# Machine Learning Prediction of Intrapartum Cesarean Delivery in Women with Obesity

## TL;DR

This study uses machine learning to predict the likelihood of cesarean delivery during labor for women with obesity, offering a more accurate tool than previous methods.

## Contribution

The study introduces a machine learning model (XGBoost) that outperforms a regression-based score in predicting intrapartum cesarean delivery in obese women.

## Key findings

- The XGBoost model achieved an AUC of 0.945, showing strong predictive accuracy.
- Key predictors included cervical dilatation, BMI, age, and inflammatory markers.
- A pre-labor model using admission data retained strong performance across BMI categories.

## Abstract

Objective: To identify risk factors for intrapartum cesarean delivery (CD) among women with obesity (BMI ≥ 30) and to evaluate whether a machine learning model (XGBoost) can improve prediction of this outcome compared with a previously developed regression-based risk score. Methods: A retrospective cohort study at a single university-affiliated tertiary medical center was conducted. All women with a pre-pregnancy BMI ≥ 30 who initiated a trial of labor between 2012 and 2024 were included. Women who underwent elective CD or had missing outcome data were excluded. Maternal, obstetric, and intrapartum characteristics were compared between women who delivered vaginally and those who required an intrapartum CD. Predictors were evaluated using extreme gradient boosting (XGBoost), and model performance was assessed using receiver operating characteristic (ROC) analysis and SHAP-based interpretability. Results: Among 146,999 women who delivered during the study period, 10,248 (7.0%) had a pre-pregnancy BMI ≥ 30. A total of 7236 obese women attempted a trial of labor, of whom 1031 (14.5%) underwent an intrapartum CD. Key predictors included limited cervical dilatation at admission, epidural anesthesia, nulliparity, maternal BMI and age, oxytocin use, birthweight, inflammatory markers (white blood count and neutrophils to lymphocytes ratio), and previous cesarean scar. The XGBoost model demonstrated excellent discriminatory ability with an AUC of 0.945 (95%CI 0.930–0.960, DeLong), and exceeded the performance of our previous regression-based score, and provided detailed insight into nonlinear effects through SHAP analysis. In a secondary analysis restricted to variables available at admission, a pre-labor model retained a strong discriminatory performance across BMI categories, supporting its applicability for early risk stratification prior to labor onset. Conclusions: A machine learning-based model accurately predicts intrapartum cesarean delivery in women with obesity and may serve as a valuable tool to support individualized counseling and delivery planning.