Till Örebro universitet

OBJECTIVE: To assess the accuracy and validity of the DEtermination of Diabetes Utilities, Costs, and Effects (DEDUCE) model, a Microsoft Excel-based tool for evaluating diabetes interventions for type 1 and type 2 diabetes.

RESEARCH DESIGN AND METHODS: The DEDUCE model is a patient-level microsimulation, with complications predicted based on the Sheffield and RECODe diabetes models for type 1 and type 2 diabetes, respectively. For this tool to be useful, it must be validated to ensure that its complication predictions are accurate. Internal, external and cross validation was assessed by populating the DEDUCE model with the baseline characteristics and treatment effects reported in clinical trials used in the Fourth, Fifth, and Ninth Mount Hood Diabetes Challenges. Results from the DEDUCE model were evaluated against clinical results and previously validated models via mean absolute percentage error (MAPE) or percentage error.

RESULTS: The DEDUCE model performed favorably, predicting key outcomes including cardiovascular disease in type 1 diabetes and all-cause mortality in type 2 diabetes. The model performed well against other models. In the Mount Hood 9 Challenge comparison, error was below the mean reported from comparator models for several outcomes, particularly for hazard ratios.

CONCLUSIONS: The DEDUCE model predicts diabetes-related complications from trials and studies well when compared to previously validated models. The model may serve as a useful tool for evaluating the cost-effectiveness of diabetes technologies.