Predicting Hypothermia in Gynecological Laparoscopic Surgery Using Machine Learning

Development and Validation of a Machine Learning Model to Predict Hypothermia in Gynecological Laparoscopic Surgery Based on Preoperative Clinical Indicators: A Multicenter Prospective Cohort Study

Brief Title: Predicting Hypothermia in Gynecological Laparoscopic Surgery Using Machine Learning

Brief Summary: This study aims to develop and validate a machine learning model for predicting intraoperative hypothermia (IOH) in patients undergoing gynecological laparoscopic surgery based on preoperative clinical indicators. This prospective, multicenter case-control study will enroll female patients aged 18 years and older who are scheduled for laparoscopic surgery across multiple hospitals from 2026 to 2027. The primary objective is to identify high-risk patients who may experience IOH, defined as a core temperature below 36.0°C during surgery.

Participants will be classified into two groups: the IOH group, consisting of patients who experience hypothermia, and the normal temperature group, comprising patients who maintain a core temperature of 36.0°C or higher. Data collection will include demographics, comorbidities, surgical details, anesthesia information, and preoperative laboratory results.

The primary outcome measure will be the area under the curve (AUC) of the model, assessing its predictive performance at various thresholds. Secondary outcomes will include sensitivity, positive predictive value, negative predictive value, and F1 score. The study hypothesizes that the developed machine learning model will significantly improve the accuracy and timeliness of predicting IOH, thereby enhancing patient safety during surgery and postoperative recovery. This research is expected to inform clinical practices related to preventative warming strategies, ultimately improving patient outcomes in gynecological laparoscopic surgery.

Background: Intraoperative hypothermia (IOH), defined as a core body temperature below 36.0°C during surgery, is a common complication with an incidence as high as 50% in gynecological laparoscopic procedures. IOH is associated with adverse outcomes including surgical site infections, increased blood loss, cardiovascular complications, prolonged recovery, and higher healthcare costs. Accurate preoperative identification of patients at high risk for IOH is crucial for implementing targeted preventative measures and optimizing resource allocation.

Objective: The primary objective of this study is to develop and validate a machine learning model that utilizes preoperative clinical indicators to predict the occurrence of IOH specifically in patients undergoing gynecological laparoscopic surgery.

Study Design: This is a multicenter, prospective case-control study. Data will be prospectively collected from participating hospitals between 2026 and 2027.

Technical Methods:

Sample Size: Based on an estimated IOH incidence of 40% and 24 predictor variables, a minimum sample size of 1500 participants is planned to ensure adequate power for model development and validation.

Data Collection: Clinical data will be collected using electronic medical records (EMR). Core body temperature will be monitored intraoperatively using a wireless temperature monitoring system.

Statistical Analysis & Model Development: Data analysis will be performed using SPSS (v25.0) and R (v4.3.1). The dataset will be randomly split into training (80%) and testing (20%) sets. The Least Absolute Shrinkage and Selection Operator (LASSO) regression will be applied to the training set for feature selection. Six machine learning algorithms-Support Vector Machine (SVM), Logistic Regression (LR), Multilayer Perceptron (MLP), Random Forest (RF), Extreme Gradient Boosting (XGB), and Decision Tree (DT)-will be developed. Model hyperparameters will be optimized via 10-fold cross-validation.

Model Evaluation: The performance of all models will be independently validated on the test set. The primary metric for model comparison and selection will be the Area Under the Receiver Operating Characteristic Curve (AUC). Secondary performance metrics include sensitivity, positive predictive value (PPV), negative predictive value (NPV), and F1-score. The optimal cutoff point for the final selected model will be determined by maximizing Youden's index.

Ethical Considerations: This study will be conducted following approval by the Institutional Review Boards/Ethics Committees of all participating centers. Written informed consent will be obtained from all participants. The study protocol will be registered in a clinical trial registry to ensure transparency. All participant data will be handled with strict confidentiality and in accordance with relevant data protection regulations.

Expected Outcomes: This study is expected to result in a validated machine learning model capable of accurately predicting IOH risk prior to gynecological laparoscopic surgery. The identification of key predictive factors and the deployment of this model aim to facilitate personalized preventative care, reduce the incidence of IOH, and improve patient safety and postoperative recovery outcomes.

Inclusion Criteria:

• Female patients aged 18 years or older.
• Patients scheduled for laparoscopic surgery.

Exclusion Criteria:

• Preoperative body temperature exceeding 37.5°C or below 36.0°C.
• History of hypothyroidism or hyperthyroidism.
• Patients with thermoregulatory dysfunction, such as severe infection or central nervous system disorders.
• Patients who refuse to sign the informed consent form.