Predicting risk of preterm birth in singleton pregnancies using machine learning algorithms.

Area of Science:

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

Background:

• Preterm birth (<37 weeks' gestation) is a leading cause of neonatal morbidity and mortality.
• Accurate prediction of preterm birth remains a clinical challenge, necessitating advanced analytical approaches.

Purpose of the Study:

• To develop, train, and validate machine learning models for predicting preterm birth in singleton pregnancies.
• To identify key predictive features for preterm birth using various machine learning algorithms.

Main Methods:

• Utilized data from 22,603 singleton pregnancies in a prospective cohort study in China (2014-2016).
• Applied algorithms including Catboost, Random Forest, DNN, SVM, and logistic regression for feature selection and prediction.
• Employed 5-fold cross-validation for internal model validation and assessed performance using Area Under the Receiver Operating Curve (AUC).

Main Results:

• The CatBoost model, applied after 26 weeks' gestation, demonstrated the best performance with an AUC of 0.70.
• Key predictors identified include number of antenatal visits, aspartate aminotransferase levels, symphysis fundal height, maternal weight, abdominal circumference, and blood pressure.
• The best model achieved an accuracy of 0.81, sensitivity of 0.47, and specificity of 0.83.

Conclusions:

• Machine learning application on pregnancy surveillance data shows promise for preterm birth prediction.
• Several modifiable antenatal predictors were identified, suggesting potential targets for early intervention strategies.
• The study highlights the utility of advanced computational methods in improving obstetric outcomes.