BMI-Stratified Risk of Cesarean Delivery Following Labor Induction: A Robson Classification-Based Cohort Study with Predictive Modeling

Objectives: Maternal obesity is associated with adverse labor induction outcomes, but real-world data stratified by the full WHO body mass index (BMI) classification and adjusted for parity remain limited. We evaluated BMI-stratified induction outcomes using the Robson Ten-Group Classification System and developed a combined multivariable model integrating BMI, Bishop score, and parity for pre-induction risk stratification. Methods: This single-center retrospective cohort study included 501 singleton term pregnancies undergoing labor induction in Ankara, Turkey (March–August 2023), stratified by five WHO BMI categories. The primary outcome was cesarean delivery. Analyses included Cochran–Armitage trend tests, multivariable logistic regression, and ROC analysis with DeLong comparison and bootstrap optimism-corrected internal validation. The study followed STROBE guidelines. Results: The overall cesarean rate was 30.3%, rising from 17.1% (normal weight) to 52.2% (Obese class III; Cochran–Armitage Z = 6.099, p < 0.001), with consistent trends across Robson Groups 2 and 4 (both p < 0.001). Failed induction increased from 2.9% to 15.2% (p < 0.001). The cervical ripening requirement rose from 32.4% to 60.9% (p = 0.003). BMI (adjusted odds ratio [aOR] = 1.489, 95% confidence interval [CI] 1.254–1.767) and Bishop score (aOR = 0.807, 95% CI 0.718–0.906) were independent predictors; nulliparity showed a non-significant trend (p = 0.090). The combined model (BMI + Bishop + nulliparity) achieved an apparent AUC of 0.715 (optimism-corrected 0.709), outperforming both Bishop score alone (DeLong p = 0.010) and BMI alone (p = 0.012). Calibration was adequate (Hosmer–Lemeshow p = 0.632). Conclusions: Higher BMI independently predicts increased cesarean risk following labor induction across parity subgroups. The combined multivariable model provides practical bedside pre-induction risk stratification superior to either individual predictor, though external validation is warranted before widespread clinical implementation.