dynnomapp

dynnomapp predicts four-year risk of type 2 diabetes mellitus (T2DM) in adults with metabolic syndrome using statistical modeling and machine learning.

Key Features:

• Prediction horizon: Provides four-year risk estimates for incident T2DM.
• Predictor variables: Uses Age, Gender, Body Mass Index (BMI), Diastolic Blood Pressure, Fasting Plasma Glucose, and Alanine Aminotransferase levels as model inputs.
• Modeling approaches: Employs statistical modeling and machine learning techniques.
• Development cohort: Developed on a large multicenter cohort study across 32 sites in China.
• External validation: Externally validated in the Henan population-based cohort study.
• Observed incidence: Reported T2DM incidence during follow-up was 17.63% in the development cohort and 18.67% in the validation cohort.
• Performance metrics: Achieved AUC 0.824 (95% CI, 0.759–0.889) in the training cohort and AUC 0.732 (95% CI, 0.594–0.871) in the external validation cohort.
• Calibration: Internal and external calibration plots showed good agreement between predicted probabilities and observed outcomes.

Scientific Applications:

• Clinical risk stratification: Identify adults with metabolic syndrome at high risk of developing T2DM for targeted monitoring or intervention.
• Preventive planning: Inform four-year preventive strategies and individualized management decisions.
• Resource allocation: Support prioritization of healthcare resources based on predicted T2DM risk.
• Epidemiological research: Facilitate validation of risk predictors and estimation of population-level T2DM risk across cohorts.

Methodology:

Statistical modeling and machine learning techniques were applied with predictor selection based on statistical significance, and model performance assessed by AUC and calibration in internal and external validation using a multicenter development cohort (32 sites in China) and the Henan population-based external cohort.