Generative artificial intelligence can identify social determinants of health within EHR notes, accurately and without bias, determining which patients might need extra support, according to Somerville, Mass.-based Mass General Brigham.
The health system's researchers fine-tuned large language models to automatically extract information about social determinants — factors such as housing, employment and family status that can affect a person's medical care — from clinician notes. The study was published Jan. 11 in npj Digital Medicine.
"Algorithms that can pass major medical exams have received a lot of attention, but this is not what doctors need in the clinic to help take better care of patients each day," said corresponding author Danielle Bitterman, MD, a faculty member at Mass General Brigham's AI in Medicine program, in a Jan. 11 news release. "Algorithms that can notice things that doctors may miss in the ever-increasing volume of medical records will be more clinically relevant and therefore more powerful for improving health."
Studying hundreds of clinician notes, Mass General Brigham's models caught 93.8 percent of patients with adverse social determinants of health, compared to 2 percent for official diagnostic codes. The study's authors say their algorithms were also less prone to bias than generalist large language models such as ChatGPT, but more research is needed to discover the origins of algorithmic bias.