Predictive tool for the risk of hypothermia during laparoscopic gynecologic tumor resection

Based on the machine learning algorithm, construct a hypothermia prediction model for gynecological tumor resection under laparoscopic general anesthesia.

This research conducted a retrospective analysis, gathering data from individuals who had undergone minimally invasive surgical procedures for gynecological tumors in a Chinese Hospital, ranging from June 2018 to August 2024. During this timeframe, a total of 308 cases were examined for analysis, with 70% of the cases allocated to the modeling dataset and the remaining 30% designated for the validation dataset. The factors associated with intraoperative hypothermia were identified within the modeling dataset. Subsequently, three predictive models were established: a Classification and Regression Trees model, a Random Forest model, and a Support Vector Machine model.

The incidence of intraoperative hypothermia in 308 cases was 30.84%. The results showed that age, body mass index, preoperative body temperature, preoperative albumin, operating room temperature and peritoneal lavage fluid volume were the influencing factors of intraoperative hypothermia. Using these variables, it was determined that the Random Forest model had demonstrated strong predictive performance.

The prediction model developed using the random forest algorithm exhibits excellent predictive capabilities, which are highly significant for pinpointing the critical factors contributing to intraoperative hypothermia in patients undergoing laparoscopic procedures.