Excess COVID Death (part 2)

Excess all cause death in heavily vaccinated nations is making global headlines. In this mini-series I derive a measure for excess COVID death.

In part 1 of this series I introduced a variable I’m calling case detection rate (CDR), this being a proxy for disease prevalence that takes into account the widely varying number of viral tests that have taken place over time. Test 1,000 folk and you’ll find a few COVID cases; test 10,000,000 folk and you’ll find a few more cases. Hence CDR, which is derived as rolling 7-day COVID cases per 100 viral tests (the ‘rolling’ bit avoids artefact arising from weekly admin patterns).

We ended-up comparing certified COVID deaths with CDR and noted an excellent agreement over time indicating COVID cases may indeed sometimes end in COVID death. The agreement was not perfect, however, and we pondered on why COVID deaths spiked at a time when CDR did not - how can large numbers of people be dying of COVID if it wasn’t rampant as a disease? I suggested we derive the ratio of rolling 7-day certified COVID deaths to CDR and take a look at this time series for clues. Here is that time series:

I nearly blurted my tea out when this plot popped up! We have an extraordinary peak centred on March 2021 whereby certified COVID death counts go through the roof with respect to the proxy for disease prevalence (CDR). What on Earth went on back then? Is this seasonal death or something else? If seasonal then why don’t we see the same hike for the 2021/2022 season?

At this point I reached for daily vaccination data for dose 1 and threw that time series into the pot for due consideration:

Is that a corker or what? We see a sudden flurry of initial dosing across England mirrored by a near identical surge in the death to disease ratio.

If we stop to think about matters this ratio is providing us with a proxy of excess COVID death; that is, death beyond what we would expect given the degree of infection within the nation.

If you want a smoking gun for initial vaccine rollout inducing COVID death among the clinically vulnerable then I guess this is as good it gets at the population level. But, of course, this smoking gun comes out of a holster brimming with confounding factors. I shall be putting all this through statistical modelling to see if I can obtain a rigorous result for what our eyes are telling us.

What is less shocking and more fascinating is that secondary dosing and subsequent boosting don’t feature and neither does the 2021/22 season. Could it be that folk who were particularly susceptible to the elixir were already dead by this point in time? I shall thus be looking at the quiet period beyond this dirty great spike just in case a delicate pattern sits waiting patiently in the shadows.

One More Cunning Thing Before Luncheon…

There’s one cunning thing we can do before we retire to the kitchen to make a fine luncheon of Welsh rarebit and down ale and that is compare the ratio of certified COVID deaths to CDR with certified COVID death on a week-by-week basis. If we are looking at significant seasonal factors by way of explanation then we should see two peaks line up. Except they don't:

What we have here is clear evidence of seasonally-influenced COVID death occurring 3 - 7 weeks before the peak in the ratio of certified COVID deaths to CDR. Thus we are obliged to rule out seasonal effects as the main driver behind this astonishing result. Right now my conclusion must be that we are looking at smoking gun evidence of a hike in COVID death during the early stages of initial dosing.

With that I shall slip into cogitation mode for a while. Until then…

Kettle On!