Impact of Food Retailers' Presence and Composition on Nutritional Equity and Health Outcomes in the United States With Machine Learning
Syed Badruddoza,Modhurima Dey Amin,Jill McCluskey
DOI: https://doi.org/10.1016/j.jneb.2023.05.231
IF: 2.822
2023-07-10
Journal of Nutrition Education and Behavior
Abstract:Objective Many studies find evidence of nutritional inequality in the United States–that is the association between low access to healthful food and low-income or non-White population. However, the findings are not causal: we do not know whether the residents of the low-income non-White neighborhoods are choosing unhealthful diets because of the lack of access, or large food retailers with a variety of healthful options choose not to locate in the neighborhood due to lack of demand. The current project conducts an in-depth analysis for policy generation. Description First, we derive the conditional probabilities of each type of food retailer, eg, grocery or fast-food restaurants, that locates in a certain region, given the demographic and regional features. Second, we calculate the rate of substitution between two types of food retailers to understand the dietary preferences of residents. Third, if one type of food retailer is present, we assess how it affects residents' dietary and nutritional choices, such as fresh vegetable consumption. Evaluation We use the COVID-19 lockdown and store closures for causal inference. Household food purchase information is obtained from scanner data. We use cellphone mobility data to track available food stores in a region, and machine learning models for improved prediction. Finally, we evaluate the impact of the presence of various food retailers and their composition in the neighborhood on diseases that might be influenced by dietary choices, such as diabetes, obesity, and cardiovascular diseases. Conclusions and Implications Machine learning results indicate that demographic features and neighborhood amenities of census tracts predict food retailer type in 72-92% of the out-of-sample cases. Store types are complements rather than substitutes: there is a positive association between the density and popularity of all store types, ie, grocery stores' presence or visitors are positively associated with those of fast food retailers. However, a large grocery's presence has a negative association with diabetes and obesity, whereas a convenience store's presence has a positive association. The estimation of the causal effects of store closure is underway. Funding Year 2022-68015-37142 Funding Acknowledgement This work is supported by the USDA National Institute of Food and Agriculture.
nutrition & dietetics,education, scientific disciplines