Introduction

African-Americans with end-stage renal disease (ESRD) face widespread disparities in access and quality of care compared to Whites [1–5]. To date, systemic differences in health-care access and outcomes have been less fully explored. Distance to a health-care facility is a potentially important determinant of individuals’ access to care. For many patients with ESRD, distance from their dialysis facility may be particularly important due to the thrice-weekly hemodialysis treatment requirement. Two US studies have shown that most people in urban areas live within 30 min of their dialysis facility, and that longer travel time was associated with greater mortality and worse health-related quality of life [6, 7]. However, to our knowledge, no study has examined racial disparities in proximity to dialysis facilities.

In addition, systematic differences in care across neighborhoods or regions may also contribute to racial disparities in quality of care in dialysis facilities. Prior investigations suggest that dialysis facilities vary significantly regarding dialysis adequacy, anemia management, and renal transplant referral [4, 8–11]. Worse dialysis outcomes and reduced access to transplantation have also been associated with neighborhood factors such as poverty, urbanicity, and the proportion of African-Americans residing in the neighborhood [11–14]. Due to ongoing residential segregation, African-Americans and Whites tend to live in separate neighborhoods. These differences could translate to differential access to high-quality dialysis facilities. While prior work has demonstrated that African-Americans are less likely to dialyze in high-quality dialysis facilities, it is unknown whether this is due to racial differences in proximity to these facilities [11, 15].

Two competing pathways could explain African-Americans’ reduced access to high-quality dialysis facilities. First, African-Americans could live further from high-quality dialysis facilities which, in turn, reduces their likelihood of choosing to dialyze there. Another potential pathway is that African-Americans live closer to high-quality dialysis facilities than their White counterparts but still do not obtain care there. In this case, African-Americans could either be less likely to select those high-quality facilities for their care or to be accepted into those facilities. We sought to determine whether African-Americans and Whites differ in proximity to dialysis facilities, both overall and those of high quality. Additionally, we sought to determine whether African-Americans and Whites differed in access to high-quality facilities and if distance from the facility was associated with differential access.

Methods

Subjects and Data

First, we created the cohort using the United States Renal Data System (USRDS), a comprehensive dataset of patients with ESRD. We included all urban, non-Hispanic Whites and African-Americans aged 18–65 initiating in-center hemodialysis between January 2005 and December 31, 2008. We defined an urban area using both the Urbanized Areas (UA) and Urban Clusters (UC) of the 2010 US Census Urban Area Cartography Files [16]. We used a USRDS crosswalk file to link individuals to their dialysis facilities in the Centers for Medicare and Medicaid Services (CMS) ESRD Quality Incentive Program (QIP) file which provided information on facility address, dialysis facility characteristics, and facility quality outcomes for 2007. We applied the 2010 CMS definition for high-quality facility scoring to the 2007 data. We excluded individuals who could not be matched to a dialysis center (n = 2,154). Finally, we used the 2007–2011 American Community Survey by Zip Code Tabulation Areas (ZCTAs) to derive the neighborhood characteristics for each individual’s home ZIP code at time of dialysis initiation. The final analytic file contains individuals nested within dialysis centers linked to a dialysis facility and residential neighborhood descriptors.

Outcomes and Covariates

Outcomes

Our outcomes were distance from patients’ home to (1) their dialysis facility, (2) the nearest dialysis facility, and (3) the nearest high-quality dialysis facility. We also examined individuals’ likelihood of dialyzing at a high-quality facility. We defined the nearest facility as the facility that was the shortest straight-line distance from patient’s home ZIP code. High-quality dialysis facilities were defined as those with a total performance score (TPS) greater than 26 (out of 30), based on the ESRD Quality Incentive Program scoring algorithm which gave a 1–10 rating for the proportion of patients with (1) a hemoglobin (Hgb) <10 g/dL, (2) Hgb >12 g/dL, and (3) urea reduction rate >65 [17].

Predictor and Covariates

The main predictor is the racial difference between non-Hispanic Whites (reference group = 0) and African-Americans (comparison group = 1). In the USRDS, subjects’ race and ethnicity was obtained from the Medical Evidence form using patient self-report.

Individual covariates include gender and age at dialysis initiation (by quartile). Neighborhood-level characteristics include percent African-American and percent of population below the Federal Poverty Level which were right-skewed and divided into quartiles to create categorical variables. Region indicator variables—Northeast, Midwest, South, and West—were created from the 18 End-Stage Renal Disease Networks, regional entities responsible for organization, health planning, and quality improvement tasks for ESRD care.

Analysis

Five-digit residential ZIP code, the most detailed patient location available in USRDS, was used as a proxy for the patient’s residence. Travel distance was the straight-line distance between the race-specific population weighted center of the patient’s residential five-digit ZIP code and the corresponding dialysis center address [18, 19].

To calculate the race-specific population center of each ZIP code, we first calculated the sum of geometric centers of blocks within the ZIP code which was weighted by the race-specific population and then divided by the total group population in the ZIP code in order to produce the population-weighted mean geometric center. The geometric centroid (latitude and longitude) of all blocks in the US ZIP code as well as the 2010 population of each race group were obtained from the Missouri Census Data Center [20]. Then, dialysis facilities were assigned with geographic coordinates based on their reported addresses using ArcGIS Streetmap Premium (ESRI, Redlands, CA). Finally, straight-line travel distance between patient’s residential ZIP code and the corresponding dialysis center was calculated in Stata using Vincenty (Haversine formula), a user-written program.

First, we examined the bivariate association between our outcomes of interest—patient distance to current, nearest, and nearest high-quality dialysis facility—and our covariates of interest stratified by race. Then we used linear regression to characterize the associations between our outcomes of interest and individual demographic characteristics. In addition, we used logistic regression to determine the association between dialysis in a high-quality facility and our covariates. Variables were determined a priori and were sequentially added to the model to adjust first for individual characteristics, then neighborhood demographics, and finally region. Interaction was assessed between neighborhood demographic variables (proportion African-American and proportion below poverty) for likelihood of attending a high-quality facility. We used the Breen scaling method to determine the effect of neighborhood on racial differences in likelihood of attending a high-quality facility [21, 22]. Breen scaling is used here to calculate the change in the coefficient of the variable of the interest attributed to the inclusion of confounding variables when employing logit and probit models. Although patients were nested within neighborhoods and dialysis centers, we were unable to use multilevel modeling because not all higher level units (dialysis units or neighborhoods) had sufficient number (n ≥ 5) of observations [23]. To account for within-unit correlation, the sandwich estimate was used with ZIP code of residence as a cluster variable. We used chi-square to assess differences for categorical variables, and t-tests and ANOVA to assess differences for linear variables. For all analyses, a two-tailed significance level of p < 0.05 was used. All analyses were conducted using Stata, version 12.0.

Results

Patient Characteristics

Our final sample included 183,190 patients, 49 % White and 51 % African-American (Table 1). On average, the African-American patients were younger (49.6 versus 52.6 %) and less likely to be male (53.0 versus 57.8 %). In general, African-American respondents were significantly more likely to live in high poverty (42.3 versus 10.7 %) and predominately African-American neighborhoods (45.6 versus 4.0 %) than their White counterparts (all p < 0.05).

Table 1 Descriptive statistics, overall, and by race
Full size table

The median distance from patients’ home to their dialysis facility was 4.11 miles (mi). African-Americans lived 3.62 mi from their dialysis facility, significantly closer than Whites (4.70 mi). African-Americans were less likely to dialyze in high-quality facilities than Whites (52.1 v 55.9 %, p < 0.05). In unadjusted analysis, African-Americans lived closer to high-quality facilities compared to Whites (2.26 versus 3.36 mi) (all p < 0.05).

Overview of Multivariable Results

In the fully adjusted models (Table 2), African-Americans lived a half-mile closer to the nearest dialysis facility (B = −0.52) but on average traveled the same distance to their own dialysis facility than their White counterparts. In addition, individuals lived closer to a high-quality facility as the proportion of African-Americans in the neighborhood increased (B = −5.92, highest versus lowest quartile). Further, initial analysis (Table 3, model 1) shows that African-Americans are 14 % less likely than their White counterparts to attend a high-quality dialysis facility (odds ratio (OR) 0.86) and that disparity is even more striking after controlling for distance to high-quality facility (model 2, OR 0.69). After controlling for individual and neighborhood characteristics and region, individuals in predominately African-American neighborhoods are 53 % less likely to receive dialysis in a high-quality facility (OR 0.47, both highest quartile versus lowest, p < 0.05, Table 3, model 5). Living in a predominately African-American neighborhood explains 24 % of African-Americans’ reduced likelihood of attending a high-quality facility (Table 3).

Table 2 Multivariable regression for dialysis facility distance between patient ZIP code and the nearest, patients’ own, and nearest high-quality dialysis facility
Full size table
Table 3 Multivariable analysis for odds of attending high-quality dialysis facility with stepwise adjustment
Full size table

Multivariable Results: Distance to Nearest Dialysis Facility

Factors associated with living closer to the nearest dialysis facility (Table 2, full model 1) include African-American race (B = −0.52), male gender (B = 0.04), having an increasing proportion of African-Americans in the neighborhood (B = −2.33, highest quartile versus lowest), and residence in the Northeast (B = −0.42, compared to the Midwest) (all p < 0.001). Individuals in the South and West lived farther from the nearest center.

Multivariable Results: Distance to Patient Dialysis Facility

In the final adjusted analysis (Table 2, full model 2), factors associated with living closer to patients’ current dialysis facility include older age (B = −2.72, over 60 versus less than 45), larger proportion of African-Americans in the neighborhood (B = −5.64, Q4 greater than 49 % versus Q1, less than 3 %), higher neighborhood poverty (B = −0.89, Q4 greater than 23 % versus Q1, less than 7 %), and residence in the Northeast (B = −2.42, compared to the Midwest) (all p < 0.05). Individuals in the South and West lived farther from their center.

Multivariable Results: Distance to High-Quality Dialysis Facility

Factors associated with living closer to a high-quality dialysis facility (Table 2, full model 3) include African-American race (B = −0.93), larger proportion of African-Americans in the neighborhood (B = −5.92, highest quartile versus lowest), and residence in the Northeast (B = −0.89, compared to the Midwest) (all p < 0.001). Whites, those in high-poverty neighborhoods and those in the South and West lived farther from a high-quality facility.

Multivariable Results: Dialysis at a High-Quality Facility

Table 3 demonstrates the odds of attending a high-quality dialysis facility with step-wise addition of variables. Examining individual characteristics (model 1) reveals that African-Americans and males were significantly less likely to attend a high-quality facility (OR 0.86 and 0.97, respectively, all p < 0.05). After controlling for distance to the nearest high-quality facility (model 2), African-Americans were 31 % less likely than their White counterparts to attend a high-quality facility (OR 0.69, 95 % confidence interval (CI) 0.66, 0.72). African-American race remained significantly associated with lower likelihood of attending a high-quality dialysis facility even after controlling for region, neighborhood percent poverty, and neighborhood percent African-American (models 3, 4, and 5 respectively). In the fully adjusted model (model 5), African-Americans were still 12 % less likely to attend a high-quality facility than their White counterparts (OR 0.88, 95 % CI 0.85, 0.91). Neighborhood racial composition accounted for 23.9 % of the racial disparity in attending a high-quality facility.

In the fully adjusted model (Table 3, model 5), living in the Midwest, in a moderate- to high-poverty neighborhood and having a lower proportion of African-Americans in the neighborhood were associated with an increased likelihood of attending a high-quality facility. Having a larger proportion of African-Americans in the neighborhood (OR 0.47, highest quartile versus lowest, 95 % CI 0.42, 0.53) was associated with decreased likelihood of dialysis in a high-quality facility, even after controlling for neighborhood poverty. Surprisingly, individuals who lived closest to a high-quality facility were 8–9 % less likely to attend a high-quality facility across all models.

Discussion

We found African-Americans’ proximity to dialysis facilities did not translate into access.

Further, African-Americans appear to bypass closer facilities to travel to their own dialysis facility. In addition, although African-Americans lived a significantly shorter distance to high-quality dialysis facilities, they were not more likely to dialyze these facilities than their White counterparts. Our work showed that this difference was explained both by where they lived (region and neighborhood) and by than who they are (race). Living in a predominately African-American neighborhood was associated with a reduced likelihood of receiving dialysis in a high-quality facility despite living significantly closer to these facilities.

In addition, we found that after accounting for differences in region and neighborhood, the effect of race was attenuated. Region and neighborhood racial composition accounted for approximately 25 % of differences by race. Our results are supported by prior work that contends that racial residential segregation plays an important role in continued health disparities for African-Americans [24–26]. African-Americans and Whites have had different migration patterns. African-Americans are clustered in urban areas in the Northeast and Midwest and in both urban and rural areas in the South. In addition, even within the same areas, due to ongoing racial residential segregation African-Americans and Whites often live in separate communities. This continued separation by race leads to differences in access to the information, opportunities, and organizations which can impact health [24–26].

Our work underscores that access to health care is related to both physical distance and, potentially, “social distance.” Nearby dialysis facilities may still be inaccessible for a variety of social reasons. Despite their proximity, nearby facilities may be “off their mental map” if they lie outside patient’s activity space, the space within which people move about or travel in the course of their daily activities [27, 28]. Alternatively, patients may reject nearby facilities because they do not feel comfortable in the dialysis facility neighborhood or the facility itself.

In addition, African-Americans may be guided toward lower-quality dialysis facilities. Patients may also rely on their social networks and word of mouth from friends and family. Due to residential and social segregation, African-Americans’ social networks are more likely to be closed and limited to those of their own race and neighborhood, which may limit information about facility quality [27, 29]. While facility quality ratings are publicly available online through Dialysis Facility Compare, patients need to know these ratings exist and have computer access for this online data [30]. African-Americans may also be guided toward lower-quality facilities by health-care providers. For many patients, dialysis occurs emergently within a hospital setting which may require patients to rely on hospital-based providers for recommendations [31, 32]. These providers may steer patients toward particular facilities based on their referral practices, knowledge, or assumptions about patients’ needs.

Even with appropriate knowledge and referral, African-Americans may be less likely to be accepted at higher-quality facilities. Unlike other medical services, patients are accepted in a dialysis facility rather than just deciding to receive care there. Due to greater patient demand, higher-quality dialysis facilities may have a longer waiting list or more stringent selection criteria. Higher-quality dialysis facilities may also seek to maintain their quality outcomes by “cherry-picking,” reducing access for patients perceived to be medically or socially complex, less well-insured or more likely to be nonadherent [33]. Our study is designed to determine association rather than causality. Our results, while important, raise additional questions. Further qualitative and quantitative work is needed to determine how individuals end up in their dialysis facilities and how we can equalize access to high quality facilities.

Access to high-quality dialysis facilities has important clinical significance. Poor performance in achieving targeted hemoglobin and greater dialysis adequacy, the dialysis facility outcomes measured by the Centers for Medicare and Medicaid Services (CMS) ESRD Quality Incentive Program (QIP), is associated with increased morbidity, mortality, and increased risk for hospitalization [34–37]. Patients whose care met quality indicators, including dialysis adequacy and achieving targeted hemoglobin, show a greater self-reported quality of life than those who do not [35].

We note several limitations to this study. First, we only had ZIP codes for patient residence so our neighborhood-level data was based on the ZIP code tabulation area. Second, we used straight-line distance to dialysis facility rather than driving distance or travel time. While some work has demonstrated that straight-line distance underestimates travel distance by 20–25 % compared to road network, others have shown that over large geographical areas straight-line distance is highly correlated with travel time. Systematic underestimation may be inherent in the methods used; however, our method is unlikely to generate any bias in statistical models so long as the errors are not biased against any subpopulation [38, 39]. Third, our measure of facility quality is based on the CMS QIP which may only partly reflect overall facility quality; however, this currently is the only comprehensive quality metric for dialysis facilities and is linked to reimbursement. Fourth, we focused our analysis to urban, African-Americans, and Whites under 65. We chose African-Americans and Whites due to their larger sample size and lower rates of misclassification in the USRDS. We selected patients under 65 because older patients may have varied residential mobility, making our initial home address and our distance measure less accurate. Older adults may have increased residential mobility as they move due medical illness, e.g., living with family or in a nursing home due to frailty or due to increased time for leisure and family, e.g., move to retirement communities or warmer areas [40]. Our choices increase our internal validity, but decrease our generalizability. Further work should examine the important issues of access to high-quality dialysis care for rural and older patients with ESRD.

Overall, individuals in urban areas in the USA live close to a dialysis facility and close to a high-quality facility. We found that patients live, on average, 4 mi from their chosen dialysis facility and less than 2 mi to the nearest dialysis facility. Our work was consistent with previous studies that showed urban patients on hemodialysis do not travel far to their dialysis centers [6, 7, 41]. Having dialysis facilities close to where patients need them is a significant accomplishment, yet to be attained in rural areas in the USA and in many other countries. However, proximity does not equal access. Although African-Americans live a shorter distance to high-quality dialysis facilities than their White counterparts, they are still less likely to receive dialysis in these facilities. This effect is explained both by location (neighborhood and region) and by race. The challenge continues to fully integrate African-Americans into both the communities and the opportunity structures in the USA. In the meantime, to improve the quality of care received by African-Americans with ESRD, further work should focus not only how to improve the dialysis facilities that primarily serve minorities but also to improve access to nearby high-quality dialysis facilities.