Case Count Heads-Up

New first episodes and new reinfections: a case count curveball (rev 1.1)

My modelling of new daily cases by specimen date using Cochrane-Orcutt estimation to account for serial correlation has thrown up some interesting results in more ways than one. One issue that got me tugging my beard is the extraordinary case count during December 2021, that may be squinted at here. This explosion of COVID gave the model a headache and so once again I ask the question… is this data real?

By ‘real’ I mean does it really mean what we may logically think or does it mean something that we haven’t fathomed? Some twist of definitions perhaps, or a methodological nudge that once again fails to get the publicity needed to avoid the public going down the wrong track.

We all know that a COVID ‘case’ isn’t a medical case proper, being a term used to denote a positive test result (which may or may not be erroneous, and may or may not point to a viral presence capable of infecting somebody else). Young folk at Mrs Dee’s school cottoned on quickly how to become a ‘case’ if they wanted to bunk off, and we receive phone calls almost daily from friends and family who’ve apparently “got COVID” for the second time this month. We always ask them how they feel, and after the usual, “well, I’m feeling fine, I’m just a bit annoyed I can’t get to the sales”, we point out that they are suffering from a positive test result. This usually goes ‘in one ear and out the other’ as my grandmother used to say.

With this pantomime in mind I rummaged through the UK GOV pantry again to see if any new variables had emerged and spotted two beauties:

• newFirstEpisodesBySpecimenDate

‘Daily numbers of new first infection episodes of COVID-19. Data are shown by the date the first sample that identified the infection was taken from the person being tested.’

• newReinfectionsBySpecimenDate

‘Daily numbers of new reinfections with COVID-19. Data are shown by the date the first sample that identified the reinfection was taken from the person being tested.’

A few minutes later I was looking at this graph:

There is that December explosion, with new daily cases exceeding the number of new first infection episodes. Interesting!

N.B. Replace the phrase ‘infection episodes’ by the phrase ‘positive test result’ and you’ll be paddling up the right creek since an episode in NHS-speak refers to a formal consultant or hospital episode that suits in admin keep track of for contractual purposes. Once again UK GOV groupspeak experts have been at the brandy.

Note the sudden appearance of reinfections during December 2021. The data record tells us these didn’t happen before this point in time, which is odd. The eagle-eyed subscriber will note the gap between new cases and first episodes looks suspiciously like the number of reinfections and they’d be right: if I sum first episodes and reinfections I get the exact case count.

Whilst there’s nothing wrong in this (it makes a great deal of sense) it does provide a fine example of how we can misinterpret what UK GOV numbers actually mean. It also explains why Cochrane-Orcutt modelling process found the winter 2021/22 season tricky to negotiate. Consequently, I shall go rummage some more to see what other goal posts have been quietly moved.

As time goes by the green line is going to bloom, and the red line may well dribble along in total obscurity. From the political point of view this matters not one jot for they will always be able to sum the two series to produce a monstrous blue line to whip the public into perpetual submission. COVID may well disappear from the radar but we wouldn’t know it!

Kettle On!