Artificial Intelligence Applications in Obstetric Risk Prediction: A Systematic Review of Machine Learning Models for...

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Biostatistics

Background:

• Preeclampsia is a significant global health concern, contributing to maternal and perinatal mortality.
• Traditional prediction models for preeclampsia have demonstrated limited accuracy.
• Machine learning (ML) offers advanced capabilities for analyzing complex, non-linear data in multidimensional datasets for improved prediction.

Purpose of the Study:

• To systematically review and evaluate the performance, methodological quality, and clinical applicability of ML models for preeclampsia prediction.
• To identify key predictors and assess the risk of bias in ML models for preeclampsia.
• To synthesize evidence on the discriminative ability of ML approaches in preeclampsia prediction.

Main Methods:

• A systematic review adhering to PRISMA 2020 guidelines was conducted across five major databases.
• Studies developing or validating ML models for preeclampsia prediction were included, with data extracted up to April 15, 2025.
• The Prediction Model Risk-of-Bias Assessment Tool (PROBAST) was used to assess the risk of bias.

Main Results:

• Eleven studies involving 116,253 pregnancies were included.
• Ensemble ML methods (XGBoost, Random Forest) achieved high performance (AUCs 0.84–0.973).
• Key predictors identified were mean arterial pressure, prior preeclampsia history, placental growth factor (PlGF), and pregnancy-associated plasma protein A (PAPP-A).

Conclusions:

• ML models, especially ensemble methods, demonstrate strong discriminative power for preeclampsia prediction.
• Heterogeneity in predictors and limited external validation hinder widespread clinical adoption.
• Future research should focus on prospective validation with standardized protocols for clinical translation.