Research using health-related social media in the Large Language Model era: methodological advancements, novel applications, and challenges

Presentation Time: 08:30 AM - 12:00 PM

Abstract Keywords: Social Media and Connected Health, Large Language Models (LLMs), Natural Language Processing, Delivering Health Information and Knowledge to the Public, Population Health
Working Group: Consumer Health Informatics Working Group
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics

Health-related social media (HSM) such as health-related tweets contain valuable information that can help improve healthcare service quality and patient experiences. Researchers have analyzed HSM concerning a wide range of health topics, including vaccination, healthcare policies, and infectious disease outbreaks. However, qualitative analysis of HSM is limited in scale, and Natural Language Processing (NLP) methods often require large, high-quality annotated datasets as well as technical expertise and computation resources. Large Language Models (LLMs) such as the Generative Pre-trained Transformer (GPT) are pre-trained on large corpora and show superior performances with little to no annotated data. This paradigm-shift will particularly benefit researchers who study patient and public-generated HSM by enabling advanced NLP-based analysis with little human input. Despite their promises, there is limited discussion of how LLMs could transform research using patient and public-generated HSM and associated challenges such as ethical and research replicability issues. This discussion is particularly crucial, as patient and public-generated HSM contain valuable information that can promote patient-centered care and incorporate patients’ voices into improving healthcare services and policymaking.

This workshop will bring together informatics researchers, public health researchers, LLM researchers, and ethicists to identify promising application areas of LLMs on HSM and pinpoint challenges of using LLMs on patient and public-generated social media data for secondary research. There are three main topics for the workshop: 1. Assessing the performance of LLMs for analyzing diverse HSM; 2. Identifying novel areas of research enabled by LLMs on HSM; 3. Discussing challenges of using LLMs on HSM and strategies to mitigate them.

Speaker(s):
Xiaoyu Liu, PhD
Saint Louis University

Lu He, PhD
University of Wisconsin-Milwaukee

Tera Reynolds, PhD, MPH, MS
University of Maryland Baltimore County

Xiaoyu Liu
Southern Illinois University Carbondale

Yong Choi, PhD, MPH
University of Pittsburgh

Vignesh Subbian, PhD
University of Arizona

Author(s):
Lu He, PhD - University of Wisconsin-Milwaukee; Tera Reynolds, PhD, MPH, MS - University of Maryland Baltimore County; Xiaoyu Liu, PhD - Saint Louis University; Yong Choi, PhD, MPH - University of Pittsburgh; Vignesh Subbian, PhD - University of Arizona;