An AI-powered app for diabetes prevention met CDC benchmarks for diabetes risk reduction at rates comparable to those in human-led programs, Baltimore-based Johns Hopkins Medicine and the Johns Hopkins Bloomberg School of Public Health found.
The study included 368 participants with a median age 58, 71% of whom were female, 61% white, 27% Black and 6% Hispanic, according to an Oct. 27 news release. They were assigned either to one of four 12-month, remote, human-led programs or used a reinforcement learning algorithm app that delivered push notifications to facilitate weight management behaviors, physical activity and nutrition.
At the 12-month mark, 31.7% of participants in the fully AI-driven program and 31.9% of those in human-led programs met the CDC’s composite benchmark for diabetes risk reduction. The AI-paired group also had higher initiation and completion rates — 93.4% and 63.9%, respectively — compared to 82.7% and 50.3% in the human-led group.
