Comparing Different Metrics for Analyzing the Rural Food Environment and Their Associations with Diabetes, High BMI, and Poor Diet Quality

Introduction and Objective: Associations between unhealthy food environments and diabetes or obesity have been studied extensively in urban areas, but not precisely in rural communities. We performed a geographically detailed study of the rural food environment using various measures to identify those strongly associated with a higher rate of diabetes, high BMI, and poor diet. Methods: We surveyed 1,310 residents of rural Sullivan County, collecting demographic, health status, and food frequency data. The food environment was assessed by cataloging restaurants and food stores into fast food and convenience stores versus supermarkets, grocery stores, and wait service restaurants. LASSO regression was used to identify food environment factors strongly associated with diabetes, high BMI, and poor diet. Proximity, density, relative proportions, and modified retail food environment indexes (mRFEI) metrics were compared. We used fixed distance bands and novel nearest-neighbor approaches for proportions. Results: Older age (LASSO coefficient: +0.30), nearest-neighbor proportion of restaurants that were fast food (+0.22), and low household income (+0.16) were associated with higher diabetes rates, whereas White race (-0.11) had lower rates. Nearest-neighbor proportion (+0.29) and density (+0.23) of fast food were the strongest predictors of high BMI, along with Hispanic ethnicity (+0.22). Low household income (-1.24) was associated with poor diet quality, whereas older age (+1.17), female gender (+1.06), and high income (+0.59) were associated with better diet quality. Nearest-neighbor metrics helped with the problem of frequent zeros (p=0.36 for skewness of fast food with 20 nearest-neighbors). Conclusion: We compared different approaches for measuring the food environment in rural settings and found that novel nearest-neighbor approaches were superior to other metrics based on fixed distance bands, proximity, or density. Disclosure: J. Gjonaj: None. H. Yi: None. T.A. Flores: None. C. So: None. H.L. Motola: None. B. Elbel: None. L. Thorpe: None. D.C. Lee: None. Funding: National Institutes of Diabetes and Digestive and Kidney Diseases (R01DK124400)