Vaccines & Death (part 3)

Further investigation into delayed death following vaccination

In part 1 of this series I seemingly confirmed the somewhat disturbing results of recent research undertaken by Steve Kirsch over in the US using data supplied by the UK GOV coronavirus dashboard and the Office for National Statistics (ONS). ARIMA time series modelling formally confirmed what our eyeballs were telling us, this being the inescapable fact that weekly excess all cause death in England & Wales was following the pattern of weekly vaccination dosing (any and all doses) with a delay of around 5 months. I say seemingly because I ended-up with a model that indicated vaccine benefit and not harm.

All cause death necessarily includes COVID cases and so I derived a time series for weekly excess non-COVID death to determine whether delta and other variants might be driving results. The results were presented in part 2.

Given that death certification isn’t as robust as we’d like to think (a confidential chat with an honest GP is highly recommended for those who dream that it is) it occurred to me that some non-COVID deaths may well have been misclassified (e.g. false negative test results), so we ought to cross-check results by running the same analysis, not with vaccine dosing, but with disease prevalence (proportion infected within the population). We may think of this as developing a baseline model against which our hypothesis may be tested. In this newsletter I shall be revealing the first tranche of this additional work.

Case Detection Rate

Disease prevalence is usually expressed as the percentage of infected persons to all persons within a population such that 1% prevalence means 1 in 100 people are infected. If we undertake 100 viral tests on a random sample of people we are thus likely to detect just 1 infected person. However, if we undertaken 100,000 viral tests on a random sample of people we are likely to detect around 1,000 infected people. The infection rate has remained the same (1% prevalence) but owing to the larger number of tests undertaken we have ended-up detecting more cases.

I am sure all this sounds trivial to subscribers but the fallacy of comparing 1 case with 1,000 and claiming an increase in the disease without adjusting for test activity is exactly what authorities have gone and done since the pandemic began. I presume this to be a deliberate act of misinformation.

What I’ve gone and done to develop a proxy measure of disease prevalence that is reasonably free from sampling bias is to sum the daily cases for England & Wales (specimen date basis), then sum the total viral tests undertaken (PCR & LFD), thence to divide cases by tests to arrive at COVID cases per 100 viral tests.

Spiky Stuff

Daily data is a mighty spiky affair due to weekends and, owing to delays in administration in one set of figures or another, we can see some well-wacko percentage swings. Whilst this can be avoided by using some sort of fancy smoothing function it can also be avoided simply by calculating a rolling 7-day total for both cases and tests and using this to derive less wacko percentages. We may call this Rolling 7-day COVID cases per 100 viral tests but that really is a serious mouthful! Before we get stuck in further let’s just have a quick squizz at this series:

There we are! The monster spike that kicks off the pandemic is misleading in that the first few cases detected were folk already ill in hospital, with few tests being undertaken across both nations back then. In essence, we are looking at the state of play in just a couple of hospitals. As the pillar 1 and pillar 2 testing schemes rolled out across the country as a whole we start to see less biased estimates of disease prevalence. Since this is not strictly a measure of disease prevalence I’m calling it the case detection rate (CDR).

When I look at this slide I always ponder over the strange ramp in CDR from week 29 of 2021 onward. I also tend to ponder on the ‘triple humper’ from week 50 of 2021 onward. What we have here is the Loch Ness Monster! Are these humps the impact of variants spreading within the population in a transient manner or something else entirely? Let’s go see…