Patient-made knowledge networks: Long COVID discourse, epistemic injustice, and online community formation

Long COVID represents an unprecedented case of patient-led illness definition, emerging through Twitter in May 2020 when patients collectively named and legitimized their condition before medical recognition. This study examines 2.8 million tweets containing #LongCOVID through topic modeling followed by content analysis of a purposively sampled 5100 tweets, and exponential random graph modeling (ERGM) to understand how contested illness communities construct knowledge networks. User roles were identified through analysis of Twitter biography fields and review of 1000 tweets (10 per user) from the 100 highest-centrality accounts. We identify seven discourse themes spanning symptom documentation, medical dismissal, cross-illness solidarity, and policy advocacy, alongside a differentiated ecosystem of four user roles. ERGM results demonstrate that tie formation centers on epistemic practices-knowledge sharing and community building-rather than policy debates, supporting characterization of this space as an epistemic community in Haas's (1992) sense. Long COVID patients achieved World Health Organization (WHO) recognition within months, contrasting sharply with decades-long struggles of similar conditions. These findings illuminate how social media affordances enable marginalized patient populations to construct alternative knowledge systems, form cross-illness coalitions, and contest traditional medical authority.