Nouveau Teaching Ventures in the Digital Era

Nadine Housri, MD, and Erin Gillespie, MD, share a passion for sharing knowledge. Each has produced an educational instrument used by tens of thousands of physicians around the United States and beyond.

Dr. Gillespie founded eContour in 2015 as a free, online, case-based interactive contouring atlas of value to trainees and seasoned practitioners of radiation oncology. Usage skyrocketed in the first five years of its existence and has continued to grow steadily since, with literally hundreds of thousands of case views annually. Founded by Dr. Housri in 2014, theMednet is an online question-and-answer forum where experts from a wide assortment of medical specialties answer clinical questions, filling the gap between textbooks and real world experience. Its database is easily searchable, and to date nearly 40,000 physicians have sought advice via theMednet. We caught up with Drs. Gillespie and Housri to learn how those initiatives came into being and became so successful.

ASTROnews: Could you share the journey that led to the creation of eContour? What specific challenges in radiation oncology education and practice were you aiming to address?

Erin Gillespie, MD

Dr. Erin Gillespie: As a resident, if I had a question about how to contour a new case, my senior residents would point me to a published guideline or a prior case contoured by my attending. Similarly, my mentor Jim Murphy, MD, MS, had received at residency graduation a CD (yes, a physical one!) of anonymized H&N cases contoured by his attending for future reference. Around this same time, our research group was identifying increasing differences in patient outcomes based on radiation oncologist case volume in the IMRT era.¹ We also were observing that contouring guidelines were becoming increasingly available, but they were often complex and not particularly easy to access at the point of care. We therefore asked if generating a database of high quality, guideline-informed cases that could provide quick reference for residents and less specialized practicing physicians, could potentially address a gap in knowledge and skills underlying differences in patient outcomes.

AN: What overarching goals have driven the development of eContour, and how do you determine which features or content to prioritize to best serve the radiation oncology community?

EG: eContour has always been driven to serve the needs of busy radiation oncologists trying to deliver the best quality of care to their patients. We have surveyed our userbase every few years which has further informed prioritizing expansion of high yield case content, as well as website speed and ease of use. We have been asked many times to add full sagittal and coronal images but have declined to do this without technology to make image loading acceptably fast. We did just this past year establish a collaboration with researchers at Massachusetts General Hospital to leverage Open Health Imaging Foundation (OHIF) technology to address this image viewer challenge, which is funded through an NIH U24 grant. So more to come on that!

AN: What business models did you consider to make eContour self-sustaining? Are there future business strategies you’re contemplating?

EG: Early on we considered establishing eContour as a company. When we approached UC San Diego’s Office of Innovation and Commercialization in 2016, they thought it best represented scholarly work of an educational nature, akin to a book or journal. This combined with a fairly limited market for rad onc education made it an easy decision to keep the site’s goal of open access education, which aligned best with our personal priorities and interests. We have been fortunate to receive educational innovation grants, especially from the Radiologic Society of North America, and to establish a partnership with the American College of Radiation Oncology to award 10 travel grants per year for case contributions, with funding provided by MIM Software. This is our current approach, though we have filed as a 501c3 non-profit to keep open the possibility of soliciting donations like Wikipedia, or implementing a fee-based model based on ability to pay. We are, though, still a couple years from needing to go in this direction.

AN: Throughout eContour's evolution, what significant challenges have you encountered, and how have you adapted your strategies to overcome them?

EG: Figuring out how to motivate physicians to participate in activities like education and quality improvement that extend outside their busy jobs has been an ongoing challenge. Early content development and review fell mostly on me and the dedicated attendings at UCSD, led by Parag Sanghvi, MD, MSPH (who was critical in facilitating our robust set of H&N cases at the start). This became unsustainable, so in line with the approach of running the website more like a journal, we developed an editorial board structure. The board enabled us to highlight the work of the dedicated few, provide professional opportunities for growth and managing others, and establish a platform for social accountability by meeting and getting to know each other. With a few early dedicated members like Mike Sherer, MD, and Jeremy Price, MD, PhD, we have been fortunate to build up a robust and committed team that curates submitted cases, seeks expert peer review, etc. And on the case submission side, we partnered with a few different professional organizations, but ultimately ACRO was the most successful and sustainable. Their funding supports travel to their conference, which serves as an educational and networking opportunity for residents. I think this speaks to the importance of professional development and interpersonal relationship building as a motivator as much if not more than the money itself.

AN: What strategies have been most effective in translating your vision for eContour into a practical tool that enhances clinical practice? Are there specific metrics or feedback mechanisms you use to assess its impact?

EG:Remaining responsive to our userbase has made it easy to feel we are achieving the vision. Our main metric to monitor progress has been website analytics (number of users and especially number of case views per day) as evidence of utility. Creating something that people actually use is hard, so as long as we continue to grow our usage, we know we are on the right track. In our initial randomized study, we asked participants to rate the website using the System Usability Scale, which enabled us to compare usability to a contouring textbook and show superiority.²

AN: Given the need to integrate clinicians and computer scientists at eContour, how do you foster collaboration among the eContour team and contributors, and what approaches have you found effective in maintaining motivation and alignment with your strategic goals?

EG: It has definitely helped to develop an editorial board, which has a standing weekly meeting. Our programmer, Scott Lundy, sometimes attends this meeting. We are very lucky to have found Scott - I was a new attending in NYC, and he was a Cornell graduate whom I found through the Cornell computer science department. Scott is an independent contractor, and one thing that has helped maintain his commitment to the project is that we almost never ask for things done urgently, which helps accommodate his other jobs and life. He also appreciates the mission and visible impact.

AN: With rapid advancements in technology, such as AI and machine learning, how do you envision these tools integrating into platforms like eContour to further support radiation oncology professionals?

EG: There are certainly increasing opportunities here, particularly for partnerships with industry to develop and pilot test solutions. One angle I’ve been thinking about for a while is how to test a physician’s ability to detect an inaccurate contour, as our role moves from active contouring to contour editing/approval.

AN: Reflecting on your experience with eContour, what key lessons have you learned about bridging the gap between strategic planning and effective execution?

EG:A wise mentor, Dave Fuller, MD, PhD, once told me it was not worth worrying about what the website would be more than five years in the future, because everything can change in that time. But Dave is also one of the biggest proponents of goal-setting. I think the balance is to establish a vision and set short-intermediate-term goals that will encourage focused efforts to accomplish that vision but still allow yourself to take risks. I think as physicians we are trained to think so long-term, for example how our specialty choice will impact our life in 30 years, which can end up being somewhat counterproductive.

AN: Can you tell us how theMednet came to be?

Nadine Housri, MD

Dr. Nadine Housri: After I started my internship, my dad was diagnosed with cancer. To be sure he was getting the best care, I started reaching out to experts at top cancer centers around the country. I sent some emails and spoke on the phone with experts, and their advice was incredibly helpful. He received excellent treatment and is doing very well years later. I would have left it at that, because this is something we do all the time as physicians: we ask experts our questions. But I happen to have an older brother whose background is in technology, and he started asking about how experts share knowledge. He pointed out two things that I had taken for granted. The first is how much expertise is simply stored in a few people’s brains! The know-how of what to do when there are no clear guidelines is trapped in their heads, because there's only so much that's published or presented at conferences. Second, when this type of knowledge is shared, it's often shared privately, not in a way that helps people who are not in the middle of this expert-to-clinician conversation. And so, we thought, let's put all this online. Ask experts questions and share answers in a private physician platform, so that previously one-on-one conversations become part of a searchable repository of answers for physicians who might not have expert access otherwise.³

AN: What is the overarching vision driving theMednet, and how do you ensure that your team stays focused on the most critical priorities amidst competing demands?

NH: Our vision is to answer every doctor's question. The way that the team really stays focused is that the Mednet's first value is “Doctors first.” And it's not just a platitude, it's not just something put up on the wall for people to think, "oh, that sounds nice." We live by this value. Every decision that we make goes back to this question: Are we helping physicians make better decisions and take better care of their patients?

AN: Can you share an example of a significant strategic pivot you've made during the history of theMednet?

NH:We’d been working since 2018 to raise awareness to oncology clinical trials through theMednet with grant support through SWOG and the National Science Foundation. Then, two things happened in 2020. First, we applied for the second part of a National Science Foundation grant, which would have given us resources to build machine learning technology to match clinical trials to a clinician’s questions. Unfortunately, we didn't get the grant — and at the same time, COVID-19 happened. To me, COVID was a wake-up call and showed how much of a need there was for theMednet outside of oncology. And so with this clarity, we started expanding to include more and more specialties within internal medicine and beyond into dermatology, neurology, psychiatry, and even a few pediatric specialties. Today theMednet exists in 15 specialties across medicine.

AN: It seems that you were thinking of some AI ideas before ChatGPT became a household word. Do you see a role for this or another emerging technology in your short- or long-term planning?

NH: Yes, the vision has always been to answer every physician’s question. We started with the most difficult questions — those that are not answered by textbooks, guidelines, or UpToDate. The idea that a clinical trial would be suggested as an answer to a clinical question came about before the emergence of large language models (LLMs) and required older machine learning technologies. With LLMs, this is not only easier, but the results are better. We were recently awarded a $2M contract from the NCI to build this technology, and we already have a prototype that matches patient cases to appropriate clinical trials. We will also be integrating traditional sources of knowledge into theMednet like guidelines and papers. So stay tuned!