The ONS have finally published data on deaths by vaccination status for the first five months of 2023.
Predictably, some have used big picture analysis to claim vaccine failure in an unreasonable way. For example this claim is flawed:
Although it is interesting to note that the percentage of deaths in the unvaccinated has fallen from 4.3% in 2022 to only 3.6% in 2023, there are confounders in the data.
Such an analysis does not account for massive differences between these groups. The primary one being age. The unvaccinated are disproportionately at no risk from covid deaths and those who had 4 doses are old or frail and at disproportionate risk. This needs to be accounted for.
The fair way to account for it is to look at age groups separately and to calculate a mortality rate for each group.
However, that requires knowing the size of the population in each group. In theory the vaccinated population can be counted when the vaccines are recorded and it is the unvaccinated population which is much more of a guess - because we do not know the size of the whole population.
The ONS tried to resolve this by matching up individuals that they held information on from the census. The idea was that a death could be linked first to the vaccination database and then to the census and only those present in both would be included in their dataset.
(There are issues around what happens when there is difficulty matching an individual’s records which could account for hugely different interpretations. I have written about this before. The ONS have promised they are investigating how many records might have been incorrectly identified and have yet to respond with an answer to that.)
The latest interation shows serious problems with inflation of the apparent number of vaccinated people in the database. The ONS do not provide a population count. Instead they give “person year”. This is a measure of people and time such that a person who changes status half way through a month can contribute to both the earlier and later status with the right weighting.
Reverese engineering the person years each month gives the number of people claimed to be in each group. Table 3 gives the number of vaccinated and unvaccinated males or females across the whole cohort.
In 2021, the total population in the cohort was static while the number of vaccinated people increased and the number of unvaccinated decreased. However there were two subsequent rises in the vaccinated population which took effect without diminishing the size of the unvaccinated population, in early 2022 and then a much larger one in autumn 2022. The number of vaccinated people in the cohort rose from under 19 million to nearly 26 million from March 2022 to May 2023.
Exaggerating the size of the vaccinated population will artificially reduce their mortality rate. This means (again) any analysis carried out using the ONS denominator for the population is going to be corrupted.
There is no point wasting time on the first tables in the dataset which rely on this error. However, there is somthing of interest in the last table - table 5. This gives the raw data for all deaths by vaccination status without trying to link them to the ONS census. A further post will look at that.
Can we rename them please? ONS Office of National Silliness? Monty Python comes to mind 😹
I was initially optimistic about this data release in the UK. Maybe we could finally see some reasonably accurate large population statistics. Canadian stats are worthless and possibly intentionally so. The surge you noted in the population size is a frustrating loss of statistical relevance. Their approach seems to be designed to never produce confident results.
A randomly selected 100k per age strata from census in 2013 - 2023 followed for vax status, hospital visits, mortality, would eliminate some of the fudging artifacts.