Using Generative AI to Highlight Social Determinants of Health in Doctors’ Notes

Large language models (LLMs) can effectively extract information on social determinants of health (SDoH) from clinicians' notes, improving patient care and addressing health disparities, according to a study by Mass General Brigham. The researchers found that LLMs could identify adverse SDoH in patients at a rate of 93.8%, compared to just 2% inclusion in official diagnostic codes. These models were also found to be less biased than generalist models. Further research is needed to understand and address algorithmic bias.