Neighborhood Walkability and Cardiovascular Risk in the United States

Abstract

Background:

Neighborhood walkability may be associated with increased physical activity and thus may confer protection against cardiovascular disease and associated risk factors. We sought to characterize the association between neighborhood-level cardiovascular diseases and risk factors with neighborhood walkability across US census tracts.

Methods:

We linked the Centers for Disease Control and Prevention (CDC) PLACES dataset which provided census-tract level prevalence of coronary artery disease (CAD) and cardiovascular risk factors (hypertension, high cholesterol, obesity, and diabetes), with census tract population-weighted national walkability index (NWI) from the US Environmental Protection Agency (EPA). We calculated the mean prevalence of each cardiovascular health indicator per quartile of the walkability score. We also fit a multivariable linear regression model to estimate the association between walkability index and the prevalence of CAD adjusting for age, sex, race, and the CDC’S social vulnerability index, an integrated metric of socioeconomic position. We additionally performed mediation analyses to understand the mediating effects of CAD risk factors on the relationship between NWI and CAD prevalence.

Results:

A total of 70,123 census tracts were analyzed nationwide. Across walkability quartiles Q1 (least walkable) through Q4 (most walkable), we found statistically significant decrease in the prevalence of CAD (7.0% to 5.4%), and risk factors including hypertension (35.5% to 29.7%), high cholesterol (34.5% to 29.2%), obesity (35.0% to 30.2%), and diabetes (11.6% to 10.6%). After multivariable adjustment, continuous walkability index was negatively and significantly associated with the prevalence of CAD (β = −0.09, p < 0.0001). The relationship between NWI and CAD is partially mediated by the risk factors. High cholesterol accounted for 45%, high blood pressure 41% and diabetes 10% of the total effect of walkability on CAD. While direct relationship between walkability and CAD accounted for 9% of the total effect.

Conclusion:

This nationwide analysis demonstrates that neighborhood walkability is associated with a lower prevalence of cardiovascular risk factors and CAD. The association between NWI and CAD seems to be partly mediated by prevalence of traditional risk factors.

Introduction:

Cardiovascular Disease (CVD) is the leading cause of death in the United States and globally^1,2. Current studies on CVD prevention focus on individual cardio-metabolic risk factor optimization, overlooking crucial, yet, understudied environmental drivers, which may be important in primordial and primary prevention. Recently, the American Heart Association released the life’s essential 8, which enhances the American Heart Association’s construct of cardiovascular health. In this presidential advisory statement, the AHA highlighted the role of neighborhood effects on cardiovascular health factors.

Built environment plays a significant role in determining population-level cardiovascular health^3,4. Previous literature suggests a relationship between environmental characteristics and cardiometabolic diseases⁵. Additionally, physical activity is recommended by professional societies to decrease cardiovascular risk⁶. As such, neighborhood walkability (how walkable the neighborhood is) may be an important determinant of cardiometabolic health. Neighborhood walkability varies significantly throughout the United States⁷, and may explain a significant proportion of the variation in cardiometabolic risk factors. Limited data exists on the association between neighborhood walkability and cardiovascular health in the US. Given the known association between physical activity and health⁸, we sought to examine the association between the walkability of a neighborhood and the prevalence of CVD and its associated risk factors throughout the United States. To our knowledge, this may be the first study to investigate the effect of walkability of a neighborhood on CVD and risk factors on a national scale.

Methods:

This is a cross-sectional analysis investigating the association between census-tract level data on self-reported prevalence of coronary artery disease (CAD), and its risk factors (diabetes, high blood pressure, high cholesterol, and obesity) and neighborhood walkability.

Walkability index:

walkability of a neighborhood (how walkable a neighborhood is) was estimated from National Walkability Index (NWI), an index describing the propensity of walk trips in a certain neighborhood. It categories the block groups according the propensity of walk trips ranging from 1 to 20 where 1 is the least walkable and 20 is the most walkable. It is based on a simple formula that ranks selected indicators from the Smart Location Database (from the Environmental Protection Agency)⁹. This index was based on three measurable features of the build environment, including factors that affect the probability of people choosing walking as method of transportation. These features are: (1) street intersection density, (2) proximity of population centers to the nearest transit stops in meters, (3) variety of land uses which is subsequently divided into: employment mix, i.e the mix of employment activities such as (office, industry, stores etc.) and employment and household mix, i.e. mix of employment activities and occupied housing units.

The selected features were weighted using this following formula: $NWI=\left(\frac{w}{3}\right)+\left(\frac{x}{3}\right)+\left(\frac{y}{6}\right)+\left(\frac{z}{6}\right)$, where w is the block group’s ranked score for intersection density; x is the block group’s ranked score for transit stops; y is the block group’s ranked score for employment mix and z is the block group’s ranked score for employment and household mix. As such, NWI ranks each block group relative to all other block groups in the United States.

Demographic features: We obtained census tract median age, sex distribution, and racial composition (the percentage of white people in the whole population) from the United States census bureau American Community Survey (ACS).

Census tracts’ shapefiles:

census tracts’ shapefiles for our maps were obtained from Cartographic Boundary Files - United States census bureau.

Census-tract level urban and rural classification:

The geographical areas with urban-rural classification were obtained from the Census Bureau. The areas were delineated based on census blocks, so we first created census-block level urban-rural classification. We summarized the census-block level urban-rural classification into census tracts based on the population. A census tract would be labeled urban if the more than 50% of summarized census-block population are urban. A total of 58,947 census tracts were classified using this method.

Cardiovascular disease and risk factors estimated prevalence:

Estimated prevalence of US census tracts’ characteristics were obtained from the Centers for Disease Control and prevention (CDC)’s PLACES dataset 2021. The dataset provides estimated prevalence of (coronary artery disease [CAD], high blood pressure, high cholesterol, obesity, diabetes etc.) for the whole US at a census tract level. Prevalence is estimated by self-reported metrics based on the Behavioral Risk Factor Surveillance System. The Behavioral Risk Factor Surveillance System is the nation’s chief system of health correlated telephone surveys, where data is collected from US residents about their health-related risk behaviors and chronic health conditions. This system collects the data from the 50 states, via > 400,000 adult interviews each year. Interviews are conducted in person or via phone calls. The states use a standardized core questionnaire, optional modules and state-added questions.

Social vulnerability index (SVI):

The CDC Social vulnerability Index (SVI) was originally developed to identify vulnerability in the context of disaster management, but has been applied across multiple healthcare contexts including cardiovascular outcomes. The SVI is available at the census tract level and is derived from census indicators. Percentiles for each of the Census indicators within a theme are summed and the sum of these variables is ranked to determine an overall census tract percentile for each theme. To generate an overall theme ranking, theme percentiles are summed and ranked to determine a census tract percentile for the overall SVI.

The following dimensions are used in calculating the SVI: socioeconomic status (SES), household composition and disability, minority status and language, and housing type and transportation. The SES theme is derived from the percentage of persons living below the poverty line, percentage unemployment, per capita income, and percentage of persons aged 25 and above without a high school diploma. The household composition and disability dimension is derived from the percentages of: persons aged 65 years and older, persons aged 17 years and younger, civilian non-institutionalized populations with a disability older 5 years of age, and single parent households with a child younger than 18 years. The minority status and language dimension is derived from the percentage of all persons except non-Hispanic white and the percentage of persons aged 5+ years who speak English “less than well.” Finally, the housing type and transportation dimension is derived from percentages of: housing structures with 10 or more units, mobile home estimates, occupied housing units with more people than rooms (crowding), households with no vehicle available, individuals living in group quartiers¹⁰.

Statistical analysis:

Census-block level walkability index was weighted by block population into census tract-level estimates in order to match it with other census tract-level variables. Census-tract level estimated prevalence of coronary artery disease (CAD) and risk factors (hypertension, high cholesterol, obesity and diabetes) were matched with census-tract population weighted walkability index. Census tracts were divided into quartiles of NWI, with each quartile containing 17,530 census tracts. The average census-tract prevalence of each variable was calculated across the four NWI quartiles Q1-Q4. We examined the prevalence of cardiovascular risk factors and CAD as continuous variables per each NWI quartile. We additionally performed spearman’s correlation test to evaluate the relationship between NWI and prevalence of CAD and its risk factor. We fit five (unadjusted) linear regression models with NWI as the predictor and each cardiovascular health indicator (CAD, hypertension, high cholesterol, obesity and diabetes) as the covariate for each model. We then repeated the same models, but including age, sex, race and social vulnerability index (SVI) as covariates for adjustment. We also sought to examine if there is mediation effect in the relationship between NWI and CAD conducted be the risk factor covariates. We performed the mediation analysis initially between three variables as a 3-variabe model between (NWI, CAD and one covariant at a time), and then we performed it again but this time with combining all the risk factors in one model. Statistical analyses were performed using RStudio version 2022.07.1. Census tract prevalence and NWI were graphically represented using QGIS version 3.26.3.

Results

A total 70,123 census tracts (with a total population of 315,221,353) were analyzed throughout the United States (figure 1). The distribution of NWI across census tracts is shown in figure 1 and supplemental figure 1. Census tracts with higher walkability had lower median age, higher percentage of females, and less white residents (table 1). The distribution of NWI, and the prevalence of CAD and risk factors are shown in Figure 1. There was a negative correlation (β= −0.147, P value< 0.001) between NWI with prevalence of CAD and risk factors (supplemental figure 2). There was a statistically-significant stepwise decrement in the prevalence of CAD and risk factors across the four walkability-based quartiles. This decrease (from Q1 [least walkable] through Q4 [most walkable]) was as follows: CAD from 7.0% to 5.4%, hypertension from 35.5% to 29.7%, high cholesterol from 34.5% to 29.2%, obesity from 35.0% to 30.2% and diabetes from 11.6% to 10.6% (Figure 2). The association was relatively similar in rural and urban census tracts (supplemental figure 2)

Figure 1.

Maps of National Walkability Index and prevalence of coronary artery disease (CAD) and Risk Factors

Table 1.

Characteristics of census tracts’ mean prevalence across walkability index quartiles

	National Walkability Index
	Q1 (least walkable)	Q2	Q3	Q4 (most walkable)
Mean NWI	4.7	7.5	11.2	14.9
Mean age (years)	41.3	39.2	37.0	35.6
Female percentage	49.7	51.3	51.6	51
White percentage	80.4	72.9	55.7	47.3
Asian percentage	1.3	3.6	5.8	8.9
Black percentage	8.4	11.2	17.9	17.7
Mean Household Income ($)	90,416	103,012	97,543	102,953
Percentage Low income	10.1	10	11.8	13.3
Percentage Household income above 200,000$	5.3	7.9	7.2	7.9
Percentage Unemployment rate	2.9	3.1	3.8	4
Social vulnerability index	0.4	0.5	0.5	0.6
Percentage Under 18 years	22.1	22.7	22.5	20.6
Percentage Above 65 years	18.8	17.6	15.5	13.9
Percentage in labor force (>16 years)	58.8	62.4	63.9	65.6
Percentage with no health insurance coverage	8.7	7.9	9.3	9.4

Figure 2:

Mean prevalence of cardiovascular health outcomes and risk factors according to quartile of walkability index. BP = blood pressure, CAD = coronary artery disease. Error bars represent standard deviations. P values are from ANOVA tests.

In univariate models, we observed a negative association between NWI and prevalence of CAD (β = −0.147, p<0.001), hypertension (β = −0.545, p<0.001), high cholesterol (β = −0.52, p<0.001), obesity (β = −0.452, p<0.001) and diabetes (β = −0.08, p<0.001), table 2. After multivariable adjustments for age, sex, race and social vulnerability index, NWI was negatively and significantly associated with the prevalence of CAD (β = −0.09, p<0.001), hypertension (β = −0.59, p<0.001), high cholesterol (β = −0.32, p<0.001), obesity (β = −0.71, p<0.001) and diabetes (β = −0.24, p<0.001), table 2. Additional adjustment for percentage of population living in urban(vs rural) setting (n=58,947 census tracts) did not significantly change the estimates: CAD (β = −0.09, p<0.001), hypertension (β = −0.55, p<0.001), high cholesterol (β = −0.31, p<0.001), obesity (β = −0.65, p<0.001) and diabetes (β = −0.22, p<0.001). The joint association between NWI with age, percentage of female, percentage of white, and social vulnerability index with CAD prevalence is shown in figure 2.

Table 2.

Unadjusted and adjusted^* linear regression models

	Univariable		Multivariable*
	β (Standard Error)	P value	β (Standard Error)	P value
Coronary Artery Disease	−0.147 (0.002)	<0.001	−0.095 (0.001)	<0.001
High Blood Pressure	−0.545 (0.007)	<0.001	−0.593 (0.005)	<0.001
High Cholesterol	−0.520 (0.004)	<0.001	−0.321 (0.003)	<0.001
Obesity	−0.452 (0.006)	<0.001	−0.72 (0.005)	<0.001
Diabetes	−0.08 (0.003)	<0.001	−0.243 (0.002)	<0.001

^*Adjusted for (age, sex, race and social vulnerability index)

Mediation analyses revealed that cardiovascular risk factors accounted for significant proportion of the association between NWI and CAD prevalence. Each risk factor contributed to the relationship as follows; High blood pressure −0.133, High cholesterol −0.177, obesity −0.07 and diabetes −0.034 (Figure 3).

Figure 3:

The relationship between National Walkability Index (NWI) and coronary artery disease (CAD) across census-tract level (a) median age, (b) percentage of females, (c) percentage of white individuals, and (d) Social vulnerability index. Color codes represent prevalence of coronary artery disease.

Multivariable mediation analysis showed that the relationship between NWI and CAD is highly mediated by high cholesterol > high blood pressure > diabetes > obesity. The β for each mediator was as follows; high cholesterol −0.067, high blood pressure −0.061, diabetes −0.015, obesity 0.01 and the direct effect was −0.014 (Supplemental Figure 3). This translated to the following percentage of effect mediated: high cholesterol accounts for 45%, high blood pressure accounts for 41% and diabetes accounts for 10% of the total effect of walkability on CAD. While direct relationship between walkability and CAD accounts for 9% of the total effect.

Discussion:

To our knowledge, this is the first national study to investigate the relationship between walkability, CVD and CVD risk factors in such a large sample size, which is the whole US population. In this analysis, we demonstrated a significant inverse association between walkability and CVD risk factors/CVD prevalence. The negative association between walkability and CAD remained significant even after multivariable adjustment for age, sex, race and SVI, and the relationship between NWI and CAD prevalence seems to be mediated by traditional risk factors.

Although we believe that living in better and healthier neighborhoods would have better impact on CVD risk factors, additional research is required to examine the effect of more walkable neighborhood on CVD overall health. Previous regional studies have explored the association between walkability and CVD and risk factors. Howell et al. analyzed >44K individuals in Ontario, Canada and showed that walkability was inversely associated with poor cardiovascular risk profile (as estimated by 10-year predicted ASCVD risk of ≥ 7.5% by the AHA/ACC pooled cohorts equation).¹¹ Jones et al. showed that higher walkability is associated with lower incidence of hypertension in ~6000 participants in the REasons for Geographic and Racial Differences in Stroke (REGARDS) Study, independently of race¹². Koohsari et al. showed an inverse relationship between walkability and CVD rates in Japan even in areas with high socioeconomic status¹³.

Our study extends and support these findings. It has shown that population with ideal cardiovascular overall health was higher in neighborhoods with high walkability index and it was decreased while we move to neighborhoods with lower walkability index. These findings were similar in the other risk factors we have studied (high blood pressure, high cholesterol, obesity and diabetes). We extended the findings of the prior studies to all of the US population with more than 70,000 census tracts (with >315 million residents) which makes it the first nationwide study of its type. Our study is unique in examining other CVD risk factors such as (high blood pressure, high cholesterol, obesity and diabetes) and how they are affected by walkability as well. It is unique also in examining the mediation effects contributed by these risk factors towards the original relationship between walkability and CAD.

Neighborhood walkability may improve risk factors via improved physical activity, access to healthy food, and access to healthcare, all of which have been linked with cardiovascular risk^14,15. Prior studies suggest an association between neighborhood walkability and walking frequency for transport per week.¹⁶ Another study has shown an independent relationship between walkability and food security in a study of seniors in New York City.¹⁷ Additionally, low neighborhood walkability may accentuate the adverse effects of other environmental exposures, leading to further increase in CVD risk.¹⁸ This subsequently plays an important role in prevention of CAD and its risk factors.

These findings have to be interpreted within the context of limitations. Because of its cross-sectional ecological design, it cannot adjust for individual level data like the genetic profile of each individual or their previous health status/condition. In addition, the source of our data was the modeled and self-reported CDC PLACES data, which may be prone to estimation mischaracterization. There is no data on multi-morbidity and co-existence of risk factors in the CDC PLACES.

Conclusion:

This nationwide study demonstrates that neighborhood walkability is associated with lower prevalence of coronary artery disease and cardiometabolic risk factors, largely mediated through traditional risk factors. These findings suggest that living in a highly walkable neighborhood may protect against cardiovascular disease and its risk factors. Our study suggests not to be confined by hospital-level individual risk control, but extending to involve urban planning, neighborhood designing and even governmental policies. It encourages investments in designing and building more walkable neighborhoods, which in turn would play a vital role in decreasing the burden of CAD and its risk factors.

Figure 4:

Uni-variable mediation analysis examining the effect of each potential mediator on the relationship between National Walkability Index and coronary artery disease (CAD).

Data sharing statement:

The data used in this analysis are publicly available. The analytic code can be made available upon request.