Groundbreaking Trial Shows AI-Driven DPP Can Rival Human-Led Programs
A new study from Johns Hopkins Medicine and the Johns Hopkins Bloomberg School of Public Health demonstrates that an AI-powered lifestyle intervention app can reduce diabetes risk in adults with prediabetes as effectively as traditional, human-led programs. Published in JAMA on Oct. 27 and funded by the National Institutes of Health, the phase III randomized controlled trial suggests AI can extend the reach of proven diabetes prevention strategies without sacrificing outcomes.
Why This Matters: Prediabetes and the DPP Landscape
Nearly 98 million adults in the United States live with prediabetes, a condition that raises the likelihood of developing type 2 diabetes within five years. The Centers for Disease Control and Prevention (CDC) has established risk-reduction benchmarks for the Diabetes Prevention Program (DPP): sustained weight loss, increased physical activity, and improvements in blood glucose markers. In traditional, human-led DPPs, participants who complete the program are about 58% less likely to develop type 2 diabetes, according to the CDC’s landmark findings. Yet access barriers often limit participation, including scheduling challenges and program availability.
The Study: AI vs. Human-Led DPP Over 12 Months
During the COVID-19 pandemic, researchers enrolled 368 middle-aged adults (median age 58) with prediabetes and assigned them to one of four remote, 12-month, human-led DPPs or to an AI-driven DPP app guided by reinforcement learning. All participants wore wrist activity monitors to track physical activity for seven days each month. Participants continued standard medical care and were not allowed to take medications that could affect weight or glucose levels.
Key metrics focused on the CDC-defined composite benchmark for diabetes risk reduction, including at least 5% weight loss, at least 4% weight loss with 150 minutes of weekly physical activity, or an absolute A1C reduction of 0.2% or more. Importantly, the study looked at not just outcomes, but engagement—how often participants started and completed the program when referred to either modality.
Engagement and Outcomes: AI Demonstrates Parity (and More)
After 12 months, 31.7% of AI-DPP participants and 31.9% of human-led DPP participants achieved the CDC benchmarks for diabetes risk reduction, illustrating similar effectiveness between the two approaches. Remarkably, the AI group showed higher engagement: initiation rates were 93.4% for AI vs. 82.7% for human-led programs, and completion rates were 63.9% versus 50.3%, respectively. These findings suggest that the convenience and accessibility of AI-led interventions can drive participation without compromising health outcomes.
Interpreting the Results
Researchers highlight that accessibility is a critical driver of success in lifestyle interventions. The AI-DPP’s always-available, fully automated design reduces barriers such as scheduling conflicts and staffing constraints that commonly hinder human-led programs. The study’s co-first author, Benjamin Lalani, notes that initiation is often the greatest hurdle; the AI platform appears to lower this barrier by offering flexible, asynchronous support.
Implications for Primary Care and Public Health
The findings indicate that AI-led DPPs can be a viable, scalable option for patients who face logistical challenges to traditional programs. Primary care providers may consider recommending AI-powered DPPs as a complementary or alternative pathway for patients with prediabetes, particularly where staffing shortages or geographic limitations exist. As AI interventions become more common, clinicians will need to address questions about data privacy, user experience, and the interpretability of AI-driven recommendations.
Looking Ahead: Real-World Applications and Further Research
The Johns Hopkins team plans to examine how AI-DPP outcomes translate to broader, underserved populations who struggle to access conventional DPPs. Ongoing secondary analyses will explore patient preferences between AI and human modalities, the relationship between engagement and outcomes, and the cost implications of AI-led DPPs. While the study was funded by NIH and supported by various Johns Hopkins resources, researchers emphasize that the AI-DPP can offer a reliable, scalable alternative in routine clinical settings.
About the Study and Acknowledgments
The trial was funded by the National Institute of Diabetes and Digestive and Kidney Diseases and the National Institute on Aging. Authors from Johns Hopkins and collaborating institutions contributed to data collection, analysis, and interpretation. Sweetch Health, Ltd. provided services to participants under compensated terms, with safeguards to ensure data integrity. The study’s outcomes open a path for AI-enabled lifestyle interventions to support diabetes prevention at scale.
Conclusion: A Step Toward Wider Access to Effective DPPs
As AI-powered interventions prove their merit alongside traditional, coach-led programs, healthcare systems have an opportunity to expand access to effective diabetes prevention. The convergence of technology, behavioral science, and public health could reshape how we reach millions with prediabetes, delivering personalized, practical guidance that fits into people’s real lives.
