Why case numbers don’t tell the whole Covid story

In an article for Computer Weekly this week, I highlighted four different official numbers for UK Covid-19 deaths on 30 April, from 548 to 769 (on the day, Boris Johnson said 674). The graph below shows how much these four measures vary.

But with deaths thankfully much lower now than in the spring, the focus is on coronavirus case rates. This is unfortunate, because as a measure they are seriously flawed. Leaving aside the loss of nearly 16,000 cases due to incompetent use of spreadsheets (this is not a new problem with Microsoft Excel) there is the issue that more testing will almost certainly reveal more cases, and UK testing has quadrupled since April. Better availability of testing in an area, a good thing, could well lead to it having a higher case rate.

Regardless of this case numbers are rising very fast in areas including Manchester, but even there the story is more nuanced than it looks. The city has a case rate of just over 500 per 100,000 people, but scaling the highly-localised case data for 28 September to 4 October by population shows a huge range, from less than 250 in Castlefield and Deansgate and East Didsbury to above 4,400 in Fallowfield Central, which records 528 actual cases.

It doesn’t take former University of Manchester staffer Alan Turing to crack this code: Fallowfield is home to the university’s main accommodation site, which one parent quoted by the Manchester Evening News called “Covid soup”. Most students in their early 20s are, thank goodness, at low risk if they catch the virus. But case numbers treat students and the much-more-at-risk elderly the same.

The experts I spoke to for Computer Weekly recommended monitoring a number of indicators to get a clearer picture. If you insist on a single source of data to guide you, expect to be misguided.

This has also been published through my monthly newsletter. Sign up below.