This came out of a discussion I had with a colleague regarding my likely termination on Dec 8th. Since said termination will result in what might be significant disruption to some number of projects, I have gotten the “it’s pretty selfish not to take one for the team” talk. During the course of one of those that was much less in that line and much more in vein of curiosity regarding why I would make such a decision, beyond the philosophical and ethical considerations, I was attempting to make a utilitarian, practical case as well. Given the quality and reliability of all reporting on vaccine efficacy and safety, I was looking for a small example with simple data. I wrote up the following for this colleague and figured, since the work was already done, I’d throw it up here.
The following is based on a simple pull of data from the NHS response to a UK FOIA request to a single hospital unit for the first half of 2021. I attach a screen shot of the letter and the data; I didn’t insert the screen shot as the formatting was shite. For this quick analysis, I’m only looking at the month to month data in the last two tables. These data cover the first half of 2021, from roughly the start of a widespread vaccination program in the UK to the end of May of this year.
To get a lay of the land, I first just plotted the data – always a good idea when starting on an unfamiliar data set. Just spend some time looking at it. These are just plots of the raw data, black representing quantities for vaccinated patients, blue, unvaccinated. First, I’m just plotting the percentage of admissions each month; this is just the raw data from the table in point 2 of the letter. On the Y-axis, I’m plotting the number admissions in each category as a percentage of total admissions. Note that the fraction of vaccinated patients grows while the unvaccinnated population falls as the vaccination program begins in January. This is exactly as expected as you draw from an increasingly vaccinated population.
Next, do a similar plot for all cause mortality (data in item 3 of the letter). In all cause mortality the same pattern is seen, with the fraction of people dying climbing rapidly in the vaccinated group and falling in the unvaccinated group. Again, this is exactly as expected in a population that is increasingly vaccinated regardless of efficacy or safety. I’ve unfortunately seen some commentators confuse, either deliberately or through carelessness (and on both sides), this sort of data by saying things like “the percentage of people dying who are vaccinated exceeds the unvaccinated”, trying to imply that this, in and of itself, indicates an issue with vaccine safety. Of course it doesn’t; without knowing the relative populations, it’s almost meaningless. One needs to know if, as a population, the vaccinated are faring worse in some relevant metric. In fact, the fact that the admissions and mortality plots are almost identical might indicate no or minimal effect on all cause mortality for the vaccines. One needs to look at population normalized numbers to tease out any effect in a multivariate problem like the above, where variables are changing with time.
To correct for the time varying relative populations, I normalized the total death from all causes in each cohort by the fraction of the population in that cohort. This tells you how many people in each cohort died each month per unit population. I’ve multiplied by 10000 here just to get the numbers up to readable. Could just as easily go per 1000 population, but…
There are 2 things to note. 1) All cause mortality is nearly Identical in January, before the start of a widespread vaccination program. That gives some confidence that we have a measure of the background rate of mortality. One could go to historical data and verify that. 2) While both curves drop, reflecting seasonal variation in all cause mortality, the un-vaccinated curve rapidly falls below the vaccinated curve, indicating that, per capita, i.e. independent of population vaccination rates, the vaccinated cohort are experiencing significantly higher mortality rates and the difference gets larger as a larger fraction of the population is vaccinated. There are of course confounders – it could be that the vaccinated population seeking hospital care is significantly less healthy than the unvaccinated population, though this seem unlikely. In fact, the straight forward explanation is that, while vaccination seems to protect against COVID-19 associated death, safety issues result in overall increased motality. The latter is consistent with analysis of the initial Pharma company data showing worse overall outcomes in the vaccine arm when overall population health is used as a metric as opposed to restricting the metric to only COVID-19 specific outcomes – and even in the trial data, at least, the COVID-19 mortality data were not statistically significant with 1 death in the vaccine arm and 2 in the control group. Note that I don’t mean to imply that the vaccines do not mitigate deaths from COVID-19; they certainly seem to do so even though, given the generally small death risk, the trials were not large enough to generate a statistically significant result there. But I also think there needs to be an analysis of overall population health and mortality to see if one can claim a net benefit from a widespread vaccination program. And of course, none of this says anything about the completely unknown long term safety profile of the vaccines, especially in the regime of continuous boosers.
Are these data an outlier? It would be nice to see this for a wider selection of sources, but those data seem very difficult to come by. It seems like this should be standard data collected and analyzed. While there’s no absolute proof here, it is consistent with the actual trial data, if buried, in the 6 month follow up after EUA. It is possible that vaccine administration will, in and of itself, increase mortality when the metric is not confined to COVID-19 outcomes; seems like something medical professionals and experts should like to know.
BTW, I apologize for slacking on “Is the Universe Getting Biggerer”; writing it has proven more difficult than I thought and I’ve been distracted. I have the “Laster” episode mostly done, but it got pretty long and I think I need to make it into two parts. Promising a “Laster” but always delaying it; sounds familiar.