Treatment options and emotional well-being in patients with rosacea: An unsupervised machine learning analysis of over 200,000 postsCapsule Summary

• Published in: JAAD international Vol. 13; pp. 172 - 178
• Main Authors: Karan Rajalingam, BS, Nicole Levin, BS, Oge Marques, PhD, James Grichnik, MD, PhD, Ann Lin, DO, Wei-Shen Chen, MD, PhD
• Format: Journal Article
• Language: English
• Published: Elsevier 01.12.2023
• Subjects: machine learning; medical dermatology; rosacea; social media
• ISSN: 2666-3287; 2666-3287

Background: Many patients with rosacea join online support groups to gather and disseminate information about disease management and provide emotional support for others. Objective: To better understand rosacea patient’s primary concerns for the disease as well as their disease search patterns online. Methods: Overall, 207,038 posts by 41,400 users were collected from June 1, 2017, to June 1, 2022, in a popular online forum. We applied Latent Dirichlet Allocation (LDA), an unsupervised machine learning model, to organize the posts into topics. Keywords for each topic supplied by LDA were used to manually assign topic and category labels. Results: Twenty-three significant topics of conversation were identified and organized into 4 major categories, including Management (50.33%), Clinical Presentation (24.14%), Emotion (21.97%), and Information Appraisal (3.57%). Limitations: Although we analyzed the largest forum on the internet for rosacea, generalizability is limited given the presence of other smaller forums and the skewed demographics of forum users. Conclusion: Social media forums play an important role for disease discussion and emotional venting. Although rosacea management was the most frequently discussed topic, emotional posting was a significantly prevalent occurrence.