Clinical Obstetrics
Poster Session 4
Cecilia B. Leggett, BA, MD
Fellow
Stanford University
Palo Alto, CA, United States
Elaine Albertson, PhD
Cedars-Sinai Medical Center
Los Angeles, CA, United States
Benison Pang, PhD
Cedars-Sinai Medical Center
Los Angeles, CA, United States
Matthew Wells, BS, MS
Cedars-Sinai Medical Center
Los Angeles, CA, United States
Melissa S. Wong, MD, MHDS (she/her/hers)
Assistant Professor, Maternal-Fetal Medicine
Cedars-Sinai Medical Center
Los Angeles, CA, United States
An automated machine learning (AML) platform was used to develop a model to predict severe morbidity (defined as occurrence of unscheduled hysterectomy, uterine artery embolization, massive transfusion protocol, ICU admission, takeback to the operating room, or in-hospital death) using continuously updated data based on electronic health record for all delivering patients at a large academic hospital from 01/01/2013 through 12/31/2022. The primary outcome is ROC-AUC in predicting severe morbidity from PPH. A sequential approach was used to develop admission, intrapartum, and postpartum models using continuously updated data throughout a patient’s hospital course. Time-series engineering was applied to transform individual timelines into statistics to describe trends during intrapartum and postpartum periods.
Results: 12,807 patients were included in the study and used in the training algorithm with 386 events of severe morbidity from PPH identified. The features of highest importance (with the highest permutation index) for each model are described in Figure 1. Each model demonstrated continuous improvement through the postpartum period which showed the highest discriminatory ability (ROC AUC 0.88).
Conclusion: AI incorporating time-series engineering has improved AUC, precision, and sensitivity when compared to relying on patient risk factors alone for prediction of severe morbidity from PPH. AI-based predictions could identify patients at risk of severe morbidity and provides opportunity for proactive intervention to reduce morbidity associated with PPH.