A new $1 million award to the University of Florida will enable advanced artificial intelligence research that could repurpose electronic health records and create content to improve healthcare services.
The wealth in health records is hidden by too much information scattered across too many healthcare systems. Previous generations of AI have a limited ability to learn from one healthcare system and use it to solve similar problems from other healthcare systems due to variations in electronic health records. Hospitals and health systems cannot trust this technology.
To build a solution, an award from the Patient-Centered Outcomes Research Institute (PCORI) was given to a team led by Yonghui Wu, Ph.D., an Associate Professor in the UF College of Medicine’s Department of Health Outcomes and Biomedical Informatics (HOBI).
Starting in November 2024, this three-year project intends to reduce costs and improve how quickly and fairly generative AI technology can learn to interpret electronic health records. The researchers must give computers the ability to learn from one healthcare system and use this knowledge to solve other complicated tasks at different healthcare systems — a process known as transfer learning.
For example, by learning to recognize disease symptoms from millions of patients in Florida, computers could accurately tell if a patient receiving routine care in New York has dementia.
Models Building on Models
This research builds upon previous work by Wu to create GatorTron, the most widely used clinical large language model, and GatorTronGPT, a computer assistant that can generate patient reports at a level equivalent to human providers. The project’s data comes mainly from two highly secure networks developed with PCORI support: INSIGHT, and the OneFlorida+ Clinical Research Network.
The research team will collaborate with patients, nurses, and physicians to overcome potential biases of generative AI and seek solutions for mitigation. The team includes HOBI researchers Jiang Bian, Ph.D., Yi Guo, Ph.D., and Elizabeth Shenkman Ph.D. Additional expertise comes from Glenn E. Smith, Ph.D., Professor in the UF Department of Clinical and Health Psychology, and Yifan Peng, Ph.D., Assistant Professor in the Department of Population Health Sciences and Department of Radiology at Weill Cornell Medicine.
Comparing Clinical Research
This study is among the latest methodology studies that PCORI has funded to address gaps in comparative clinical effectiveness research methods. These studies provide results that guide researchers in planning future studies and improve the strength and quality of their evidence.
“This study was selected for its potential to address a high-priority methodological gap in patient-centered comparative clinical effectiveness research,” said PCORI Executive Director Nakela L. Cook, M.D., M.P.H. “Improving methods for conducting comparative clinical effectiveness research helps ensure this research generates sound, trustworthy evidence to help patients and those who care for them become more empowered decision makers. We look forward to following the study’s progress and working with the University of Florida to share the results.”
Wu’s award has been approved pending completion of a business and programmatic review by PCORI staff and issuance of a formal award contract. To follow this project, sign up for updates by visiting the PCORI project page: Transfer Learning Natural Language Processing to Improve Adoption of Clinical Text in Multisite Studies.
PCORI is an independent, nonprofit organization authorized by Congress with a mission to fund patient-centered comparative clinical effectiveness research that provides patients, their caregivers and clinicians with the evidence-based information they need to make better-informed health and health care decisions.