Weight Loss in a Digital Diabetes Prevention Program Powered by Artificial Intelligence: Retrospective Cohort Study (Preprint)
BACKGROUND The National Diabetes Prevention Program (DPP), governed by the Centers for Disease Control and Prevention (CDC), reduces the incidence of diabetes and diabetes-associated medical costs. Typically, providing this program is staffing-intensive, limiting the ability to scale the DPP and keep pace with the growing incidence of prediabetes. OBJECTIVE We investigated the average weight loss of users of a program called Lark DPP that has full CDC recognition and is powered by conversational artificial intelligence (AI). METHODS We analyzed weight loss of 674 users who met CDC qualifications (completed ≥3 lessons in months 1-6 with ≥9 months between first and last lessons). In addition to the weight loss expected from the CDC curriculum, we investigated whether user characteristics and engagement with AI coaching increased the likelihood of achieving the CDC’s benchmark of ≥5% weight loss at 12 months using logistic regression. RESULTS We observed that 279 users met CDC qualifications and achieved an average of 5.2% (SE=.4) weight loss at 12 months (46% achieved ≥5%). CDC qualifiers completed an average of 20.7 (SE=.4) of 26 available educational missions/lessons. The number of weeks with >2 weigh-ins (standardized coefficient β=.39; P<.001); days with an exchange with the AI coach (β=.24; P=.016); and days since last coaching exchange at final weigh-in (β=-.45; P<.001) were significantly associated with the likelihood of achieving ≥5% weight loss. CONCLUSIONS The Lark DPP resulted in weight loss and sustained engagement for 12 months that was equal to or greater than in-person or hybrid-digital DPPs. Beyond the association between educational mission completion and weight loss, the synchronous personalized feedback and exchanges with the AI coach increased the likelihood of achieving ≥5% weight loss. An AI-powered program is one method to deliver DPPs in a scalable and resource-effective manner to keep pace with the prediabetes epidemic.