Gamifying Medical Education

Mayo Clinic, 2017

How building an educational game using lean product practices changed the Mayo Clinic and established it's digital innovation lab.

Table of Contents

  1. Background
  2. Planning
  3. Designing
  4. Impact
  5. Challenges
  6. Learnings
  7. Closing

Background

In 2018 the Mayo Clinic came to thoughtbot, where I was the NYC Design Director, to learn to innovate. They wanted to learn and ship quickly and solve real problems with software. Myself and Trace Wax thoughtbot NYCs Managing Director, established an engagement with the Mayo Clinic focused on building an innovation practice.

We would develop the innovation practice through building a product. That product would be OnPar; an educational game to improve the continuing education methods for doctors, nurses and residents. This is the story of how we built the product and helped the Mayo Clinic build an innovation practice.

Planning

In our discovery phase with Mayo, we uncovered a problem: current continuing education methods for doctors, nurses and residents were not engaging, were expensive to produce, and mildly effective. With some assumptions, I put together a customer development interview (based off the template below), and set off to interview ~25 individuals in, and tangental to, what we thought our target market was: health care professionals.

The continuing educational space had competition. There was qualitative research that supported our assumptions around engagement, expense, and effectiveness. We aligned that this was a good starting point for our project. As the project, lead I took on many responsibilities:

My Role

  • Lead and execute customer development and research.
  • Facilitate solution exploration and testing.
  • Lead interface and experience design.
  • Establish experiment process, setup and analysis.
  • Develop the product roadmap and manage the project.
  • Communicate with, align, and present to stakeholders.
  • Build, manage, and mentor the design team.

As with most successful projects, it takes a village. I was lucky to be surrounded with amazing thoughtbot designers and engineers, and a Mayo Clinic administrative duo. The Mayo Clinic also provided us with world class Doctors to create the case content, as well as a board to lobby for investment.

Contributors

  • Board
  • Mark Warner, MD
  • Barbara Baasch Thomas
  • Mayo Clinicians
  • Richer Berger, MD
  • David Cook, MD
  • Jane Linderbaum, NP
  • Rozalina McCoy, MD
  • Farrell Lloyd, MD
  • Mayo Admin
  • Jeannie Poterucha Carter
  • Abhi Bikkani
  • Directors
  • Trace Wax
  • Designers
  • Ward Penney
  • Tyson Gatch
  • Engineers
  • Sean Doyle
  • Eric Collins
  • Christina Entcheva
  • George Brocklehurst

For this project I chose a lean product approach. We would discover a problem, identify the business, develop the customer, identify risks and assumptions, then build measure and learn from the solutions you ship.

The first phase of the project was dedicated to refining the problem space, testing the business idea, establishing goals, and honing in on a specific problem set to solve. We used the Lean Canvas method made popular by Ash Maurya.

The Problem

Current continuing education methods for doctors, nurses and residents are not engaging, expensive to produce, and only somewhat effective.

The Goals

  • Create an engaging method of continuing education.
  • Improve effectiveness of continuing education.
  • Reduce the cost of creating content.
  • Have a proven viable business model.

We did a competitive audit to understand what we were up against, and potentially the size of our market. The audit validated our problem space, gave us a sense of market size, and illuminated our unique value proposition: world class doctors creating content.

With an understanding of our problem, our goals, and our competition, we kicked off a design sprint to test our assumptions. The Google Venture Design Sprint emphasizes effective problem identification and iterative testing. It accelerates learning and provides clear direction.

An abstraction of the 5-day Design Sprint Process made famous by Jake Knapp. Created while Jake Knapp was at Google Ventures. This was used to demonstrate the similarities to design thinking and get alignment on using the process for vague problem spaces.

Designing

For our first sprint, we tested on a card game as a solution. We incorporated elements we'd heard from our target customers about what might engage them: solving real cases, pitting doctors against an expert clinician, and professionalism. The initial prototype was a physical card game made using index cards.

We set out to test the card game with 15 doctors in the Mayo Clinic. We ended up testing it with around 40. Doctors were not only eager to play, but they were finding other doctors in the clinic to share their experience with, and recommend they play. We had real world referrals to a prototype. Aside for the engaement, the game answered many of our assumptions and helped us understand short term needs and a potential long term direction.

Our next task was to create a simple digital version of the game and get it out to our target customers to get more feedback. Our team had proficient Ember engineers, and capable of building powerful javascript apps ripe with interaction, so we decided we'd build a web based app..

We prioritized a mobile platform as continuing education for doctors was generally done "between things". Often times on a commute, or between appointments. But rarely was this work done at home. When doctors got home, they were tired and wanted to rest.

Here is the case picking screen. You'll notice the first version is quite bare bones, but offers a few "nice-to-have" features our test groups asked for; the last score a doctor got was displayed, and unplayed cases were separated.

By version 4, we'd made many changes to the case picking screen. The words “real case” meant a lot to the users, so we emphasizing it. By doing so we saw an increase in the number of cases started. However, it also caused a a small reduction in retention. It turns the users now had concerns about PHI (Protected Health Information). We later addressed this on the "meet the patient" screen. Another change was using the Mayo Logo to improve trust. Adding it showed an increase in cases started and retention. Finally, we highlighted "par" (the score to beat) for each case. Doctors told us they were very competitive, and emphasizing the score signaled what to beat. We saw an increase in cases started with lower par numbers, and a reduction of cases started with higher par numbers.

Another important aspect of the game was being introduced to the patient. We followed the Grand Rounds methodology and maintined that language. On the screen there is an image of a patient, and some basic information about that patient. From here the user goes on to ask questions to this patient.

The initial version uncovered some significant learnings; people were concerned the case might contain PHI (Protected Health Information), images of people were not helpful unless providing diagnostic clues, and finally proper domain language was a must. In our iterations, we addressed these issues in several ways. We highlighted the fact that we removed any PHE. We removed all of the images of people and only used relevant information. And finally, we updated all of our language to be domain language

After meeting the patient, we were on to game play. Asking a question (playing a card) added 1 to your score. You'd ask a question by dragging a card to the playing field (middle). Once a question was asked, the card would flip, and the answer exposed. You could freely page through all the questions, and diagnose the patient whenever you wanted..

Through interviews and data we learned a lot, and made adjustments. Doctors were not happy that all questions had the same cost to play - they felt some questions or test should cost more. Adding variable costs to cards increased retention. Our original game mechanics were janky; it wasn't obvious where to drop the card, once you'd asked 4 questions it was nearly impossible to scroll up, and not being able to easily revisit the patient information provided uneccesary friction.

The climax of the game was the diagnosis screen. Here was the chance for the user to make their diagnosis and solve the case. The original screen was quite simple, the user could see par, review the questions they asked, and make a diagnosis. Once they made the diagnosis, they were either correct and celebrated, or wrong and shown the correct answer.

During our in-person interviews there were a lot of “I knew its” and “oh no, it can’t be that” on this screen. There was a lot of emotion from the users. The question most often asked when the diagnosis was wrong was "how did the expert do it?". They were curious what questions would have been a better line of thought. The biggest win we had in the product was adding adding a section that showed the line of questioning an expert took to solve it. By adding this, we say a significant increase in cases played and overall retention.

Ultimately the iterations proved successful as we saw steady engagement, an upwards trend in acquisition, good 30 day retention (adding new cases was the only driver of continued retention), and signal of customers willing to pay.

The Impact

After extensive testing and pivoting, we discovered a successful business model for OnPar. Our shift from targeting individual users to focusing on those invested in learning and credentialing led us to two types of paying customers.

Firstly, nurse organizations desired our platform for ongoing education, curated content dissemination, and tracking participant engagement.

Secondly, larger organizations like the CDC saw the value in leveraging our platform for urgent messaging, improved communication, and more accurate tracking compared to their existing systems.

These distinct customer perspectives guided our final pivot, resulting in a problem-solution fit and a group of paying customers. Moreover, the Mayo Clinic team gained valuable insights in product development, decision-making, customer engagement, iteration, and innovation. This successful product launch marked the establishment of the Mayo Clinic's digital innovation lab.

The Challenges

The biggest challenge with this project was changing the mindset of the Mayo Clinic from shipping something perfect, to shipping something to learn. The two things we struggle with was shipping something scrappy under the Mayo brand, and keeping the board aligned with learning as progress. We focused on metrics often to show progress, but struggled with expressing how important the qualitative learnings were, and that a failed version of an A/B test was a path forward, not a set back. This took a lot of hand holding, weekly meetings, and education. I leveraged the Stop, Pivot, or Preserve framework framework to disclose learings and progress, as well as force alignment and decision making as a team.

Learnings

Our key realization was that we initially targeted the wrong paying customer for the project. While doctors and residents found value in the product, they were reluctant to pay for it. Nurses were willing to pay, but we couldn't sustain their engagement beyond 60 days. Ultimately, department heads emerged as the paying customer, benefiting from the product's insights into residents and nurses, enabling them to address knowledge gaps and track performance in their organizations.

Closing

The collaboration between thoughtbot and the Mayo Clinic resulted in OnPar, an innovative educational game for healthcare professionals. Through a lean product approach and iterative design, the team addressed the shortcomings of traditional continuing education methods. OnPar successfully launched, attracting paying customers such as nurse organizations and larger institutions like the CDC. project established the Mayo Clinic's digital innovation lab and provided valuable insights into product development, customer engagement, and innovation in healthcare. OnPar represents a cultural shift in the Mayo Clinic.

😷