Machine Learning Models for Predicting Mortality Risk and Survival Time in Lung Cancer Patients Treated with...

This study develops machine learning models to predict patient mortality and estimate survival time using electronic health record (EHR) data from three Taipei Medical University-affiliated hospitals (TMU Hospital, Wan-Fang Hospital, and Shuang Ho Hospital). We built and evaluated predictive models for (1) binary mortality risk and (2) time-to-event survival. In testing, the best classification model achieved a high area under the ROC-curve (AUC ≈ 0.82), while the best survival model yielded a concordance index (C-index) on the order of 0.66. These results demonstrate that multi-institution EHR data can support accurate prognostic modeling, combining both classification and survival analysis to inform clinical decision-making.