Abstract
BACKGROUND: YouTube has become a popular source of health care information, reaching an estimated 81% of adults in 2021; approximately 35% of adults in the United States have used the internet to self-diagnose a condition. Public health researchers are therefore incorporating YouTube data into their research, but guidelines for best practices around research ethics using social media data, such as YouTube, are unclear. OBJECTIVE: This study aims to describe approaches to research ethics for public health research implemented using YouTube data. METHODS: We implemented a systematic review of articles found in PubMed, SocINDEX, Web of Science, and PsycINFO following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. To be eligible to be included, studies needed to be published in peer-reviewed journals in English between January 1, 2006, and October 31, 2019, and include analyses on publicly available YouTube data on health or public health topics; studies using primary data collection, such as using YouTube for study recruitment, interventions, or dissemination evaluations, were not included. We extracted data on the presence of user identifying information, institutional review board (IRB) review, and informed consent processes, as well as research topic and methodology. RESULTS: This review includes 119 articles from 88 journals. The most common health and public health topics studied were in the categories of chronic diseases (44/119, 37%), mental health and substance use (26/119, 21.8%), and infectious diseases (20/119, 16.8%). The majority (82/119, 68.9%) of articles made no mention of ethical considerations or stated that the study did not meet the definition of human participant research (16/119, 13.4%). Of those that sought IRB review (15/119, 12.6%), 12 out of 15 (80%) were determined to not meet the definition of human participant research and were therefore exempt from IRB review, and 3 out of 15 (20%) received IRB approval. None of the 3 IRB-approved studies contained identifying information; one was explicitly told not to include identifying information by their ethics committee. Only 1 study sought informed consent from YouTube users. Of 119 articles, 33 (27.7%) contained identifying information about content creators or video commenters, one of which attempted to anonymize direct quotes by not including user information. CONCLUSIONS: Given the variation in practice, concrete guidelines on research ethics for social media research are needed, especially around anonymizing and seeking consent when using identifying information. TRIAL REGISTRATION: PROSPERO CRD42020148170; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=148170.