Machine Learning Models for the Prediction of Preterm Birth at Mid-Gestation Using Individual Characteristics and...

Area of Science:

• Obstetrics and Gynecology
• Perinatal Medicine
• Computational Biology

Background:

• Preterm birth (PTB) before 37 weeks is a leading cause of neonatal mortality.
• PTB is classified into spontaneous and medically indicated (iatrogenic) subtypes with different causes.
• Accurate prediction models differentiating PTB subtypes are lacking.

Purpose of the Study:

• To develop and validate predictive models for spontaneous and iatrogenic PTB.
• To predict PTB at <32, <34, and <37 weeks' gestation.
• To utilize medical history and second-trimester data for prediction.

Main Methods:

• Retrospective cohort study of 9805 singleton pregnancies (2012-2025).
• Predictors included maternal characteristics, obstetric history, and second-trimester ultrasound markers.
• Four algorithms (Logistic Regression, Random Forest, XGBoost, Neural Network) were evaluated using Area Under the Curve (AUC).

Main Results:

• Models performed significantly better for iatrogenic PTB (AUC up to 0.862 for <32 weeks) than spontaneous PTB (AUC up to 0.749 for <32 weeks).
• Iatrogenic PTB prediction was driven by placental dysfunction markers (e.g., estimated fetal weight, uterine artery pulsatility index).
• Spontaneous PTB prediction was associated with a history of PTB and short cervical length.

Conclusions:

• Mid-gestation data models effectively predict iatrogenic PTB, with improved accuracy for earlier gestations.
• Predictive performance for spontaneous PTB was modest.
• Logistic Regression performed comparably to machine learning algorithms, emphasizing subtype-specific modeling.