1.07)	0.79	1.03 (0.81 to 1.30)	0.83	1.12 (0.94 to 1.33)	0.19	0.95 (0.79 to 1.14)	0.58
SVI theme a
Socioeconomic status	1.11 (1.02 to 1.20)	0.01	0.95 (0.88 to 1.03)	0.21	1.06 (0.86 to 1.30)	0.60	1.07 (0.94 to 1.22)	0.28	0.91 (0.79 to 1.05)	0.22
Household composition and disability	0.90 (0.76 to 1.07)	0.23	1.02 (0.96 to 1.08)	0.58	0.94 (0.82 to 1.08)	0.40	1.09 (0.94 to 1.28)	0.27	0.96 (0.83 to 1.11)	0.60
Minority status and language	1.22 (1.07 to 1.38)	0.002	0.97 (0.88 to 1.07)	0.52	1.11 (0.94 to 1.31)	0.21	0.99 (0.87 to 1.13)	0.89	1.02 (0.69 to 1.51)	0.93
Housing type and transportation	1.05 (0.95 to 1.16)	0.34	1.01 (0.93 to 1.10)	0.77	1.04 (0.90 to 1.21)	0.59	1.11 (1.01 to 1.22)	0.03	1.07 (0.95 to 1.21)	0.28
SVI component a
Housing crowding b	1.24 (1.11 to 1.37)	<0.001	0.98 (0.89 to 1.08)	0.72	0.96 (0.78 to 1.17)	0.67	1.07 (0.92 to 1.24)	0.38	1.08 (0.74 to 1.57)	0.69

Overall SVI, SVI themes, and SVI components calculated at the census tract level. Bolded values are statistically significant at a threshold of P<0.05.

^aPresented as unadjusted ORs of COVID-19 positivity per 10-percentile increase in overall SVI, SVI theme, or SVI component.

^bHousing crowding percentile is defined as the percentage of occupied housing units in a census tract with more people than rooms.

Discussion

In this cohort study of patients on in-center hemodialysis, we found that non-Hispanic Black and Hispanic individuals were substantially more likely to acquire symptomatic COVID-19 during the first wave of the pandemic in New York City. Census tract–level socioeconomic status, minority status and language, and housing crowding were associated with COVID-19 among non-Hispanic White patients on hemodialysis but did not explain racial/ethnic disparities. Neighborhood-level COVID-19 prevalence was associated with COVID-19 cases among patients on hemodialysis, suggesting community transmission.

Our results contribute to a growing literature documenting racial/ethnic disparities in COVID-19 across different United States geographies and clinical populations. Our results are significant because patients on hemodialysis are a highly vulnerable patient population that faces unique, excess risks of acquiring COVID-19 and high mortality rates.^6,8 Thus, stark racial/ethnic disparities in kidney failure incidence²¹ and COVID-19 incidence represent a situation of multiplicative risk, placing non-Hispanic Black and Hispanic individuals on hemodialysis at high risk of acquiring COVID-19 and subsequent adverse outcomes, including mortality. Another analysis of 2178 patients on dialysis in New York City also showed a high prevalence of COVID-19 among patients on dialysis and higher incidence in Black, Hispanic, and Asian patients, contributing to a greater population burden of mortality in these racial/ethnic groups.²² Structural and historical racism, such as redlining (race/ethnicity-based discriminatory practices, including discriminatory lending practices, that place financial and other services out of reach for residents of certain areas), results in neighborhood segregation and lack of economic opportunity, and it likely contributes to racial/ethnic disparities in COVID-19 incidence.^{23,24,25 (preprint)}

In our study, racial/ethnic disparities in COVID-19 incidence in the hemodialysis population mirrored trends seen in the general population of New York City, but they were only partially explained by neighborhood COVID-19 cumulative incidence and were not explained by census indicators of social vulnerability, suggesting that additional unmeasured social variables contribute to excess risk. Our findings that race/ethnicity modifies the association between neighborhood social vulnerability and COVID-19 suggest that neighborhood-level factors contributed to COVID-19 incidence among non-Hispanic White patients, whereas other residual factors, such as unmeasured household exposures, accounted for excess COVID-19 cases in the Black and Hispanic patients in our study sample. For example, in the general population, Black individuals are more likely to live in households with health care workers, and Hispanic persons are more likely to live in households with essential workers unable to work from home.²⁶ Similarly, patients on hemodialysis who are Black or Hispanic may be more likely to live with essential workers unable to work from home, increasing their risk of acquiring COVID-19 from household contacts. Limited English proficiency and immigration status are additional unmeasured social factors that contribute to excess risk of acquiring COVID-19.²⁷ A greater understanding of patients’ household composition and other community exposures may be useful for risk-stratifying patients and providing individualized education about risk mitigation. Additionally, exposure to dialysis facility staff and other patients who may be traveling from COVID-19–prevalent areas may contribute to racial/ethnic disparities.

Our findings provide evidence that both neighborhood and dialysis-related exposure contribute to COVID-19 risk among patients on hemodialysis. Formal guidance has been disseminated to dialysis facilities on infection control practices,^28,29 but less attention has been provided to patient-oriented education within dialysis facilities to decrease risk of community acquisition. As the COVID-19 pandemic continues, qualitative and survey-based research on the perceptions and practices of patients on dialysis and their household contacts surrounding social distancing, mask wearing, and vaccination will be highly informative.

The lack of association between overall SVI and COVID-19 in our data differs from findings from ecologic studies in the general population, which show a strong association between county-level SVI and positive tests per capita.^{5,15 (preprint),16} There are several explanations for this discrepancy. First, patients on hemodialysis in our cohort disproportionately lived in high–social vulnerability neighborhoods and face excess risk from dialysis-related exposures, so the same relationship between overall SVI and COVID-19 may not apply. Second, certain variables in the SVI, such as multiunit housing, mobile homes, and vehicle ownership, may perform differently or not be relevant in New York City in relation to COVID-19, so similar analyses should be investigated in other geographies. The magnitude of our findings and interactions between race/ethnicity and neighborhood-level social vulnerability may not fully generalize to other geographies, depending on urbanicity, population demographics, and timing/patterns of the COVID-19 pandemic.

Strengths of our analysis include the substantial size of our cohort and use of patient-level address data that allowed for granular geocoding. Limitations include missing data for some variables, particularly individual-level social factors, which may contribute to misclassification bias. Misclassification of race/ethnicity data may have occurred if race/ethnicity was entered by dialysis facility staff by presumption and not on the basis of patient self-report. Our results may also be subject to misclassification of neighborhood-level social vulnerability if patients moved or there were lags in updating addresses in the EHR. Importantly, we did not systematically test for SARS-CoV-2 using PCR or antibody testing, so our results do not incorporate asymptomatic cases, which may be considerable among patients on hemodialysis.³⁰ Lastly, our results may have been subject to residual confounding and/or were underpowered to detect some associations between neighborhood-level characteristics and COVID-19, particularly in stratified analyses.

In summary, racial/ethnic disparities in COVID-19 incidence among patients on hemodialysis largely mirror community transmission patterns and likely reflect neighborhood spread to this vulnerable population. Neighborhood-level socioeconomic status, minority status and language, and housing crowding were positively associated with acquiring COVID-19 among non-Hispanic White patients but did not explain racial/ethnic disparities. Our findings suggest that socially vulnerable patients on dialysis face disparate COVID-19–related exposures, calling for targeted risk-mitigation strategies.

Disclosures

D. Cukor reports research funding from the National Institutes of Health. D.M. Levine reports patents and inventions with The Rogosin Institute. J. Silberzweig reports consultancy agreements with Alkahest Biotech, Bayer Pharmaceuticals, and Kaneka Pharma and scientific advisor or membership with the American Society of Nephrology: COVID-19 Response Team, Emergency Partnership Initiative. S.L. Tummalapalli received consulting fees from Bayer AG unrelated to the submitted work. All remaining authors have nothing to disclose.

Funding

S.L. Tummalapalli is supported by funding from National Institute of Diabetes and Digestive and Kidney Diseases grant F32DK122627 and National Kidney Foundation Young Investigator Grant.