Scholarlink

Objective <p>This study aims to develop multiple machine learning models for predicting hypothermia risk in laparoscopic surgery and to perform interpretability analysis of the best-performing model. Our goal is to provide robust decision support for clinicians and ensure safe and effective patient care during surgical procedures.</p> Methods <p>This study included 1,030 patients who underwent laparoscopic surgery at Shandong Provincial Cancer Hospital, affiliated with Shandong First Medical University, between July 2023 and June 2024. We employed the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm for feature selection. We explored the performance of five machine learning algorithms—logistic regression (LR), decision tree (DT), random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGBoost)—to predict hypothermia risk during laparoscopic surgery. Finally, we conducted an interpretability analysis of the top-performing model using Shapley Additive Explanations (SHAP).</p> Results <p>LASSO identified nine risk factors: BMI, ASA classification, total volume of intravenous fluids, irrigating fluids during the operation, volume of CO<sub>2</sub>, blood loss, ambient temperature, long-term alcohol consumption, and type of surgery. Performance comparison among the five models revealed that the XGBoost model performed the best, with an accuracy of 0.762 (95% CI: 0.717–0.807) and an area under the curve (AUC) of 0.835 (95% CI: 0.794–0.872). The model achieved a specificity of 0.749 (95% CI: 0.683–0.816) and a sensitivity of 0.773 (95% CI: 0.713–0.824). The F1 score was 0.778 (95% CI: 0.735–0.828). SHAP analysis revealed that the four most influential factors for hypothermia risk during laparoscopic surgery were operating room temperature, rinsing fluids during the operation, volume of CO2, and blood loss.</p> Conclusions <p>This study developed an efficient and interpretable predictive model for the risk of hypothermia in patients undergoing laparoscopic surgery. This model supports clinical decision-making and contributes to the overall goal of providing personalized care in the surgical environment.</p>

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

Explainable prediction of hypothermia risk in laparoscopic surgery: a retrospective cross-sectional study using machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月3日

查看全文
  • Wenyu Su,
  • Xiaoli Wang,
  • Huiyu Jia,
  • Wenjing Chang,
  • Shan Jiang,
  • Huaiju Ge,
  • Shihong Dong,
  • Jie Yu,
  • Guifeng Ma,
  • Yingtao Meng

摘要

Objective

This study aims to develop multiple machine learning models for predicting hypothermia risk in laparoscopic surgery and to perform interpretability analysis of the best-performing model. Our goal is to provide robust decision support for clinicians and ensure safe and effective patient care during surgical procedures.

Methods

This study included 1,030 patients who underwent laparoscopic surgery at Shandong Provincial Cancer Hospital, affiliated with Shandong First Medical University, between July 2023 and June 2024. We employed the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm for feature selection. We explored the performance of five machine learning algorithms—logistic regression (LR), decision tree (DT), random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGBoost)—to predict hypothermia risk during laparoscopic surgery. Finally, we conducted an interpretability analysis of the top-performing model using Shapley Additive Explanations (SHAP).

Results

LASSO identified nine risk factors: BMI, ASA classification, total volume of intravenous fluids, irrigating fluids during the operation, volume of CO2, blood loss, ambient temperature, long-term alcohol consumption, and type of surgery. Performance comparison among the five models revealed that the XGBoost model performed the best, with an accuracy of 0.762 (95% CI: 0.717–0.807) and an area under the curve (AUC) of 0.835 (95% CI: 0.794–0.872). The model achieved a specificity of 0.749 (95% CI: 0.683–0.816) and a sensitivity of 0.773 (95% CI: 0.713–0.824). The F1 score was 0.778 (95% CI: 0.735–0.828). SHAP analysis revealed that the four most influential factors for hypothermia risk during laparoscopic surgery were operating room temperature, rinsing fluids during the operation, volume of CO2, and blood loss.

Conclusions

This study developed an efficient and interpretable predictive model for the risk of hypothermia in patients undergoing laparoscopic surgery. This model supports clinical decision-making and contributes to the overall goal of providing personalized care in the surgical environment.