AI-led diabetes prevention matches clinical outcomes but lacks patient preference

A randomized trial has found that while AI-driven diabetes prevention programs can match human-led coaching in clinical outcomes, they may be less satisfactory for patients. The study, which involved 368 adults with prediabetes and obesity, compared a fully automated AI program via the Sweetch app against traditional coach-led interventions.

While the AI group demonstrated faster start times and more consistent engagement over 12 months, the technology fell short regarding patient preference and acceptability. Clinical results, including weight loss, physical activity, and A1C reduction, were found to be non-inferior between the two models.

These findings coincide with discussions at the International Diabetes Federation Western Pacific Region Congress in Melbourne, where experts are examining the shift from monitoring current glucose levels to using digital health and artificial intelligence to anticipate future risks and reduce patient uncertainty.