Interpretable Prediction of Mortality Risk in Elderly Patients With Type 2 Diabetes Mellitus and Cerebral Infarction Based on the Medical Information Mart for Intensive Care-Ⅳ Database.

Key Points

• The aim is to predict 30-day and 365-day mortality risk in elderly patients with type 2 diabetes and cerebral infarction using a machine learning model.
• Utilized the Medical Information Mart for Intensive Care-Ⅳ database with elderly ICU patients
• Applied the XGBoost model for prediction of mortality risk
• Conducted SHAP analysis to identify important predictive factors.
• Identified key predictive factors for 30-day mortality: Oxford Acute Severity of Illness Score, length of hospital stay, congestive heart failure, peripheral capillary oxygen saturation, and heart rate.
• Top predictors for 365-day mortality included blood urea nitrogen, age, and respiratory rate.
• The XGBoost model demonstrates significant predictive capability for mortality risk in this population.