Patterns in Geographic Access to Health Care Facilities Across Neighborhoods in the United States Based on Data From the National Establishment Time-Series Between 2000 and 2014.

This cross-sectional study examines geographic access to health care facilities across neighborhoods in the United States over a 15-year period and assesses whether racial/ethnic composition and socioeconomic status are associated with change in neighborhood-level presence of health care facilities over time.


Introduction
Geographic access to health care is associated with increased use of preventive care and improved health outcomes for certain chronic conditions. [1][2][3][4][5][6][7] Although geographic access is one of several components that can alter an individual's overall access to health care, including insurance status, out-of-pocket costs, facility hours, appointment wait times, and linguistic services, prior research has shown increased geographic access is associated with greater use and improved outcomes.
Neighborhoods with more income inequality and residential segregation along sociodemographic lines may not attract or may underinvest in institutions that benefit the general population, resulting in unequal geographic health care access. 8 Previous analyses of geographic access to health care services, including trauma centers, specialty care for neonatal populations, and mental health care, have indicated that neighborhoods with predominantly minority residents, lower socioeconomic status, and high residential turnover have less geographic access to care. 9,10 This observation was confirmed by Smiley et al, 11 who reported that health-related resources are not equally distributed across space and that disadvantage often clusters with residential racial/ethnic patterning. Although recent data indicate access to health care, as measured by insurance coverage or self-report of having a usual source of care, has improved since implementation of the Patient Protection and Affordable Care Act, 12 few sources are available to understand geographic health care environments, including the presence of ambulatory care facilities, retail clinics, and pharmacies and drugstores, beyond county-level geographies.
Despite increasing demographic change in racial/ethnic composition and household income and aging subgroups over the last few decades, few studies have assessed temporal change in the geographic access to or the presence of health care facilities across neighborhoods in the United States. A study 13 conducted in Illinois from 1990 to 2000 found an overall improvement in geographic access to health care over time, with worsened geographic accessibility primarily concentrated in rural areas along with a few urban pockets. Areas that experienced decreasing geographic access had higher levels of socioeconomic disadvantage, sociocultural barriers, and health care needs. Similarly, in a 2011 study, Busingye et al 14 found substantial increases in the proportion of the population with geographic access to cardiac facilities from 1999 to 2010, with disparities still existing in rural communities. Hospital closures over the last decade and increased consolidation across hospital systems may also have altered geographic access for certain neighborhoods over time. [15][16][17][18][19] Insights into long-term temporal trends on the availability of health care facilities nationally, particularly with respect to nonhospital facilities and attention to changing neighborhood-level sociodemographic characteristics of residents, are lacking.
This gap in the literature is addressed herein by examining change in the presence of ambulatory care facilities and pharmacies and drugstores across neighborhoods (ie, census tracts [CTs]) as a measure of geographic access in the United States over a 15-year period. Specifically, the objectives of this study were (1) to examine patterns in neighborhood-level presence of health care facilities across the United States by neighborhood-level sociodemographic characteristics and (2) to assess whether neighborhood-level population characteristics (racial/ethnic composition and socioeconomic status) were associated with change in neighborhood-level presence of health care facilities over time. We hypothesized that socioeconomically disadvantaged neighborhoods would continue to experience limited local presence of health care facilities over time compared with more advantaged neighborhoods. We also hypothesized that neighborhoods undergoing demographic compositional change across time from disadvantaged to advantaged would experience increased presence of health care facilities.

Study Sample
Using longitudinal business data from the National Establishment Time-Series (NETS), this crosssectional study compiled health care environment, demographic, and socioeconomic data between 2000 and 2014 for all CTs in the continental United States (n = 72 538). Of these, 292 CTs were excluded because they contained no land area (ie, were water tracts), leaving 72 246 nonwater CTs.
For consistency over time despite boundary changes, all health care environment, demographic, and socioeconomic measures were assigned to 2010 US Census geographies.
This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 20 The study is part of a larger study (Communities Designed to Support Cardiovascular Health for Older Adults 52 ) that includes human participants not included in this analysis and was approved by the Drexel University Institutional Review Board.

Dependent Variable of Health Care Facilities
To characterize neighborhood-level geographic access, the presence of health care facilities was measured using 2000 to 2014 business data from the NETS database, licensed from Walls & Associates (Denver, Colorado) in January 2017. Detailed methods on the creation and cleaning of the NETS data can be found elsewhere. 21,22 Briefly, the NETS pulls annual snapshots of Dun & Bradstreet (Short Hills, New Jersey) business data to create time series information on all names of US businesses, years active, and industrial classification using Standard Industrial Classification (SIC) codes. The NETS data represent a census of all businesses across the United States, and the NETS is considered one of the most comprehensive databases of establishments available. Prior studies 2,23,24 have used the NETS data to examine health care facilities and specific chronic conditions. From the NETS, records were categorized as ambulatory care facilities or as pharmacies and drugstores using SIC codes (eTable 1 in the Supplement). Ambulatory care was a category designed to capture locations able to provide outpatient care, including screenings and other preventive measures. As such, ambulatory care captures offices or clinics of health practitioners, mental health outpatient and continuous care facilities, behavioral health outpatient and continuous care facilities, urgent care locations, retail clinics, physical therapists, kidney centers, and dental care facilities. The pharmacy and drugstore category was designed to capture locations where medications and medical supplies could be purchased. To capture national chain pharmacies and drugstores otherwise missed because of incorrect SIC code, we searched a broader set of SIC codes for any company or trade name that was on the Nielsen (New York, New York) TDLinx list for trade channel "drug" and subchannels "conventional drug store" or "Rx only and small drug store." geographies and measurements over time. Demographic and socioeconomic characteristics were selected to represent a range of domains while minimizing collinearity. To classify neighborhood demographic characteristics, we used the proportion of residents identifying as non-Hispanic (NH) white individuals, NH black individuals, Hispanic/Latino individuals, NH Asian/Pacific Islander individuals, and non-US born individuals, and those aged 75 years or older. Racial/ethnic composition of neighborhoods was assessed by predominant (>60%) racial/ethnic group into the following racial/ethnic categories: predominantly NH white, predominantly NH black, predominantly Hispanic/ Latino, or predominantly NH Asian/Pacific Islander. Places with no predominant group were classified as racially/ethnically mixed areas. These categorizations were based on prior use in the literature. 27 To represent socioeconomic conditions, we used the proportion of residents living at 100% of the federal poverty level, the proportion with a high school (HS) diploma or less, and home ownership.  those 75 years or older, those living below poverty level, and the population density increased between 2000 and 2010. In contrast, the proportion of residents with a HS diploma or less and those who owned a home decreased.

Description of Change in Health Care Facilities
Census tracts had many more ambulatory care facilities than pharmacies and drugstores, and both facility types increased between 2000 and 2014. The mean (SD) count of pharmacies and drugstores

Characteristics in 2000 Associated With Change in Health Care Facilities Between 2000 and 2014
Consistent with our hypotheses about demographic characteristics, CTs of areas with a racial/ethnic composition classified as predominantly NH black, predominantly Hispanic/Latino, or racially/ ethnically mixed in 2000 were more likely to never have any or to lose ambulatory care facilities between 2000 and 2014 than predominantly NH white tracts (   Results for socioeconomic variables were more mixed (

Change in Characteristics Between 2000 and 2010 Associated With Change in Health Care Facilities Between 2000 and 2014
In general, CTs of areas with a racial/ethnic composition that remained predominantly NH black, predominantly Hispanic/Latino, and racially/ethnically mixed between 2000 and 2010 were more likely to never have any or to lose health care facilities between 2000 and 2014 than CTs of areas with a composition that remained predominantly NH white (  Overall, decreases in neighborhood-level socioeconomic status were associated with never having or losing health care facilities (Table 4). Census tracts that had increases in the percentage of residents living below poverty level and having a HS diploma or less were less likely to gain health care facilities. However, CTs that had increases in the percentage of individuals living below poverty were also less likely to never have a health care facility. Results for change in home ownership were more mixed: an increase in the percentage of individuals who own homes was associated with higher odds of losing or gaining ambulatory care facilities and never having or losing pharmacies and drugstores (vs consistent presence), but an increase in the percentage of individuals who own homes was also associated with lower odds of never having (vs always having) ambulatory care facilities.

Discussion
In this cross-sectional study of neighborhoods across the continental United States over a 15-year period, we found differential change in the presence of health care facilities across neighborhoods, with more socioeconomically disadvantaged neighborhoods never having or losing facilities.  [35][36][37] Furthermore, emerging data specifically on Asian American residential density suggest bimodal distributions in socioeconomic characteristics and other patterns that are unique from other racial/ ethnic communities. 38 More nuanced understanding of neighborhood racial/ethnic composition and more complex measurement are warranted, but more complex measurement was beyond the scope of this study. Therefore, increased research is needed on how the geographic presence of health care facilities and use of services are operationalized differently across population subgroups.
In addition, we observed that higher neighborhood socioeconomic status was associated with an increased number of health care facilities across neighborhoods. Medically underserved areas and populations are identified as geographic areas and populations with a lack of geographic access to primary care services. 39 These areas are eligible for federal grants and health programs, such as predominantly minority residents continue to disproportionately lack health care facilities. These trends over time may indicate a need for more targeted efforts to address disparities in access to ambulatory care services. Although prior studies [40][41][42][43][44][45] have focused on geographic barriers to hospitals and tertiary care, few studies 3,46,47 have examined trends in access to ambulatory care within neighborhoods across the nation; therefore, a critical understanding of these patterns is warranted.

Strengths and Limitations
To our knowledge, this cross-sectional study is one of the first studies to examine longitudinal change in the presence of health care facilities across neighborhoods in the United States over a 15-year period. We focused on the presence of nonhospital facilities, thus giving a context for health care services that would provide primary care and care across the life span, and used detailed data on businesses for more accurate geographic location data and dates of operation of each facility.
However, some limitations should be noted. First, we focused on the presence of health care facilities within CTs and did not examine use of services among populations within CTs; therefore, we were unable to directly link availability with use. Second, the context of health care markets and concentrations of health care providers vary across the United States, and state and regional policies, market entry forces, and patterns of health care consolidation may have shaped the patterns observed 48 but were beyond the scope of this study. Third, we used a broad classification of racial/ ethnic composition at CTs, which limited our ability to examine more nuanced associations within and across population groups of high and low density. Fourth, we focused on 2 broad categories of health care facilities, including ambulatory care facilities (which encompass a wide variety of outpatient care services) and pharmacies and drugstores, to limit the consequences of misclassification; however, errors may have remained. Our study specifically spanned a period when urgent care clinics and retail pharmacies and drugstores were increasing, potentially accounting for the longitudinal change in specific areas. 49,50 Limitations of the NETS data have also been noted in our group's prior work. 51

Conclusions
Given the importance of geographic access to care on health outcomes, it is critical to monitor the spatial distribution of health care resources within the context of population health disparities. The Patient Protection and Affordable Care Act expanded overall health care access to primary care through insurance coverage, including Medicaid expansion in several states. However, even insured populations may face geographic barriers to accessing ambulatory care. Therefore, it remains important to understand neighborhood context and geographic access to health care resources when designing population health programs and policies.