Looking inside your shopping bags: The use of retail data to capture health lifestyle

Doctors usually diagnose patients based on their symptoms or medical investigation results, yet useful supplemental information such as living style behaviors and daily nutrition intake prior to the clinic visit is hardly known. This study proposes a statistical model of time-dependent-weighted maximum likelihood estimator and constructs the Caloric Activity Index (CAI) at an individual level to objectively monitor and predict consumers' health status and dietary behavior by using retail scanner data. The sample set is comprised of 2,362 membership customers' transaction records of a Taiwanese chain supermarket over a two-year period. Their transaction records of rice, milk, and potato/corn chips were systematically calculated into individual members' caloric-consumption amount based on the quantity of purchased food items and the corresponding nutritional facts of each product. Our results show that the CAI model correctly predicted about 70% of the customers who were identified as at-health-risk, regardless of food categories. The study contributes to the healthcare literature showing that the proposed CAI model can inspect individual's health status. It could be helpful for doctors to uncover patients' daily nutrition intake and make early detection of health risks possible.