Machine Learning Model Developed to Predict Postoperative Pneumonia in Aneurysmal Subarachnoid Hemorrhage Patients

A new study has developed a machine learning model to predict postoperative pneumonia (POP) in patients undergoing surgical treatment for aneurysmal subarachnoid hemorrhage (aSAH). The internal cohort study involved 706 aSAH patients, randomly divided into 80% training and 20% testing sets. Using perioperative data, six machine learning models were established, with logistic regression demonstrating the highest predictive accuracy (AUC: 0.91). Independent predictors of POP included mechanical ventilation time, Glasgow Coma Scale, smoking history, albumin level, neutrophil-to-albumin ratio, and c-reactive protein-to-albumin ratio. External validation from the MIMIC-IV database confirmed the model's efficacy (AUC: 0.89). This predictive model could facilitate early identification and intervention for high-risk POP patients.