A novel electronic health record-based prediction model successfully identified patients who were at the highest risk of developing type 2 diabetes up to 10 years later. Researchers presented the findings at the 2026 Scientific Sessions of the American Diabetes Association, which took place from June 5-8 in New Orleans, US.
The retrospective cohort study included 3,365,464 adults aged 18–70 receiving care at Kaiser Permanente Northern California, US, from 2012 to 2024. The median patient age was 39 years and 55 per cent of patients were female. The study used a hazard-based super learning approach that combined multiple survival-analysis models to estimate each patient’s risk of developing type 2 diabetes over the next one, three, and 10 years. The model used clinical and demographic information routinely collected at medical visits, such as age, weight, blood glucose (blood sugar) levels, medical history, and medications, along with publicly available data, such as access to healthy food and walkable areas.
During a median follow-up of 5.4 years, the study found a type 2 diabetes incidence of 10.7/1,000 person-years. The training model effectively identified adults at high-risk for type 2 diabetes with an area under the curve of 0.886 (95% CI: 0.883–0.888). The validation model scored 0.883 (95% CI: 0.88–0.886). The one-year follow-up resulted in a near-ideal calibration (mean predicted risk 1.03% versus observed 1.01%). At the threshold defining high risk (>1.2% risk), the model had a sensitivity of 74 per cent and a specificity of 82 per cent over up to 10 years of follow-up.
“These findings represent a potential advancement over existing approaches for identifying individuals at risk of developing type 2 diabetes by enabling earlier, more precise detection and supporting a more targeted, proactive approach to prevention,” said Dr Luis A Rodriguez, lead author of the study.
Leave a Reply
You must be logged in to post a comment.