Development and temporal validation of a nationwide prediction model for cesarean delivery after induction or augmentation of labor in Japan

OBJECTIVE: To develop and temporally validate a nationwide prediction model for cesarean delivery following induction or augmentation of labor in Japan. METHODS: We conducted a retrospective cohort study using the Japan Society of Obstetrics and Gynecology Perinatal Database. Women with singleton pregnancies who underwent induction or augmentation of labor at ≥37 weeks of gestation between 2020 and 2022 were included. Cases from 2020 to 2021 (n = 113 572) formed the development cohort, and cases from 2022 (n = 59 045) served as the temporal validation cohort. Predictors were selected based on clinical relevance. Variable selection used least absolute shrinkage and selection operator logistic regression with the one standard error rule, followed by multivariable logistic regression. Model performance was assessed with the area under the receiver operating characteristic curve (AUROC) and calibration plots. RESULTS: A total of 172 617 women were included in the study. Thirteen predictors were selected: maternal age, height, pre-pregnancy body mass index (BMI), gestational BMI gain, multiparity, gestational age, assisted reproductive technology pregnancy, mechanical cervical ripening, pregestational diabetes mellitus, hypertensive disorders of pregnancy, epidural analgesia, birthweight, and neonatal sex. Discrimination was good in the development cohort (AUROC 0.757, 95% confidence interval [CI] 0.754-0.761) and temporal validation cohort (AUROC 0.767, 95% CI 0.762-0.772). Multiparity and epidural analgesia were associated with lower risk, whereas all other predictors increased cesarean risk (all P < 0.001). CONCLUSION: This nationwide prediction model demonstrated robust performance and might support individualized counseling, risk assessment, and perinatal care planning.