By: Shiv Gaglani, CEO of Osmosis, and Almanac Editor Sapana Panday, Director of Educational Development & Grant Writing, The France Foundation
Osmosis is a medical health education company with an audience of over 300,000 current and future health professionals, as well as patients and family members. The former Khan Academy Medicine team joined Osmosis in 2015 and has produced more than 325 concise animated videos that are featured on Wikipedia, YouTube, Merck Manual, Medscape and other sites.
Shiv and his co-founder, Ryan Haynes, began Osmosis when they were medical students at Johns Hopkins University and have since grown the company to 30 people, bringing on an impressive list of angel investors including Peter Frishauf, the founder of Medscape, and Dr. Lois Nora, the CEO of the American Board of Medical Specialties.
What exactly do Netflix and Facebook do that keeps their users going back to them?
There are a number of strategies these tech companies apply to hook us on their products and services. At the Alliance presentation in January 2017, I shared six case studies and will describe two of them here.
One theme is personalization. One reason Facebook’s newsfeed is so addictive is because Facebook makes decisions about which updates to share with you based on how interested you are in a given individual or page based on your stated preferences and, more importantly, how often you click on or engage with that individual's/page’s updates. This means that there is more relevant content every time you visit the site. Netflix and Amazon do something similar with their recommended content engines that aim to, respectively, keep you watching their programming or upsell you on more products.
A second strategy is focused updates. If you’ve used Redfin to search for real estate, you’ve probably received emails with price updates that are specific to properties you were looking at, without necessarily even flagging those properties. Algorithms track which listings you visit and how frequently you visited them, and can use that to influence which updates they send you. Similarly, LinkedIn can send you updates on people in your network who have changed positions so you can congratulate them (and, when you do, be served up ads since it’s increasing their monthly active user numbers).
These are just two examples — there are a number of other tactics such as gamification, user-generated content, and A/B testing.
Have you seen this concept applied in medical education? If so, can you tell us an example?
Yes. A few of these tactics are being applied by education companies to improve learner engagement and retention. Of course we’re working to apply many of these techniques at Osmosis, e.g., through our mobile app that features a game similar to QuizUp/TriviaCrack called “Osmose with your Friends.” I’ve also received personalized updates from companies such as Medscape and Merck Manual based on what emails I click on and pages I visit, which is a step in the right direction. These groups have also incorporated user-generated content as a way to improve engagement among their audiences.
I’ve heard you talk about “ed retargeting” before. Can you tell me more about that?
This is a term we coined by adapting it from a ubiquitous marketing tactic called “ad retargeting.” If you search for a term on Google, say “pants,” and then click on a result from the retailer Nordstrom, you’re more likely to see a Nordstrom ad in your Facebook newsfeed later that day. That’s not a coincidence, but rather a tactic that can improve the chances of you converting into a purchaser of Nordstrom pants (and while you’re there, maybe other apparel). The key is to keep that item or brand top of mind.
Osmosis has borrowed this strategy and calls it “ed retargeting.” Essentially our platform “knows” what you’re learning based on the content you’re clicking on or uploading to the system (e.g., a calendar, lecture document, etc.) and then can automatically ping you with relevant associated content. So, if you read something about therapies for managing cystic fibrosis today, tomorrow you’ll get a multiple choice question about CF therapies sent to your phone via a push notification. We have a provisional patent on this process and are particularly excited about its implications for continuing education and “just-in-time” learning at the point of care.
Shouldn’t we just hit our learners with everything constantly? I mean, we do want them to learn everything.
Medical education is too vast and dynamic for anyone to try learning all there is to know, even within a specific specialty. We do need to make sure that clinicians have an up-to-date corpus of knowledge because it would otherwise take too long for them to make decisions — time that they don’t have given the volume of patients they have to see and paperwork they now have to do. Thus, our vision is to provide a lightweight “personal learning assistant” that knows about the clinician’s background, practice population, interests and resources, and can automatically give them bite-sized, active learning snippets. A short video here, a quiz there — all relevant to the types of patients they’re seeing and articles they’re reading. They also need to develop meta-cognition — understanding how they learn and what their limitations are, so for example they know when to consult external resources including reference databases, papers, and colleagues.
Can you tell our readers about the types of metrics you recommend collecting beyond the basic number of visitors to a website?
It depends on the application, but I can share some of the data we collect.
We recently published a paper in Medical Teacher entitled “Using ‘big data’ to guide implementation of a web and mobile adaptive learning platform for medical students” that describes our approach. In addition to daily and monthly active users we track — for example — session length, pages visited, videos watched, items answered, time spent to answer, time spent to move to the next item and confidence of the answer. We used this last metric to analyze confidence versus accuracy (called calibration) in roughly 1,000 medical students, we found that male and female medical students were equally accurate, though male medical students tended to be far more overconfident. This is described in an Annals of Internal Medicine report we published in 2015 entitled “The Association Between Confidence and Accuracy Among Users of a Mobile Web Platform for Medical Education.” We continuously aim to use our data not just to improve engagement metrics, but, most importantly, to improve learning outcomes and, ultimately, patient outcomes.
Learner-generated content seems like a cool idea, but I’m not sure it works in our industry. I mean, everything we do has to be scientifically proven. We are not Wikipedia.
Crowdsourcing has several advantages, such as the ability to develop large amounts of content in relatively little time, though it requires solid quality-control mechanisms (just read the comments on unmoderated forums and you’ll see what I mean). For example, we’ve had 5,000 learners on Osmosis generate more than 3 million practice questions and flashcards using guidelines and videos we’ve produced.
There are many success stories such as Mechanical Turk, FoldIt (a game that turns puzzle-lovers into protein folders that have succeeded where super computers have failed), and Recaptcha (which helps prevent credit card fraud while digitizing books). Wikipedia is also a success story, with many peer-reviewed papers showing remarkable accuracy and detail, as well as the ability to be more up-to-date than sources that go through a cumbersome peer-review process. Indeed, there is a dedicated team of about 200 health professional volunteers who have formed a group called WikiProject Medicine that helps coordinate and review improvements to these health science articles. They are led by an emergency physician named James Heilman who is also on the Board of Trustees of the Wikimedia Foundation.
Earlier this year, Osmosis received a grant from the Hewlett Foundation as part of our partnership with UCSF Professor, Dr. Amin Azzam, to turn medical trainees into Wikipedia editors so they can further enhance the quality of the health articles on Wikipedia while simultaneously learning. This concept has been picking up steam and we have more than half-a-dozen implementations at medical and health professions schools. I think there will be applications to continuing medical and health education as well!
Is there anything else you’d like to share?
Though Osmosis got its start in the medical student space (which makes sense given our backgrounds as medical students), we now reach more than 300,000 current and future clinicians across the health professions. These include physicians, nurses, pharmacists, dentists, physician assistants, physical therapists, dieticians and others. We have partnered with more than 25 companies and organizations such as Kaiser Permanente, Rush Medical College, and the American Board of Medical Specialties to develop high quality video and assessment content and then deliver these via our unique technology and distribution channels. We’d welcome any readers to contact us at firstname.lastname@example.org if they’re interested in learning more.