Scarcely a day goes by without fresh statistics on the number of Covid-19 infections in the UK. This morning saw the publication of Round 6 of Imperial College’s REACT-1 study of coronavirus infection rates. Interesting and concerning reading it is too,
But the REACT-1 methodology is based on sampling. Round 6 found “863 positives from 85,971 swabs”, so the error bars are wide, especially at a regional level. For example, the R for London was found to be 2.86, but with a confidence interval between 1.47 and 4.87. Hold the front page!
You hear epidemiologists saying they don’t rely on the daily counts of cases to judge the progression of the epidemic. They give more weight to hospitalisations and so forth, and sometimes burble about changing numbers of tests.
But other indicators are never as up to date, It seems to me that the full daily count has got to give the earliest indication of in changes in the infection rate, R. The availability and uptake of tests has certainly changed over the course of the whole epidemic, so that spring and autumn figures are not comparable, but it doesn’t change much from week to week.
So I’ve been looking very closely at the daily case numbers. Here’s a snapshot of them as of 4pm yesterday, Wednesday 28th October:
A bit fuzzy, sorry, but hopefully clear enough to see what’s going on.
The headline writers focus on the bottom graph, cases by date reported, But that does fluctuate depending on whatever is going on behind the scenes. For example, the spike to 22,961 on Sunday 4th October was due to the famous error copying data into an Excel spreadsheet. More mysterious is the spike to 26,288 on Wednesday 21st October. I noticed that a larger proportion than usual of that Monday’s (19th) cases were reported that Wednesday than usual. Because of the spike on 21st there was a sudden dip yesterday, Wednesday 28th, in the UK headline 7-day case growth rate on the summary page, to 13.7% (when Wednesday 21st suddenly became “last week” and not “this week”!):
The figure was 21.5% the previous day! I’d very much like to know what happened to cause the blip in reported test results last Wednesday – perhaps a backlog was cleared somehow – but that’s not the point of this post.
It seems to me that test results by specimen date, rather than reported date, give a more reliable indication of short-term changes in infection rates. So I’m going to focus on the top part of Fig 1.
We need to note a couple of other points about the data. First, Fig 1 includes data from all UK countries. I want to focus on England, the blue part of the bars in the top part of Fig 1. Second, the UK’s testing is slow, a massive problem in itself, but again, not the point of this post. Test results take up to a week to come through.
So, to answer the question in the heading, we can estimate the weekly growth rate in cases in England by comparing the blue parts of the bars in the top part of Fig 1 (you can do this yourself at home by hovering over the bars at https://coronavirus.data.gov.uk/cases).
Here’s what we get:
|7th Oct (Wed)||15,182||14th Oct||16,420||8.2|
|8th Oct||15,247||15th Oct||15,237||(-)o|
|9th Oct||12,999||16th Oct||14,517||11.7|
|10th Oct||10,515||17th Oct||12,324||21.4|
|11th Oct||9,643||18th Oct||11,823||22.6|
|12th Oct||16,178||19th Oct||22,030||36.2|
|13th Oct||15,429||20th Oct||21,276||37.9|
|14th Oct||16,420||21st Oct||21,323*||29.9*|
I’ve asterisked the figures for 21st October as provisional, since there may be a few test results to come through for that specimen date. [Later note: actually not enough to change the percentage increase significantly].
It seems, though, that there’s been a dramatic increase in cases of Covid-19 detected in England, starting in mid October. The rate of increase may be slowing a bit, though it’s too soon to say for sure.
The period analysed in Table 1 overlaps with the latest React 1 data, which is for swabs collected between 16th and 25th October (and estimates R over that period).
Why this has happened, and why the number of cases was apparently growing more slowly in the first half of October is hard to say. It could be that more older people were catching Covid-19 later on, with more symptoms and therefore more likely to be tested, Or it could be that government restrictions reduced case growth rates in some areas of high incidence, but that as time went on this was swamped by the continuing increase in cases in the rest of England. Or it could be the weather!
It’s a shame there are no longer daily press conferences to help us all understand what is really going on.