AI-Driven Mortality Prediction in COVID-19 Patients Using Advanced Feature Selection

COVID-19 has caused significant global mortality, with early risk stratification being critical for effective clinical management. Using a dataset of 8,032 COVID-19 hospitalized patients from a multicenter UK study, we developed and evaluated seven AI models, including deep and machine learning techniques, to predict in-hospital mortality. Key predictors were identified through a rigorous feature selection process combining statistical analysis, clinical expertise, and literature review. The Support Vector Classifier (SVC) achieved the best performance with 84% accuracy, 86% precision, and an AUC of 0.858, outperforming other methods in robustness and predictive accuracy. This study presents a novel application of AI on a large and diverse dataset, offering valuable insights for managing future pandemics/other clinical setting and improving clinical decision-making to reduce mortality.