Read on Twitter
Since March Joe (+ a team) have been modelling the spread of COVID, I feel like ‘the model says’ has been said every day on the news but they hardly ever explain how they work so here is my (very basic not science) explanation if you’re interested (please be lol) A THREAD 1/16
So, firstly they created a computer version of what the England looks like normally, they looked at census data to create households, worked out where people commuted to each day and how many contacts they had with other people etc. (this came from BBC surveys) 2/16
They then added in what people do in their spare time and how many people they see then. This data came from national free time surveys . Once this was done they had a pretty good idea of what everyone of the 60 mil people were doing each day. 3/
16
Highlights from this time for me included learning lots of fun facts about census data (did you know the East Midlands is the only place with male midwives?) and hearing about how hard it was to get commuting right in the code. 4/16
Then they had to add in the virus and look at the probability of people getting infected based on the number of interactions they have with people each day. 5/16
From this they can see how quickly people will get infected based on one person having it and going about their every day life. 6/16
Other things were added into the code here like the number of people who were likely to die from COVID if they caught it (this must have been really emotionally challenging to code) and the fact that children are less likely to be infected. 7/16
Then they used the model to see if the government policies were working. So they added in lockdown, it was important to check how many key workers there were and how many people couldn’t work from home. This was added into the code. 8/16
They also used survey data to add in the compliance rate of people following the guidelines, self isolating when asked etc etc. 9/16
This was all fine (from my perspective anyway) until policies became a bit more varied and this all had to be adapted in the code - eat out to help out? Apparently quite difficult to code (although they managed it bc dream team) 10/16
Then policies changed again and then they adapted their code again, lots of adapting, lots of coding (not much sleep tbh). 11/16
The code was able to tell them how many people were likely to be infected, how many people would end up in hospital, how many in ICU and ultimately if there would be enough hospital beds for those people. 12/16
It was then really important to check the results of their code against the hospital data, it fits almost exactly - the black line is data and the colours are the model. This means that they can use the model to predict what will happen in the future 13/16
This code is, in my opinion, amazing and it has taken so much time (Joe puts so much work in & its only a fraction of the team overall) it also must be so hard - they are literally predicting how many people will die. So I'm v proud and I hope you’ll humour me this thread. 14/16
For a more sciencey thread on this with more graphs see here https://twitter.com/arnauquera/status/1339581857147469824 15/16
And for even more science, the paper is here https://www.medrxiv.org/content/10.1101/2020.12.15.20248246v2.full.pdf (16/16)
