Machine learning effectively predicts mortality risk in elderly patients

A robust mortality risk prediction model was developed for geriatric patients over 65 undergoing non-cardiac surgery. Utilizing data from 1960 patients and machine learning techniques, the CatBoost model outperformed traditional risk assessments with an AUC of 0.96 and 0.98 in validation datasets. Key factors influencing the model’s accuracy included anion gap, age, prothrombin time, and weight. This web-based application supports clinicians in improving risk assessments and decision-making for high-risk elderly patients.