A major update for this week’s analysis of the UKHSA Vaccine Surveillance report, with corrections to previous data. In our last post on week 5’s data we noted that it appeared the data for infections/hospitalisations/deaths for two doses of vaccine also included data for three doses of vaccine within two weeks of vaccination i.e., people were not counted as triple-dosed until 14 days post-jab. As a result we tried this week to obtain further data on infections/hospitalisations/deaths that might include the missing information. Unfortunately this was unsuccessful, but we did manage to get hold of the vaccination data used by UKHSA in its calculations (this differs from the official ‘vaccinations given’ figures because it only counts people that are still alive and resident in the country). This new data provides some surprising additional information that necessitates revisiting prior data.
The main issue is it transpires that the UKHSA has been using the number vaccinated at the end of the four week period to calculate its rates of infection/hospitalisation/death. As a result, its estimates of the boosted population have been excessively high; it’s been using a boosted population that is higher than the actual boosted population over the four-week data period.
To illustrate this, see the figure below.
The data for the week 4 report is for up to week 3 (indicated by blue vertical line), as there is a delay of a week in their collation of the data. As the data is made up of four weeks’ worth of data, this covers the period between the end of week 51 2021 and the end of week 3 2022 (indicated by purple bar below the graph). However, the data is for 14 days after vaccination, so the relevant vaccination data should be taken from the blue square shown in the figure, from the end of week 49 2021 to the end of week 1 2022. Instead, the UKHSA took its vaccination data as that at the end of this period, shown by the red vertical line. As is apparent from the graph, because of the rush to boost the population during this period, the number vaccinated at the point of the red line is significantly greater than at the start of the period in question (considerably so for younger age groups). Really, the estimate of the population boosted during this period should be a weighted average across the study period; this will lie somewhere within the green rectangle (exactly where is dependent on the shape of the curve between over the period in question).
As a result of this error, the UKHSA rates of infection/hospitalisation/death will have been smaller than they were in reality, giving the impression the vaccines were performing better than they were. Whether this was done on purpose to exaggerate the benefits of the vaccines or was the result of a simple mistake is unclear.
Unfortunately, because our estimate of vaccine effectiveness of dose 2 alone was based on matching existing vaccination data to the UKHSA’s data which used its inflated population estimates, it resulted in our estimates being too pessimistic. So much for trusting official sources of information.
But, of course, if there is a mistake the right thing to do is to try to rectify it – so this week we’ve reanalysed the data for this year using a more appropriate estimate of the proportion of the population that has received two and three doses. As of today we’ve got data for the over-40s population, and next week we’ll aim to expand this.
There is an upside to all this, which is that in obtaining the same vaccinated population data as the UKHSA uses we can also estimate the vaccine effectiveness for those who only received one dose of vaccine.
So, on to the analysis. Vaccine effectiveness against infection first.
This is a busy graph, but it splits broadly into two sections:
- We now have the data for dose 1 – these are shown in the dashed lines near the top of the graph. Those having taken only one dose of the vaccine appear to have approximately 50% more chance of infection compared with the unvaccinated (note that all the values are negative, meaning the unvaccinated had lower infection rates than all the other vaccination categories).
- Data for dose 2 and dose 3 of the vaccines appear to show a somewhat similar level of protection, at around minus-100% – that is, the vaccinated appear to be approximately twice as likely to become infected with Covid as the unvaccinated. The pronounced downward trend in the data is likely a result of the vaccines’ protection being even worse for Omicron variant. There is a hint in the data that in the weeks following the booster (but after the 14 day post vaccine period) there is an increased risk of infection (indicated by the upward trend in the data between the first and second data point for those aged 40-49 and, to a lesser extent, those aged 50-59).
The finding that the vaccine effectiveness isn’t so negative for those with only one dose of vaccine offers some hope that the increased risk of infection might wane with time. Alternatively, the data might simply suggest that two doses ‘seals the deal’ with even more increased risk. It would be nice if there were official investigations into this effect, but, as with all potentially negative aspects of the vaccines, our authorities appear to prefer not to know. It’s worth mentioning that these are unadjusted vaccine effectiveness estimates of course.
The estimate for vaccine effectiveness against hospitalisation is shown below.
The data again can be split into two:
- Three doses of vaccine appear to have offered some protection against hospitalisation during the Omicron wave. Interestingly, the data suggest that those aged 40-49 obtained the least benefit. We hope to update this chart with data for those aged under 40 later in the week; it will be interesting to see if this trend is also followed by data for younger age groups.
- Two or one dose of vaccine appears to offer very little benefit against hospitalisation. All of the estimates of vaccine effectiveness for one dose of vaccine are slightly negative (with the data for those aged 40-49 at minus-100% possibly an outlier – we should find out more with this Thursday’s update); the estimates of vaccine effectiveness for two doses of vaccine are slightly positive.
Again, an interesting aspect of the graph is the pronounced downwards trend, possibly due to Omicron’s vaccine evasion.
On vaccine effectiveness against death, there is an important additional consideration. While the UKHSA data for infections and hospitalisations refer to ‘by specimen date’, the data for deaths refer to ‘by date of death’. Thus for the deaths data we also need to consider that people currently infected don’t get vaccinated, and that the time from infection to death is usually at least 14-21 days. As a result, it is necessary to consider the vaccination figures not at the point of death, but at a point some weeks earlier. We are undertaking a sensitivity analysis to work out the ‘right’ delay to use, but for now the data shown is for a delay between infection and death of two weeks – thus it has somewhat but not fully compensated for this effect and as a result the estimates given will likely be slightly too high.
With that in mind, data for the protection offered by the vaccines against death shows a similar trend as seen in the hospitalisation data:
Three doses of vaccine appear to offer a reasonable protection against death, two doses show significantly reduced protection (especially for those aged 70-79 where it is negative), while a single dose appears to result in an increased risk of death, compared with the unvaccinated. Again, the downward trend in the data suggests that Omicron variant has made the vaccines’ job much more difficult.
I should explain why I concentrate on ‘deaths within 60 days’ rather than the more conventional ‘deaths within 28 days’. There is some evidence that the time course of infection in the vaccinated is more drawn out than in the unvaccinated. It isn’t clear whether this is a delay between infection and serious symptomatic Covid or between hospitalisation and eventual death, although data on hospital beds taken up by Covid patients suggests the former (there are fewer than you’d otherwise expect). Up until now the evidence has been limited to the deaths within 28 and 60 days of infection data in the UKHSA Vaccine Surveillance report. However, in recent weeks a new source of supporting evidence has emerged – the official deaths data published every day by the U.K. government:
An aspect of the Omicron wave I found disturbing was how the deaths didn’t decline in synchrony with the case numbers; there was a clear lag of about two weeks before deaths started to rise, but they didn’t fall as might be expected and remained elevated. In the past week they have declined rather steeply, which has resulted in the exclamations of ‘vaccine success’ by the usual suspects, but note that the precipitous decline in deaths has tracked the decline in cases with a lag of around 28 days (it appears to be between 25 and 31 days; close enough). Is this, however, because Covid deaths in fact remain elevated, but they’re not counted as Covid deaths as they’re beyond 28 days after the original infection appeared? We should know more about this effect in the deaths data from the Vaccine Surveillance reports over the next few weeks, as they include data for deaths within 60 days of vaccination.
Amanuensis is an ex-academic and senior Government scientist. Find him on his Substack page, Bartram’s Folly.