Development of an Interpretable Machine Learning Model to Predict Pregnancy Outcomes Following Cervical Cerclage

Area of Science:

• Obstetrics and Gynecology
• Reproductive Medicine
• Medical Informatics

Background:

• Cervical insufficiency is a leading cause of spontaneous preterm birth.
• McDonald cerclage is used to manage cervical insufficiency but has frequent adverse events.
• Predicting post-cerclage outcomes is crucial for personalized care.

Purpose of the Study:

• To evaluate the efficacy of machine learning (ML) models in predicting outcomes after McDonald cerclage.
• To identify key predictors of spontaneous preterm birth following cerclage.
• To develop a clinically useful tool for risk stratification in patients undergoing cerclage.

Main Methods:

• Retrospective analysis of 462 pregnant women undergoing McDonald cerclage.
• Development and comparison of multiple ML models: logistic regression, random forest (RF), support vector machines (SVM), decision trees (DT), and extreme gradient boosting (XGBoost).
• Evaluation of model performance using discrimination, calibration, and clinical utility metrics, with SHAP analysis for predictor interpretation.

Main Results:

• The random forest (RF) model demonstrated balanced performance, accuracy, interpretability, and reliability.
• Elevated C-reactive protein, increased white blood cell count, and amniotic fluid sludge were identified as strong predictors of adverse outcomes.
• Conception method, maternal weight, and cerclage subtype also influenced risk.

Conclusions:

• The RF model provides a clinically useful framework for predicting preterm birth risk after cerclage.
• Inflammatory markers, maternal characteristics, and cerclage indication are key determinants of outcomes.
• An online prediction tool was developed; further multicenter prospective validation is warranted.