In Episode 38 of 15 Minutes with the Doctor, Vinay is joined by George Batchelor. He is the Co-Founder and Director of Edge Health, a data-driven agency that helps healthcare providers be more efficient through better and more intelligent use of their data. The company has various data products that have supported multiple hospitals, charities and supported nationally with Covid-19 analytics.
Edge Health works with each client to understand their problems and tailor data solutions. Output can be customised to be displayed in a way that makes the most sense for the organisation, whether that is an interactive dashboard, intelligent scheduling, or reports. Today, we share an unreleased clip from the episode for Once Daily readers about the approach they take on using artificial intelligence in their company:
So tell me about the AI and the machine learning being used in your services. Could you share any information on that?
Yeah, what type of information, like the type of approach we take, or…
Yeah, a bit more about the approach. How long it took you to develop or how much of a dataset there is at the moment? Is it something you’re continuously developing with hospitals from around the UK?
I think 70% of the time is spent cleaning and linking the data so that you can then analyse it quickly. Again, with building trust in the data. In terms of the analytical approach to making the predictions, I think that was something we went through a few iterations with the hospital. Initially, we were looking at the data and using medians and averages so that they could look at it and say, “That’s about right.”
But what we found is that, as you said, there was just so much complexity behind that. A machine learning algorithm that took into account all of that data and made sense of that actually gave a better output.
The approach that we took is not one of the more complex neural network approaches. It’s all very intuitive. It takes all the information, but it doesn’t do any weird transformations with it. So in that sense, it’s very auditable – We can explain to a clinician or a booking team why one thing has predicted this and that another thing has predicted that. So, they can see and understand it.
Usually, we’re very conscious of all the data and information governance requirements. So when we do one of these projects, it’s very much for a specific hospital trust, and everything we developed with that trust would sit with them. Their data would never leave beyond the work that we do with them. But, because most hospitals are collecting these datasets for so many years, there’s always been enough data in all the hospital trusts for us to develop it afresh. It enables us to take the learnings of the methodology with us but apply it somewhere else.
I guess one of the things that we always hope is that this would be a hugely scalable approach to work. In reality, what we find is that most hospitals have many different systems configured in many different ways, and the data is recorded in slightly idiosyncratic ways, even between teams, let alone hospitals. So, there’s always a fair amount of data cleaning and data processing that’s needed for every new kind of project we do.