PCORI Awards $1 Million to UB Researcher to Use AI in Analyzing Social Determinants of Health

The potential impact of this work affects the management of multiple medical conditions, including opioid use disorder. The work also aims to develop novel methods for the analysis of social determinants of health.

The World Health Organization defines social determinants of health as the “non-medical factors that influence health outcomes.” These include income, neighborhood conditions, food insecurity and social supports.

Marianthi Markatou, PhD, SUNY Distinguished Professor in the Department of Biostatistics in UB’s School of Public Health and Health Professions and associate chair of research and health care informatics, is the principal investigator on the project. Andrew H. Talal, MD, MPH, professor of medicine in the Jacobs School of Medicine and Biomedical Sciences, is a co-investigator. 

The project will leverage large language models (LLMs), an extremely large type of artificial intelligence that is pre-trained on vast amounts of data, to extract diverse data — such as health care provider notes, discharge summaries, patient responses, blood pressure, weight and height — collected from opioid treatment programs.

“This research is an example of use-inspired research. That is, research that combines the development of scientific knowledge with the goal of creating solutions to solve societal and scientific challenges,” Markatou explains. “A fundamental component of such research is the improvement and creative use of technology. In our context, large language models fulfill that role.”

The team will also develop analytic methods for identifying groups that are characterized by the same conditions.

“One of the objectives of our work is to reveal patterns in the data to identify novel social determinants of health, filling a methodological gap to permit analyses of mixed-type datasets and provide new methods to analyze clinical data,” says Markatou, adding that the team will develop and disseminate software solutions to implement the new methods.

Their goal is to inform and facilitate patient-centered care by evaluating several parameters of social determinants of health. “These findings have the potential to be incorporated into new patient-centered management approaches,” Markatou says.

The project team will incorporate feedback from patients and partners into their methodology via a human-in-the-loop approach, so that their LMM is continually refined to ensure that information is extracted and analyzed accurately, according to a description of the project on the PCORI website.

“We will also engage our stakeholders to understand how clustering techniques can be used to identify patient subgroups that might benefit from an intervention and to identify social determinants of health to target for intervention. Our long-term objective is to provide evidence to support interventions to address social determinants of health among underserved populations,” according to Markatou.

Project collaborators represent an interdisciplinary team of researchers with expertise in a variety of areas, including statistical sciences, liver disease, computer science, linguistics, addiction medicine, opioid treatment program administration and data management.

The team will also be collaborating with the New York State Office of Addiction Services and Supports.

In addition to acknowledging the support of the research team, Markatou also thanked the Jacobs School, especially the Office of the Vice President for Health Sciences, for their support.