Prediction of Maternal Hemorrhage Using Machine Learning: Retrospective Cohort Study
Jill M Westcott1, Francine Hughes2, Wenke Liu3,4
1Division of Maternal-Fetal Medicine, Department of Obstetrics and Gynecology, New York University Langone Health, New York, NY, United States.
Journal of Medical Internet Research
|July 18, 2022
Summary
Machine learning accurately predicts postpartum hemorrhage risk using electronic health data. Models available before delivery show high performance, aiding early intervention for maternal safety.
Area of Science:
- Medical informatics
- Machine learning in healthcare
- Obstetrics and Gynecology
Background:
- Postpartum hemorrhage (PPH) is a leading cause of maternal mortality and morbidity.
- Accurate risk identification is crucial for timely intervention and improved patient outcomes.
Purpose of the Study:
- To develop and evaluate machine learning models for identifying patients at risk of PPH.
- To assess model performance using data available at different stages of labor and delivery.
Main Methods:
- Analysis of electronic medical records from 30,867 deliveries.
- Utilized supervised learning (regression, tree-based, kernel-based methods) with 497 variables.
- Models were trained, validated, and tested, with performance assessed by accuracy and AUROC.
Main Results:
- Gradient boosted decision trees demonstrated superior discrimination (AUROC 0.979 for the overall model).
- Models using data available prior to the second stage of labor achieved high accuracy (98.0%) and sensitivity (0.737).
- Models stratified by delivery mode showed good discrimination but lacked sufficient sensitivity for clinical use.
Conclusions:
- Machine learning effectively identifies women at risk for PPH, enabling personalized preventative care.
- Early-stage data models show significant clinical utility, supporting integration into practice.
- Further research is needed to refine models for specific delivery modes and validate findings.
