Using Patient-Reported Health Data From Social Media to Identify Diverse Lupus Patients and Assess Their Symptom and Medication Expressions: A Feasibility Study.

Abstract

Patient communities use social media for peer support and information seeking. This study assessed the feasibility of using public patient-generated health data from the social network Twitter to identify diverse lupus patients and gather their perspectives about disease symptoms and medications. We extracted public lupus-related Twitter messages (n = 47,715 tweets) in English posted by users (n = 8,446) in the US between September 1, 2017 and October 31, 2018. We analyzed the data to describe lupus patients and the expressed themes (symptoms and medications). Two independent coders analyzed the data; Cohen's kappa coefficient was used to ensure interrater reliability. Differences in symptom and medication expressions were analyzed using 2-tailed Z tests and a combination of 1-way analysis of variance tests and unpaired t-tests. We found that lupus patients on Twitter are diverse in gender and race: approximately one-third (34.64%, 62 of 179) were persons of color (POCs), and 85.47% were female. The expressed disease symptoms and medications varied significantly by gender and race. Most of our findings correlated with documented clinical observations, e.g., expressions of general pain (8.39%, 709 of 8,446), flares (6.05%, 511 of 8,446), and fatigue (4.18%, 353 of 8,446). However, our data also revealed less well-known patient observations, e.g., possible racial disparities within ocular manifestations of lupus. Our results indicate that social media surveillance can provide valuable data of clinical relevance from the perspective of lupus patients. The medical community has the opportunity to harness this information to inform the patient-centered care within underrepresented patient groups, such as POCs.