Lupus & antiphospholipid syndrome
Harnessing AI and social media to understand real-world patient experiences in SLE
Authors' conclusion
LLM-enabled social media listening offers a scalable method for synthesising large volumes of unstructured patient narratives, providing timely insights into lived experiences and unmet needs among individuals discussing lupus online.
Study context
- Population
- Reddit posts referencing SLE from 10 SLE-related Reddit communities (October-November 2025)
- Participants
- 2,603 posts
- Exposure
- GPT-5.2 vs Gemini 3.0 for automated thematic labelling
- Outcome
- Theme prevalence across posts
Methodological detailssocial media listening, LLM-based thematic labelling, qualitative review, Thematic codebook developed through manual review of 300 posts and refined iteratively
- Social media listening
- LLM-based thematic labelling
- Qualitative review
- Thematic codebook developed through manual review of 300 posts and refined iteratively
Main finding
As reported in the abstract
- Advice Seeking (most frequent subtheme)
- 84.1%
- Emotional Coping
- 55.6%
- Pain
- 37.2%
- Other Symptom Presentations
- 37.6%
- Fatigue
- 24.7%
- Acute or Worsening Flares
- 30.2%
- Testing and results confusion
- 24.0%
- Emotional impact of uncertainty and disease progression
- 33.0%
Observational data: associations, not effects.
Further findings
| Outcome | GPT-5.2 | Gemini 3.0 |
|---|---|---|
| LLM labelling performance vs human-annotated reference set (n=100) | ||
| F1 | 0.844 | 0.811 |
- LLM labelling performance vs human-annotated reference set (n=100) — Intermodel agreement across main themes, mean κ
- 0.71
Original publication
Harnessing AI and social media to understand real-world patient experiences in SLE. Lupus Sci Med. Published 21 July 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →