Geographic Disparities in Access to Neurologists and Multiple Sclerosis Care in the United States

Abstract

Background and Objectives

A shortage of neurology clinicians and healthcare disparities may hinder access to neurologic care. This study examined disparities in geographic access to neurologists and subspecialty multiple sclerosis (MS) care among various demographic segments of the United States.

Methods

Neurologist practice locations from 2022 CMS Care Compare physician data and MS Center locations as defined by the Consortium of Multiple Sclerosis Centers were used to compute spatial access for all U.S. census tracts. Census tract-level community characteristics (sex, age, race, ethnicity, education, income, insurance, % with computer, % without a vehicle, % with limited English, and % with hearing, vision, cognitive, and ambulatory difficulty) were obtained from 2020 American Community Survey 5-year estimates. Rural-urban status was obtained from 2010 rural-urban commuting area codes. Logistic and linear regression models were used to examine access to a neurologist or MS Center within 60 miles and 60-mile spatial access ratios.

Results

Of 70,858 census tracts, 388 had no neurologists within 60 miles and 17,837 had no MS centers within 60 miles. Geographic access to neurologists (spatial access ratio [99% CI]) was lower for rural (−80.49%; CI [−81.65 to −79.30]) and micropolitan (−60.50%; CI [−62.40 to −58.51]) areas compared with metropolitan areas. Tracts with 10% greater percentage of Hispanic individuals (−4.53%; CI [−5.23 to −3.83]), men (−6.76%; CI [−8.96 to −4.5]), uninsured (−7.99%; CI [−9.72 to −6.21]), individuals with hearing difficulty (−40.72%; CI [−44.62 to −36.54]), vision difficulty (−13.0%; [−18.72 to −6.89]), and ambulatory difficulty (−15.68%; CI [−19.25 to −11.95]) had lower access to neurologists. Census tracts with 10% greater Black individuals (3.50%; CI [2.93–10.71]), college degree holders (−7.49%; CI [6.67–8.32]), individuals with computers (16.57%, CI [13.82–19.40]), individuals without a vehicle (9.57%; CI [8.69–10.47]), individuals with cognitive difficulty (25.63%; CI [19.77–31.78]), and individuals with limited English (18.5%; CI [16.30–20.73]), and 10-year older individuals (8.85%; CI [7.03–10.71]) had higher spatial access to neurologists. Covariates for access followed similar patterns for MS centers.

Discussion

Geographic access to neurologists is decreased in rural areas, in areas with higher proportions of Hispanics, populations with disabilities, and those uninsured. Access is further limited for MS subspecialty care. This study highlights disparities in geographic access to neurologic care.

Introduction

An increasing number of people are affected by neurologic disorders in the United States (US). In 2011, nearly 100 million Americans had at least one neurologic disease.^1,2 The existing shortage of neurologists is projected to grow to 19% by 2025.³ In addition neurologist availability differs significantly across the US with a 5-fold variation between states, ranging from >20% demand to a >20% surplus.³ Furthermore, individuals are often referred to tertiary care centers, which may be even less accessible. The increasing number of individuals with a neurologic condition, the shortages of neurologists, and referral to subspecialty centers affect access to neurologic care.

Health disparities are also known to affect health outcomes and access to care, which was highlighted during the COVID-19 pandemic. US racial and ethnic minority populations had substantially higher rates of infection, hospitalization, and death compared with White populations.⁴ Several of the factors that contributed to these disparities were related to equitable access to health care. In one study, once mortality rates were adjusted for age, sex, insurance, comorbidities, neighborhood deprivation, and site of care, there were no significant differences in mortality between Black and White patients.⁵ In addition to race and ethnicity, individuals residing in rural areas account for 17% of the population and are known to have reduced access to care, worse health outcomes, and often receive care from providers with less specialized training despite similar prevalence in diseases compared with urban areas.^6-8

Inequities in access to neurologists and disparities in clinical outcomes have also been well described for neurologic conditions.^9,10 Black and Hispanic individuals with neurologic conditions are 30% less likely and 40% less likely, respectively, to see outpatient neurologists compared with White and non-Hispanic individuals.¹⁰ Studies have reported higher mortality for stroke, epilepsy, and Parkinson disease for racial and ethnic minorities.¹¹ Furthermore, disparities in receiving treatment for Parkinson, dementia, headache, and muscular dystrophy have been observed.¹¹

For this study, access to multiple sclerosis (MS) centers was further explored for several reasons including it is a leading cause of nontraumatic disability in young adults, has a substantial burden for the healthcare system, requires complex multidisciplinary care, and has known disparities in health outcomes in underrepresented populations.^12-14 Inequities in MS care access have been described for Black and Hispanic patients leading to diagnostic delays and underutilization of necessary healthcare services.^15,16 In addition, underrepresented populations are known to have increased MS-related morbidity including more rapid disability accumulation, increased ambulatory disability, and increased age-adjusted MS mortality.^17-19 These studies highlight systemic, institutional, and interpersonal factors that prevent equitable access to health care, leading to disparate outcomes.

The purpose of this study was to examine disparities in geographic access to neurologists in US populations. In addition, to understand differences in specialty care access, this study also examined spatial access to MS centers to elucidate differences for general and subspecialty care. Geographic proximity is not the only factor that affects access to health care but is an important aspect that needs better characterization. In the US, there were a variety of historical zoning practices that shaped the geographic distribution of individuals in the country and have been described to affect health outcomes.^20,21 The goal of generating these data is to identify the characteristics of underserved areas and communities within the US that could inform interventions to improve access to neurologic care.

Methods

Overview

This is a cross-sectional analysis of geographic access to neurologists and MS centers for U.S. census tracts. The study used Medicare data for neurologist location, 2020 American Community Survey 5-year estimates, 2010 rural-urban commuting area (RUCA) codes, and MS center locations as defined by the Consortium of Multiple Sclerosis Centers (CMSC) to define geospatial access to neurologists, geospatial access to MS centers, and demographic and community characteristics that examine the geospatial access.

Standard Protocol Approvals, Registrations, and Patient Consents

The study was approved as minimal risk to human subjects and exempt from continued review by the Cleveland Clinic IRB (IRB 22-464). The study was exempt from obtaining written informed consent.

Data sets and Study Population

Data were obtained from (1) neurologist street addresses from the 2022 Care Compare national downloadable file from the Centers for Medicare and Medicaid Services (CMS),²² (2) US MS center addresses from the CMSC directory,²³ (3) census tract population counts from the U.S. Census Bureau's 2010 Census Summary File 1, (4) 2010 census tract distance matrix data from the National Bureau of Economic Research's Public Use Data Archive,²⁴ and (5) 2010 census tract polygons from the U.S. Census Bureau's TIGER/Line shapefiles database. For analysis of covariates of geospatial access, data were obtained for census tract-level community characteristics (sex, age, race/ethnicity, education, income, insurance, area deprivation index, % with computer, % without a vehicle, % with limited English, and % with hearing, vision, cognitive, or ambulatory difficulty) from 2020 American Community Survey 5-year estimates. To account for the effects of urban/rural status, we used 2010 census tract-level RUCA codes from the US Department of Agriculture. The categories of race, ethnicity, and sex used are the definitions from the US Census. Responses are voluntary and self-identified. Race categories include White, Black or African American, Asian, American Indian and Alaska Native, Native Hawaiian and Other Pacific Islander, or some other race. Only % Black was included in the below models to avoid multicollinearity in tracts with a low percentage of several race categories. Ethnicity categories are Hispanic/Latino and not Hispanic/Latino. Sex categories are female and male.

Identification of Neurologists and MS Centers

Primary specialty was used to derive active neurologists from 2022 Care Compare data, which is updated monthly for healthcare providers who practiced and billed Medicare or Medicaid in the previous year. MS centers were identified by membership with the CMSC. To be a member of the CMSC, the Center is required to demonstrate evidence of comprehensive care services for MS including medical, psychological, social, and rehabilitative services. Geocoding of neurologist and MS Center street addresses was primarily completed via Census Geocoder batch geocoding and OpenStreetMap Nominatim single address geocoding in the ‘tidygeocoder’ R package,²⁵ with ArcGIS World Geocoding Service in ArcGIS Pro version 2.8 serving as a reserve method to geocode unmatched addresses. Spatial points were spatially joined to a geodatabase containing population counts linked to census tract polygons. For those physicians with multiple practice locations, all locations were retained. CMS Care Compare was selected over other physician location data because of frequent updates on practicing physicians, relative completeness because of the ubiquity of Medicare/Medicaid billing, and previous use in the physician spatial access literature.^26-28

Spatial Access Measures

Using the “access” package version 1.1 from the PySAL Python library,²⁹ we constructed neurologist and MS center spatial access measures using spatial access ratios derived from 2 separate floating catchment area (FCA) methods: the enhanced two-step FCA method (E2SFCA) and the three-step FCA (3SFCA) method. E2SFCA and 3SFCA are more advanced methods for measuring spatial access to care that were developed to build upon the basic two-step FCA (2SFCA) method.^30,31 The original 2SFCA method is a special form of gravity model first created for measuring spatial access to primary care providers (PCPs).³² Further updates included the addition of distance decay functions (E2SFCA) and spatial impedance to account for more realistic healthcare supply and demand (3SFCA).^30,31 A key consideration in building spatial access measures is the maximum distance or time that is reasonable for patient travel, the catchment size.³³ Catchment sizes for primary care are commonly set to 30 minutes.^30,32 Generally, patients travel further for specialty care, and previous work examining distance to specialty care acknowledged that rural patients travel further for care than urban patients.³⁴ Considering these factors, we designated 60 miles (proxy for minutes) as the catchment size for both neurologists and MS care centers. Detailed methods are presented in our previous work.³⁵ In addition, because of the use of open-source tools, there are limitations in existing spatial access to care computation for data sets with low counts of locations leading to missingness at the national level; therefore, the number of computed spatial access measures will differ between neurologists and MS centers.

Statistical Analysis

A two-step modeling approach was used. Logistic regression models were used to assess for any access to a neurologist or MS Center within 60 miles, using all demographics and community characteristics as covariates.³⁶ Characteristics in the final models were chosen based on previous evidence of known association with healthcare disparities or based on specific hypotheses. Because teleneurology has been proposed as an access solution, we included “% with computer” as an indication for access to a device that would be necessary for this intervention. Finally, we included several other variables that had potential to directly affect accessing health care (%without a vehicle and % with limited English) and several disability measures that may be important to consider in neurologic conditions. We next applied linear regression models with the same characteristics to examine log-transformed 60-mile E2SFCA spatial access ratios (SPARs) and 3SFCA SPARs. Odds ratios (OR) were calculated from logistic regression model estimates to assess how the odds of having access differed with each community characteristic. Using estimates from the log-linear models, the percentage change in the degree of spatial access (neurologists to individuals) associated with differences in each community characteristic was calculated. Variance inflation factors were calculated to confirm no excessive multicollinearity among these large multivariable models, and model residuals were visually assessed for normality and heteroscedasticity. Robust standard errors of regression coefficients were estimated using HC3 heteroscedasticity-consistent covariance matrix estimation.³⁷ An additional analysis was conducted to further explore the relationship of rural-urban status with race and ethnicity. Interactions between race and RUCA and between ethnicity and RUCA were added to each model. All statistical analyses were completed in R studio version 4.2. Significance was set at p < 0.01. This level of significance was chosen as a conservative measure. Given the large sample size, we wanted to minimize the possibility of false positive results while still being able to detect meaningful associations.

Data Availability

All data are publicly available from the Centers for Medicare and Medicaid Services, the 2020 American Community Survey, Census Bureau's 2010 Decennial Census, and United States Department of Agriculture. Anonymized data will be shared with qualified investigators by request from the corresponding author for purposes of replicating procedures and results.

Results

Geographic Distribution and Spatial Access Measures

The study area contained 15,113 neurologists at 34,143 practice locations and 185 MS centers. The population-weighted mean number of neurologists in the study area was 11.95 per 100,000 population for E2SFCA and 12.06 per 100,000 population for 3SFCA (Figure 1). The population-weighted mean number of MS centers in the study area was 0.061 centers per 100,000 population for E2SFCA and 0.062 centers per 100,000 population for 3SFCA. To provide another assessment of access separate to the spatial access measures, we used previously published estimates of MS prevalence³⁸ to estimate the number of individuals each MS center would need to care for. For the entire US, the number of individuals per MS center would be 3,664–3,932 (185 centers).³⁸ When applying regional MS prevalence estimates, the number of individuals per MS center would be west 4,749–5,096 (29 centers), midwest 4,086–4,384 (41 centers), northeast 2,330–2,500 (65 centers), and south 4,423–4,746 (50 centers).

Figure 1. Geospatial Access to Neurologists as Measured by (A) E2SFCA and (B) 3SFCA Methods.

Census Tract Community Characteristics and Spatial Access Measures to Neurologists

Of 70,858 tracts (96.9% of all US census tracts) with available covariates and spatial access measure for neurologists, 70,470 had some (nonzero) degree of access to neurologists. We first assessed the odds of any access to a neurologist within 60 miles and then assessed the level of access within 60 miles. Only 388 census tracks lacked any spatial access. The results of the logistic regression are presented in Supplemental Material and were similar to the linear model (eTable 1, links.lww.com/WNL/D268).

The degree of access to neurologic care, as measured by the 3SFCA (% change in ratio of neurologists to individuals), was lower for rural (−80.49%) and micropolitan (−60.50%) areas compared with metropolitan areas (Table 1). In addition, decreases in the estimated degree of access were found in tracts with a 10% higher percentage of Hispanic individuals (−4.53%), men (−6.76%), uninsured (−7.99%), individuals with hearing difficulty (−40.70%), vision difficulty (−13.00%), and ambulatory difficulty (−15.68%). Increases in estimated spatial access to neurologists were found in census tracts with a 10% higher percentage of Black (3.50%), college degree holding individuals (7.49%), individuals with computers (16.57%), individuals without a vehicle (9.57%), individuals with cognitive difficulty (25.63%), individuals with limited English (18.50%), and 10-year older median age (9.85%). The results were similar for the E2SFCA spatial access measures and are presented in Table 1.

Table 1.

Linear Model Examining the Degree of Spatial Access for 3SFCA and E2SFCA

	E2SFCA linear model		3SFCA linear model
Independent variable	% change in access	99% CI	% change in access	99% CI
Total population (1,000 person increase)	1.104	0.693 to 1.517	0.688	0.1 to 1.279
*Percentage male	−3.501	−5.035 to −1.943	−6.755	−8.956 to −4.5
Median age (10 y increase)	9.544	8.306 to 10.795	8.849	7.025 to 10.705
*Percentage Black	3.977	3.607 to 4.347	3.496	2.934 to 4.06
*Percentage Hispanic	−4.774	−5.253 to −4.292	−4.532	−5.227 to −3.832
RUCA: micropolitan vs metropolitan	−43.881	−45.434 to −42.284	−60.502	−62.4 to −58.509
RUCA: rural vs metropolitan	−52.261	−53.796 to −50.675	−80.489	−81.646 to −79.259
*Percentage with college degree	1.528	0.994 to 2.065	7.492	6.673 to 8.318
Median income ($10k increase)	0.679	0.409 to 0.948	−1.831	−2.199 to −1.462
*Percentage with computer	10.467	8.826 to 12.133	16.574	13.82 to 19.395
*Percentage without vehicle	7.892	7.305 to 8.482	9.574	8.689 to 10.467
*Percentage with hearing difficulty	−30.378	−33.503 to −27.106	−40.718	−44.624 to −36.536
*Percentage with vision difficulty	−12.79	−16.541 to −8.87	−13.002	−18.716 to −6.887
*Percentage with cognitive difficulty	15.077	11.754 to 18.499	25.634	19.774 to 31.781
*Percentage with ambulatory difficulty	−14.093	−16.495 to −11.622	−15.68	−19.254 to −11.948
*Percentage with public insurance	−1.445	−2.285 to −0.598	−0.201	−1.478 to 1.092
*Percentage without insurance	−6.343	−7.521 to −5.15	−7.986	−9.721 to −6.216
*Percentage with limited English	18.501	16.855 to 20.17	18.499	16.304 to 20.734
Observations	70,350	70,350
R2	0.338	0.381

*Change in the level of access associated with a 10% increase.

To further explore the relationship between race, ethnicity, and rural-urban location, interactions between race and RUCA and between ethnicity and RUCA were added to the models. The effects of both race and ethnicity percentages were modified by rural-urban designation. Metropolitan census tracts with 10% larger Black populations continued to have higher spatial access (5.25%), but micropolitan and rural census tracks had lower access (−5.57% and −8.24%, respectively) (Figure 2, Table 2). Metropolitan, micropolitan, and rural tracts with 10% larger Hispanic populations all had varying degrees of decreased access (−2.11%; −22.92%, −21.39,%, respectively) (Figure 2, Table 2).

Figure 2. Race (A) and Ethnicity (B) RUCA Interaction.

Table 2.

Linear Model Examining the Degree of Spatial Access for 3SFCA and E2SFCA With Race, Ethnicity, and RUCA Interaction

	E2SFCA linear model		3SFCA linear model
Independent variable	% change in access	99% CI	% change in access	99% CI
Total population (1,000 person increase)	1.132	0.723 to 1.542	0.74	0.159 to 1.325
*Percentage male	−2.842	−4.367 to −1.294	−5.654	−7.852 to −3.404
Median age (10 y increase)	8.736	7.523 to 9.963	7.377	5.609 to 9.174
Race* RUCA interaction
*Percentage Black in metropolitan	4.939	4.584 to 5.296	5.251	4.748 to 5.757
*Percentage Black in micropolitan	−1.035	−2.649 to 0.606	−5.573	−8.437 to −2.619
*Percentage Black in rural	−2.646	−4.876 to −0.364	−8.244	−12.262 to −4.042
Ethnicity*RUCA interaction
*Percentage Hispanic in metropolitan	−3.45	−3.918 to −2.979	−2.108	−2.737 to −1.475
*Percentage Hispanic in micropolitan	−14.437	−15.975 to −12.871	−22.921	−25.726 to −20.01
*Percentage Hispanic in rural	−15.824	−17.979 to −13.613	−21.39	−25.13 to −17.463
*Percentage college degree	2.182	1.651 to 2.716	8.757	7.951 to 9.57
Median income ($10k increase)	0.793	0.526 to 1.061	−1.622	−1.986 to −1.257
*Percentage with computer	8.995	7.421 to 10.592	13.71	11.124 to 16.357
*Percentage without vehicle	7.174	6.605 to 7.746	8.234	7.399 to 9.075
*Percentage with hearing difficulty	−30.019	−33.162 to −26.728	−40.143	−44.094 to −35.913
*Percentage with vision difficulty	−12.309	−16.050 to −8.400	−12.197	−17.91 to −6.086
*Percentage with cognitive difficulty	14.62	11.345 to 17.991	24.742	19.021 to 30.738
*Percentage with ambulatory difficulty	−13.006	−15.423 to −10.52	−13.762	−17.371 to −9.996
*Percentage with public insurance	−1.653	−2.479 to −0.819	−0.552	−1.798 to 0.709
*Percentage without insurance	−6.519	−7.677 to −5.346	−8.306	−9.998 to −6.581
*Percentage with limited English	18.079	16.492 to 19.687	17.788	15.735 to 19.878
Observations	70,350		70,350
R2	0.350		0.398

*Change in the level of access associated with a 10% increase.

Census Tract Community Characteristics and Spatial Access Measures to MS Centers

Of 69,425 tracts (95% of all tracts) which had a complete set of covariates and spatial access measures for a MS center, 51,488 had any spatial access within 60 miles. Counts of tracts differ between neurologists and MS centers because of limitations in existing spatial access to care computation for data sets with low counts of locations (185 MS centers) and large numbers of geographic areas (nationwide census tracts). A two-step modelling approach was used to examine any access to a MS center within 60 miles and the level of access within 60 miles. Because many census tracts (17,937) had a zero value for spatial access, the logistic regression models are presented in the main manuscript and the linear regression models are provided in the Supplemental Material (eTable 2, links.lww.com/WNL/D268). Overall, the results were similar.

Census tracts with a 10% higher percentage of men (OR 0.83), college degree holders (OR 0.83), Hispanic individuals (OR 0.97) individuals with a hearing difficultly (OR 0.26), individuals with a vision difficulty (OR 0.55), individuals with an ambulatory difficulty (OR 0.74), and individuals without insurance (OR 0.70) had lower odds of having any access to a MS center. Micropolitan and rural census tracts had lower odds of having access to a MS center compared with metropolitan tracts (OR 0.37 and OR 0.27, respectively). Census tracts with a population that had a 10-year higher median age (OR 1.36), $10k higher median income (OR 1.32), a 10% higher percentage with a computer (OR 1.44), without a vehicle (OR 2.25), with cognitive difficulty (OR1.17), and with limited English (OR 1.61) had higher odds of having access to MS centers. Census tracts with a 10% higher percentage of Black individuals (OR 0.99) had no difference in access. This finding was in contrast to neurologist access in which census tracts with a higher proportion of Black individuals had different access compared with census tracts with higher proportions of White individuals. The results were similar for the E2SFCA spatial access measures and are presented in Table 3 along with 99% CI for all estimated ORs. The relationship between race, ethnicity, and rural-urban location was also explored for MS centers through the addition of an interaction between race and RUCA and between ethnicity and RUCA. The effects of both race and ethnicity percentages were modified by rural-urban status. Metropolitan census tracts with 10% larger Black populations showed no difference in MS center access (OR 1.01), whereas micropolitan and rural census tracks 10% larger Black populations showed lower access (OR 0.86 and OR 0.90, respectively). Metropolitan census tracts with 10% larger Hispanic populations continued to have similar odds of MS center access (OR 1.01), whereas in micropolitan and rural tracts with 10% larger Hispanic populations showed decreased access (OR 0.72 and OR 0.81, respectively).

Table 3.

Odds of Having Any Access to an MS Center Within 60 Miles for the 2-step FCAM (2SEFCA) and 3-step FCA (3SEFCA)

Logistic model 2SEFCA and 3SEFCA
Independent variable	Odds ratios	99% CI
Total population (1,000 person increase)	1.065	1.044–1.087
*Percentage male	0.833	0.780–0.890
Median age (10 y increase)	1.364	1.300–1.430
*Percentage Black	0.987	0.971–1.003
*Percentage Hispanic	0.973	0.952–0.994
RUCA: micropolitan vs metropolitan	0.373	0.345–0.404
RUCA: rural vs metropolitan	0.272	0.248–0.297
*Percentage with college degree	0.828	0.806–0.850
Median income ($10k increase)	1.323	1.298–1.350
*Percentage with computer	1.441	1.362–1.524
*Percentage without vehicle	2.252	2.136–2.375
*Percentage with hearing difficulty	0.262	0.222–0.308
*Percentage with vision difficulty	0.549	0.460–0.654
*Percentage with cognitive difficulty	1.173	1.039–1.325
*Percentage with ambulatory difficulty	0.740	0.660–0.831
*Percentage with public insurance	1.027	0.990–1.065
*Percentage without insurance	0.697	0.662–0.734
*Percentage with limited English	1.612	1.504–1.730
Observations	69,425
R2	0.244

*Change in odds of access associated with a 10% increase.

Discussion

Lower spatial access to neurologists was seen for census tracts that were micropolitan and rural, had a higher proportion of underrepresented minorities, uninsured people, and disabled individuals. For MS centers, covariates of access were similar to neurologists, but there were fewer MS centers leading to poorer geospatial access across the US, indicating further barriers to receiving subspecialty care.

Previous studies have found variation in neurologist density across the US when evaluating at the level of hospital referral region (HRR) or state.^3,39 These studies estimated 5.2 neurologists per 100,000 in 2012 and 22.3 neurologists per 100,000 Medicare beneficiaries in 2021. This study's estimate of 12 neurologists per 100,000 individuals is an overall improvement from the 2012 estimate but not as promising as the 2021 Medicare beneficiary estimate. A limitation of these previous study estimates is that the geographic unit of HRR is a large region and focused on tertiary care centers. For this study, the goal was to better quantify the spatial availability of both outpatient general neurology and subspecialty MS center care at the census tract level to provide a granular assessment of community characteristics associated with poor access. There are >70,000 census tracts in the US as opposed to only 306 HRRs.⁴⁰ This study uniquely identified demographic characteristics at a granular level that should be considered when developing healthcare delivery interventions.

Black and Hispanic individuals with neurologic conditions are 30% less likely and 40% less likely, respectively, to see outpatient neurologists compared with White and non-Hispanic individuals.¹⁰ In this study, census tracts with a higher percentage of Hispanic individuals consistently had poorer access to neurologists even when accounting for rural/urban status. Census tracts with a higher percentage of Black individuals had discrepant results when accounting for rural/urban status. When an interaction between race and RUCA is not included, tracts with a higher percentage of individuals identifying as Black were associated with greater access to neurologists. When accounting for RUCA designations, tracts with larger Black populations only had better access in metropolitan areas. Micropolitan and rural areas with larger Black populations were associated with lower access. Our results have several important implications: (1) poor geospatial access likely potentiates healthcare disparities in underrepresented populations and (2) healthcare delivery interventions to target poor access to care should be tailored to demographics and community characteristics. This need is specifically highlighted in the census tracts with a higher percentage of Black individuals. Although our analysis suggests that communities with higher Black populations may have better access to care, when RUCA designation is accounted for, micropolitan and rural Black communities have poorer access than White communities. Our results indicate that healthcare delivery interventions for rural Black communities may need to be different than urban Black communities where proximity to neurologists is not the main driver to poor access to care. It is important to highlight that better geospatial access does not ensure better access to care. Although urban Black communities may be geographically close to neurologists, there are barriers to healthcare access that are independent of proximity of care and should be further explored. Future studies and interventions should include strategies that incorporate culturally appropriate communication and outreach, expand free communication platforms, address social needs (e.g., housing, food), and connect patients with existing community resources for these needs.⁴

Better geospatial access to health care has been associated with improved health outcomes for several chronic conditions.^41-43 In oncology, increased distance from cancer treatment location was associated with more advanced disease at diagnosis, inappropriate treatment, worse prognosis, and worse quality of life.⁴² Similarly, better geospatial access to preventive care through primary care physicians and cardiologists is associated with fewer hospitalizations.^41,43,44 This study and those for other chronic conditions demonstrate the importance of geospatial access as another dimension to addressing healthcare disparities.^41-43 In addition, patterns of geographic access to healthcare facilities change. and these changes disproportionally affect racial/ethnic minorities and communities with higher levels of poverty. In one study that assessed change in access to ambulatory care facilities between 2000 and 2014, census tracts with predominantly non-Hispanic Black individuals, Hispanic individuals, racially/ethnically mixed individuals, and higher levels of poverty had higher odds of healthcare facilities closure.⁴⁵ Altogether, shifting landscape of access to care and associations with poor outcomes highlight the need to consider geospatial access when developing healthcare delivery interventions for underrepresented populations. In addition to racial and ethnic minorities, census tracts with higher percentages of individuals with hearing, vision, and ambulatory difficulties were found to have poorer geospatial access to neurologic care. It has previously been described that individuals with physical disabilities have up to 75% higher odds of having unmet medical needs.⁴⁶ These observed health disparities are because of a combination of individual, environmental, and societal factors that affect health outcomes.⁴⁷ Neurologic conditions frequently result in physical disabilities, further indicating the importance of spatial access.

Finally, disparities in geospatial access are compounded when subspecialty neurologic care is needed. Only 388 US census tracts had no neurologists within 60 miles, whereas a staggering 17,937 census tracts had no MS center within 60 miles, suggesting particularly poor spatial access to the subspecialty care provided at MS centers. This is not necessarily surprising because there are only 185 MS centers in the US, yet most of these centers are located in metropolitan areas. Based on our results, vulnerable populations, such as rural patients with MS with disabilities, likely have difficulty traveling to these centers, greatly reducing their effectiveness in addressing disparities. Currently in rural communities, many people living with MS are managed by local general neurologists that are expected to care for a wide range of conditions. In the field of MS, there have been rapid changes including the approval of greater than 10 new disease-modifying therapies in the last decade. Furthermore, if a patient is referred to a specialist, there are physical and financial barriers that may limit their ability to receive this care. The indirect costs of MS care are estimated to be 22.1 billion, which includes costs for transportation and time off work that are substantially higher if rural patients are expected to travel several hours for MS care.¹² In addition, there is geographic variation of MS center location. Although the northeast region has the highest estimated MS prevalence (377.4 per 100,000), there are substantially more MS centers (n = 65), leading to a potentially better access ratio of 2,330–2,500 individuals with MS per MS center. Conversely, the west region has a lower MS prevalence estimate (272.7 per 100,000), but only 29 MS centers with a doubled ratio of individuals per MS center (4,749–5,096 individuals per center). Beyond specialty MS care, there are trends that will likely lead to further centralization of neurologic care. A recent editorial argued that “modern neurology training is failing outpatients.”⁴⁸ The authors highlighted only 20% of neurology residency is dedicated to outpatient training, the quality of a residency program is often judged by fellowship placements of their trainees, and general neurology mentorship is limited. Furthermore, 90% of neurology residents have reported plans to pursue a fellowship, and a higher proportion of neurologists work at academic medical centers in urban areas.^39,49 As neurologic care advances and has the potential to be further centralized, it is important to consider the impact on care access. There are several potential policy considerations and interventions that could be considered. Teleneurology could help address some of these needs but would require modifications to state licensure requirements, allowing clinicians to treat patients regardless of their physical location. Ensuring equitable access to affordable broadband internet through acts such as the “Accessible, Affordable Internet for All Act.” In addition, continued policy interventions such as the Conrad 30 ARC J-1 visa waiver to promote clinicians practicing in rural locations.

This study had several limitations. The analysis was performed at the census-tract level and not the individual level. A census tract can provide excellent granularity, but it does not provide insight into individuals. Similarly, access to care was defined by the location of neurologists and MS centers, but this study did not account for where individuals with neurologic conditions are located. Previous studies have shown that the prevalence of neurologic conditions does not vary by region; therefore, this study still aides with a better understanding of access to neurologic care.³⁹ The use of CMS Care Compare data has the potential to over or underrepresent density of neurologists. Data used are estimated to capture 97% of neurologists, leading to a slight underrepresentation, but other sources including the AMA and ABPN have similar limitations. Conversely, since the data set used includes all neurologist practice locations, but does not weight “primary” location, it is possible that access is overestimated in certain areas. This can also be viewed as a strength given that additional practice locations are not ignored, which could have led to further underestimation of access. In addition, this study only accounted for geographic access to care and did not account for other individual factors that influence healthcare delivery and access. We acknowledge that proximity to a neurologist does not ensure care access and future studies will need to investigate other unique barriers. An additional limitation is that this study only quantified access to neurologists and did not account for the location and care provided by advanced practice providers. As care teams expand and more frequently include a variety of clinicians, this should be considered in future studies. Regarding subspecialty care, only the location of MS centers in the US was used to quantify access to subspecialty care. Future studies should be performed to evaluate unique geographic access barriers for other subspecialties. Also, there was more missingness for MS center spatial access to care measures because of limitations in existing computational tools for data sets with low counts of locations (185 MS centers) and large numbers of geographic areas (nationwide census tracts). In our case, it was compulsory to use new open-source tools for computation of spatial access to care measures because more formalized tools are both extremely costly and generally unable to handle large data sets (e.g., nationwide census tracts). Finally, the use of Euclidean distance matrices has the potential to overestimate access to care. This method does not account for transit systems (road networks or public transportation), which may be especially important for urban areas with higher traffic volumes and complex transit systems. Advanced methods that account for these complexities require significant computational requirements but should be considered in future studies.

As recently highlighted by Robbins et al.,¹¹ the fundamental cause of “neurodisparities” are complex and rooted in structural inequity, socioeconomic disadvantage, unequal treatment, and racism. The impact of location on racism and health disparities is well described. In the US, there is a longstanding history of radical zoning practices such as “redlining” that shaped the geographic characteristics of the country today. These historical practices continue to affect health outcomes, including for neurologic conditions.^20,21 This study supports the impact of geographic location on healthcare access specifically for underrepresented racial and ethnic groups and those with physical disabilities. To successfully address healthcare disparities, geographic factors need to be considered. This study identified spatial access disparities to neurologic care that can aide clinicians, researchers, and policymakers in the development of equitable healthcare delivery models.

Glossary

CMS

Centers for Medicare and Medicaid Services

CMSC

Consortium of Multiple Sclerosis Centers

FCA

floating catchment area

HRR

hospital referral region

MS

multiple sclerosis

OR

odds ratios

PCPs

primary care providers

RUCA

rural-urban commuting area

SPARs

spatial access ratios

Appendix. Authors

Name	Location	Contribution
Marisa McGinley, DO	Cleveland Clinic	Designed and conceptualized the study, interpreted the data, drafted and revised the manuscript, and major role in data collection and management
Tucker Harvey, MS	Cleveland Clinic	Statistical analysis and revised the manuscript for intellectual content and major role in data collection and management
Rocio Lopez, MS, MPH	University of Colorado	Statistical analysis and revised the manuscript for intellectual content
Daniel Ontaneda, MD, PhD	Cleveland Clinic	Designed and conceptualized the study, interpreted the data, and revised the manuscript for intellectual content
R. Blake Buchalter, PhD, MPH	Cleveland Clinic	Designed and conceptualized the study, analyzed the data, drafted and revised the manuscript and major role in data collection and management

Study Funding

This manuscript has been funded by National Center for Advancing Translational Sciences (KL2TR002547).

Disclosure

M.P. McGinley has received consulting fees from Genentech, EMD Serono, and Octave and has received research support from Novartis, Biogen, and Genentech. She also receives funding from the NIH and a KL2 (KL2TR002547) grant from the Clinical and Translational Science Collaborative of Cleveland and from the National Center for Advancing Translational Sciences (NCATS) component of the NIH.; T. Harvey has nothing to disclose; R. Lopez has nothing to disclose.; D. Ontaneda has received research support from the NIH, National Multiple Sclerosis Society, Patient-Centered Outcomes Research Institute, Race to Erase MS Foundation, Genentech, Genzyme, and Novartis. He has also received consulting fees from Biogen Idec, Genentech/Roche, Genzyme, Novartis, and Merck; R.B. Buchalter has nothing to disclose. Go to Neurology.org/N for full disclosures.