Development and Validation of a Novel Household Grocery Food Purchase Quality Score

Abstract

Burdensome surveys and convenience sampling limit the ability to assess the quality of grocery food purchases. Our goal was to develop a metric for assessing such purchase data unobtrusively and at scale. We obtained de‐identified 2012 sales transaction data for 92,314 households in 4 geographic locations and re‐classified 1,888 higher‐order food descriptors into the 29 categories of USDA's Food Plan market baskets, which comply with the Dietary Guidelines. The standardized expenditure share for each category was the weighted average of all sex‐age groups used in the Food Plans. Quality of food purchases was evaluated by comparing the observed to the standardized expenditure share for each category. The 29 categories were then grouped into 10 food groups related to the Healthy Eating Index‐2010‐‐Total Vegetables, Greens and Beans, Total Fruit, Whole Fruit, Whole Grains, Dairy, Total Protein Foods, Seafood and Plant Proteins, Refined Grains, and Sweets‐‐and scored. Households who never purchased tobacco (n=71,475) had higher mean total scores (39.6 out of a possible 70) than those who did (36.3) (p<0.01) as well as higher scores for 8 of the 10 food groups (p<0.01). Thus, construct validity was established. This approach can be applied using longitudinal data to describe population‐level food purchase quality patterns and trends.Support: NIH grant T15‐LM007124; University of Utah Funding Incentive Seed Grant