In a previous post, I attempted a very rough estimate of when the peak for COVID-19 in the UK would be over. I reckoned it would be around April 22nd.
I calculated this by doing a linear regression on the effective reproduction number as calculated using UK case data and the R0 R package. I could then work out the intercept with the R=1 line, which is the inflexion point at which the epidemic should begin to recede.
It’s a calculation that’s full of caveats and assumptions: that the trajectory of the reproduction number is the same, that the data is accurate in the first place. I have just run the numbers again and the predicted intercept has been pushed out the May 1st. What has changed?
It can just about be seen that the downward slope of effective R has flattened slightly toward the end. This will push out the intercept with the R=1 line further into the future.
The daily increase has flattened out. Naturally, we’re all about flattening the curve, but without mass testing can we be sure?
One thing that seems clearer to me, looking at this and daily increase charts for other countries, is that the idea of a peak is an idealised picture. More realistically there is a plateau. The estimate I presented above points at the end of the plateau rather than the single day with the most number of cases.
If we look at Germany we can see that the peak may be past, even though there are still thousands of cases a day:
Running the numbers gives us an estimate of 8 days before the peak has passed. Indeed Germany is planning to allow small shops to re-open on April 20th.
Anyway, so that I don’t need to keep recalculating this figure with the latest data I have added it to the COVID-19 Explorer, which presents all these calculations broken down by country. It now includes a section on the estimated end of the peak.