Disparities in Geographic Access to Cardiac Rehabilitation in Los Angeles County

Abstract

Background

Exercise‐based cardiac rehabilitation (CR) is known to reduce morbidity and mortality for patients with cardiac conditions. Sociodemographic disparities in accessing CR persist and could be related to the distance between where patients live and where CR facilities are located. Our objective is to determine the association between sociodemographic characteristics and geographic proximity to CR facilities.

Methods and Results

We identified actively operating CR facilities across Los Angeles County and used multivariable Poisson regression to examine the association between sociodemographic characteristics of residential proximity to the nearest CR facility. We also calculated the proportion of residents per area lacking geographic proximity to CR facilities across sociodemographic characteristics, from which we calculated prevalence ratios. We found that racial and ethnic minorities, compared with non‐Hispanic White individuals, more frequently live ≥5 miles from a CR facility. The greatest geographic disparity was seen for non‐Hispanic Black individuals, with a 2.73 (95% CI, 2.66–2.79) prevalence ratio of living at least 5 miles from a CR facility. Notably, the municipal region with the largest proportion of census tracts comprising mostly non‐White residents (those identifying as Hispanic or a race other than White), with median annual household income <$60 000, contained no CR facilities despite ranking among the county's highest in population density.

Conclusions

Racial, ethnic, and socioeconomic characteristics are significantly associated with lack of geographic proximity to a CR facility. Interventions targeting geographic as well as nongeographic factors may be needed to reduce disparities in access to exercise‐based CR programs. Such interventions could increase the potential of CR to benefit patients at high risk for developing adverse cardiovascular outcomes.

Nonstandard Abbreviations and Acronyms

ACS

American Community Survey

CR

cardiac rehabilitation

LA

Los Angeles

SPA

service planning area

Clinical Perspective.

What Is New?

• When compared with non‐Hispanic White individuals, racial and ethnic minorities more frequently live ≥5 miles from a CR facility.

• This finding is most pronounced for non‐Hispanic Black individuals, who are nearly 3 times more likely to live >5 miles from the nearest CR facility.

What Are the Clinical Implications?

• Clinicians should consider distance to CR facilities as a potential barrier to accessing CR, particularly for racial and ethnic minorities.

• Public policy should consider geographic distribution of CR facilities or telehealth resources as an approach to reducing racial and ethnic disparities in CR participation.

Cardiovascular disease, including heart failure, myocardial infarction, and valvular heart disease, affects >80 million Americans and remains the leading cause of death in the United States. ¹ Long‐term management and secondary prevention efforts provide the greatest benefit for affected patients when they include enrollment in cardiac rehabilitation (CR), a program that incorporates graduated cardiovascular activity, education, risk factor modification, and social support services. ² Compared with eligible patients who do not complete CR, those who do complete CR experience improved health‐related quality of life, fewer hospitalizations, and reduced cardiovascular mortality. ³ , ⁴ , ⁵ , ⁶ , ⁷ In light of these data, expert consensus guidelines have emphasized the importance of CR for patients with heart failure and coronary artery disease, among other conditions, with the highest level of recommendations. ⁸ , ⁹

Despite the well‐established benefits of CR, enrollment of eligible patients is low at 24%. ¹⁰ , ¹¹ , ¹² , ¹³ Even lower enrollment rates are seen among racial and ethnic minority patients and those of lower socioeconomic status, ¹² , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ contributing to the adverse outcomes experienced by these vulnerable patient groups recovering from cardiovascular disease. ¹⁸ , ¹⁹ Previously reported barriers to CR participation include, among several factors, the time and distance required for a patient to travel to a CR facility. ²⁰ , ²¹ , ²² In fact, prior research has suggested that patients living at least 3 miles from a facility are 22% less likely to engage with CR. ²³ The Fundamental Cause Model of Health Disparities highlights the role that social conditions, including living environments, may play in health and disease. ²⁴ To the end, the extent to which persistent racial, ethnic, and socioeconomic disparities may be related to the distance between where patients live and where CR facilities are located remains unclear. ²⁵ , ²⁶ To investigate this possibility, we evaluated the association between sociodemographic characteristics and geographic proximity to CR facilities across a large and diverse urban county.

Methods

Analyzed data were obtained from public sources. Requests to access the used data set may be sent to Cedars‐Sinai Medical Center at biodatacore@cshs.org.

We performed a geospatial analysis of CR facilities across Los Angeles (LA) County that involved mapping neighborhood‐level sociodemographic characteristics with the location of CR facilities. Study protocols were reviewed by the Cedars‐Sinai Institutional Review Board and received a determination of not human subject research and an exemption from informed consent.

Data Acquisition

We identified CR programs that were actively operating in LA County as of 2020, using 3 sources: (1) the American Association of Cardiovascular and Pulmonary Rehabilitation program directory ²⁷ ; (2) the California Office of Statewide Health Planning and Development list of licensed health care facilities ²⁸ ; and (3) the California Society for Cardiac Rehabilitation program directory. ²⁹ For each identified program, we contacted the facilities by telephone to confirm active operational status and the location where CR services are provided. We also obtained census tract‐level demographic and socioeconomic characteristic data from the 2019 American Community Survey (ACS) release of summary data for 2014 through 2018. The entire geography of the United States is divided into census tracts, which are small geographic subdivisions of a given county, within a given state or region. These “neighborhood”‐like geographic subdivisions allow for estimating the distribution of sociodemographic characteristics within and across specific regional areas. More important, the census tract approach to national data collection permits aggregation and analysis of regional variation in sociodemographic characteristics that can be calculated at the neighborhood level without the need to access or manage individual‐level data. ³⁰ Demographic and socioeconomic data from the ACS include both point estimates and margins of error for proportions as well as counts. We then linked ACS data census tract polygons, on top of which we geocoded the operational CR facilities. ³¹ We determined proximity to a CR facility for each census tract based on whether any portion of that census tract was located within a 5‐mile Euclidian distance from an active CR facility. ³² Five miles was selected on the basis of results from prior research examining the relationship between distance and CR participation. ²³ Given that sociodemographic characteristics were captured at the census tract (ie, neighborhood) level, we sought to examine how sociodemographic variations by census tract compared and contributed to variations seen in CR accessibility at the level of larger geographic regions, and particularly those larger regions designated as LA County Service Planning Areas (SPAs) given these are used for planning the organization and distribution of services for area‐specific residents. For this reason, we mapped both census tracts and CR facilities onto maps of the municipally designated SPA regions to provide a contextual framework for interpreting results of all analyses. ³³ LA County is geographically composed of 8 total SPA regions, which are distinct geographic areas designated for planning the organization and distribution of services relevant to the specific needs of residents living in each region. We performed all geocoding analyses using ArcGIS. ³⁴ We defined “regions of marked disparity” as census tracts with a population density of ≥6000 per square mile but located ≥5 miles from the nearest CR facility.

Statistical Analysis

We generated summary statistics for each census tract using weights based on population size. We excluded census tracts with missing population values (n=16) and those without available data for geographic information system mapping (n=18). We compared select sociodemographic variables between census tracts that were proximate to a CR facility (within 5 miles) versus distant from a CR facility (5 miles or farther), using Wilcoxon rank‐sum test for weighted median values.

Because the data derived from the ACS are provided at the census tract level rather than the individual person level, we used multivariable Poisson regression models to identify demographic and socioeconomic characteristics of census tracts associated with proximity to (<5 miles) or distance from (≥5 miles) CR facilities. In prioritizing a parsimonious statistical model for identifying the potential factors associated with lack of proximity to a CR facility, we included median household income as the measure considered most representative of socioeconomic status. ³⁵ Because the ACS provides estimated counts of residents within each census tract, we also calculated the proportion of residents lacking proximity to CR facilities across key demographic and socioeconomic characteristics. We used these proportions to calculate prevalence ratios, although we were unable to adjust for covariates as the data were available only at the aggregate level. Calculation of 95% CIs for combined estimates was performed in accordance with ACS guidelines. ³⁶ All P values were adjusted using the Bonferroni method for multiple comparisons. ³⁷ We performed all statistical analyses using R (v3.6.1) and considered differences to be statistically significant at a threshold of P<0.05.

Results

We identified a total of 26 CR facilities across LA County that we also confirmed to be actively operational (Figure 1). The highest number of these facilities were located in SPA region 2 (San Fernando Valley) and SPA region 4 (Metro LA), each containing 7 CR facilities. No CR facilities were located in SPA region 1 (Antelope Valley) or SPA region 6 (South). Notably, these 2 SPA regions had the lowest percentage of census tracts with median household incomes ≥$60 000 a year; SPA region 6 had the highest percentage of census tracts for which ≥95% of residents identified as non‐White (those identifying as Hispanic or a race other than White) (Table 1). SPA 6 also contained the highest proportion of census tracts qualifying as regions of marked disparity. A geocoded map of CR facilities in LA County, overlaid on SPAs and regions of marked disparity, is represented in Figure 1.

Figure 1. Regions of marked disparity in access to cardiac rehabilitation facilities relative to population density in Los Angeles County.

Portions of this figure are adapted from the Los Angeles County Community Health Services website. ⁵⁸

Table 1.

Sociodemographic Characteristics of the LA County SPA Regions

	Estimated population (No. of residents)*	No. (%) of census tracts with ≥95% non‐White residents	No. (%) of census tracts with ≥15% residents aged ≥65 y	No. (%) of census tracts with median household income ≥$60 000	No. (%) of census tracts representing areas of marked disparity†
SPA
1. Antelope Valley	392 120	0 (0)	15 (2)	35 (3)	19 (23.2)
2. San Fernando Valley	2 221 100	38 (7)	186 (25)	332 (27)	7 (1.4)
3. San Gabriel Valley	1 784 370	100 (17)	180 (24)	268 (22)	26 (6.7)
4. LA Metro	1 152 503	71 (12)	71 (10)	102 (8)	0 (0)
5. LA West	650 667	1 (0)	88 (12)	140 (11)	11 (6.9)
6. LA South	1 037 211	183 (31)	24 (3)	21 (2)	54 (23.8)
7. LA East	1 305 648	115 (20)	62 (8)	140 (11)	25 (8.7)
8. South Bay	1 508 223	76 (13)	118 (16)	186 (15)	64 (19.1)

LA indicates Los Angeles; and SPA, service planning area.

Non‐White residents includes those identifying as Hispanic or a race other than White.

^*Based on American Census Survey Data. ³⁰

^† Areas of marked disparity defined as being located at least 5 miles from a cardiac rehabilitation facility and having a population density of >6000 residents per square mile.

Census Tract Demographic Characteristics

In total, 2346 census tracts were identified in LA County, of which 2312 were included in our analysis, with 85.0% of residents living within 5 miles of a CR facility. Across all census tracts, the median percentage of residents aged >65 years was 12.1% (interquartile range [IQR], 8.8%–16.2%); the proportion of individuals aged >65 years was similar in census tracts located closer to CR facilities compared with in census tracts located farther from CR facilities (median [IQR], 12.1% [8.9%–16.3%] versus 11.3% [8.2%–15.4%]; P=1.00). Hispanic/Latinx individuals represented the highest median ethnic or racial subgroup proportion across LA County census tracts (47.4% [IQR, 21.9%–72.9%]), followed by non‐Hispanic White (16.5% [IQR, 4.8%–45.5%]), Asian (9.4% [IQR, 3.6%–18.5%]), and non‐Hispanic Black (3.6% [IQR, 1.2%–8.9%]) individuals (Table 2). Census tracts located within 5 miles of a CR facility were composed of a larger proportion of non‐Hispanic White (17.6% [IQR, 5.3%–47.1%] versus 11.7% [IQR, 2.5%–32.5%]; P<0.001) and Asian (10.2% [IQR, 4.7%–19.1%] versus 4.3% [IQR, 1.2%–13.1%]; P<0.001) residents than census tracts located ≥5 miles from a CR facility. Conversely, census tracts located within 5 miles of a CR facility were composed of smaller proportions of non‐Hispanic Black residents than those located ≥5 miles from a CR facility (3.2% [IQR, 1.1%–7.0%] versus 9.9% [IQR, 3.1%–24.5%]; P<0.001).

Table 2.

Demographic Characteristics of LA County Census Tracts, Overall and Categorized by Geographic Distance From a CR Facility

Characteristics	LA County (n=2312)	Within 5 miles of nearest facility (n=1973)	Not within 5 miles of nearest facility (n=339)	P value*
% of Residents in racial and ethnic group, median (IQR)
Non‐Hispanic Black	3.6 (1.2–8.9)	3.2 (1.1–7)	9.9 (3.1–24.5)	<0.001
Asian	9.4 (3.6–18.5)	10.2 (4.7–19.1)	4.3 (1.2–13.1)	<0.001
Native Hawaiian or other Pacific Islander	0 (0–0)	0 (0–0.1)	0 (0–0)	1.000
Hispanic/Latinx	47.4 (21.9–72.9)	48.2 (21.4–74.7)	45.3 (24.0–69.1)	1.000
Non‐Hispanic White	16.5 (4.8–45.5)	17.6 (5.3–47.1)	11.7 (2.5–32.5)	<0.001
Other†	2.2 (0.9–4)	2.2 (0.8–4.1)	2.0 (1–3.9)	1.000
% of Residents aged >65 y, median (IQR)	12.1 (8.8–16.2)	12.1 (8.9–16.3)	11.3 (8.2–15.4)	1.000
Household income, median (IQR), $	63 184 (45 964–85 462)	63 756 (46 488–85 594)	58 059 (43 127–84 049)	0.770

Summary statistics are weighed by the population of each census tract. Census tracts with missing values for population are excluded (n=16). CR indicates cardiac rehabilitation; IQR, interquartile range; and LA, Los Angeles.

^* P values for weighted median income come from Wilcoxon rank‐sum test. All P values adjusted using the Bonferroni method for multiple comparisons.

^† Other includes Alaska Native/American Indian, multiple races, and other race.

An estimated 11.9% (95% CI, 11.76%–12.11%) of non‐Hispanic White patients lived ≥5 miles from the nearest CR facility (Table S1). Compared with non‐Hispanic White individuals, individuals in certain racial and ethnic minority groups more frequently lived ≥5 miles from a CR facility. This difference was largest for non‐Hispanic Black (prevalence ratio, 2.73 [95% CI, 2.66–2.79]), American Indian/Alaska Native (prevalence ratio, 1.70 [95% CI, 1.36–2.04]), and Hispanic/Latinx (prevalence ratio, 1.19 [95% CI, 1.16–1.21]) individuals. Similarly, individuals with annual household incomes of <$60 000 more frequently lived ≥5 miles from a CR facility compared with those with annual household incomes of ≥$60 000 (prevalence ratio, 1.08 [95% CI, 1.06–1.11]) (Figure 2A and 2B). Conversely, those aged >65 years (prevalence ratio, 0.98 [95% CI, 0.96–0.99]) more frequently lived within 5 miles of a CR facility compared with those aged ≤65 years.

Figure 2. Lack of access to cardiac rehabilitation facility by sociodemographic characteristics.

A, Differences by race and ethnicity are shown, with non‐Hispanic White individuals considered the referent group. B, Differences by socioeconomic characteristics are shown, with the referent groups being individuals aged <65 years and household income ≥$60 000/year. *AI/AN, American Indian/American Native. ^†NHPI, Native Hawaiian/Pacific Islander. PR indicates prevalence ratio.

Recognizing that CR facilities are more likely to be located in more densely populated areas, we performed 2 sensitivity analyses, excluding less densely populated regions. First, we calculated the proportion of residents lacking proximity to CR facilities after excluding the least densely populated SPA (SPA 1). We subsequently repeated this analysis excluding all census tracts with <6000 residents per square mile. The racial and ethnic disparities in proximity to CR facilities persisted, and in fact were more pronounced, in both sensitivity analyses (Figures S1 and S2).

Multivariable Poisson Regression

Stratification of key census tract characteristics demonstrated an increasing prevalence of census tracts located ≥5 miles from a CR facility as the proportion of non‐White residents increased. Compared with census tracts with <50% non‐White residents, tracts with ≥50% non‐White residents were more frequently located ≥5 miles from a CR facility, with an increasing prevalence ratio as the percentage of non‐White residents increased: 50% to 80% non‐White residents (prevalence ratio, 1.46 [95% CI, 1.01–2.10]), 80% to 95% non‐White residents (prevalence ratio, 1.64 [95% CI, 1.11–2.42]), and ≥95% non‐White residents (prevalence ratio, 2.19 [95% CI, 1.44–3.33]) were more frequently located ≥5 miles from a CR facility (Table 3). There were no appreciated differences by age or median household income.

Table 3.

Census Tract‐Level Characteristics Associated With Lack of Proximity to a CR Facility

Characteristic	Unadjusted prevalence ratio (95% CI)	Adjusted prevalence ratio (95% CI)
Proportion of non‐White residents, %
<50	Referent	Referent
50–80	1.31 (0.91–1.88)	1.46 (1.01–2.10)
80–95	1.45 (1.04–2.02)	1.67 (1.14–2.46)
≥95	1.93 (1.41–2.65)	2.25 (1.48–3.40)
Proportion of residents aged >65 y, %
<15	Referent	Referent
≥15	0.83 (0.66–1.05)	0.96 (0.74–1.26)
Median household income, $
<45 000	Referent	Referent
45 000–65 000	0.87 (0.66–1.14)	0.97 (0.74–1.27)
65 000–85 000	0.68 (0.49–0.94)	0.88 (0.62–1.26)
≥85 000	0.80 (0.60–1.06)	1.29 (0.88–1.88)

CR indicates cardiac rehabilitation.

Discussion

Our study results indicate that sociodemographic characteristics are significantly associated with geographic proximity to CR facilities across the diverse urban populous of LA County. In particular, we found that racial and ethnic minorities, compared with non‐Hispanic White individuals, more frequently live geographically far from a CR facility; this disparity was greatest for non‐Hispanic Black individuals, who are nearly 3 times more likely to live ≥5 miles from the nearest rehabilitation establishment. Exercise‐based CR is well known to reduce morbidity and mortality for patients with cardiac conditions, and our findings suggest that persistent sociodemographic disparities in the use of guidelines recommended CR could well be related to the distance between where patients live and where facilities are located.

Our findings expand from a well‐established body of evidence documenting racial and ethnic disparities in CR use. Prior studies have demonstrated that racial and ethnic minority patients with cardiac conditions are 20% to 50% less likely than White patients to be referred for CR ³⁸ , ³⁹ and, among those referred, non‐White patients are only half as likely as White patients to initiate CR ¹⁶ for reasons that have remained unclear. ¹⁰ , ¹⁶ , ¹⁷ , ³⁸ More important, the potential role of geographic distance in contributing to these disparities has not been explored previously, despite prior research having shown that greater distance to a CR facility is related to lower participation. ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ Over half of patients referred to CR report that they would not have attended the program if the facility was located farther from home. ⁴⁶ Furthermore, greater distance from a patient's home to CR facilities reduces referral rates from physicians. ²² , ⁴⁷ Our study findings now indicate that racial and ethnic disparities may well arise, at least in part, from geographic disparities and structural inequities.

Our analyses focused on understanding how sociodemographic and geographic factors related to CR enrollment may intersect and disproportionately affect racial and ethnic minority populations. Indeed, we found that some municipal regions, including the communities represented by SPA regions 1 and 6, had no CR facilities at all. Although SPA region 1 is a less densely populated area, SPA region 6 is the fourth most populous region in the county, with >1 million residents. Given that our analyses incorporated population weights, the lack of a CR facility in the smaller‐sized but more densely populated SPA 6 region was a far more substantial contributor to measured disparity than lack of a CR facility in the larger‐sized but less densely populated SPA 1 region. The regional disparity seen in the population‐dense areas was even more profound when we conducted sensitivity analyses excluding the least population‐dense areas. In effect, our results highlight a population‐resource mismatch particularly for the municipal region, represented by SPA 6, that had not only the largest disparity in local access to CR but also happens to include the county's highest proportion of census tracts with >95% non‐White residents. Notably, both the municipal regions represented by SPAs 1 and 6 have the highest percentage of census tracts with low median household incomes. Given that both non‐White and lower‐income individuals are known to experience greater risks for primary cardiovascular disease outcomes, ⁴⁸ , ⁴⁹ the relative lack of access to CR for certain area residents represents an even greater mismatch of resource to need rather solely of resource‐to‐population density, and one that may well be contributing to disparities in secondary cardiovascular outcomes. ⁵⁰ , ⁵¹ , ⁵²

There are several potential mechanisms underlying our findings. In a mixed methods analysis, including aggregated descriptive statistics and provider interviews, Mead et al identified that potential drivers of disparities in access to CR could include misaligned financial incentives limiting investment in CR programs. ⁵³ Others have also proposed that flawed reimbursement models of care lead certain geographic areas to lack the financial resources needed to attain CR facility profitability or even simple viability. ⁵³ These economic factors, among others, likely contribute to the observed wide geographic variability in location of CR programs, with the number of CR facilities per square mile varying nearly 600‐fold by state. ⁵⁴ Neighborhood‐level socioeconomic factors are also likely contributors, with individuals living in high deprivation areas nearly 60% less likely to initiate CR than those in nondeprived areas. ⁵⁵ Our results demonstrate geographic clustering of CR facilities near communities that are predominately non‐Hispanic White race and ethnicity and wealthier. Taken together, our data suggest that although the juncture of sociodemographic and geographic factors is likely not the sole driver of disparities, policy efforts to more optimally structure economic incentives may represent at least one method by which equitable access to CR could be improved for socioeconomically disadvantaged populations. Such interventions are not limited to the geographic positioning of CR facilities and should consider the equitable expansion of telemedicine‐based CR, particularly to traditionally underserved patients and those living farther away from facilities.

Several limitations of our study warrant consideration when interpreting results. Our analyses were conducted for a single metropolitan county, and additional analyses conducted in separate geographic locations are needed to validate our findings. Although LA County is among the most populous counties in the United States, with racial and ethnic diversity that is representative of the population as a whole, the extent to which our findings may be generalizable to other regions (eg, rural communities and locales with more homogeneous sociodemographics) is not known and may be limited. Our study also relied on data gathered from the ACS survey and publicly available listings of CR facilities, potentially leading to selection bias. More important, the use of aggregated data at the census tract does not fully account for the heterogeneity in sociodemographic factors within a given tract. Fortunately, ACS data have been used in multiple previously published studies and have been found to be reliable for population studies of this nature. ³³ , ⁵⁶ , ⁵⁷ In addition, we identified CR facilities from 3 unique data sources and contacted each via telephone to confirm active status, as well as the location at which services are provided, to ensure accuracy. Finally, our study design was cross‐sectional and, therefore, limited in the ability to establish causality or temporality of associations. Recognizing the logistical challenges in conducting prospective and controlled studies in this field, our analyses of the available data can yet offer new information with relevance for addressing persistent knowledge gaps about disparities in cardiovascular disease outcomes.

In summary, we found that racial, ethnic, and socioeconomic characteristics are significantly associated with lack of geographic proximity to a CR facility. Despite evidence‐based and clinical guideline recommended benefits of CR for patients recovering from common cardiac conditions, we found that neighborhoods in our county with a greater proportion of non‐White and lower‐income residents are significantly more likely to be located farther from CR facilities, effectively limiting access to CR for many patients and potentially rendering participation infeasible. Interventions targeting geographic and nongeographic factors are likely needed to reduce persistent disparities in access to CR programs and, in turn, increase the potential of CR to benefit all patients at risk for the sequelae of common cardiac conditions.

Sources of Funding

This work was supported in part by K23‐HL153888. Additional support provided by The Barbra Streisand Women's Cardiovascular Research and Education Program, the Linda Joy Pollin Women's Heart Health Program, the Erika Glazer Women's Heart Health Project, and the Adelson Family Foundation, Cedars‐Sinai Medical Center (Los Angeles, CA).

Disclosures

None.

Supporting information

Table S1

Figures S1–S2