MediDT: Digital TwinBased Predictive Analytics for Personalized Patient Recovery

Abstract

In this paper, we present MediDT, a Digital Twinbased predictive analytics framework designed to forecast patient recovery trajectories and assess individualized risk profiles. Unlike conventional clinical prediction models that provide static estimates, MediDT dynamically simulates patient health progression by integrating real-world clinical records with digital twin forecasting. Our framework employs supervised machine learning models trained on synthetic patient datasets, incorporating features such as demographics, comorbidities, activity levels, sleep quality, and vitals. MediDT generates forward-looking "what-if" scenarios, allowing clinicians to evaluate the impact of lifestyle changes (e.g., increased physical activity or improved sleep) on recovery outcomes. Furthermore, the system merges predicted recovery curves with actual patient data in unified Excel exports, enabling physicians to compare real and simulated trajectories for clinical validation. Experimental evaluation shows that MediDT achieves a mean absolute error (MAE) of 2.3 days, an RMSE of 3.8 days, and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score of 0.91 in recovery prediction, while risk classification attains 89.7% accuracy and an F1-score of 0.88. These results demonstrate that MediDT provides reliable, interpretable, and actionable forecasts for personalized healthcare monitoring. By offering a scalable and interactive decision-support tool, MediDT bridges the gap between predictive modeling and practical clinical application.

Ähnliche Arbeiten