AI-Driven Predictive Models for Patient Readmission in Post-surgical Care
Main article
Abstract
Post-surgical readmission remains a persistent challenge in healthcare, which contributes significantly to patient morbidity, healthcare costs, and system inefficiencies. Accurate early prediction of readmission risk is crucial for enabling proactive interventions and improving surgical outcomes. In this study, we developed and evaluated AI-driven predictive models, including XGBoost, random forests, and neural networks, utilizing structured clinical data from sources such as MIMIC-III and hospital databases. Through comprehensive feature engineering encompassing clinical indicators and socioeconomic factors, and the application of explainable AI techniques like SHAP, we identified key predictors of readmission and achieved superior model performance compared to conventional statistical methods. Our findings demonstrate that AI models not only enhance prediction accuracy but also provide clinically interpretable insights and facilitate personalized post-operative care strategies. However, challenges such as model interpretability, data privacy, generalizability, and ethical considerations must be addressed for successful real-world deployment. This research contributes a validated predictive framework and envisions intelligent post-operative management systems that leverage AI to optimize patient outcomes and healthcare resource utilization.
