Vaccinated folks are fine. We're all fine here now, thank you. How are you?
The CDC tells us: "Fully vaccinated people with a breakthrough infection are less likely to develop serious illness than those who are unvaccinated and get COVID-19."1 To what extent is this true now?
Well: Definitely some food for thought here.
So let’s just see if we can find out how things look in the U.S. Currently the most vaccinated state is2 (drum roll): Vermont, at 68% fully vaccinated. But runner up is Massachusetts, 66% and maybe 10x the population. Let’s go hunt up what numbers they have available there, and particularly what they have been seeing regarding breakthrough cases. A breakthrough case is one that happens in somebody that has been vaccinated against it.
Hmm. Googling pops up immediately with:
"Two days ago" in this case was August 25. Between these two search hits we have numbers on either end of a two week span. So then roughly:
“over” 15,000 - “nearly” 10,000 ≥ 5,000 new COVID cases among vaccinated people in MA, and
“more than” 130 - “just over” 100 ≈ 30 vaccinated people died during those two weeks.
OK. So right away we have a couple numbers we can use to get a recent death rate among Massachusetts vaccinated people and tested positive. Some data appears to be findable, let’s dig a bit more and see if we can understand what we’re sitting in.
Admittedly taking media quotes out of a google search is pretty slapdash, but lets compare this to the general numbers for the same time period (less slapdash: officially from the State of Massachusetts, dates reflect their report date)3:
OK -- this is interesting and a bit worrisome. 82 people actually pretty comparable to what we quickly googled up for vaccinated, and in fact includes the vaccinated people. We should probably not rely google snapshots here, go find what the real numbers were from the state government if we can.
From the Massachusetts government we can find the following numbers4, well directly at least the first column of:
Turns out the numbers from prior weeks though are not available (at least not obviously) on the government site (this doesn't seem atypical, have found this for most other states). We can go check the news articles above, if they’re reputable, they hopefully quote and in some sense have archived the actual state numbers. The first NBC article does actually contain the 8/24 number from the state in the text5. Some level of validation there. If we go look through the two week old article6, we find they also quote the official state numbers from that time. So we will try them out. Both are cumulative: adding up all cases since vaccinations began, so to find the number of new incidents between those two weeks, you need to subtract them. Thats already in the table above in the 3rd column.
If we divide deaths/cases in each population we get virtually the same ratio, last lines in both tables.
We also in here should have done some freshman level error propagation at least. They are actual measurements, so in that sense they are what they are, but when you use one to try to predict the other (i.e. compare the two) you need to consider the possible statistical fluctuations in the observed numbers. If you have very few data, predictions based on them are less certain.
In this case for the vaccinated population we have a group of only 25 people. There are certainly long discussions to be had about systematic uncertainties, counting biases, all kind of problems one might run in to with these numbers, but we’re going to take this as a pure simple counting experiment. A simple Poisson estimate for the uncertainty on that number is
So roughly a 20% uncertainty on what we're doing here. The death rate for all people in Massachusetts in the last two weeks looks completely consistent with the rate for only vaccinated people.
In case you're wondering, lets make a table for only unvaccinated:
OK we get a slightly higher death rate for just the unvaccinated population, ignoring uncertainties. Within uncertainties though this is completely consistent with the other numbers. Meaning at least based on the last couple weeks in Massachusetts you can’t really make the CDC statement. It may be that a year ago, with variants that were around when trials were happening, vaccinated people fared better, but this data doesn't seem to back that up.
In fact. It perhaps says something worse. We have been told two basic statements about the vaccines:
The vaccines are effective, they will prevent >90% of COVID-19 infections
If you did happen to get infected, your chance of severe infection is much lower if vaccinated (see CDC statement above).
So this quick check of 2 recent weeks of data from one of the most vaccinated states in the US seems to contradict that. It looks like:
The vaccines appear less effective than they possibly used to be. Glancing at case numbers in the last couple weeks, it appears they may only prevent around half the number of infections. But maybe we actually can calculate this given what we’ve looked up. The vaccine efficacy7 (i.e. the 90+% number we hear all the time) is basically the difference between unvaccinated and vaccinated “attack rates” divided by the unvaccinated attack rate. If I actually do this using the overall vaccination ratio of the state to define the two populations, with the two week case numbers and the vaccination rates we looked up above, I get a number around 70%. Maybe not terrible, a path to herd immunity it is not though.
If the vaccines were really still 90% effective at preventing infection, vaccinated, infected people must be MUCH MORE likely to encounter severe reactions, in this case death, to reach the observed 25 people in a couple weeks. If they were 90% effective, you should only expect to see something like 2000 breakthrough cases in this time, and if they were then also reducing serious illness you should probably be seeing less than ~8-9 deaths in the 2 week period. There were 25.
It seems like you can't claim fully effective vaccines for the variants in circulation now and be able to explain the vaccinated deaths we are seeing without making alarming conclusions about their effect on disease severity. Something is either horribly wrong with calculating vaccine efficacy based on week to week case numbers, or vaccines are horribly less effective (or both). The more conspiratorial among us might question if this is why its hard to come up with public data that correlates covid cases with vaccination status, and existing reporting only presents cumulative numbers extending back to the beginning of the year, when more vaccine friendly variants dominated. (and are also quick to point out a week’s case/death numbers with respect to the entire population of the state or millions vaccinated there) There do not seem to be public trend plots of breakthrough cases or ratios w.r.t. unvaccinated cases? To do this right the data needs to be available — need to know the vaccination status of each case, so one can try to consider proper correlations between age, dates of vaccination, etc.
This deserves a proper treatment, especially since it appears we're seeing significant breakthrough infections among a population of people who have been told that by being vaccinated, they are protected. Who were told in the spring by the highest level of government that they can safely de-mask and go and hug their family and friends, can go freely canoodling in bars and concerts as in times of old.
This fall and winter is going to suck. But of course, no worries. The vaccines are effective.