Mortality by Vaccination Status in UK Data Sets

By Tore Gulbrandsen - @patternsofsignificance

A guest post by Tore Gulbrandsen

Introduction

Most of Europe, if not most of the world is currently facing extensive excess mortality in a period when the pandemic is supposedly over and, importantly, when many countries have already seen previous waves of excess deaths. Although the patterns have become exceedingly clear during the last year and winter, this anomaly is by no means new. I have previously written about excess death anomalies in the Nordics, starting early summer 2021, in a study paper on Finland, and X posts on Nordics+Baltics and Norway.

The UK has also faced such unexpected excess in summer 2021. While most countries' health authorities have been worryingly secretive with their data, the UK's ONS and the NIMS, subsequent to the persistent badgering by the host of this Substack, have provided the public with a modicum of insight in information that can shed light on a link between mortality and vaccination status.

On correlation vs causation

But before we dive into that data, I want to clarify something. Call it a fact check. An actual real fact check. Whenever someone makes an observational connection between a negative event and a vaccine, the mainstream media and the majority of society responds with the almost eerily predictable phrase "correlation does not equal causation". As such, this phrase is correct. But it is missing some crucial nuance. A more complete statement would be "Correlation between A and B does not mean that A causes B. But it does mean there is a causal connection between them, one way or another." Why is that? Because variables don't correlate if they're not connected. And the only way to be connected is by cause and effect. So, if random occurrence is off the table there is only one of three possibilities:

1. A causes B

2. B causes A

3. C causes A and B (through any number of steps D, E, F, etc.)

This is universally true. In the case of vaccines (A) and death (B), which we'll be talking about here, the death can hardly cause the vaccine now can it? So that leaves option 1. or 3. So in the lingo of the fact-checkers, the initial statement would probably deserve the label "misleading".

The Data

The following is based on three sources of data. The NIMS dataset that Joel Smalley acquired, EuroMOMO's excess deaths and the ONS mortality by vaccination status data set. The latter is arguably flawed as described by Profs. Fenton and Neil, but it's the best we've got.

The NIMS data shows the deaths of the 1-dose-or-more-vaccinated per week since the rollout of the vaccination campaign. Since we don't have the calendar weeks, I used EuroMOMO's z-scores to "calibrate" the data. That is, I adjusted the data sets so that the major mortality peaks aligned, thus translating the "weeks since roll-out" of the NIMS dataset to calendar weeks from EuroMOMO. This is shown in Figure 1, upper graph.

Figure 1. Upper: Weekly deaths from NIMS data set (solid red line) and EuroMOMO Z-scores (dashed blue line). Lower: Cumulative % change in cohort size from July (ref.=0%) to October for age groups 60+.

The Hypothesis

If nothing peculiar happened in the summer of 2021, we should expect to see normal death patterns. Sadly, EuroMOMO clearly shows that this was not the case. Instead, we see an upward trend from May 2021, long after the winter peaks. Now, if this excess mortality was exclusively driven by vaccine adverse events, then we should see both a strong correlation between vaccinated mortality and excess mortality and no correlation between unvaccinated mortality and excess mortality (because, obviously, the unvaccinated wouldn't die in excess if the excess was driven by the vaccine). Conversely, if whatever agent of death sweeping the country in early summer was indiscriminately ravaging all people, then the excess should be visible regardless of vaccination status.

The problem is, when the official agencies are hoarding the data with the same kind of ferocity with which Gollum clings to his Precious, we have to wrangle the data a lot to get to real information, and even then the quality is lower than would have been possible with full insight. We don't have data stratified by health status, and we don't have good stable comparison levels. What we do have, however, is change over time. Change that EuroMOMO says, is a change toward excess. So, let's take a period where EuroMOMO says there is excess, and where the overall death numbers are increasing. If there is no connection between the vaccine and this increase, the increase should be equally visible in both vaccinated and unvaccinated. So, let's look at a period where (a) there is an increase in deaths, (b) EuroMOMO shows excess deaths, which (c) is shortly after the roll-out of the vaccines.

The Results

The period June-October 2021 fits all of these criteria. But first let's consider some possible biases. Since it is not possible to "unvaccinate" oneself (sadly, some might argue), the unvaccinated cohort is likely to decrease with time, especially since the government has goaded, coerced and "scared the pants off of people" to make them leave this cohort. So, one possible kind of bias in the data would be that absolute number of deaths in the cohort inevitably will decrease with time when the cohort itself is decreasing. If the death rate was increasing but the cohort size decreasing at the same rate, it would appear as if the number of deaths was stable.

So, we need to correct for this effect. Luckily, we have the ONS data, which shows us how many people are in the different cohorts as a function of time. We can see that from July to October, the biggest decrease in size of the unvaccinated cohorts 60+ is -9% (Figure 1, lower). In other words, this change could mask an increase in deaths of up to 9%. In reality, the effect is likely even lower, since the people who constitute the majority of deaths are in the cohorts 70+ and 80+, where the change from July to October is only negative 6-7%.

Now to the correlations in death patterns. The vaccinated comprise the vast majority of the population, so any pattern affecting the population as a whole would be expected to be visible among a group that is >90% of the population. This is what we see, with a correlation coefficient of r=0.99 (Figure 2, lower). It is worth keeping in mind that when correlating a metric for >90% of the population to the same metric for the whole population, r is expected to be very high.

But the big deal in the data is not that deaths patterns in the population correlates with death patterns among the vaccinated majority. The big issue is that there is no correlation whatsoever between the death patterns of the population as a whole and the death patterns of the UNvaccinated. The number of deaths in the unvaccinated goes down approximately 10% in the period July-October. Even when we correct for the maximum 9% reduction in the cohort size of the unvaccinated, there is still no increase in the number of deaths among the unvaccinated, as that would land the total change for July to October at -1% (Figure 2, upper).

Figure 2. Correlations between death patterns among unvaccinated (upper left and right) and ever vaccinated (lower left and right).

The Conclusion

So, what can we say for sure with this data? We can say for sure that the increase in deaths from July to October in 2021 happens exclusively among the vaccinated. What does that mean? It means that whatever agent of death swept the country, it did NOT indiscriminately ravage all cohorts. Instead, it selected out the vaccinated and left the unvaccinated unscathed. Now, what kind of agent could affect the vaccinated but leave the unvaccinated alone? What kind of agent have only the vaccinated been exposed to, not the unvaccinated? Maybe it was heat waves - the selectively deadly kind that strikes in late autumn? Lockdowns? Having one’s pants scared off?

Scientists are puzzled. Except the ones who aren’t.

The words of a certain British gentleman announcing a late start basketball career on YouTube come to mind: "slam dunk".

---

Tore A. Gulbrandsen

19 September 2023

I am a Data Scientist with a passion for the truth, unafraid to ask questions that might lead to uncomfortable answers. I apply statistical concepts to evaluate the significance of real-world data patterns.

X:

Substack: