Somerville, Mass.-based Mass General Brigham has spun out an AI startup it says represents a “breakthrough” for clinical trial patient identification.
The academic health system said Dec. 12 it launched AIwithCare, which deploys AI in the EHR to match patients with clinical trials. An estimated third of the cost of clinical trials is tied to screening.
“To be able to do this in an automated way using a large language model is really breakthrough technology,” Mass General Brigham Chief Innovation Officer Chris Coburn told Becker’s. “No one wants clinical trials to take as long as they take or be as expensive as they are, but the fact is it’s painstaking to find people who meet the eligibility criteria.”
While the health system spins off about 30 companies a year, this is the first to employ retrieval-augmented generation AI, which looks up relevant information from trusted data sources and then uses that information to generate accurate, contextual answers. This differs from typical large language models that rely on training alone. In this case, the RAG AI scans EHR data and clinical trial protocols to find patient matches.
Researchers at Mass General Brigham’s Accelerator for Clinical Transformation developed the RECTIFIER tool, nearly doubling the number of heart failure trial participants in a February JAMA study. (RECTIFIER stands for RAG-Enabled Clinical Trial Infrastructure for Inclusion Exclusion Review.)
RECTIFIER has since expanded across Mass General Brigham to cardiology, gastroenterology, neurology, oncology, pathology and psychiatry. Pediatric gastroenterologists, for example, used the tool for patient referral triage with 94.7% accuracy and identified critical labs and symptoms buried in clinical notes with 98% accuracy.
The technology is now being commercialized as AIwithCare, led by CEO and co-founder Alexander “A.J.” Blood, MD, associate director of the Accelerator for Clinical Transformation and a Mass General Brigham cardiologist, and President and Chief AI Officer Sandy Aronson, a digital research principal architect at the health system.
“This is exactly what you think about when you think about innovation — the practicing, resourceful clinician, with the technologist who can actually build the platform,” Mr. Coburn said.
It also illustrates how large language model technology is quickly moving beyond AI scribes, which has been the popular use case in healthcare thus far.
“Generative AI is taking root, real solutions are being created, and they’re starting to hit the market — and they’re going to grow fast,” Mr. Coburn said. “We expect many more combinations of our clinical leaders addressing long-held unmet needs in their field and working in a very focused way to create solutions for them.”
