ECPN Student Poster Contestant: The Rate of Hospitalizations Among Adults with Diabetes Residing in Food Swamps

Introduction: Prevention and management of diabetic complications is dependent on dietary intake, but diet can be impacted by the local food environment, including the availability of both healthy and unhealthy foods. Recent work suggests that the environment’s impact on diet and obesity has more to do with whether the food environment is net-negative or net-positive rather than with a single type of outlet (Cooksey-Stowers et al. 2017, Mezuk et al. 2016). Rose et al. coined the term “food swamps” to describe net-negative areas, defining them as places in which large numbers of unhealthy energy-dense offerings inundate or “swamp out” the relatively few existing healthy offerings (Rose et al. 2009). This study analyzes the association between food swamp residence and the rate of complications experienced by adults with diabetes across counties over time, as measured by inpatient hospitalizations.

Data: USDA Food Environment Atlas 2009-2014, AHRQ Health Care Cost and Utilization Project (HCUP) state inpatient databases 2010-2014, CDC Behavioral Risk Factor Surveillance Survey (BRFSS) 2010-2014, HHS Area Health Resource File (AHRF) 2010-2014

Methods: This study is a longitudinal county-level analysis with an analytic sample of 836 counties across 16 states with observations in 2010, 2012, and 2014. Food swamp severity is measured by the percentage of the total outlets in a county that are unhealthy (unhealthy = fast food restaurants & convenience stores; total = fast food restaurants, convenience stores, grocery stores & full-service restaurants). Hospitalization rates are measured as the sum of all hospitalizations experienced by adult patients with diabetes in a calendar year per 1,000 adult county residents with diabetes. Analysis uses multi-level linear regression with county random intercepts. The model also includes a polynomial term for the food swamp score and indicator variables for state and year. Controls variables include the percentage of diabetic patients admitted in the emergency room, the percentage on Medicaid, average comorbidity burden, primary care physicians per population, recreational facilities per population, median household income, population density, and the percentage of the population that is non-Hispanic black, Hispanic, female, and over 65.

Results: The mean food swamp severity is 53.65% (min: 11.11%, max: 94.12%, IQR: 47.34%-60.71%). Results from the random intercept model suggest that food swamp severity is significantly and positively associated with hospitalization rates among adults with diabetes (β=2.24, p=0.014). The polynomial term is also significant and negative (β=-0.02, p=0.034), which suggests the association is curvilinear. The association appears to be stronger at lower swamp percentages but attenuates as percentages increase, eventually leveling off after a certain point of saturation by unhealthy outlets (~65%).

Conclusions: Food swamps appear to be associated with hospitalizations for complications among adults with diabetes. Policies that limit oversaturation of the environment with unhealthy outlets, such as zoning, may help prevent complications and service utilization among adults with diabetes, at least up to a certain point of saturation.