A Comparison of Data Sampling Techniques for Predicting Postoperative Delirium in Neurosurgery.

This research created an XGBoost model to predict postoperative delirium in neurosurgical patients, addressing data imbalance through various sampling methods. Utilizing data from 2017-2023, the study found the SMOTETomek under-over sampling technique yielded the best performance, with an AUROC of 0.97. Significant risk factors identified include preoperative hospital stay, ICU admission, multiple diagnoses, fall risk, and prior neurological procedures. This model facilitates early detection and tailored prevention for patients at high risk of postoperative delirium.