Healthy user bias is obvious in booster data as well
Although I did not use the term at the time, a post I wrote six months ago was essentially about the so-called healthy-user bias.
Seems like our AI overlords at ChatGPT do a pretty good job of defining this bias:
In that post, I argued that the figures we hear from people on how effective the COVID vaccines are against hospitalization and death should not be taken at face value because they do not likely account for the fact that those who get vaccinated are likely to be healthier to begin with vs. those that don’t. Looking at vaccination rates by US County vs. a variety of statistics about those counties, I showed that higher vaxxed counties tended to be higher income, fewer smokers, less obese, more physically active, longer life expectancies, and lower death rates.
Recently, it seems that I am hearing a renewed effort to promote boosters & bivalent boosters. This tweet from the FDA commissioner is one example
Given this, I thought it would be a good idea to check whether the same healthy user bias is likely with the boosters. Below I plot the % of population boosted & bivalent boosted vs. the same factors mentioned above for each US county. Each circle represents a county, sized by population, and colored by Region. I have filtered out very small counties (<25K people) because I suspect this data is less reliable for them.
Starting with % of Smokers, % Physically Inactive, and % Adults with Obesity (**update- for clarity, the graphs on left are the “regular” booster %’s, the graphs on the right are the bivalent (omicron) booster %’s) :
We see that there is a strong relationship (p-value <0.001) whereby counties with higher rates of boosters also have less smokers, less people physically inactive, and less people obese.
Moving on, let’s look at Median Household Income, Life Expectancy, and Age-Adjusted Death Rate.
Again, strong relationships (p-value <0.001) across the board. Counties with higher rates of boosters also have higher incomes, longer life expectancies, and lower age adjusted death rates.
It is clear that there is a healthy user bias (at least in the US) with regards to outcomes of boosted vs. non-boosted populations. Randomized control trials (RCTs) are designed to eliminate this (and other) form of bias. There are no RCTs that demonstrate these vaccines to be effective against hospitalization and death. Claims about their effectiveness in this area are not based on sound statistical principles. It may be true, but we truly don’t know & it does not seem that the powers that be are interested in finding out.