Prediction model for intrapartum labor analgesia efficacy based on preoperative multidimensional indicators.

Background: Effective labor analgesia is paramount for maternal well-being and positive birth experiences. However, the efficacy of intrapartum labor analgesia exhibits considerable inter-individual variability. Identifying reliable preoperative predictors for labor analgesia efficacy is crucial for optimizing pain management strategies and enhancing patient outcomes. This study aimed to develop and validate a prediction model for intrapartum labor analgesia efficacy utilizing preoperative multidimensional indicators. Methods: This retrospective study enrolled 137 parturients who received labor analgesia. Preoperative multidimensional indicators, encompassing demographic data, clinical characteristics, and psychological assessments, were systematically collected. Logistic Regression (LR), Naive Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGB) models were developed and comparatively analyzed. Feature selection was conducted using LASSO regression. Model performance was rigorously evaluated using metrics such as Area Under the Curve (AUC), sensitivity, specificity, F1 score, positive predictive value, negative predictive value, calibration curves, and decision curve analysis. Results: Initial logistic regression analysis revealed several significant preoperative predictors, including pregnancy-induced hypertension (PIH), binary ultrasound cervical length, and continuous ultrasound cervical length. LASSO regression further refined the feature set, illustrating the dynamic changes in coefficients across varying lambda values. Among the developed models, the XGB model consistently demonstrated superior discriminative ability, achieving an AUC of 0.875 (95% CI: 0.802 to 0.948) in the training set and 0.759 (95% CI: 0.606 to 0.912) in the validation set, thereby indicating robust performance. Calibration curves provided insights into the agreement between predicted and observed probabilities across models. Decision curve analysis elucidated the clinical utility of these models, underscoring the potential net benefit of employing these predictive tools. Conclusion: This study successfully developed and validated a prediction model for intrapartum labor analgesia efficacy grounded in preoperative multidimensional indicators. The XGB model exhibited promising performance, suggesting its potential for clinical application in identifying parturients who could benefit from individualized pain management strategies. Further prospective validation in diverse cohorts is warranted to confirm these findings and facilitate clinical implementation.