Earlier this month, IMO Health experts traveled to San Francisco to attend the AMIA 2024 Annual Symposium.
AMIA, or the American Medical Informatics Association, is a community of professionals and students who are committed to transforming care through the practice of informatics. The organization perceives informatics as the key to accelerating current healthcare reform initiatives and the “bridge for knowledge and collaboration across a continuum, from basic and applied research to the consumer and public health arenas.”
Informaticians play a critical role in helping to promote intelligent care decisions through the collection, analysis, and application of data. In fact, quality healthcare data fuels the entire practice of informatics, enabling vital advancements that directly benefit individuals and populations.
AMIA’s Annual Symposium is an event dedicated to the research and practice of biomedical and health informatics. It draws thousands of experts from all corners of the world to share insights for improving human health by leveraging health information and advanced technologies. This year’s theme focused on “Informatics in the Age of Generative Artificial Intelligence” —a critical and timely issue with far-reaching effects, both positive and negative.
Clinical AI and the acceleration of drug development in life science
IMO Health’s first AMIA event of the year was a panel discussion moderated by Xiaoyan Wang, PhD, Chief Scientist and Sr. Vice President of Life Science Solutions, featuring speakers from Pfizer, Epic, Regeneron Pharmaceuticals, and Merck. The session—Integrating Generative AI into the Life Sciences Industry to Move the Needle from Drug Development to Regulatory Decisions: A Realistic Prospect or a Pipe Dream? —provided a detailed overview of the state of artificial intelligence (AI) in life sciences today from multiple perspectives, including health system, biotechnology, and phamaceutical industries.
Speakers explored the potential of industry-specific clinical AI, including large language models (LLMs) and generative AI like GPT-4, to interpret vast datasets at unprecedented speeds, effectively accelerating drug development. They also discussed the challenges of integrating and implementing AI technologies in the final stages of deployment and examined the varying stances of regulatory agencies on the use of AI in healthcare.
LLMs and the extraction of SDOH data from medical problem lists
At IMO Health’s second AMIA event, Surabhi Datta, PhD, Sr. Staff NLP Scientist, presented a poster titled Toward Granular Social Determinants of Health (SDOH) Coding: A Semantics AI Framework to Extract and Encode SDOH Enabled by Large Language Models. In addition to Datta, the study’s authors included Hunki Paek, PhD, Kyeryoung Lee, PhD, Liang-Chin Huang, PhD, Jingqi Wang, PhD, and Xiaoyan Wang, PhD, FAMIA.
By leveraging granular IMO Health terminologies related to social determinants of health (SDOH), the team was able to successfully identify patients’ SDOH elements from free-text electronic health record (EHR) notes and subsequently map these elements to standard IMO Health concepts.
The proposed LLM framework was based on the GPT-4 model and achieved promising results, notably an F1-score of 90.18 in identifying the SDOH elements. Among the correctly identified elements, the accuracies in classifying the broad SDOH category and mapping to granular IMO Health concepts were 86.63% and 95.45%, respectively.
The team’s manual analysis of the results also revealed common mistakes made by the system and underscored areas for future improvement. Ultimately, this research can assist researchers and clinicians in predicting health outcomes, providing personalized care, and identifying patients who may require social support.
