On Simpson’s paradox and vaccine effectiveness
Why vaccine effectiveness is so hard to prove or disprove, and what to do about it
Some Twitter users got very upset with me for not mentioning Simpon’s paradox in a recent post, in which I observed a correlation between more boosters and more deaths with Covid in the NSW health data. I then speculated on the notion that negative vaccine effectiveness may be a factor in play.
As I touch on in the post, it’s likely that the most boosted are also the most vulnerable (elderly, immunocompromised) and that this is a factor in the over representation of the boosted in hospitalisations and deaths with Covid.
Simpson’s paradox goes further, suggesting that a positive trend (eg: vaccine effectiveness) seen in each cohort separated out (eg: age group by decade) can appear in the reverse (eg: showing negative vaccine effectiveness) when those groups are aggregated together. This is possible, but it needs to be demonstrated in the data to be taken as fact in any given situation. See this example in which El Gato addressed the possibility of Simpson’s paradox but disproved the hypothesis subsequent to deeper analysis.
As NSW does not publish (with) Covid hospitalisation and deaths data showing age x vaccination stratification, we can’t prove or disprove Simpon’s paradox. Thus at this time, any suggestion of Simpson’s paradox must be taken as speculation. To take it as fact is an act of faith. This will require that the believer ignore or adequately explain away the mounting evidence of immune suppression, vulnerability to reinfection, waning immunity and the impact of dose effect on adverse events.
To be clear, my speculation on negative vaccine effectiveness is just that - speculation. Hence my use of the word if and my efforts to make observations of correlation without attributing causation.
It is challenging to prove or disprove vaccine effectiveness at this stage for all sorts of reasons, some of which Norman Fenton laid out in an article called Paradoxes in the reporting of Covid19 vaccine effectiveness, from September last year. Some key takeaways from the article:
A demonstration of Simpon’s paradox in UK Delta data, in which age was the confounding factor
Misclassification of freshly vaccinated people as unvaccinated, giving a false impression of vaccine effectiveness (such misclassification was recently spotted in NSW Health data by Twitter user @LCHF_Matt)
Differences in the way a Covid case is classified can skew conclusions
Fenton proposes that the best way to circumvent these issues is to use all-cause mortality as the appropriate measure of vaccine effectiveness. He published the below graphic in a follow-up article the following month:
Over the weekend, Fenton posted another insightful article on the illusions of vaccine efficacy, along with a video explainer on how efficacy can be ‘demonstrated’ in trials without the product offering any benefit whatsoever (ie: a placebo).
If it’s so hard to determine vaccine effectiveness or lack thereof, is there any point to publishing data updates as I do from time to time?
I think so, yes.
The principal reason I write about this at all is to draw attention to the fact that, from the data published by our own health departments, there is not sufficient evidence to support the claim that these vaccines reduce severe illness or death.
It is curious to me that Australian government and health officials are so adamant that the vaccines are as effective as promised and yet cannot (or will not) produce data to demonstrate it. If they have the data, why won’t they allow the public to access it? It would shut all those pesky ‘anti-vaxxers’ up and would reduce vaccine hesitancy. Isn’t that what they want?
And if they don’t have the data, then on what grounds are they basing their claims that these vaccines are doing what they are supposed to do? Are we just supposed to take it on faith that they know what they’re talking about, like when they locked the unvaccinated out of employment and society because the vaccines were going to prevent transmission?1
There are some who cry ‘misinformation!’ at any reference to the available Australian Covid data, not to mention any hint of speculation. To these people, I have two things to say.
First, take the log out of your own eye. As Fenton so clearly elucidates, speculation and data misrepresentation abound on all sides, and not necessarily deliberately. To parse out what is true and useful from what is not requires robust discussion, analysis, reanalysis and further discussion. It requires an attitude of humility, a willingness to constantly reassess one’s own position, and curiosity when holes are poked in one’s limited understanding (we are all limited in our understanding, for this is the nature of being human).
Second, your proclivity for clutching at pearls when encountering oppositional view points on the internet reveals an underestimation of the role of healthy debate in the process of collective meaning-making. The internet has democratised knowledge creation and dissemination. Oligarchic technocrats and elitist ‘experts’ may think this a bug, but I argue that it’s a feature.
We have seen a sustained effort to impose a top-down model of knowledge creation and dissemination in recent years, a somewhat feudal structure that appeals to authoritarian types (looking at you, Jacinda ‘single source of truth’ Ardern). Until such time that we become subject to global technocratic feudalism (hopefully never, but the creep is real), democratic debate maintains on the internet, albeit seriously hampered by big tech censorship.
Iron sharpens iron through combat. Healthy debate advances knowledge. There is no need to clutch pearls. Simply mount a well thought out argument in a respectful manner and enjoy the debate. You may even come away with a better argument for it.
I’d like to personally thank Twitter user @AndrewWalkom for challenging me to refine my thoughts on this topic.
See this article by Frank Chung for a comprehensive list of instances when Australian premiers and health officers said, repeatedly, that the vaccines would prevent transmission. Also note the dates of these statements, most of which were made after the claim to transmission prevention (or reduction of any significance) had been seriously brought into question (mid to late 2021) or altogether disproven (early 2022).
Let’s make meaning together.