In an abstract presented at the ISPOR 2026 Annual Meeting, researchers examined the feasibility of using large language models (LLMs) to analyze social media data and generate insights into unmet needs among patients with systemic lupus erythematosus (SLE). Patient experience and unmet need insights are critical inputs for patient-focused drug development; however, traditional qualitative approaches are often resource intensive and difficult to scale in complex diseases such as SLE. Social media represents an emerging real-world data source for capturing the patient voice.

In 2022, the US Food and Drug Administration (FDA) issued patient-focused drug development guidance that recognized social media as a potential source of patient experience data and outlined ethical considerations for its use. In this study, LLMs were applied to analyze Reddit posts about SLE, identify patient-expressed unmet needs, symptom experiences, and health care challenges, and demonstrate how artificial intelligence (AI)-enabled social media listening can complement traditional patient experience research.

Between October 14, 2025, and November 25, 2025, investigators collected 4633 Reddit posts from 10 lupus-related subforums. After duplicates, promotional content, and incomplete posts were removed, 2603 posts remained, most of which were published since 2020 (98.4%). Two LLMs, Google Gemini 3.0 Pro and OpenAI GPT-5.2, demonstrated high performance comparable to that of human evaluators. The average percentage agreement was 94.4%, with a Cohen κ of 0.67 for both Gemini 3.0 Pro and GPT-5.2.

Findings from the thematic analysis identified advice-seeking (84.7%) and emotional coping (53.8%) as dominant areas of discussion. Patients frequently referenced pain (40.9%), systemic symptoms (49.6%), and flares (28.4%), whereas medication-related concerns centered on effectiveness (27.5%) and tolerability (19.2%). Diagnostic uncertainty (26.3%) and the emotional impact of diagnosis (27.8%) were also prominent. Reported challenges included provider dismissal (18.2%) and access barriers (10.0%), along with meaningful lifestyle implications, including changes to daily living (25.1%) and workplace accommodation needs (10.8%).

LLM-powered social listening demonstrated the ability to generate large-scale insights into patient experiences and unmet needs from SLE-related Reddit discussions. This framework has the potential to augment traditional qualitative research and support more scalable, patient-centered evidence generation across disease areas.

Reference:
Yang S, Hawryluk C, Liu J, Otoo J, Erkert N, Yao L. Listening at scale: AI-powered insights into systemic lupus erythematosus patient needs from social media. Value Health. 2026;29(suppl 6). Presented at: ISPOR 2026 Annual Meeting; May 17-20, 2026; Philadelphia, PA.