Article Text
PDF
Abstract
Background Efforts to reduce socioeconomic inequities in cardiovascular disease include interventions to change the built environment. We aimed to explore whether socioeconomic inequities in coronary heart disease (CHD) incidence are ameliorated or exacerbated in environments supportive of physical activity (PA).
Methods We used data from the Reasons for Geographic and Racial Differences in Stroke study, which recruited US residents aged 45 or older between 2003 and 2007. Our analyses included participants at risk for incident CHD (n=20 808), followed until 31 December 2014. We categorised household income and treated it as ordinal: (1) US$75 000+, (2) US$35 000–US$74 000, (3) US$20 000–US$34 000 and (4) <US$20 000. We operationalised PA-supportive environments using characteristics within a 1 km residential buffer: walkable destinations density, PA facility density and proportion green land cover. We used Cox proportional hazards models to estimate the adjusted association of income with incident CHD, and tested effect modification by PA-supportive environment variables.
Results We found a 25% (95% CI 1.17% to 1.34%) increased hazard of CHD per 1-category decrease in household income category. Adjusting for PA-supportive environments slightly reduced this association (HR=1.24). The income–CHD association was strongest in areas without walking destinations (HR=1.57), an interaction which reached statistical significance in analyses among men. In contrast, the income–CHD association showed a trend towards being strongest in areas with the highest percentage of green land cover.
Conclusions Indicators of a PA supportive environment show divergent trends to modify socioeconomic inequities in CHD . Built environment interventions should measure the effect on socioeconomic inequities.
- neighborhood/place
- cardiovascular disease
- socio-economic
Data availability statement
Data may be obtained from a third party and are not publicly available. Data may be obtained from the REGARDS cohort committee at the University of Alabama at Birmingham on request.
Statistics from Altmetric.com
Introduction
Cardiovascular diseases (CVD) are a leading cause of death among low-income and middle-income and developed nations.1 In the USA, CVD accounted for 840 678 deaths in 2016, approximately one of every three deaths,2 and 121.5 million American adults had some form of CVD between 2013 and 2016.2 Yet, the distribution of CVD burdens within the US population highlights persistent inequities.
Socioeconomic position has been associated with CVD incidence and mortality, whether operationalised based on education,3 income4 or social class.5 In the Reasons for Geographic and Racial Differences in Stroke (REGARDS) cohort study, participants with low income and low education have a 42% higher risk of incident coronary heart disease (CHD) (95% CI: 1.14 to 1.76) than those with high income and high education.3 There are several pathways through which socioeconomic inequities in health could emerge and be maintained,6 including through barriers to residing in neighbourhoods with health-supportive built environments.7 Indeed, studies have assessed whether and how built environments can prevent CVD,8 including through support of physical activity (PA).9 However, exposure to built environment characteristics do not necessarily follow a pattern suggestive of deprivation amplification.10 People from lower socioeconomic positions live in neighbourhoods may disproportionately lack high quality green spaces to use as settings for PA,11 yet tend to live in neighbourhoods with greater walkability (an ‘advantage in disadvantage’).12 13
Complementary to the possibility that built environments represent a pathway connecting socioeconomic disadvantage to cardiovascular risk is the less commonly explored possibility that socioeconomic inequities are themselves modified by neighbourhood conditions.14–18 Individual-level longitudinal studies in particular have potential to complement the existing evidence by allowing for measurement of socioeconomic characteristics and neighbourhood conditions prior to CVD onset.
This study uses data from the REGARDS study to investigate differences in the magnitude of socioeconomic inequities in incident CHD across neighbourhood built environments that differ in indicators of support for PA. We hypothetise that the association between socioeconomic position and CHD will be ameliorated in environments more supportive of PA.
Methods
Study population
The REGARDS cohort is a US national, population-based study of 30 239 community-dwelling participants aged 45 or older who self-identified as non-Hispanic black or white. Recruitment between 2003 and 2007 was designed to oversample participants from the southeastern US-Stroke Belt (North Carolina, South Carolina, Georgia, Alabama, Mississippi, Tennessee, Arkansas and Louisiana)19 and non-Hispanic black people; the study design and objectives are described elsewhere.20 At enrolment, participants completed detailed surveys and clinical measures, collected via phone and in-home visits performed by trained technicians.20 Standardised follow-up telephone contacts were conducted every 6 months to detect change of home residence, hospitalisations and deaths. The study team additionally queries the National Death Index annually to assess whether any participants have died.
This study was restricted to those free of CHD at baseline and those with complete data for socioeconomic and neighbourhood characteristics. We excluded participants with self-report of myocardial infarction (MI) or coronary revascularisation procedure at baseline or evidence of prior MI on the baseline electrocardiogram (N=5314). We also excluded those with no information on baseline household income (N=3730) or neighbourhood built environment (occurring because we were not able to geocode a home address within the 48 contiguous USA and characterise the surrounding enviornment (N=884)).
The final sample size for complete case analyses was 20 808 participants. As sensitivity analysis, we used multiple imputations by chained equations with 10 datasets to assess robustness of our findings to missing at random data on income.21 22
All participants provided written informed consent.
Outcome variable: incident CHD
The outcome for this analysis was adjudicated incident CHD defined as MI or acute CHD death. During the biannual follow-up contacts, any report of a heart disease-related event (including emergency department visit or hospitalisation) or participant death prompted retrieval of medical records. For fatal events, medical history, hospital records, interviews with next of kin or proxies and death certificate or National Death Index data were reviewed to adjudicate the cause of death, with definite or probable CHD death used in the analysis.23
Cases were assigned to two adjudicators and disagreements were resolved by a committee, following published guidelines by the American Heart Association.23 24 Briefly, MI was determined by: (1) considering clinical signs and symptoms suggestive of ischemia; (2) ECG findings consistent with ischemia or MI, guided by the Minnesota code and classified as evolving diagnostic, positive, non-specific or not consistent25 and (3) a rising and/or falling pattern in cardiac troponin level or creatine phosphokinase MB level over six or more hours, with a peak level greater than twice the upper limit of normal (diagnostic cardiac enzymes). Definite MI was defined based on diagnostic cardiac enzymes or ECG. Adjudication of an MI as probable relied on elevated but not diagnostic (ie, equivocal) enzymes; a positive but not diagnostic ECG; or, if enzymes were missing, a positive ECG in the presence of ischaemic signs or symptoms. Agreement between adjudicators for the outcomes included in this study (definite or probable MI or definite or probable acute CHD death) yielded a κ greater than 0.80.
Exposure variable: socioeconomic position
The primary measure of socioeconomic position was household income at baseline. We divided participants in four categories: (1) US$75 000 and above, (2) US$35 000–US$74 000, (3) US$20 000–US$34 000 and (4) less than US$20 000. For sensitivity analysis, we used a dichotomous socioeconomic position variable combining household income and education previously used for analyses of CHD in REGARDS,3 26 with low socioeconomic position defined as either low annual household income (<US$35 000) or low education (less than high school).
PA supportive environments
We used three previously studied indicators of environmental support for PA: density of walking destinations,27 density of PA facilities,28 and proportion green land cover.29 These characteristics were estimated for a 1 km buffer surrounding each participant’s geocoded home address at baseline and in each year of follow-up. Environment variables were constructed as time-varying to incorporate address updates during follow-up and changes captured in longitudinal geographical data.
Measures of walking destinations and PA facilities were created using the National Establishment Time-Series (NETS)30 31 data. Commercially licensed NETS provided annual establishment-level data for each January from 1990 to 2014 for US businesses and many non-profit and government establishments. Walking destinations and PA facilities were classified informed by previous studies,32 33 with further details on data cleaning and categorisation provided elsewhere.31 Walking destinations were defined to include both potential destinations for daily living and those judged as amenities adding to pedestrian interest and overall walkability. PA facilities included a variety of venues for PA or fitness, such as indoor gyms, that have been previously associated with both gym membership and objective PA.28 Measures of green land cover are derived from the National Land Cover Dataset (NLCD). NLCD describes the visible features that cover the Earth’s surface, using land cover data derived from satellite imagery or aerial photography. We created measures of green land cover per km2 by summing across vegetation-related categories (41-95).
Covariates
We included demographics such as age, sex and race (Black or white) as well as geographical variables including region of residence (Stroke Belt, Stroke Buckle or other) and urbanicity (categorised as urban, large rural, small rural or isolated small rural tract using Rural and Urban Commuting Areas categories).
Analysis
Descriptive analysis used baseline individual-level and neighbourhood variables. We used Cox proportional hazards regression models to examine the association of income category with incident CHD. Observation started at REGARDS baseline, and events through 31 December 2014, were included in the analysis. Individuals were censored at the time of their event, death or the end of the follow-up. An ordinal grouped linear income variable was used, coded such that HRs above one indicate that a lower income group is associated with an increased hazard of CHD.
Initial models (model 1) were adjusted for sociodemographics and geographical covariates. After this, we adjusted for the three PA environment variables (model 2). We calculated the proportional change in HR estimates with additional adjustment for PA environment using the following formula34 35 :
To test effect modification, we introduced an interaction term between income (treated ordinally) and categorical versions of each PA environment exposures (walking destinations, PA facilities, green land cover). For each environment characteristic, we grouped participants into five categories. Because of a large number of zeros, the first category (lowest level) included those with no destinations (either walking destinations, PA facilities or green land cover) within their 1 km buffer, and among the participants with >0 categories were defined using quartiles.
To formally test the global statistical significance of each income–environment interaction, we compared a model with and without the interaction terms using a likelihood ratio test. Results are presented overall, stratified by sex and race, and for sensitivity analyses with a combined measure of education and income as the exposure or including imputed data on household income. Analyses were conducted in R V.3.6.3.
Results
Table 1 shows the main descriptive analyses. Participants with lower reported income were older, were disproportionately female and Black, and had a greater number of walking destinations in the 1 km buffer surrounding their baseline home address. Higher-income participants were more likely to live in urban areas, report higher education, have better self-reported health and have a greater percentage of green land cover in their 1 km buffer.
Descriptive characteristics of included regards participants by income category
Table 2 shows the results of the Cox regression models on the relationship between household income and CHD before (model 1) and after adjustment by PA environment (model 2). There was a strong association between household income and the risk of CHD, with residents of low income being more likely to experience incident CHD. Each 1-category decrease in household income was associated with a 25% increased hazard of CHD (HR 1.25; 95% CI: 1.17% to 1.34%). In stratified analyses, the income–CHD association was strongest among women (HR=1.31) and among black people (HR=1.30), though CIs for all stratified analyses overlapped.
Association between household income and incident coronary heart disease, overall and stratified by sex or race
After adjustment for PA environment variables (model 2), the association between income and CHD was shifted toward the null, but in all cases remained statistically significant. Overall, there was a 4.8% reduction in the HR after adjusting by PA environment variables, and none of the sex or race strate showed attenuation with adjustment by more than 8.5%.
Figure 1 shows patterns in how the strength of the relationship between income and CHD varies across categories of each indicator of a PA supportive environment. The left panel shows the interaction by walking destinations; the association between income and CHD was highest in areas with no walking destinations (HR=1.57) but did not show a clear trend across increasing quartiles of walking destination density (Q1=1.18, Q2=1.16, Q3=1.39, Q4=1.20). This statistical interaction reached significance only among males (interaction p value: 0.02). We did not find a statistically significant variation of the association between income and CHD within PA facilities groups (middle panel), though among males there was a trend suggestive of a stronger income–CHD association in contexts without PA facilities (interaction p=0.15). For green land cover (right panel), the association between income and CHD was stronger in areas in the highest quartile of green land cover, although the trend was not significant (interaction p=0.189). Among Blacks, there was a gradient suggestive of dose-response, such that areas with incrementally higher green land cover had stronger associations between income and CHD. On the other hand, both areas with the lowest and highest green land cover a relatively strong income-–CHD association (interaction p=0.07).
Effect modification by physical activity supportive environment for the association between income and coronary heart disease (CHD), considering and stratification by sex and race.
Sensitivity analyses using a combination of education and income gave similar results in both adjusted (online supplemental file 1) and interaction models (online supplemental file 2). Sensitivity analyses using multiple imputations show no changes to our main inferences (online supplemental files 3 and 4).
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Discussion
In this study, residents reporting lower household income had a greater incidence of CHD compared with those with higher household income. We explored the contribution of the PA environment to these socioeconomic inequities. Adjustment for indicators of a PA supportive environment (walking destinations, PA facilities and green land cover) led to a small reduction in the effect of socioeconomic position. Analyses investigating potential effect modification of the income–CHD association by PA supportive environment (and for combined socioeconomic status (SES)–CHD association in a sensitivity analysis) noted persistence of the association across strata. Trends in the magnitude of the association suggested that socioeconomic inequities in CHD were most pronounced in areas with no walking destinations or with the greatest green land cover, providing mixed support for our hypotheses.
Both socioeconomic position and PA environment features (such as walking destinations, PA facilities and green land cover) have been associated previously with risk of CHD.3 9 36 From an environmental justice perspective, previous studies have shown an unequal distribution of PA environment features within cities.11 Thus, targeted investments to increase support for PA in the built environment might mitigate the CHD inequities. However, adjusting for selected indicators of a built environment supportive of PA did not appreciably reduced the relationship between socioeconomic position and CHD in our analyses. Previous research has studied how PA environment can mitigate inequities in PA,37 38 while other have looked at whether other neighbourhood factors (such as violence and or neighbourhood disorder) mediate the association between SES and CHD risk factors.14 16–18 For instance, Turrell et al 37 found that, in Brisbane (Australia), there was a partial mediation by walkability features (residential density, street connectivity and land use mix) for the association between neighbourhood disadvantage and walking for transport. Pascual et al 38 studied PA facilities as potentially mediating the association between income and physical inactivity in Madrid (Spain).
In this present study, we were able to juxtapose patterns of effect modification across three indicators of an environment supportive of PA. Overall, socioeconomic inequities in CHD were noted across all strata, though a pattern of effect modification provided mixed support for our hypothesis; we found that socioeconomic inequities in CHD were greater in areas with no walking destinations and in areas with greater green land cover. One explanation is that walking destinations might act as an ‘advantage in the disadvantage’,12 where low-SES areas have greater availability of walking destinations.13 Another possibility is that destination density and green land cover diverge in their implications for perceived safety, and CHD inequities are themselves exacerbated by safety concerns. In particular, areas with greater walking destinations might attract more pedestrian activity and be perceived as safer (following Jane Jacobs concept of ‘eyes on the street’39) while areas with greater green land cover might be perceived as less safe for PA.40
Our work should be interpreted keeping in mind that that SES inequities in leisure-time PA in PA facilities and green spaces tend to be higher than those in transport-related PA (related with walkability).41 Previous studies have shown an interaction between SES and aspects of the built environment, such as walkability, and the risk of different health conditions and behaviours, especially PA.42 Following this, Adkins et al suggest that unprivileged populations respond differently to built environment features due to safety, time to do PA or working conditions, among others.43
This study has several strengths. First, we were able to do a longitudinal study of 20 808 individuals from a national US sample, taking into account time-varying exposures to the PA environment. Second, we were able to use objective measures for both PA environment variables and CHD outcomes, avoiding the possibility of same source bias.44 Third, we used multiple types of PA resources which map onto different types of PA that impact CHD. Lastly, we were able to use both adjusted and effect modification analyses to get a better sense of the role of the PA environment in CHD’s socioeconomic inequities.
We acknowledge that this study has several limitations. Household income was measured by self-report and entered as ordinal categories which may be subject to measurement error or bias; there were also missing income data, though results were similar in sensitivity analyses imputing instead of excluding all missing income values. Our indicators of PA supportive environment included heterogeneous walkable destinations (which represent only one component of walkability45), PA facilities which might require fees to enter, and green land cover that differs in accessibility and vegetation type. Despite the point-level annual resolution of the NETS source for walking destinations and PA facilities,30 misclassification is possible and validity checks to refine our categorisation were limited to the most recent year (2014). Additionally, the focus on built environment measures within 1 km buffers around home addresses might not reflect where people participate in PA (including activity spaces around work, around children’s schools or in areas accessed by vehicle). The REGARDS study ascertainment of CHD events requires notification by participants or via linkage to the NDI, so despite careful adjudication, there might be under-reporting of the CHD outcome or CHD events that have been erroneously included a probable. Finally, the impact of the built environment on CHD and CHD inequities may be small, creating challenges as we attempt to disentangle true causal effects from bias, including bias emerging from uncontrolled confounding. In the US context, race-based segregation and historic patterns of income inequities likely contribute to residual confounding by experiences of structural discrimination that connect place and race.
Conclusions
We found that REGARDS participants with lower socioeconomic position had a higher incidence of CHD. These inequities were relatively wide in areas with no walking destinations. In contrast, the income–CHD association showed a trend towards being strongest in areas with the highest percentage of green land cover. Indicators of a PA supportive environment show divergent trends to modify an income–CHD association, and future evaluation efforts should monitor the implications of built environment change for the magnitude of health inequities.
What is already known on this subject?
Previous studies have demonstrated that individuals of low socioeconomic position have more risk of developing cardiovascular disease. However, there is scarce information about how the built environment can ameliorate or exacerbate socioeconomic inequities in cardiovascular disease.
What this study adds?
Low-socioeconomic position individuals have greater risk of developing coronary heart disease and the built environment may modify this association. Socioeconomic inequities in coronary heart disease were noted to be larger in areas with no walking destinations.
Data availability statement
Data may be obtained from a third party and are not publicly available. Data may be obtained from the REGARDS cohort committee at the University of Alabama at Birmingham on request.
Ethics statements
Ethics approval
The REGARDS study protocol was reviewed and approved by the University of Alabama at Birmingham Institutional Review Board and all participating institutional review boards.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @pgullon, @usama_bilal
Contributors PG and GL conceived the idea. JAH, AR, MMS and SJ developed the cohort and contextual data. PG and UB carried out the statistical analysis. PG drafted the manuscript. All authors provided critical intellectual contributions. All authors read and approved the final manuscript.
Funding The REGARDS research project is supported by cooperative agreement U01 NS041588 cofunded by the National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), National Institutes of Health, Department of Health and Human Service. Representatives of the NINDS were involved in the review of the manuscript but were not directly involved in the collection, management, analysis or interpretation of the data. The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at: https://www.uab.edu/soph/regardsstudy/. Additionally, the integration and analyses of geographic data were supported by the National Institute of Aging (grants 1R01AG049970, 3R01AG049970-04S1), Commonwealth Universal Research Enhancement (C.U.R.E) program funded by the Pennsylvania Department of Health - 2015 Formula award - SAP #4100072543, the Urban Health Collaborative at Drexel University, and the Built Environment and Health Research Group at Columbia University. PG was supported by Fundación Alfosno Martín Escudero 2018 Postdoctoral fellowship program. UB was supported by the Office of the Director of the National Institutes of Health under award number DP5OD26429.
Disclaimer The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or the NIA.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.