Scholarlink

Postinduction hypotension (PIH) is one of the complications that may cause organ injury, prolonged hospital stays and postoperative mortality. Conventional monitoring fails to predict PIH early, thus a proactive approach based on high technology is needed. This study evaluates the use of time-weighted mean arterial pressure (TWA-MAP) to predict PIH. A total of 184 patient records were extracted from the open database, VitalDB were analysed using Orange Data Mining software. Six ML algorithms were tested, logistic regression algorithm performed best with the highest in AUC value (0.844), making it the most reliable model for real clinical applications. Feature importance analysis identified that TWA-MAP was the most significant predictor of PIH risk, with the largest decrease in AUC value (0.062). Further studies are recommended using prospective data for investigation medical data from other surgical risk classes.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Data Mining and Machine Learning Approach for Predicting Postinduction Hypotension Using Time-Weighted Average MAP (TWA-MAP)

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zulfadhli Osman,
  • Fatanah Mohamad Suhaimi,
  • Ahmad Fakrurrozi Mohamad

摘要

Postinduction hypotension (PIH) is one of the complications that may cause organ injury, prolonged hospital stays and postoperative mortality. Conventional monitoring fails to predict PIH early, thus a proactive approach based on high technology is needed. This study evaluates the use of time-weighted mean arterial pressure (TWA-MAP) to predict PIH. A total of 184 patient records were extracted from the open database, VitalDB were analysed using Orange Data Mining software. Six ML algorithms were tested, logistic regression algorithm performed best with the highest in AUC value (0.844), making it the most reliable model for real clinical applications. Feature importance analysis identified that TWA-MAP was the most significant predictor of PIH risk, with the largest decrease in AUC value (0.062). Further studies are recommended using prospective data for investigation medical data from other surgical risk classes.