
Methods: An inquiry for the search term ‘Ebola’ was made on YouTube. The first 100 results were arranged in decreasing order of “relevance” using the default YouTube algorithm. Videos 1-50 and 51-100 were allocated to a high relevance (HR), and a low relevance (LR) video group, respectively. Multivariable logistic regression models were used to assess the predictors of a video being included in the HR vs. LR groups. Fourteen videos were excluded because they were parodies, songs or stand-up comedies (n=11), not in English (n=2) or a remaining part of a previous video (n=1). Two scales, the Video Information and Quality and Index (VIQI) and the Medical Information and Content Index (MICI) assessed the overall quality, and the medical content of the videos, respectively.
Results: There were no videos from hospitals or academic medical centers. Videos in the HR group had a higher median number of views (186,705 vs. 43,796, p<0.001), more ‘likes’ (1,119 vs. 224, p<0.001), channel subscriptions (208 vs. 32, p<0.001), and ‘shares’ (519 vs. 98, p<0.001). Multivariable logistic regression showed that only the ‘clinical symptoms’ component of the MICI scale was associated with a higher likelihood of a video being included in the HR vs. LR group (OR 1.86; 95%CI, 1.06 to 3.28, p=0.03).
Conclusion: YouTube videos presenting clinical symptoms of infectious diseases during epidemics are more likely to be included in the HR group and influence viewers behavior. Additionally, credible organizations such as academic medical centers should prioritize the development of multimedia for public education during acute epidemics as a means to reduce unnecessary health care encounters and costs.

D. Mukhija,
None
A. Karimianpour, None
D. Mohan, None
A. Brateanu, None