Open educational resources (OERs),1 which can be accessed and shared by anyone with Internet or cellular data access, have the potential to overcome local resource constraints and put high-quality medical education content within reach of any educator or learner.
Online videos are especially valuable OERs in medical education. The cognitive theory of multimedia learning and a growing body of evidence suggest that videos enhance learning by activating visual and auditory pathways and presenting words and pictures in a congruent manner,2,3 which can help students more efficiently consolidate medicine’s vast body of knowledge. Videos can function as “reusable learning objects” (RLOs), digital curricular units that can be accessed and used by individuals across contexts.4 RLOs can reduce faculty effort, improve the quality of resources that learners access, and fit into multiple pedagogical formats.5,6 As one example, videos are crucial RLOs in massive open online courses, where participants spend most of their time viewing videos7 and many may only view videos, without navigating through other elements of the course.8 In increasingly popular “flipped classroom”9 models, videos are also often viewed before interactive classroom activities. Additionally, videos are easier than ever to capture, create, and share, with the ubiquity of mobile devices, improved accessibility of video-editing and graphics software, and wide availability and popularity of media-sharing sites.
YouTube, which began in 2005, is the world’s largest media-sharing site, boasting hundreds of millions of hours of videos watched per day by over 1 billion users, reaching 95% of the world that has Internet access.10 YouTube also almost certainly constitutes the world’s largest repository of medical education videos, with dedicated medical education channels attracting hundreds of thousands of subscribers and millions of video views. Large majorities of medical students use YouTube videos to supplement their studies,11,12 and some advocate for faculty to become more engaged with YouTube.13 YouTube can expand the impact of faculty-created videos14,15; in one report, 15 clinical procedure videos received only 200 views over 6 months on a custom video-hosting site, but when shared on YouTube those same videos got 1.7 million views over 33 months.14
Given YouTube’s potential to reach a global audience, its popularity among learners, the potential for faculty to have a broad impact by creating YouTube videos, and the ease with which videos can be used in medical curricula, to our knowledge surprisingly few studies have examined use of YouTube videos for medical education. No studies have identified whether YouTube videos reach those in low-resource settings or are confined to regions local to their producers, and no studies have determined whether topics that viewers select are relevant to local health needs. Having an understanding of who watches free, open medical education videos and what they watch would be important to those seeking to broadly affect medical education and to funders who seek to invest in interventions that effectively reduce disparities in access to medical education resources. Therefore, the goal of this study was to describe the global reach and topic-viewing patterns of medical education videos by analyzing viewer data for one medical-education-focused YouTube channel over a 12-month period.
Osmosis YouTube channel
From December 20, 2015, through January 31, 2017, 189 disease-focused videos were created by a North American team that included experts in health sciences, online pedagogy, and video production (including K.S., R.D., V.W., T.M.) and were placed in the Osmosis YouTube channel (https://www.youtube.com/osmosis). A YouTube channel must be created by anyone wanting to upload videos to YouTube. Each channel has its own username and password. The Osmosis channel was chosen because of our team’s access to its full YouTube Analytics data.
Osmosis also created a web and mobile platform (www.osmosis.org), which at the end of the study had been used by 90,000 students from over 500 medical schools. The platform enables students and faculty to upload their own curricular documents, which natural language processing algorithms can read and then recommend relevant content as well as flashcards and multiple-choice questions.16–18 As the videos were created, they therefore were recommended to learners in the platform when appropriate. Many Osmosis videos also appeared on relevant Wikipedia pages during the study period.
Osmosis YouTube videos were on average 6.9 minutes in length (standard deviation 2.6 minutes) and done in an informal tutorial style with voiceover animation. All scripts were written and revised by Osmosis medical illustrators and a clinical faculty script writer, which ensured consistent format and quality across all videos. In some cases, faculty or staff from academic institutions or professional organizations that sponsored Osmosis videos served as subject matter experts and participated in video production. Most of these faculty used the videos in their course sessions. Funding for videos came from a combination of grants, donations from users, sponsorship by academic institutions or professional organizations, and investment by Osmosis of revenue generated by its web and mobile platform subscriptions. During the study time period, Osmosis provided videos for free and without advertisements, so they did not directly generate revenue for Osmosis, although Osmosis YouTube video viewers may have been more likely to pay for a subscription to the Osmosis platform.
Videos were continuously being posted, typically one to two new videos each week during the study period. Video content was based on core content covered in health science textbooks, clinical journals, and board preparation resources for United States Medical Licensing Examination Step 1 and Comprehensive Osteopathic Medical Licensing Examination of the United States Level 1 exams. The order in which new videos were developed was primarily driven by the desire to complete an entire organ systems block of content before moving to another. However, the Osmosis team also wanted to demonstrate responsiveness to user feedback—which was collected on the YouTube channel via a link to a feedback form, through Osmosis platform correspondence, and during user interviews—and would create videos based on high user demand. Some video topics were also chosen according to the availability and interests of subject matter experts or funders who wanted to collaborate. After a video was posted, the Osmosis team consistently reviewed comments on videos and replied when appropriate, and revised and updated videos as needed.
Videos were explicitly intended for a global audience, so American jargon, references, and images were avoided, and the metric system was used when referring to measurements. All videos were in English, and subtitles for non-English languages were created for some videos through a collaboration with Wikimedia ProjectMedicine and Translators Without Borders and uploaded to videos on YouTube when completed. After language tracking was enabled in November 2016, 96% of videos were viewed in English, with the remainder being translated; Arabic, French, Spanish, and German were the most common translated languages.
Data for this study came from YouTube Analytics for the Osmosis channel. Variables included watch time, number of views, number of “likes” and “dislikes” for each video, and the number of channel “subscribers.” Subscribers are individuals who choose to receive notifications about new content in the channel, and YouTube reports that they generally drive greater viewership within a channel.19 Viewer country and video topic were available for each view. General demographic data (i.e., viewer age and sex) were available in aggregate percentages for the channel. YouTube uses proprietary algorithms to ensure that view counts cannot be artificially inflated (e.g., by refreshing a video page multiple times), which prevents analysis of individuals’ viewing habits. We aggregated daily viewing data by month and report over a 12-month time frame from February 1, 2016, when there were 39 videos online, to January 31, 2017, when there were 189 videos.
We conducted descriptive analyses for the 12-month period and analyzed trends in the number of views and subscribers by viewer income level and region, according to World Bank classifications. For income level, the World Bank has four categories: 79 countries are high income (HIC), comprising a population of 1.2 billion people and 32% of the world’s Internet users; 56 countries are upper middle income (UMIC), comprising 2.6 billion people and 44% of the world’s Internet users; 52 countries are lower middle income (LMIC), comprising 2.9 billion people and 22% of the world’s Internet users; and 31 countries are low income (LIC), comprising 650 million people and 2% of the world’s Internet users. The World Bank also classifies countries into seven geographic regions: North America; Europe and Central Asia; South Asia; East Asia and Pacific; Middle East and North Africa; Latin America and Caribbean; and Sub-Saharan Africa. Simple linear regressions were run to determine trends in proportion of views over the 12 time points in the yearlong period. Stata statistical software, version 13 (StataCorp LP, College Station, Texas) was used for data analysis.
To determine which topics were relevant in a given setting, we analyzed view rankings by country income level for the 50 most-viewed videos during January 2017. Institutional review board approval was not necessary because data contained no information that could be linked to specific individuals.
From February 1, 2016, to January 31, 2017, there were a total of 5,226,405 views for a total of 20,153,093 minutes (38.3 years), with a mean of 3.9 minutes per view. Viewers were located in 213/218 (97.7%) World Bank economies. The five economies that did not have a recorded viewer were the Channel Islands, Nauru, North Korea, Turkmenistan, and Tuvalu. Around half (51%) of views came from men, and 81% of views were from those aged 18–34.
A total of 105,117 individuals were subscribed to the Osmosis channel as of January 31, 2017. Across the time period, subscribers accounted for 1.2 million (24%) views. Of 68,970 ratings from subscribers and nonsubscribers, videos were “liked” 68,516 times (99.3%) and “disliked” 454 times (0.7%). Of the 5,032,053 views for which YouTube provided device information, 2,768,450 (55%) views were on a computer; 1,714,132 (34%) were on a mobile phone; 465,448 (9%) were on a tablet; and 84,023 (2%) were on a television, game console, or other device.
During the time period, there was exponential growth in the number of views and subscribers. While the number of videos increased 4.8-fold between February 2016 and January 2017, the number of monthly views increased 50-fold, and the total number of subscribers increased 117-fold (Table 1).
Distribution and growth of views and subscribers by country income level
Figure 1 shows trends in views by country income level. During the 12 months, 3.0 million (57.4%) views were from HICs, 710,828 (13.6%) from UMICs, 1,403,268 (26.8%) from LMICs, 75,339 (1.4%) from LICs, and 21,378 (0.4%) from unknown areas (percentages throughout may not add to 100% because of rounding). Greater than 40-fold growth in monthly views was seen for each income level. Across the 12-month interval, the proportion of views in LICs and UMICs remained statistically unchanged (P > .05), while there was an average 0.8% decrease per month in the proportion of views from HICs (P < .001), with a 0.9% increase per month from LMICs (P < .001).
At the end of January 2017, there were 52,502 (50%) subscribers from HICs, 19,012 (18%) from UMICs, 32,615 (31%) from LMICs, and 1,315 (1.3%) from LICs. Growth in subscribers was more than 20-fold for each income level. Across the time period, statistically significant gains were seen in proportions of subscribers from LICs (0.07%; P < .001) and LMICs (0.8%; P = .009), statistically significant decreases were seen for HICs (0.8%; P = .01), and no change was seen for UMICs (P = .07).
Distribution and growth by country region
Across the time period, 1,508,218 views (29%) were from North America, 1,147,295 (22%) from Europe and Central Asia, 866,422 (17%) from South Asia, 796,191 (15%) from East Asia and Pacific, 485,517 (9.3%) from Middle East and North Africa, 251,319 (4.8%) from Latin America and Caribbean, 150,065 (2.9%) from Sub-Saharan Africa, and 21,378 (0.4%) from unknown areas (Figure 2). A greater than 25-fold increase in monthly views was seen for each region (with South Asia having 105-fold growth). South Asia had a 1.0% increase per month in its proportion of views (P < .001), while there was no statistically significant trend for proportion of views from other regions (P > .05).
By the end of the time period, 24,101 subscribers (23%) were in North America, 21,752 (21%) in Europe and Central Asia, 17,039 (16%) in South Asia, 17,372 (17%) in Middle East and North Africa, 16,212 (15%) in East Asia and Pacific, 6,325 (6.0%) in Latin America, and 2,643 (2.5%) in Sub-Saharan Africa. There was over 80-fold growth in subscriber numbers in each region from the beginning to the end of the time period. The proportion of subscribers from South Asia increased 0.9% per month (P < .001); trends were not significant (P > .05) for other regions.
Video topic viewing by income level
When ranking video topics by number of views in the final month of the study, viral hepatitis, epilepsy, congestive heart failure, and diabetes were the only topics that ranked within the top 10 for all income groups (Supplemental Digital Appendix 1, available at https://links.lww.com/ACADMED/A519). Many conditions had variable rankings across groups. For example, tuberculosis was the most-viewed topic for LICs and LMICs, and the 3rd most-viewed for UMICs, but the 18th most-viewed for HICs. Rheumatic heart disease was ranked between 13th and 19th for non-HICs, but 44th for HIC viewers. Psychiatric conditions were watched more in HICs. For example, schizophrenia was the most-viewed video for HICs, but rankings for other income levels were 12th or lower. Depression was the 4th most-viewed by HICs, but 24th for UMICs, 45th for LICs, and 48th for LMICs. Anxiety was the 10th most-viewed for HICs, but 42nd for UMICs, 43rd for LICs, and 47th for LMICs.
To our knowledge, this is the first study to examine viewing patterns for a collection of medical education videos on YouTube. We describe the ability for these videos to reach large audiences around the world. Topic-viewing patterns also shed light onto how viewing behaviors relate to viewers’ local health needs.
For years, improving online access to relevant medical information has been seen as a way to improve knowledge and performance irrespective of geography.20 Video OERs can be especially valuable because they can stand alone or be integrated into blended or fully online educational models, which have been shown to be feasible even in low-resource settings.21 Here we show the potential for media-sharing sites to disseminate OERs. Even early in the life of the Osmosis channel, videos were being viewed in all regions and country income levels. Although the fractions of views (1.5%) and subscribers (1.3%) that came from LICs were small, approximately 2% of the world’s population with Internet live in LICs, indicating that there was reasonably substantial representation in these areas. Moreover, growth was steady across all groups throughout the time period, suggesting that medical education content on YouTube can immediately and consistently reach a global audience. Indeed, of the five World Bank economies where there were no recorded viewers, two (Nauru, Tuvalu) are island nations with less than 15,000 inhabitants, two (North Korea, Turkmenistan) do not permit YouTube access, and views from the Channel Islands are not independently recorded by YouTube Analytics.
Contributing to social media platforms such as YouTube is increasingly being seen as a legitimate form of scholarship,22 and data such as those in this study can demonstrate the international reach of faculty-produced videos and thereby contribute to faculty professional advancement. Yet, how best to maximize one’s impact in creating and disseminating medical education videos requires consideration. Although data are limited, it seems clear that posting to YouTube has advantages. One study showed that transitioning 15 procedural videos from a custom website to a new “Clinisnips” YouTube channel achieved a 2,500-fold increase in viewership, from an average of 2 views per video per month to 5,000 views per video per month; the most popular video, on central line placement, received over 13,000 views per month.14 By contrast, the most-viewed video on MedTube23—which may have the largest single collection of medical education videos in the world, at over 13,000, but places them on a custom website—was viewed 72,000 times in 6 years, approximately 1,000 times per month. Conversely, placing medical education videos on YouTube does not guarantee high rates of viewership. For example, 22 geriatrics-related videos made in the United Kingdom for physicians and placed on YouTube received on average 26 views per video per month.24 Thirty-six videos that were used in a flipped classroom model for preclinical microbiology, immunology, and infectious disease courses at three U.S. medical schools and placed on YouTube during the same time period as this study in 201625 received an average of 30 views per video per month.
By the end of the time period in this study, the average Osmosis video received over 6,000 views per month, with the most popular videos approaching 20,000 views per month. At the end of the study, on the basis of data provided by the website Social Blade, which makes public trends in views and subscribers for YouTube channels, the overall growth in subscribers for Osmosis during this time period was greater than for other medical education video channels on YouTube.26 One reason for these high levels of viewership may be that videos were continuously being posted, so that previous viewers and subscribers returned to view new content. Without new videos, viewership can decline, as the Clinisnips central line video, which had 13,000 views per month in 2011, has had on average 2,400 views per month since then. A second reason may be that by encouraging and responding to feedback, the channel fostered an interactive, social atmosphere, shown to enhance online engagement.27 A third reason may be the process by which Osmosis videos were created: The content aligned with consensus core topics in textbooks and board preparation materials, and video development followed evidence-based guidelines, such as careful preproduction, an average length of 6 to 9 minutes, and an informal tutorial style, with voiceover animation and use of personal handwriting.28 Finally, the existing Osmosis platform and user base and visibility of videos on Wikipedia likely increased their exposure. Such exposure or name recognition may be important; the Khan Academy Medicine YouTube channel, which is well known because of the popularity of Khan Academy, has a growing viewership without posting new videos for over a year.
In addition to describing trends in overall channel viewing, we described topic-viewing patterns. In general, topic viewing aligned well with available morbidity and mortality data.29 The most recent Global Burden of Disease Study identified ischemic heart disease as most responsible for premature mortality. Acute myocardial infarction, angina, and congestive heart failure, all of which can link to ischemic heart disease, were among the most-watched topics across Osmosis viewers. Some viewing differences also potentially reflect differences in health needs: the fact that depression, anxiety, and schizophrenia were preferentially watched topics in HICs corresponds to self-harm being the fifth most important cause of years of life lost in HICs, but outside the top 10 for non-HICs. Tuberculosis, which causes a substantial burden of illness in non-HICs, was most watched in LMICs and LICs but less commonly watched in HICs. Such findings raise the potential that just as Google search trends30 and Wikipedia access logs31 can be used to identify infectious disease epidemics, YouTube viewing habits may inform the diagnosis of medical education needs and guide curriculum and assessment enhancements. Of course, the fact that epilepsy, endocarditis, and systemic lupus erythematosus, none of which are within the top 10 causes of years of life lost, were among the top 10 most-viewed videos in our study population, indicates that future work would be needed to fully understand topic-viewing data. For example, viewers may preferentially watch videos on topics that they find challenging, that were not well covered or were overrepresented in their local curriculum, or that are emphasized on board examinations.
Important limitations must be considered when reviewing our data. First, all data came from YouTube Analytics. Individual-level data were not available for viewers, and so we are unable to obtain individual demographics, such as socioeconomic status. We also cannot be sure what proportion were medical students, other health professionals, or laypeople viewing out of personal interest. Although it was beyond the scope of this study to systematically analyze comments on videos, we noticed that most comments seemed to come from medical students as they often referred to studying for medical school or made requests for additional content. However, we also saw some similar comments from other health professions students and some comments from patients or their loved ones, which typically described appreciation for helping them understand an illness better. Second, individuals may repeatedly view a video, so that the number of views surpasses the number of individuals accessing the channel; YouTube Analytics data do not permit determining how many videos were watched by individual subscribers or nonsubscribers for videos on our channel or other channels. Third, at the time of evaluation, some common conditions, such as stroke, malaria, and several cancers, did not have a related video, limiting inferences that can currently be made from our topic analysis. Finally, merely viewing a video does not indicate what has been learned, and we were unable to assess the impact that videos had on learners. Given the massive audience that YouTube can reach, YouTube may consider making it easier to assess learning, such as by permitting embedded assessments after a video or for videos between playlists.
In conclusion, our analysis demonstrates that medical education videos posted to YouTube can reach individuals virtually anywhere in a short amount of time with relevant content. Educators seeking to have a global impact may consider posting videos to YouTube. Future work is needed to determine how to optimize viewership of videos posted to media-sharing sites, how assessment of learning can occur, and the potential impact on patients who encounter these videos.
1. UNESCO. Communication and information. Open educational resources. www.unesco.org/new/en/communication-and-information/access-to-knowledge/open-educational-resources/
. Accessed November 20, 2017.
2. Mayer R. The Cambridge Handbook of Multimedia Learning (Cambridge Handbooks in Psychology). 2014.Cambridge, UK: Cambridge University Press.
3. Brame CJ. Effective educational videos: Principles and guidelines for maximizing student learning from video content. CBE Life Sci Educ. 2016;15:es6.1es6.6.
4. Wiley DA. The learning objects literature. Handb Res Educ Commun Technol. 2007;16:345354. http://opencontent.org/docs/wiley-lo-review-final.pdf
. Accessed November 13, 2017.
5. Ruiz JG, Mintzer MJ, Issenberg SB. Learning objects in medical education. Med Teach. 2006;28:599605.
6. Harden RM, Gessner IH, Gunn M, Issenberg SB, Pringle SD, Stewart A. Creating an e-learning module from learning objects using a commentary or “personal learning assistant.” Med Teach. 2011;33:286290.
7. Breslow L, Pritchard DE, DeBoer J, Stump GS, Ho AD, Seaton DT. Studying learning in the worldwide classroom: Research into edX’s first MOOC. Res Pract Assess. March 2012:515.
8. Kizilcec RF, Piech C, Schneider E. Deconstructing disengagement: Analyzing learner subpopulations in massive open online courses. In: Proceedings of the Third International Conference on Learning Analytics and Knowledge. 2013:New York, NY: Association for Computing Machinery; 170179.
9. Prober CG, Khan S. Medical education reimagined: A call to action. Acad Med. 2013;88:14071410.
10. YouTube statistics page. https://www.youtube.com/yt/about/press/
. Accessed November 13, 2017.
11. Barry DS, Marzouk F, Chulak-Oglu K, Bennett D, Tierney P, O’Keeffe GW. Anatomy education for the YouTube generation. Anat Sci Educ. 2016;9:9096.
12. Rapp AK, Healy MG, Charlton ME, Keith JN, Rosenbaum ME, Kapadia MR. YouTube is the most frequently used educational video source for surgical preparation. J Surg Educ. 2016;73:10721076.
13. Rabee R, Najim M, Sherwani Y, et al. YouTube in medical education: A student’s perspective. Med Educ Online. 2015;20:12.
14. Topps D, Helmer J, Ellaway R. YouTube as a platform for publishing clinical skills training videos. Acad Med. 2013;88:192197.
15. Carlos WG, Goss K, Morad M. YouTube-based critical care educational videos go viral. Med Educ. 2014;48:526.
16. Gaglani SM, Haynes MR. What can medical education learn from Facebook and Netflix? Ann Intern Med. 2014;160:640641.
17. Haynes MR, Gaglani SM, Wilcox MV, Mitchell T, DeLeon VB, Goldberg HR. Learning through Osmosis: A collaborative platform for medical education. Innov Glob Med Health Educ. 2014;2:18.
18. Menon A, Gaglani S, Haynes MR, Tackett S. Using “big data” to guide implementation of a web and mobile adaptive learning platform for medical students. Med Teach. 2017;39:975980.
20. Aronson B. Improving online access to medical information for low-income countries. N Engl J Med. 2004;350:966968.
21. Frehywot S, Vovides Y, Talib Z, et al. E-learning in medical education in resource constrained low- and middle-income countries. Hum Resour Health. 2013;11:4.
22. Sherbino J, Arora VM, Van Melle E, Rogers R, Frank JR, Holmboe ES. Criteria for social media-based scholarship in health professions education. Postgrad Med J. 2015;91:551555.
23. Medtube. https://medtube.net/
. Accessed November 13, 2017.
24. Garside MJ, Fisher JM, Blundell AG, Gordon AL. The development and evaluation of mini-GEMs—Short, focused, online e-learning videos in geriatric medicine [published online ahead of print April 6, 2016]. Gerontol Geriatr Educ. doi: 10.1080/02701960.2016.1165217.
25. Adam M, Chen SF, Amieva M, et al. The use of short, animated, patient-centered springboard videos to underscore the clinical relevance of preclinical medical student education. Acad Med. 2017;92:961965.
26. Social Blade. https://socialblade.com/
. Accessed November 13, 2017.
27. Dixson MD. Creating effective student engagement in online courses: What do students find engaging? J Scholarsh Teach Learn. 2010;10:113.
28. Guo PJ, Kim J, Rubin R. How video production affects student engagement. 2014:In: Proceedings of the First ACM Conference on Learning. New York, NY: Association for Computing Machinery;4150.
29. Wang H, Naghavi M, Allen C, et al. Global, regional, and national life expectancy, all-cause mortality, and cause-specific mortality for 249 causes of death, 1980–2015: A systematic analysis for the Global Burden of Disease Study 2015. Lancet. 2016;388:14591544.
30. Dugas AF, Jalalpour M, Gel Y, et al. Influenza forecasting with Google Flu Trends. PLoS One. 2013;8:e56176.
31. Generous N, Fairchild G, Deshpande A, Del Valle SY, Priedhorsky R. Global disease monitoring and forecasting with Wikipedia. PLoS Comput Biol. 2014;10:e1003892.