Large language model shows promise in helping clinicians identify postpartum hemorrhage

Postpartum hemorrhage is the leading cause of maternal mortality and morbidity worldwide and a common complication of pregnancy. This serious illness is little studied and is not universally defined or well represented in health records. A new study led by researchers at Brigham and Women’s Hospital used the large Flan-T5 language model to extract medical concepts from electronic health records to better define and identify populations affected by postpartum hemorrhage.

The study found that the model was 95% accurate in identifying patients with the disease and was able to identify 47% more patients than using the standard method of tracking the disease via codes billing. The tool has shown great promise in helping clinicians identify subpopulations who are at higher risk for postpartum hemorrhage and predict those who are more likely to develop it.

The results are published in npj Digital Medicine.

“We need better ways to identify patients with 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 Medicine at pain. “There are so many great, amazing language models being developed right now, and this approach could be used with other conditions and diseases.”

The emergence of artificial intelligence tools in healthcare has been revolutionary and has the potential to positively reshape the continuum of care.

Because conditions such as postpartum hemorrhage include a wide range of patients, symptoms and causes, the research team used the Flan-T5 model to analyze comprehensive information from electronic health records to to help them better categorize patient subpopulations.

They asked the Flan-T5 model for lists of concepts known to be associated with postpartum hemorrhage, then asked it to extract them from the discharge summaries of a cohort of 131,284 patients who delivered at Mass General hospitals. Brigham between 1998 and 2015. This method obtained rapid and accurate results without the need for manual labeling.

“We looked at all the patients identified by Flan-T5 as having postpartum hemorrhage and what fraction of them also had the corresponding billing code. It turns out that Flan-T5 was 95% accurate and we “Ideally, we would like to be able to predict who will develop postpartum hemorrhage before it does, and this is a tool that can help us get there.”

Next, the team plans to continue using this approach to examine other pregnancy complications and hopes their work will help address the growing maternal health crises in the United States.

“This approach can be applied to many future studies,” Kovacheva said. “And this could be used to guide medical decision-making in real time, which is very exciting and valuable to me as a clinician.”