Mapping Five Years of #FOAMed: Trends, Engagement, and Shifting Topics on Twitter/X.

INTRODUCTION

Free Open Access Medical Education (FOAMed) has emerged as a prominent component of online medical communication, with X (formerly Twitter) serving as an active hub for professional exchange among clinicians. Despite its reach and influence, few longitudinal studies have examined how FOAMed content and engagement patterns evolve over time. In this study we aimed to analyze thematic shifts and user interaction trends in #FOAMed tweets over a five-year period.

METHODS

We conducted a retrospective bibliometric and natural language processing (NLP) study of 6,000 high-engagement, English-language tweets tagged with #FOAMed, posted between January 1, 2020-December 31, 2024. Each month, the 100 tweets were selected from Twitter's "Top" tab and manually curated. We used latent Dirichlet allocation (LDA) to identify thematic clusters. Hashtag usage and engagement metrics were assessed using descriptive statistics and linear regression.

RESULTS

We identified 10 distinct topics were identified through LDA modeling: point-of-care ultrasound (POCUS) education; neuro-radiology, cardiology-electrocardiogram (ECG); nephrology; and intensive care unit; ultrasound; prehospital/policy; webinars and learning; resuscitation scenarios; pediatric imaging; medical student education; and critical care and publications. Topic prevalence shifted over time: Early tweets focused on COVID-19 and critical care, while later years showed increasing attention to prehospital care, diagnostics, and POCUS. Mean tweet engagement peaked in 2023 (236.9 ± 914.6). Notably, hashtags such as #POCUS and #MedEd showed substantial increases in both usage and engagement, with #MedEd reaching a peak mean engagement of 287.7. In contrast, COVID-19 declined steadily, both in frequency (from 126 tweets in 2020 to just six in 2023) and in engagement (mean: 67.1 → 18.5). Spearman correlation analysis revealed that hashtag count had a weak but statistically significant correlation with engagement (ρ = 0.047, P < .001), suggesting that content quality, rather than volume, was the primary driver of visibility.

CONCLUSION

FOAMed discourse on Twitter/X remains dynamic, responsive to clinical priorities and shaped by peer interaction. Natural language processing and topic modeling are valuable tools to uncover longitudinal trends in digital medical education, reinforcing Twitter/X's role in informal, real-time learning communities.