AvaMD talk about their winning idea from the Jisc Summer hackathon….
“As medical students, we face difficulties in our training. When it comes to integrating the theory of conditions we learn from textbooks into the real world practice of communicating to patients, taking an effective history and formulating a list of differential diagnoses. After all, patients do not attend hospital or GP with a diagnosis label on their head! However, the only way to get good at this is through years of clinical exposure with patients in wards and clinics. For students with limited exposure to patients during their placements, we wondered if there was a way to augment and hone this clinical skill through digital means.
Both of us are medics and self-taught developers, with a particular interest in Natural Language Processing (NLP). We have tinkered around with a few possible solutions of recreating the patient interaction experience for a student to use as a training tool before, but unfortunately, it was only paper napkin ideas until we got an email from our university newsletter advertising a hackathon with Jisc at CAN 2019. We jumped at this opportunity in excitement, and applied in the hope that this could be the chance for us to spend a focussed 48 hour period to finally hack together a solution!
We were thrilled to be accepted and arrived at the hackathon with eagerness to begin. The Jisc
team were fantastic in getting us to settle in and provided a supportive environment from the very start.
During the first day, progress was painfully slow. We realised the problem we faced was bigger than anticipated. We had to spend hours breaking down the key components of a patient consultation so that we could simulate a realistic conversation flow between clinician and patient to then implement into an AI chatbot. We quickly discovered that this was a huge challenge; conversations are highly complex and doctors and patients can say a huge variety of things! In 48 hours, we simply could not account for all the different possible questions a student can ask. We went back to the drawing board, and asked ourselves, how can we create something that is less complex to be able to demo in the next 24hrs yet still be of immediate benefit to students?
We found our answer to be one of the core skills all doctors need: dealing with emergency scenarios. There are national algorithms that outline how clinicians should approach and assess critically ill patients. There are key steps to be taken, that cannot be missed. We decided to translate these algorithms into our chatbot and create a simulation of a critically ill patient where the student must progress through key steps of assessment in order to successfully stabilise and treat their patient.
On day 2, we carefully designed and built our chatbot to be able to prompt and correct the student should they make an error in their assessment, which is essential for trainees to improve their skills. After vigorous training of the chatbot’s AI engine, we finally felt ready to demo our product! After some quick brand designs and a slide deck, we took to the floor and shared our vision for a future where interactive and intelligent clinical medical education is accessible to all.
… And we won! We were overwhelmed by the response from the audience, with many people approaching us after to try the product themselves. We were up against the fierce competition, so this was a huge but pleasant surprise for us. We were so excited that we continued our pace straight on after the hackathon and put together a site (talktoavamd.com). We then fleshed out our chatbot with a few more scenarios and headed down to the medical school library to get some real user feedback of medical students using our product. We filmed their reactions and experience which you can view on our website.
What we learnt hacking Ava MD:
- Starting simple is a good strategy. Obtain proof of concept before exploring and building complex scenarios.
- Testing what you build with your target market is the next step post-hackathon. We learnt a great deal after giving it to a bunch of students for a few hours
- Ava simulating algorithmic emergency scenarios could help medical students hone the skills that save lives.
We believe the use of voice as a training tool is underused, and we at Ava MD hope to bridge this gap and rejuvenate a stagnant medical training sector.
If you’d like to learn more, visit https://www.talktoavamd.com/
Twitter @abdxl_m and @yasmin_abedin
Abdel Mahmoud and Yasmin Abedin