Objective: Implement in an electronic health record (EHR) a predictive model for people with prediabetes that provides individualized benefit estimates for taking metformin or participating in the Diabetes Prevention Program (DPP). This model stratifies people with prediabetes by the potential to reduce their risk of progression to diabetes based on regression models developed directly on the DPP trial and recalibrated to Optum data. With 1/3 of adults having prediabetes, health systems need a way to prioritize.

Study Design: Pre/post implementation study including provider and patient surveys and data on use of the predictive model. The model estimates risk of progression to diabetes with usual care, the DPP, or metformin, based on 11 demographic, biometric and diagnosis variables available in the EHR.

Population Studied: Patients with prediabetes at 10 pilot primary care clinics, where providers access the model via an EHR click and the data elements are automatically populated from the EHR.

Principal Findings: The predictive model was used on 68% (n=2,304) of patients with prediabetes between 5/1/18 and 1/31/19. A total of 43% of patients were classified as high-risk, 54% as moderate-risk, and 4% as low-risk. Treatment - either referrals to the DPP or prescriptions for metformin - was provided for 63% of high-risk, 15% of moderate-risk, and 4% of low-risk patients. DPP referrals and metformin prescriptions for these high-risk patients increased substantially after implementation of the calculator (DPP: 0% to 44%; Metformin: 2% to 19%).

Conclusions: A predictive model, embedded in the EHR, that predicts individual patient risk for developing diabetes at the point of care improved treatment for patients with prediabetes. Use of individualized risk estimates resulted in the prioritization of treatment for patients at greatest risk of developing type 2 diabetes.


E.L. Ciemins: None. J.E. Powelson: None. J.P. Nelson: None. F.R. Colangelo: None. J.K. Cuddeback: None. D.M. Kent: None.


Patient-Centered Outcomes Research Institute

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.