JMIR Publications

ABSTRACT

Background:

Gestational diabetes mellitus (GDM) affects over 5% of pregnancies globally, elevating risks of type 2 diabetes postpartum and complications such as fetal death, miscarriage, and congenital abnormalities. Effective GDM management is essential to balance glycemic control and pregnancy outcomes.

Objective:

To develop interpretable machine learning models using GDM datasets for predicting adverse pregnancy outcomes and identifying key factors through the SHAP algorithm, thus supporting improved maternal and infant health.

Methods:

Data preprocessing and feature selection were performed, with ADASYN used to address class imbalance. Classification models, including logistic regression, random forest, SVM, and XGBoost, were built and enhanced through stacking method. Model interpretability was assessed with SHAP to quantify feature contributions.

Results:

Among 1,670 patients, 200 experienced adverse outcomes. The stacked model achieved 97.9% accuracy and an AUC of 0.96 in external validation. SHAP analysis highlighted key predictive factors, such as gestational age, fasting glucose, and blood pressure, supporting model reliability.

Conclusions:

This study underscores the potential of machine learning in predicting adverse outcomes in GDM, with interpretable features offering valuable clinical insights to enhance pregnancy management and maternal-infant health.