Abstract
BACKGROUND: Diabetes-related costs are the highest across all chronic conditions in the United States, with type 2 diabetes accounting for up to 95% of all cases of diabetes. A healthy diet is strongly associated with lowering glycated hemoglobin A(1c) (HbA(1c)) levels among individuals with diabetes, which can help curtail other health complications. Digital health platforms can offer critical support for improving diet and glycemic control among individuals with diabetes. Less is known about the characteristics of people with diabetes who use digital health platforms (specifically, a platform that integrates personalized healthy meal plans and food ordering) and changes in their HbA(1c) levels. OBJECTIVE: The aim of this study is to characterize Foodsmart users with diabetes and evaluate the longitudinal impact of Foodsmart-a personalized digital nutrition platform with meal planning, food ordering, and nutrition education features-on changes in HbA(1c) levels. METHODS: We retrospectively analyzed data collected from 643 adults with at least two self-reported HbA(1c) entries in the Foodsmart platform between January 2016 and June 2021. Participants self-reported their HbA(1c) levels, height, weight, health conditions, and diet in a 53-item food frequency questionnaire. Diabetes was defined as HbA(1c) ≥6.5%. We analyzed distributions of characteristics by baseline diabetes status and examined the association of characteristics with the likelihood of having diabetes at baseline. To evaluate the change in HbA(1c) levels among Foodsmart users, we calculated mean changes (absolute and percent) in HbA(1c) among participants with diabetes and by length of follow-up. We also compared changes in HbA(1c) and weight between participants with diabetes at baseline who achieved a normal HbA(1c) level and those who did not. RESULTS: We found that 43.5% (280/643) of the participants with at least two HbA(1c) level entries had diabetes at baseline. Participants with diabetes at baseline were more likely to be male, have a higher weight and BMI, report high blood pressure, and have a poorer diet in comparison to participants without diabetes. Using a multivariable logistic regression model, we found that being male and obese were statistically significantly associated with baseline diabetes. Among participants with diabetes at baseline, HbA(1c) was reduced, on average, by 0.46%. In addition, 21.4% (60/280) of participants with diabetes achieved a normal HbA(1c) level (<6.5%) in their last HbA(1c) level entry; this percentage increased with longer follow-up time (39% [7/18] at >24 months). In a sensitivity analysis, users with an HbA(1c) ≥7.0% at baseline had an average absolute change of -0.62% and 31.2% (62/199) of these participants achieved HbA(1c) levels of less than 7.0%. CONCLUSIONS: This study assessed characteristics of individuals enrolled on the Foodsmart platform with HbA(1c) levels and found that users with diabetes had lower HbA(1c) levels over time and a sizable percentage of participants were successful in achieving normal levels.