AI-Powered Diabetes Prevention: A Milestone in Prediabetes Care
A landmark randomized trial from Johns Hopkins Medicine and the Johns Hopkins Bloomberg School of Public Health shows that an AI-powered lifestyle intervention app can reduce the risk of developing type 2 diabetes in adults with prediabetes at rates comparable to traditional, human-led programs. Published in JAMA on Oct. 27, the study suggests that an AI-driven diabetes prevention program (DPP) can meet the CDC’s risk-reduction benchmarks with similar effectiveness to yearlong group-based coaching.
Study Overview: How the Trial Was Conducted
The national, phase III randomized controlled trial enrolled 368 middle-aged adults with prediabetes who were referred to one of four remote, 12-month programs. Participants were randomly assigned to either conventional human-led DPP coaching or an AI-driven DPP app. During the study, all participants continued routine medical care through their primary care providers and were monitored with wrist activity devices for seven days each month to track physical activity.
The AI-DPP used a reinforcement learning algorithm to deliver personalized push notifications and guidance related to weight management, exercise, and nutrition. The human-led programs relied on trained coaches guiding participants through lifestyle changes. The trial design did not actively promote engagement; instead, it measured initiation and completion rates alongside clinical outcomes.
Key Findings: Comparable Outcomes, Higher Engagement for AI
At the 12-month mark, the researchers reported that 31.7% of AI-DPP participants and 31.9% of human-led DPP participants met the CDC-defined composite benchmark for diabetes risk reduction. The benchmarks include criteria such as at least 5% weight loss, or at least 4% weight loss with 150 minutes of weekly physical activity, or an absolute A1C reduction of 0.2%.
Crucially, the AI-powered group demonstrated higher engagement. Initiation rates were 93.4% for AI-DPP participants versus 82.7% for those in human-led programs, and completion rates stood at 63.9% for AI-DPP compared with 50.3% for the traditional programs. The results suggest that AI interventions can be as effective as human coaching in achieving meaningful risk-reduction outcomes while offering easier access and greater reach.
Why This Matters for Diabetes Prevention
Prediabetes affects an estimated 97.6 million adults in the United States and increases the risk of developing type 2 diabetes within five years. Prior research has shown that human-led DPPs reduce diabetes risk by about 58%, but access barriers such as scheduling conflicts and limited availability have constrained reach. With AI-DPPs now demonstrated to provide comparable clinical benefits and higher initiation rates, health systems could extend the reach of evidence-based prevention to more patients, including those facing logistical hurdles.
Implications for Primary Care and Health Systems
Physicians and care teams may consider incorporating AI-led DPP options as a viable alternative or complement to traditional coaching. AI-powered programs can run continuously, are less dependent on staff availability, and can be scaled to serve underserved populations who struggle to attend in-person or live-coach sessions. The study’s authors note that ongoing real-world analyses will help determine how AI-DPP outcomes translate across broader patient groups, including those with limited time and resources.
Future Directions and Considerations
Beyond confirming clinical efficacy, researchers are pursuing secondary analyses to understand patient preferences between AI and human modalities, how engagement drives outcomes, and the cost implications of AI-led DPPs. While AI offers clear advantages in accessibility and scalability, the field recognizes the importance of maintaining patient trust and addressing concerns about the “black-box” nature of some AI systems. The Johns Hopkins team emphasizes that AI-DPPs can provide reliable, personalized interventions that bolster, rather than replace, patient-centered care.
About the Study and Funding
The study was funded by the National Institutes of Health, including the National Institute of Diabetes and Digestive and Kidney Diseases and the National Institute on Aging, with additional support from Johns Hopkins institutions. Sweetch Health, Ltd. contributed as a service provider in the trial, while several co-authors represent Johns Hopkins and participating DPPs. The findings contribute to an evolving landscape of digital health in diabetes prevention and may shape how clinicians discuss preventive options with patients facing prediabetes.
What this means for you
For patients with prediabetes, an AI-powered DPP app could offer a practical, scalable path to reduce diabetes risk, especially where traditional programs are hard to access. As digital health tools mature, they may become a standard part of primary care, helping more people make sustainable lifestyle changes.
