Retrospective Machine Learning Approach for Forecasting In-Hospital Death in ICU Patients After Cardiac Arrest

Accurate identification of patients at high risk of in-hospital mortality in intensive care units (ICUs) is vital for enhancing clinical decision-making and improving patient care strategies. As traditional statistical models often fall short in modeling nonlinear and multifactorial clinical variables, this study explores a machine learning (ML) approach to overcome these limitations. Our research performed a retrospective study using the MIMIC-IV database, focusing on 2,385 ICU patients who met predefined eligibility criteria. Numerical features were summarized through statistical aggregations (maximum, minimum, mean), while categorical attributes underwent structured encoding. The dataset was split into 70% for training and 30% for validation. We applied a combination of regularization techniques (LASSO, Ridge, ElasticNet) and Random Forest-based importance ranking for feature selection. Multiple supervised ML algorithms, including CatBoost, XGBoost, and Support Vector Machines, were benchmarked using metrics such as AUC-ROC, calibration plots, and decision curve analysis. SHAP values were employed to enhance model explainability. The CatBoost algorithm achieved the most favorable results with AUC scores of 0.904 and 0.868 on the training and test sets. These findings suggest that the proposed model offers a reliable, interpretable, and potentially integrable solution for ICU mortality risk prediction.