Predicting Diabetes Risk with the Inertia-Based P_diab Model: A Dynamic AI-Driven Framework for Personalized Prevention

Type 2 diabetes (T2D) remains one of the most significant global health challenges. Conventional risk-prediction tools such as FINDRISC and ADA criteria rely on static, snapshot-based parameters that fail to capture the dynamic interaction between metabolic, behavioral, and environmental factors driving disease progression.

This study introduces P_diab, a model that uses inertia as an analytical construct to reflect how biological and behavioral systems resist or respond to change over time. By integrating metabolic inertia (insulin resistance, adiposity, glucose dynamics), cognitive-behavioral patterns (activity levels, stress, sleep), and environmental context, the model provides a unified multidimensional assessment of diabetes risk.

Calibrated using large-scale datasets (NHANES, DPP, UK Biobank), the P_diab model demonstrated high predictive accuracy (AUC 0.95–0.96), outperforming traditional scoring methods. The inclusion of dynamic inertia-related indicators, threshold effects in glucose regulation, and discrepancy measures between perceived and actual response (ΔIdiab) enables earlier detection of destabilizing trends and more adaptive prevention strategies.

An AI-based clinical tool was built on top of this model, allowing users to:

– compute personalized risk profiles,

– simulate the impact of lifestyle adjustments,

– receive adaptive, evidence-informed recommendations.

Rather than replacing existing frameworks, P_diab bridges static risk scores and dynamic health monitoring. By incorporating time-dependent metabolic and behavioral patterns, the model offers a more precise understanding of how diabetes risk accumulates, accelerates, or stabilizes.

Future work includes testing across diverse populations, integration with wearable monitoring devices, and deployment in real-time digital health applications.