10 things we learned from choosing chatbots as a theme for our rare disease hackathon
Description of the natural history of ultra-rare paediatric tumours
This weekend the Share4Rare hackathon RareHacks took place in Barcelona. During these three days, over 45 data scientists, computer scientists and clinicians joined forces in response to the needs of the rare disease community, focusing on paediatric melanoma. During this short, intense period of time, 6 teams created the fundamentals of a chatbot to solve rare disease challenges.
Let’s take a moment just to recap all that we have learned:
1. Choose your time wisely
Usually, rare disease parents can make time on Friday and on the weekends to get online and answer questions when you reach out to them. This generates a fast turnaround and quick answers to your questions.
2. Time tag everything
The field of rare disease research advances fast. It can occur that a therapy already is approved but not listed as such, or new trials are missing. It is essential to date every piece of rare disease data you find, so patients know the information is up to date.
3. Validate your entry points
One of the biggest challenges in rare disease hackathons is to find enough and validated sources of data. Although Orphanet is a large rare disease database, it misses rare early onsets of adult non-rare conditions. Unfortunately, this makes up 80% of paediatric melanomas.
4. Be aware of possible mismatches
Especially in the case of melanoma –the rare condition we were working with– symptoms can go anywhere. This makes symptoms hard to classify. Rule of rarity: if it walks like a duck and talks like a duck, it still might be something else.
5. Do not reinvent the wheel
Especially in a pressure cooker hackathon where time and resources are scarce, you do not want to invest too much in writing your own code. Tap into Python’s existing libraries or use scrapping to collect and combine current knowledge to advance faster.
6. You don’t know what you don’t ask
Rare disease parents sometimes do not know what to ask. For example, you cannot ask for clinical trials if you are not aware they exist. A chatbot also serves an educational purpose: to anticipate the next step and pave the way.
7. Build on empathy
Yes, rare disease families want answers. However, if you push an answer too fast (for example, by providing information on clinical trials when they just received diagnosis), a chatbot might come over as insensitive.
8. Define roles & set expectations
You need at least one data scientist, clinical researcher and a computer scientist to collaborate in one team. To avoid hitting the development wall, a clear task definition or appointing a project manager might come in handy.
9. Engage everyone
Whether this is a hackathon newbie who might think they are too inexperienced to contribute or programmers who usually work in another language, there is always a way to involve the whole team in the process.
10. Incorporate clinical knowledge
A patient representative or clinical practitioner can help define which part of the challenge your team can focus on and can help to deal with medical information or terms. Especially at the start of the hackathon, ask as many questions as possible to gain a deeper understanding of the subject.