Reddit and Rare Diseases: What Myositis Communities Tell Us About Support and Struggle

ABSTRACT Myositis is a rare autoimmune condition associated with muscle weakness, systemic involvement, and long-term disability. People living with rare conditions frequently use online platforms to share experiences and seek information beyond formal healthcare settings. This exploratory study applied automated natural language processing (NLP) methods to analyse public posts and comments from the myositis subreddit between 2020 and 2024. The dataset comprised 1,223 unique posts and 18,453 unique comments contributed by 3,987 users. Seven sentiment models and one ensemble were evaluated, with RoBERTa-based transformer models selected for contextual analysis. Topic modelling was conducted using Latent Dirichlet Allocation and BERTopic. While BERTopic initially generated highly granular clusters, hierarchical merging produced stable, clinically interpretable themes. Six dominant discussion themes were identified: (1) autoimmune diagnosis and systemic symptoms; (2) organ-specific concerns; (3) fatigue, mobility, and sleep; (4) navigating care pathways; (5) life impact; and (6) coping strategies, lifestyle changes, and anxiety. Topic-linked sentiment analysis showed predominantly neutral sentiment overall. More negative sentiment was associated with unresolved symptoms and diagnostic uncertainty, and sentiment became increasingly negative at deeper comment levels, suggesting escalation toward unmet needs. Highly active users mainly contributed neutral, informational content. Temporal analysis identified episodic increases in engagement, including a peak in mid-2024 linked to discussions involving dermatomyositis, COVID-19, and treatment decisions. These findings demonstrate the feasibility of scalable NLP analysis of rare-disease online discourse and highlight how topic-linked emotional signals may indicate periods of uncertainty or emerging concerns.