J Matern Fetal Neonatal Med. 2026 Dec;39(1):2724778. doi: 10.1080/14767058.2026.2724778. Epub 2026 Aug 31.
ABSTRACT
BACKGROUND: Operative vaginal delivery (OVD) is an important alternative to cesarean section, but the relative outcomes of vacuum, forceps, and sequential vacuum + forceps approaches remain uncertain. Moreover, predictors of instrument choice are incompletely understood. This study aimed to compare maternal, intrapartum, and neonatal outcomes across OVD modalities and to integrate conventional statistics with machine learning to identify key predictive factors.
METHODS: We retrospectively analyzed 967 singleton deliveries at a tertiary center: vacuum (n = 208), forceps (n = 706), and sequential vacuum + forceps (n = 53). Maternal characteristics, intrapartum variables, and neonatal outcomes were compared using ANOVA and chi-square tests. A Random Forest classifier was trained to rank predictors of delivery mode, with performance assessed by accuracy, AUC, and Cohen’s κ.
RESULTS: Forceps deliveries were more frequent among younger, primiparous women, while vacuum was used in older, higher-parity mothers with better base excess (BE). Vacuum + forceps deliveries had the longest decision-to-delivery interval (26.9 vs. 17.7 and 16.8 min, p < 0.001), highest rates of severe perineal laceration (73.6% vs. 34.1% and 47.5%, p < 0.001), and poorest neonatal outcomes, including lower Apgar scores and higher cephalohematoma, hyperbilirubinemia, and transfer rates. Random Forest achieved 73.7% accuracy (AUC 0.758) and identified second-stage duration, decision-to-delivery time, lactate, and BE as top predictors, some not significant in univariate tests.
CONCLUSIONS: Vacuum and forceps each have distinct maternal and neonatal risk profiles, while sequential vacuum + forceps delivery confers the greatest morbidity, largely reflecting case complexity. Traditional statistics and machine learning offer complementary insights, highlighting both established and novel predictors. Integrating these approaches may support individualized decision-making and improve OVD outcomes.
PMID:42675011 | DOI:10.1080/14767058.2026.2724778
