An Explainable AI-Based Scoring System for Mode-of-Delivery Decision Support: Integrating Clinical Thresholds with Secondary Infertility Risk Prediction

Background: The global rise in elective cesarean section (CS) rates, often exceeding 50%, poses significant maternal and neonatal risks. Beyond immediate surgical complications, traumatic births are increasingly linked to long-term reproductive challenges, such as secondary infertility.Objective: This study develops an explainable, evidence-synthesized, proof-of-concept machine learning (ML) framework for clinical decision support, aimed at optimizing delivery mode selection and mitigating future infertility risks.Methods: Utilizing data synthesized from 70 high-impact studies (PubMed, Web of Science, PMC), this study developed a dual-stage predictive model using light gradient boosting machine (LightGBM). The framework incorporates 17 biopsychosocial features. Class imbalance was addressed via synthetic minority over-sampling technique (SMOTE), and model interpretability was ensured through shapley additive explanations (SHAP) analysis. An optimized clinical threshold was established using Youden’s J statistic.Results: The primary model achieved high diagnostic accuracy, identifying an optimal decision threshold of 0.594. SHAP analysis revealed that fear of childbirth (FOC), Parity, and BMI were the most influential predictors. The secondary infertility model demonstrated a robust AUC-ROC of 0.876 and 81% accuracy, projecting a mean population risk of 51.54%. Findings indicate that psychological trauma and unmanaged labor pain are significant shared drivers for both surgical intervention and long-term fertility challenges.Conclusion: The explainable AI scoring system provides a standardized tool to identify high-risk patients before the golden time elapses. Utilizing the 0.594 threshold enables targeted interventions (such as mindfulness and doula support) to reduce unnecessary cesarean rates and safeguard maternal reproductive longevity.