A month or two ago I started modelling the reproduction number of COVID-19 using the R0 package. I then used this model to create a Shiny app to break down the data by country. Can we use this information to predict the peak in the UK?
For a full detailed explanation see the original article here. But the gist of the process is:
- model the UK case data using the R0 package
- fit a linear model to the tail
- use the model the estimate when the effective reproduction number reaches 1
Here is the current case data as of the 9th of April:
One might assume that we’d peaked from this but that would be entirely guesswork given the uncertainties of testing and reporting.
Let’s look at effective R. This value is greater than 1 before the peak and less than 1 after the peak. Here’s our estimate of R over time:
At the time this snapshot was taken it’s 1.5225587 which is still above 1.
So, when will the estimated R line cross the 1 line? Now the tail of the graph looks pretty linear, so let’s try a linear fit on the last 10 days:
From this fit, we can estimate that R will be 1 on day 81 or thereabouts. So about 12 days from the time of the snapshot, which would be roughly Wednesday 22nd April.
This is, of course, a very rough estimate based on several assumptions, particularly that the current measures keep the UK on its current trajectory. Also, it is assumed that the reporting of case numbers is accurate — given the scarcity of testing in the UK that is in some doubt, and even accurate numbers aren’t always reported in a timely fashion.
On top of that, this is under lockdown conditions. What happens when the lockdown is lifted?
Originally published at https://cosmo-opticon.com on April 10, 2020.