Statistical and AI-Based Hybrid Modeling for Predicting Chronic Disease Progression

Non-communicable diseases such as diabetes, hypertension, and cardiovascular diseases are a growing worldwide health concern. Early-stage prediction of disease progression is critical for improving clinical outcomes and reducing healthcare costs. Conventional statistical models may assist in highlighting significant risk predictors, yet cannot suitably model high-dimensional, nonlinear construct the large clinical datasets. Machine learning methods, such as Random Forests, Support Vector Machines and Neural Networks, provide very high prediction power but tend to provide low interpretability.This study aims to develop hybrid model for chronic illness progression forecasting by supervised learning methods using statistical and AI-based approaches. We will use freely available datasets such as the UCI Chronic Disease, NHANES, and MIMIC-III. The study will compare the predictive performance of logistic regression and Cox regression against the existing standard of care (the AI models) but also against a new model that integrates aspects of both approaches.In this study, feature encoding techniques and imbalance data fixing methods such as SMOTE, etc. have been used. The evaluation metrics that will be used to assess model performance include accuracy, precision, recall, F1-score, and ROC-AUC. The result is expected to show that hybrid models can offer better predictive performance while keeping statistical interpretability.This research helps develop transparent and efficient clinical decision-support systems that provide medical staff with realistic opportunities for early detection of risk and the tailored preparation of care