It’s about relative, not absolute, vaccine benefit

I find the degree to which people will misread data to support their narrative such a depressing part of the pandemic. As a data scientist, I’ve failed if ego trumps data truth.

This latest offering by some armchair sceptic, has lit my wick enough for me to pen this. They claim that vaccines are killing more people than those unvaccinated, and when challenged on their analysis, they have the gaul to tell me absolute numbers, not rates, matter despite there now being 7x as many people vaccinated.

They’ve looked at this table from this publication by PHE, and concluded that because 402/742, 54.2%, of deaths were vaccinated, vaccines are more dangerous. Which is a bit like saying being white is 30x more deadly than being black, when in reality it is that just 3% of the population is black.

Yes, that daft.

Criminally, 97.7% of unvaccinated cases are <50, but just 54.3% of double vaccinated cases are. So the real difference is the age distribution of spread, and he’s just massively diluting the rate of death ≥50 with lots more cases in the young to make his case.

Considering cases & deaths <50 & ≥50 separately, we see that entirely the opposite of what he is reporting is true.

A classic example of “Lies, damn lies, and statistics”.

Before we can really get to the bottom of the numbers there is a very important issues to address.

Are all 11% unvaccinated really lacking immunity?

This is the biggest part of all the fraud going on around vaccine efficacy now that so few are unvaccinated, with @toadmeister via Will Jones leading the charge in this kind of fraud.

Simply, you are more likely to not have a vaccine if you know you had a clear bad case of C19, which means as people have got vaccinated, the proportion of those remaining that already have immunity has just become more concentrated, and 25% had natural immunity after Jan to start with.

The ONS antibody study have about 94% with antibodies at the moment.

Bear in mind, this is for the entire population, and 18% are <16. Which means the 11% unvaccinated 16+ must contain many that have antibodies, easily 50–75%, so we are actually only looking at about 5% of the population 16+ that are without immunity.

So have vaccines help with spread?

Looking <50 and ≥50 separately, what we basically want to assess is, per unit of the population, what are the case rates.

To do this we 1st need to know from here that 44.4% of the population is 16–49, and 37.6% is ≥50. Add to this vaccination rates from this PHE data as a guide for the whole of the UK, we can work out that:

<50: 16% 1 dose, 63.9% 2 doses, 20.1% unvaccinated

≥50: 2.2% 1 dose, 96.3% 2 doses, 1.5% unvaccinated

With this and the table above we can then work out the case rate per 1% of the population in each of the groups.

<50

• Unvaccinated: 147612/(44.4%*20.1%) = 16540
• 1 dose: 40449/(44.4%*16%) = 5693
• 2 dose: 25536/(44.4%*63.9%) = 900

≥50

• Unvaccinated: 3440/(37.6%*1.5%) = 6099
• 1 dose: 5640/(37.6%*2.2%) = 6818
• 2 dose: 21472/(37.6%*96.3%) = 593

The unvaccinated >50 look to be as immune as those with 1 dose, they likely all have had it now, but the rate in both is 10.3x higher than those double dosed.

The key reveal here though is that <50 the unvaccinated are 18.4x more likely to have an infection than those double dosed. However, this could easily be twice that if half those unvaccinated have natural immunity meaning 95% of the infections are in just 10% not 20% of the population.

How about deaths? Have vaccines helped?

Given 90% of deaths are ≥50, we are best looking at the impact in these higher risk groups.

• Unvaccinated (no vulnerable): 205/3440 dead, 6%
• 1 dose: 65/5640, 1.2%
• 2 doses (inc. vulnerable): 389/21472, 1.8%

But, if you remove deaths in vulnerable groups from the double dosed figures, to make it a like for like comparison, we are talking just 0.6% of cases dying, about 1/12 of those without any immunity.

So like cases, a dramatic reduction in risk.

Basically, instead of demonstrating vaccines haven’t worked, the PHE publication shows an order of magnitude change to risk post double dose. 1/12 the deaths ≥50 and 1/10 the cases ≥50 and 1/18 cases <50, very much in line with vaccine trials. People writing such drivel as Chris Waldburger should stop courting popularity with such fraudulent data science. I don’t say any of this to judge those that choose not to have a vaccine, or to wear masks, as a libertarian I entirely support free choice, but only if those making that choice are aware of reality of the data. So I do judge those that try to frustrate the reasoning behind their free choice by presenting such bad evidence.