Abstract: PURPOSE: To develop and validate a machine learning-based prediction model for postoperative pain in pediatric patients undergoing head and facial day surgery. METHODS: A total of 893 children who underwent head and facial day surgery under general anesthesia were retrospectively included, the patients were categorized into pain (FLACC score ≥4) or non-pain (FLACC score <4) group based on assessments in the postanesthesia care unit (PACU). Variables were selected via the least absolute shrinkage and selection operator (LASSO) regression with fivefold cross-validation. Six machine learning algorithms—including logistic regression (LR), K-nearest neighbor(KNN), support vector machine (SVM), decision tree (DT), bidirectional recurrent neural network (BRNN), and artificial neural network (ANN)—were trained on a 7∶3 split dataset. Model performance was evaluated using the area under the receiver operating characteristic curve (AUROC). RESULTS: The incidence of postoperative pain was 23.63%. Key predictors included preoperative m-YPAS scores, parental SAI scores, surgical history, procedure type, age, gender and induction with sufentanil/ketamine. Four out of 6 different machine learning algorithms (LR, SVM, BRNN, and ANN) were able to maintain good performance in the test set (AUROC>0.70). Among the models, logistic regression achieved the highest AUROC (0.78, 95%CI: 0.72-0.84) in the test set. CONCLUSIONS: The prediction model based on machine learning in this study can effectively predict postoperative pain in pediatric patients undergoing head and facial day surgery.

Key words: Head and facial day surgery, General anesthesia, Pediatric patients, Postoperative pain, Risk prediction model