Artificial intelligence in prediction of postpartum hemorrhage: a primer and review

Abstract

Postpartum hemorrhage (PPH) is a leading cause of maternal mortality worldwide, and the ability to predict PPH may help address preventable causes of morbidity and mortality such as delays in care. Understanding the importance of standardized approaches to PPH, the National Partnership for Maternal Safety Consensus Bundle on Obstetric Hemorrhage outlines four critical domains for safe and effective PPH care: 1) Readiness; 2) Recognition and Prevention; 3) Response; and 4) Reporting and System Learning. The Recognition and Prevention domain includes recommendations for standardized methods of PPH risk prediction, and The Joint Commission now requires use of an evidence-based PPH prediction tool. Postpartum hemorrhage risk predictions can be accomplished via checklist tools completed manually by healthcare providers or via machine-assisted calculations in the form of logistic regression or machine learning populated by automated electronic health record data. The latter examples of machine-assisted calculations of PPH risk are a form of artificial intelligence.

The purpose of this review is to describe the current state of AI-based PPH risk assessment, including the application of logistic regression and machine learning. A primer on interpretation of such models is provided, along with identification of research gaps and future directions.