Development of a Machine Learning Model to Predict Epidural-Related Maternal Fever During Labor Analgesia: A Multi-Algorithm Comparative Study with Prospective Implementation Framework.

Objective: To develop and validate a machine learning-based predictive model for assessing the risk of epidural-related maternal fever (ERMF)- a common complication during labor analgesia. Methods: A prospective cohort study was conducted among 500 parturients with term singleton pregnancies who received epidural labor analgesia between September 2022 and August 2023. Key variables collected include maternal demographic characteristics, anesthesia-related indicators for complications, and obstetric features. Following application of exclusion criteria, 422 parturients were included and allocated into a modeling cohort (n = 337) and a validation cohort (n = 85) using stratified random sampling at an 8:2 ratio. Eleven machine learning algorithms were utilized to construct predictive models. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), area under the precision-recall curve (AUPRC), accuracy, precision, recall, and F1-score. Results: The incidence of ERMF was 28.1% (119/422). Among the evaluated algorithms, Categorical Boosting (CatBoost) demonstrated the highest performance, with an AUC of 0.94 (95% CI: 0.86-0.98), accuracy of 90.59%, precision of 0.88, and average precision (AP) of 0.86 in the validation cohort. Analysis using SHapley Additive exPlanations (SHAP)-an interpretable artificial intelligence method- identified prolonged duration of rupture of membranes, higher maternal body mass index, and nulliparity as the top predictors of ERMF risk. An interactive web-based interface was developed to facilitate real-time clinical risk evaluation. Conclusion: A machine learning model with high discriminative ability was constructed to predict the risk of ERMF. The CatBoost algorithm effectively identified parturients at elevated risk, and the accompanying visual tool offers evidence-based support for stratified management of intrapartum fever in clinical practice.