Prediction Of Interoperative Hypertension Using An Interpretable Deep Learning Model With Automatically Generated Featur

Abstract

The surgical complication rate increases when patients experience Intraoperative Hypotension during operations. Predicting IOH onset early enables swift medical actions that lead to better surgical outcomes for patients. The research implements the Xgboost algorithm to forecast IOH because it exhibits both success in prediction and clear interpretability. Our analysis includes detailed characteristics which merge patient demographic data with surgical data and critical measurements together with anesthesia measurements. These variables are automatically derived from actual surgical data. A performance evaluation of the XgBoost model occurred using standard classification metrics after it processed multi-type data from surgical patients throughout various procedures. The model achieved improved prediction accuracy and decision clarity through implementation of SHAP values. The XgBoost algorithm shows effectiveness in predicting IOH initiation and provides stable clinical decision support. Implementation of this model in surgical operations enables healthcare teams to prevent and manage intraoperative hypotension alongside reduced negative clinical results and increased patient protection.