Economist Writing Every Day

A recent analysis by the Kaiser Family Foundation of CDC data suggests that about 234,000 COVID deaths in the US could have been prevented if everyone was vaccinated. That’s about 25% of all COVID deaths throughout the pandemic, and about 60% of COVID deaths since June 2021 (roughly the time when most older adults in most states had had a chance to be vaccinated).

The first way to think of that death rate is tragic, given that so many lives could have been saved. Rather than being the high-income nation with the highest COVID death rate, the US could have been more in line with countries like Italy, the UK, and France. The US actually had a lower COVID death rate than Italy and the UK when the vaccine roll-out began, and today we could be at about France’s level with better vaccination rates.

But there’s a flipside to the KFF numbers. If 60% of COVID deaths since June 2021 were preventable, that means 40% weren’t preventable. Furthermore, the same data show that about 40% of COVID deaths in January and February 2022 were fully vaccinated or had boosters. That sounds like the vaccines might not work very well! So what does this all mean? Let’s dig into the data from the CDC a little bit.

The first, and most important thing, to recognize is that most American adults are vaccinated (about 78%), so unless vaccines are 100% effective (and they aren’t, despite some public officials overenthusiastic pronouncements early in the vaccine rollout), there are still going to be a lot of COVID deaths among the vaccinated. If 100% of the population was vaccinated, 100% of the deaths would be among the vaccinated. The key question is whether vaccines lower the chance of death.

And they do. Let’s see why.

Keep in mind also that the figure cited above, that 40% of COVID deaths in January and February of 2022 were vaccinated, does not control for age differences. Young people are less likely to be vaccinated, but also much, much less likely to die if they get COVID.

If we focus on one age group, such as the elderly (age 65+), we can get a better picture of how effective the vaccines are. As of the end of February 2022, the most common vaccine status for this group was to be fully vaccinated plus have at least one booster shot. Out of the almost 28 million elderly Americans, 16.5 million (or 59%) had been boosted. Another 8.6 million had been fully vaccinated but not yet boosted, and 2.7 million (less than 10%) were unvaccinated. But about 19,000 of the 34,000 COVID deaths during the first two months of 2022 were among the less than 10% unvaccinated group.

In other words, the unvaccinated elderly died at a rate of about 22 times the boosted group, and about 6 times the unboosted/vaccinated group. Keep in mind that’s 22 times as many deaths, not 22 percent higher or something like that!

Now, a few caveats. This data is what we call “observational.” It’s not from a randomized control trial. People are self-selecting into the treatment group. Also, the data isn’t controlling for anything, such as prior health status. But notice that both of the issues with the data are highly likely to understate the effectiveness of the vaccines. Those that are self-selecting into the vaccine treatment group are more likely to have prior health conditions. The actual rate of protection against death from the vaccine is probably much higher than the 22-fold difference for the boosted versus the unvaccinated.

Finally, are some vaccines better than others at preventing death? A recent working paper summarizes the randomized trial data on each of the vaccines from the clinical trials (before they were in the general population). The paper has what seems to some a shocking finding when pooling all the results by type of vaccine: the mRNA vaccines (Pfizer and Moderna) didn’t reduce total mortality in the trials (though they did reduce COVID deaths). But the adenovirus-vector vaccines (Johnson and Johnson in the US) did reduce total morality. For the mRNA vaccines, there were roughly the same number of deaths in the treatment and control groups (31 and 30 deaths), while the adenovirus-vector vaccines had statistically significantly fewer deaths in the treatment (vaccinated) group (16 deaths versus 30 deaths in the control group).

So what gives? Are the mRNA vaccines ineffective? While usually we preferred randomized clinical data over observational real-world data, in this case we might not. The participations in the trial groups were probably not the most vulnerable, elderly populations. This is evidenced by the fact that the number of deaths was so small. And small numbers can do funny things.

When we look at the individual studies, we see just one of the 5 trials for adenovirus-vector vaccines showed a big difference: the Johnson and Johnson trials, which had 3 deaths in the vaccinated group and 16 in the unvaccinated group. The other adenovirus-vector vaccines trials (3 for AstraZeneca and one for Sputnik) found no difference in overall mortality. With such small numbers, and only one study showing a huge difference, it could be due to some imperfect randomization, or just the inevitable “things you can’t really control for.” For example, the J&J trial had a suicide and an overdose in the placebo group, while the Moderna trial had a suicide and a head trauma death in the vaccine group (and one suicide in the placebo group). We probably don’t think the trials had anything to do with the suicides, but they are counted in the overall mortality.

So what does the observational real-world data look like for each of the vaccines? The CDC also gives us this data. Unfortunately they don’t break it down by age ranges, but they do provide age-adjusted relative death rates for each vaccine. The age-adjusted death rates overall (for all vaccines) look similar to the 65+ data I discussed above: unvaccinated died at about 20 times the rate of boosted, and about 7 times the rate of the unboosted/vaccinated group in January and February 2022.

For the Johnson and Johnson vaccine in January and February 2022, the unvaccinated died at about 13.5 times the rate of the boosted, and 5.5 times the unboosted/vaccinated group. For the Moderna vaccine, the unvaccinated died at about 22.5 times the rate of the boosted, and 7 times the unboosted/vaccinated group. For the Pfizer vaccine, the unvaccinated died at about 18 times the rate of the boosted, and 7 times the unboosted/vaccinated group.

From the observational data, it seems that the mRNA vaccines perform better than J&J, though all of the vaccines are very effective at reducing COVID deaths. And part of that difference is because we generally gave the J&J vaccine to lower risk individuals, which isn’t being controlled for here. We also don’t have total mortality data for each of the vaccine groups in this observational data, which ideally we would have to make a better comparison with the trial data in the working paper I discussed above. But given that the correlation between excess deaths and COVID deaths for each US state by month is about 95%, it seems highly likely that when we do have this data, all the vaccines will look very effective at reducing total mortality.