Interpreting the risk of postoperative post-neuraxial puncture-site pain after cesarean section: a machine learning approach evaluating nalbuphine infiltration along the epidural puncture tract in a prospective observational cohort.

BACKGROUND: Postoperative post-neuraxial puncture-site pain following combined spinal-epidural anesthesia can impede early maternal recovery. METHODS: In this prospective observational cohort study (n = 473), parturients undergoing cesarean section were categorized into Group L (ropivacaine infiltration) or Group NR (nalbuphine-ropivacaine infiltration along the epidural tract). We evaluated early clinical outcomes and developed machine learning models to predict post-neuraxial puncture-site pain risk, utilizing SHapley Additive exPlanations (SHAP) to interpret feature associations. RESULTS: Group NR exhibited lower incidences of acute post-neuraxial puncture-site pain, reduced opioid consumption, and attenuated early postpartum depressive symptoms compared to Group L (all P < 0.05). Among the predictive models, XGBoost demonstrated superior discriminative performance (AUC = 0.912). SHAP analysis indicated that Nalbuphine infiltration, body mass index, and maternal age were the primary contributors associated with the predicted risk, revealing non-linear dependencies. CONCLUSION: Local epidural infiltration of nalbuphine and ropivacaine may decrease acute post-neuraxial puncture-site pain incidence and improve early postpartum recovery. The XGBoost model, combined with SHAP analysis, suggests robust capabilities in predicting post-neuraxial puncture-site pain risk and elucidating complex clinical associations.