Risk Stratification for Postpartum Hemorrhage in Vaginal Delivery: A Nationwide Cohort Study of Predelivery and...

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Public Health

Background:

• Postpartum hemorrhage (PPH) is a major cause of maternal morbidity and mortality.
• Accurate risk stratification is crucial for timely intervention during vaginal delivery.
• Existing prediction models often lack real-time intrapartum data integration.

Purpose of the Study:

• To develop and validate machine learning (ML) models for PPH risk stratification in vaginal delivery.
• To evaluate the impact of incorporating intrapartum information on PPH detection rates.
• To assess if enhanced detection maintains a clinically relevant screening burden.

Main Methods:

• Nationwide retrospective cohort study of 1,224,990 vaginal deliveries in Japan (2014-2023).
• Development of ML models using predelivery clinical variables and an updated model with intrapartum data.
• Models were validated using a temporal external dataset, targeting an alert rate of ~300 high-risk classifications per 1000 deliveries.

Main Results:

• PPH occurred in 7.5% of deliveries (9.7% in validation cohort).
• The intrapartum ML model improved PPH detection (sensitivity 0.596) compared to the predelivery model (sensitivity 0.537).
• The intrapartum model identified 5-6 additional PPH cases per 1000 deliveries with a modest increase in screening burden and improved discrimination (0.676 to 0.703).

Conclusions:

• Machine learning models utilizing routinely available clinical data can aid PPH screening and preparedness.
• Integrating intrapartum information into ML models modestly enhances PPH detection without significantly increasing the screening burden.
• These findings support the clinical utility of ML-based PPH risk stratification during vaginal delivery.