LINKS
Machine learning in medicine
I really enjoyed this webinar, featuring Andrew Ng, Eric Topol, Lily Peng and Pranav Rajpurkar. Included some pretty great points about current state of AI in medicine.
Stanford released an “AI for Medicine” course on Coursera this week. I haven’t tried it yet, so can’t vouch first hand, but it’s taught by Andrew Ng’s PhD student and expect it will be great. (I’m hoping it’s accessible for people without much technical experience, but will have to check it out first to see)
I’ll be talking at the London.AI livestream on 30th April about AI in healthcare for those interested - tickets available here.
General
One of my favourite newsletters is The Hustle. It’s a daily fun read that updated updates you on business-ongoings. I’ve found it particularly helpful recently to keep up with the business impacts of coronavirus.
Maximising our positive impact
I’m a big advocate of the idea that we should seek to maximise our positive impact. This was a big factor for me when deciding to pursue skills outside of clinical medicine (in my case, data science skills).
I’ve been quite happily moving down this track for the last 3 or so years, but with a change in our current global context, I feel it’s worth re-visiting the subject.
The tricky thing with seeking to ‘maximise our positive impact’ is just how hard it is to predict the outcomes of our actions. The best-meaning actions may inadvertently lead to positive outcomes and vice versa.
One principle that is often raised is that of scalability. The argument goes like this: as a competent doctor, I can look after a ward of around 30-40 patients (as part of a team). As a doctor with excellent data science skills, I can work on data science projects that may affect many thousands, or perhaps magnitudes more, of patients.
However, as per the challenge of quantification, I can’t guarantee that my data science project will have a positive impact.
(If interested, I’ve written more about the scalability argument here.)
In the context of COVID-19
Now, with a global pandemic at hand, I would argue that two of the most valuable skillsets are clinical medicine (to treat patients) and data science (to track trends, and guide decision-making). (A key third would be to research treatments and a vaccine.)
The question that I want to figure out in this email is what is the best use of my skillset at the current time. Of course, this will be specific to me, but I hope there are some general useful principles. (I would also love to hear from anyone reading this who has any thoughts or suggestions.)
As I see it, the main ways that I can help are:
Come up with original ideas (such as this idea [https://www.telegraph.co.uk/news/2020/04/11/doctors-create-pioneering-reusable-facemasks-solve-ppe-shortage/])
Pick up shifts as a doctor
Contribute to projects applying data science to tackle COVID
Continue to develop my skills with a view to working on things later
Some thoughts on each:
(1) Coming up with original ideas can be difficult to force. I find spending time brainstorming can help, but that the best ideas tend to arise ‘organically’. So going forward, I will try to think of ideas, and pursue any that I do, but until I do this can’t be my primary strategy.
(2) There will be an increased demand for doctors, but there have also been many more recruited from other specialties, so the marginal impact of one additional doctor is unclear. (My impression, and those of colleagues, is that equipment, ventilators and number of ICU staff are greater limiting factors)
(3) I’ve seen a huge number of projects advertised but (i) I’m not confident that my data science skills are sufficiently advanced to have a huge impact on these projects and (ii) it can be challenging to identify which projects will have an impact and which won’t really. There’s a pareto principle at play; a small number of high-quality pieces of work will impact governmental policy, etc.
Deciding between these based on expected outcomes presents the challenge of quantification that I mentioned earlier. Given this, I feel the best approach would be to combine working part-time as a doctor, whilst also continuing to develop my data science skills so that I can work on useful projects in the future. It would be short-term thinking to stop developing the skills which I hope can scale up my impact in the future. (And, where possible, I can seek to develop my skills while working on COVID-related problems.)
On non-scalable impact
One other thing I believe worth noting, is that we shouldn’t underestimate the impact we can have in “non-scalable” ways; by being supportive friends and family members, and generally being there for people when they need us. Many people will be having a tough time for months and years to come, and we can do our part by being there for them.
What are your thoughts? How are you hoping to use your skills to help in the ongoing pandemic? As always, feel free to reply.
Have a great week!
Chris