One of the problems caused by a disease going hot is hot stuff bias. This is simply human nature at work. If everyone is talking about covid, then every respiratory patient starts to look like a covid patient. Add in a covid test that is so sensitive it can detect viral dandruff left behind from past infections, and the conditions are right for a form of diagnostic drift, which happens when a diagnosis accumulates patients that don’t properly belong to it. Is there any evidence this might be happening with covid?
The answer is yes, even if there is another plausible explanation, which Dr No will come to shortly. ONS have reported weekly deaths for England and Wales from all pre-covid respiratory deaths for the last decade, and the image for this post charts them for each year. Note that these are all respiratory deaths, but a good many — perhaps one third to getting on for half or more in a bad flu year in winter — will be influenza and pneumonia deaths — see footnotes for some background details. What we see is that for every year apart from 2020, the pattern is similar and familiar. Putting aside the year start/end effects caused by incomplete weeks, the first 20 weeks of the year are ravaged by flu, then things settle down over the summer, only for mortality to rise as autumn and then winter arrive.
But if we look at the 2020 line, red and a bit heavier than the other lines, something different happens. Up until week 20, it is lost in the general noise, but over the summer it is a little bit below normal, and then, come the autumn, the normal and expected upturn does not appear. Respiratory deaths continue to flat line, and then show a very unexpected fall from about week 47. Very recent weeks may have incomplete data, but most of that fall is likely to be real.
There are two plausible explanations. The first is hot stuff bias and associated diagnostic drift. Patients dying from influenza and pneumonia are incorrectly being classified as covid deaths, perhaps based on clinical suspicion fuelled by hot stuff bias, or by a loosening of diagnostic criteria, for instance, by finding some historical viral dandruff, without properly considering the clinical picture. Given that there are, by mid December, some 500 or so ‘missing’ weekly respiratory deaths, with many of them likely to be ‘missing’ excess winter flu and pneumonia deaths, it is entirely possible these normal winter 2020 excess non-covid deaths have been incorrectly given a covid cause of death, so inflating covid deaths.
The other plausible explanation is that they are indeed all or mostly covid deaths. What has happened is that covid, for want of a better phrase, has grabbed a number of patients who would in pre-covid years have been influenza and pneumonia deaths, but have in 2020 become covid deaths. These deaths, though from covid, are part of the normal winter excess deaths, and not excess covid deaths. They are part of a normal pattern, who just happened to succumb to covid, rather than another respiratory infection.
We can speculate until we are blue in the face — perhaps a rather too apposite turn of phrase, given these are respiratory deaths — about what happened to those desaparecidos respiratory deaths in late 2020. Some will even suggest they weren’t true desaparecidos, they were instead really disappeared, or rather prevented, by anti-covid measures, which worked to reduce all contagious respiratory deaths. But Dr No thinks that is unlikely, partly because he doubts the general effectiveness of the measures, and partly because old frail folk have an awkward habit of dying, and so the 500 or so missing respiratory deaths each week will have turned up somewhere, just not as respiratory deaths. Dr No’s hunch — and Dr No allows himself the occasional hunch — is that both the plausible explanations given above are in play, but by far the bigger effect comes from hot stuff bias and diagnostic drift.
Average numbers of December (so a month, not a week) deaths 2015-2019 for leading ICD-10 Codes J00-J99 respiratory causes of death in England (not E&W), from here:
(1) ~40% of all respiratory deaths; for weeks 48 to 51 in 2020, average weekly respiratory deaths were ~30% less than in the previous ten years