Do COVID Vaccines Work? (part 2)
In this article I utilise in-hospital death data for 2020 – 2021 from an unknown NHS Trust to examine the impact of vaccination rollout on respiratory & COVID death
In part 1 of this series I provided a basic analysis of vaccine efficacy for respiratory illness in terms of admissions to the A&E departments for an unknown NHS Trust for those aged 12 years and over. In this article I continue in a similar vein by looking at incidence of respiratory in-hospital death and incidence of COVID in-hospital death for unvaccinated and vaccinated individuals aged 18 years and over. In my book ‘unvaccinated’ should mean exactly that – no needle at any point prior to death – and that is precisely what is meant in slides presented below.
Before we start eyeballing those slides I must stress that we are dealing with very small sample sizes indeed. These are not robust analyses by any means and are provided largely for interest.
Acute Respiratory In-Hospital Death
We must note the very small weekly sample sizes which ranged from 1 to 106 weekly unvaccinated acute respiratory in-hospital deaths (median sample size = 8 deaths per week). Vaccinated acute respiratory in-hospital deaths ranged from 1 to 38 persons (median sample size = 16.1 deaths per week):
A Mann-Whitney U-test indicated that these series should be treated as two different distributions at the 95% level of confidence (p=0.006); that is, the difference your eyeballs is seeing isn’t likely to be some chance effect but is most likely to be real. That being said that enormous spike for vaccinated acute respiratory death is swinging things, but even if we start the analysis at 2021/w1 we still arrive at a statistically significant difference using classical methods (p=0.010).
What this needs is a bit of Bayes to help us decide whether that unvaccinated red hump during the first few weeks of 2021 means anything given the small sample sizes and the fact that we haven’t adjusted for any of a zillion confounding factors. Setting that hump aside I would venture to suggest that there is likely no difference between vaccinated and unvaccinated cohorts beyond the 2020/21 seasonal peak.
Chronic Respiratory In-Hospital Death
We must note the small weekly sample sizes which ranged from 3 to 104 weekly unvaccinated chronic respiratory in-hospital deaths (median sample size = 15 deaths per week). Vaccinated chronic respiratory in-hospital deaths ranged from 0 to 59 persons (median sample size = 39.5 deaths per week):
A Mann-Whitney U-test indicated that these series should not be treated as two different distributions at the 95% level of confidence (p=0.776); that is, any difference your eyeballs may be seeing is not likely to be real. We observe a fair amount of scatter so shouldn’t read too much into things other than gain an impression that there probably isn’t a difference, though in doing so we’re ignoring many confounding factors.
If I were to stick my neck out I’d say there appears to be a swing from unvaccinated to vaccinated death because the two curves are going in slightly different directions. This may well be because of demographics, attitudes, basic health and things like survivor bias - the list of confounding variables is endless!
Asymptomatic COVID In-Hospital Death
We must note the very small weekly sample sizes which ranged from 0 to 95 weekly unvaccinated asymptomatic COVID in-hospital deaths (median sample size = 3.5 deaths per week). Vaccinated asymptomatic COVID in-hospital deaths ranged from 0 to 22 persons (median sample size = 2 deaths per week):
A Mann-Whitney U-test indicated that these series should not be treated as two different distributions at the 95% level of confidence (p=0.203); that is, any difference your eyeballs may be seeing is not likely to be real. Ditto the reading too much into things and ditto the confounding factors.
We’re now into COVID territory and it is most striking that vaccine benefit is not showing up to the party as it should. We really shouldn’t be searching for scraps of benefit - it should be smacking us right between the eyes, baking us a thin crust pizza, pouring us a cold beer and singing us a love song. I don’t see any love but I do smell money, and lots of it.
Symptomatic COVID In-Hospital Death
We must note the very small weekly sample sizes which ranged from 0 to 92 weekly unvaccinated symptomatic COVID in-hospital deaths (median sample size = 4 deaths per week). Vaccinated symptomatic COVID in-hospital deaths ranged from 0 to 16 persons (median sample size = 3.5 deaths per week):
A Mann-Whitney U-test indicated that these series should be treated as two different distributions at the 95% level of confidence (p=0.006); that is, the difference your eyeballs is seeing isn’t likely to be some chance effect but is again most likely to be real. This really is the crunch slide for me but sample sizes are far too small to do much with without resorting to Bayesian methods (and gathering more data from NHS Trusts that are sitting on mountains of this stuff).
There’s a suggestion of a hump for unvaccinated cases during early 2021, but even if we tentatively assume some sort of vaccine benefit then it was obviously short-lived. But, tbh, we shouldn’t draw any hard and fast conclusions from small sample sizes like this, and especially when COVID designation through reliance on the PCR test is known to be an extremely dodgy affair.
I wonder how many so-called symptomatic COVID deaths actually arose from bacterial pneumonia and bizarre ICU protocols resulting in withdrawal of antibiotics in favour of anti-virals? Are we looking at yet more evidence of iatrogenic death when we look at that red hump?
Kettle On!