Development and validation of a spontaneous preterm birth risk prediction algorithm based on maternal bioinformatics:...

Area of Science:

• Obstetrics and Gynecology
• Data Science in Healthcare
• Neonatal Health

Background:

• Spontaneous preterm birth (sPTB) is a leading cause of adverse neonatal outcomes.
• Identifying predictive factors for sPTB is crucial for improving maternal and infant health.
• Big data analytics offers a novel approach to understanding and predicting sPTB risk.

Purpose of the Study:

• To analyze factors influencing spontaneous preterm birth (sPTB) in pregnant women.
• To develop and validate a predictive model for sPTB risk using clinical and laboratory data.
• To identify the optimal machine learning algorithm for sPTB prediction.

Main Methods:

• Retrospective analysis of clinical data from 3,082 pregnant women.
• Comparison of five machine learning models using AUC, accuracy, sensitivity, specificity, and precision.
• Selection of top 10 predictive variables and validation using training, validation, and external datasets.

Main Results:

• XGBoost algorithm achieved the highest performance with an AUC of 0.89 (95% CI: 0.88-0.90).
• Top 10 predictive indicators identified: ALP, AFP, ALB, HCT, TC, DBP, ALT, PLT, height, and SBP.
• The model demonstrated robust performance across training (AUC 0.93), validation (AUC 0.87), and external (AUC 0.79) sets.

Conclusions:

• ALP, AFP, ALB, HCT, TC, DBP, ALT, PLT, height, and SBP are significant factors influencing sPTB.
• The XGBoost model, utilizing these factors, provides a high-performing tool for sPTB risk prediction.
• This data-driven approach has the potential to enhance clinical decision-making and improve neonatal outcomes.