Association Between Availability of Fruits and Vegetables in Neighborhood Food Stores and Weight Among Residents of Low-Income Urban Public Housing: Cross-Sectional Study

Introduction

Though food environments are recognized as determinants of socioeconomic diet–related health disparities, there is less evidence delineating specific elements of food environments that contribute to bodyweight [1]. Previous literature is limited by variability in measurement approaches and a lack of capturing multiple food retail environments [1]. Among adults, a systematic review or meta-analysis assessing impact of food environments on obesity included only two studies with general merchandise stores as part of their analyses [1]. Furthermore, much research about general merchandise stores focuses on rural areas and children or adolescent populations, prompting evaluation among adults in urban areas [2,3]. This study’s objective was to examine associations between two measures of food access: (1) presence of food stores (overall and by store types) and (2) availability of fruits and vegetables with weight at study enrollment among residents of urban public housing developments participating in a clinical trial for weight management [4].

Methods

Design

This is a cross-sectional analysis of baseline data from a clinical trial (Table 1) [1,4] along with neighborhood food store audit data collected by research personnel.

Measures

Weight

Research personnel measured participant (n=286) weight via scale during in-person, in-home baseline study visits [4].

Food Store Audit Instrument

Pairs of research personnel conducted in-store audits using the Food Environment Audit for Diverse Neighborhoods to identify food availability [5].

Data Analysis

Food access measures (number of food stores; total count of available fruits and vegetables or audited food store) were assigned to individual participants in the clinical trial, based on their public housing development of residence. We used separate generalized linear models examining associations between each access metric and participant weight at baseline. Specifically, we used generalized estimating equations to account for clustering of participants within housing developments (adjusted for resident age, gender, Hispanic ethnicity, and education). The model coefficient represents the difference in weight at baseline enrollment with a one-unit increase in count (ie, one additional store, or one additional fruit and vegetable available). Additional models adjusting for height were conducted (Multimedia Appendix 1). See Table 2 footnotes for details.

Ethical Considerations

Boston University Medical Campus Institutional Review Board approved this analysis (H-41590). The original written consent process covered this analysis without additional consent. Privacy was ensured by deidentifying data with code numbers. Participants were paid US $50 for completing baseline assessment activities.

Results

There was no association between overall number of food stores and baseline weight nor between average number of fruits and vegetables in audited food store locations and baseline weight (adjusted difference in weight: 0.33, [95% CI −0.05,0.71 and 0.08, 95% CI: −0.26, 0.41 pounds, respectively). When food store type was examined, there were positive associations between number of convenience stores and baseline weight and between number of general merchandise stores and baseline weight (adjusted difference in weight: 0.87, 95% CI: 0.02, 1.72 and 3.08, 95% CI: 1.31, 4.85 pounds, respectively) (Table 2). Additional adjustment for height resulted in associations in a similar direction, with a small attenuation in magnitude (Multimedia Appendix 1).

Discussion

The number of convenience stores and general merchandise stores within a one-mile walking distance radius of public housing developments were associated with higher baseline weight, after adjusting for potential confounding factors. Unlike findings from systematic reviews or meta-analyses [1], we did not observe associations with number of grocery stores and baseline weight or availability of fruits and vegetables at audited food stores and baseline weight. Weight status is multifactorial and therefore, auditing a larger variety of foods may have been necessary to observe an association with weight. Others also noted lack of associations between new grocery stores and healthful eating among public housing residents [7]. Overall, neighborhood environment deprivation factors like lack of safe physical activity areas can all contribute to adverse health behaviors and outcomes.

Dollar stores are the fastest growing food retail type by share of household expenditure in the United States [3], raising concerns about nutrition and health-related outcomes [8] and impact on local businesses, staffing shortages, and overall cleanliness [9]. Municipalities have taken policy actions to address concerns including: moratoriums (prohibit new store openings for specified period) and conditional use ordinances (require local government to assess how proposed store aligns with certain conditions) [9]. Obtaining perspectives from key constituents and forming partnerships (eg, community members or policy makers workshops) can lead to creation of fair policies for equitable benefit [10].

This study benefited from a standardized validated audit data collection instrument and included diverse food stores. Limitations included: lack of consideration for sales volume; audits represented snapshots of food available that particular day or season; and inclusion of clinical trial participants potentially limited generalizability.

Our study suggests cross-sectional relationships between presence of convenience stores and general merchandise stores and higher weight among residents that were overweight or with obesity living in nearby public housing developments; findings may inform design of community-level food store policies.