A recent study conducted by researchers at Brigham and Women’s Hospital shows that a large language model can accurately identify postpartum hemorrhage, a leading cause of maternal mortality and morbidity worldwide. The investigators used the Flan-T5 large language model to extract medical concepts contained in electronic health records (EHRs) to better understand which patient populations are more likely to experience postpartum hemorrhage.
The team’s findings, reported in the journal npj Digital Medicine, found that use of the model was 95% accurate in identifying patients with postpartum hemorrhage and identified 47% more patients than current methods of tracking patients through billing codes in the EHR.
“We need better ways to identify the patients that have this complication, as well as the different clinical factors associated with it,” said corresponding author Vesela Kovacheva, MD, of the department of anesthesiology, Perioperative and Pain Medicine. “There are so many amazing large language models being developed right now, and this approach could be used with other conditions and diseases.”
The researchers noted that this medical condition is both understudied and not universally defined or well represented in current health records. But the hope is that the development and emergence of more powerful artificial intelligence (AI) tools can help break new ground in this area to better identify patients with postpartum hemorrhage and improve the continuum of care.
AI, which Mass General Brigham is increasingly using for research to inform clinical care noted that that AI-driven Flan-T5 model was used in this research since postpartum hemorrhage includes a broad spectrum of patients and it’s range of symptoms are caused by a spectrum of different symptoms.
To help the model’s performance, the Brigham team prompted Flan-T5 with lists of concepts known to be associated with the condition. It then had the model extract these concepts associated with postpartum hemorrhage from 131,284 patient records of women who had given birth at Mass General Brigham between the years of 1998 to 2015. Using Flan-T5 in this way rapidly produced accurate results without the need for human intervention of manually labeling.
“We looked at all of the patients that Flan-T5 identified as having postpartum hemorrhage and looked at what fraction of those also had the corresponding billing code,” noted first author Emily Alsentzer, PhD, a research fellow at Brigham and Women’s. “It turns out that Flan-T5 was 95 percent accurate and allowed us to identify 47 percent more patients than we would have from the billing codes alone. Ideally, we would like to be able to predict who will develop postpartum’ hemorrhage before they do so, and this is a tool that can help us get there.”
Based on their success in determining postpartum hemorrhage, the investigators will now look to leverage this same approach to examine and identify other pregnancy complications to address what they term a “growing maternal healthcare crisis in the United States.”
Kovacheva noted that “this approach can be applied to many future studies and it could be used to help guide real-time medical decision making, which is very exciting and valuable to me as a clinician.”
