Derivation and Validation of Prediction of Preterm Preeclampsia Using Machine Learning Algorithms

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Preterm preeclampsia is a leading cause of maternal and neonatal morbidity.
• Early prediction of preterm preeclampsia is crucial for timely intervention.
• Current prediction models often lack accuracy or are not widely adopted.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting preterm preeclampsia using information available before 23 weeks gestation.
• To identify key predictive features for preterm preeclampsia.
• To create an accessible tool for clinical use.

Main Methods:

• Secondary analysis of the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-Be (nuMoM2b) cohort.
• Utilized 131 features including demographics, medical history, and early ultrasound data.
• Developed and compared multiple ML models (glmnet, multilayer perceptron, random forest, XGBoost, LightGBM), selecting XGBoost for final model development.

Main Results:

• The XGBoost model achieved the highest Area Under the Curve (AUC) of 0.749.
• A final model with eight key features (including uterine artery pulsatility index, chronic hypertension, diabetes, blood pressure, BMI, and maternal age) demonstrated strong predictive performance (AUC 0.741 in training, 0.779 in validation).
• An online application was developed for the final prediction model.

Conclusions:

• Machine learning algorithms utilizing information available before 23 weeks gestation can accurately predict preterm preeclampsia.
• The developed model, incorporating early ultrasound and clinical data, offers a promising tool for early risk identification.
• The online application facilitates the integration of this predictive model into clinical practice.