An epidemiologist wrote a letter to the Strib that was published today, critiquing my analysis primarily for use of proportions only. I sent her the below email. My primary point is that at the time we wrote the op-ed, we had to use the data the state had made available. We have better data now and are working on additional analyses. I have several times, including yesterday, outlined the benefits and limitations of various approaches, and the need to have data to provide all the analyses. And let me further point out that even on the proportion approach you get some interesting data. For example, in the four week period from early September to early October in 2020 there were a total of 233 deaths, obviously all in unvaccinated persons. For the similar period in 2021, there were 151 deaths in fully vaccinated persons alone. There were an additional 214 deaths in the unvaccinated (once more, let me emphaxize that fully vaxed events are undercounted and that people with one dose are considered unvaxed). That should be a concern. Here is the email I sent:
“I saw your letter and just wanted to give you a full context and hopefully you will even help me gather and present better information to the public. I want to say first, which should have been in the column, that I am fully vaxed and boosted, and I am not anti-vax in any way, but I am pro-truth and it is not true to say that this is even predominantly an epidemic of the unvaxed.
If you read the blog at all, and especially the posts about Minnesota’s breakthrough data, you will see that I understand the complexity of statistical analysis pretty clearly and have caveated all my analyses with explanations of data limitations and the virtues and flaws of various methodological and analytical approaches. It is not possible in an op-ed aimed at the general public to go into a detailed explanation of those.
If I had not pushed, and in fact insisted through Data Practices Act requests, that the state begin releasing breakthrough data, we wouldn’t have any at all. And I am still pulling teeth to get complete data. After the column was written for example, the colleague who works with me received a data file from the State which for the first time had actual dates of cases and ages. We are pushing for dates of hospitalizations and deaths as well.
We got that file because the per capita files originally released by the state are corrupted and inaccurate, they don’t foot to or reconcile with any other data, a problem the state has now acknowledged.
In the context of breakthrough events, all we were initially given was data by date of report. Since that is all we had, we could only focus on comparing the breakthrough events reported to all events reported in the same time period. We tried to guess about relative lags in reporting, but we really had no good sense of that. Then about a week ago the DOH indicated that the lag was roughly four weeks and we were able to give a better sense of actual proportion of events over time. That was startling, not just for the percent of events that were among the fully vaxed, but how it was growing over time. Finally we got some information with actual dates of events, although it is as yet incomplete, and we got some age data, which is critical. More analysis to come.
The look at proportions of events really has very limited value. It serves mostly as a check on the notion that this is an epidemic of the unvaxed. Far more revealing are case rates–what proportion of breakthrough cases result in hospitalization or death and compare those to non-breakthrough case rates. And this should ideally be broken down by age. This helps understand vaccine effectiveness and since age is obviously such a factor in serious disease, you can understand vaccine effectiveness by age group. Our early work on the new file indicates that those rates are indeed much higher in the unvaxed, but that effect greatly lessens with age.
Per capita rates or rates of events in a given population are also important, but obviously the data has to be accurate, and as I indicated above, the state has acknowledged that those original files they released are screwed up. So comparing rates in the vaxed and unvaxed populations, comparing rates in various age groups, comparing across vaccines, this has tremendous utility in understanding vaccine effectiveness and epidemic trends. Variation in those rates over time is also important. And one very important and usually missing piece in analyses of vaccine effectiveness is the role of time since last dose. Ideally you would have that variable for every person. The same is true in trying to compare the effect of prior infection.
So what I would love to have, and DOH has every bit of this data or can easily access it, is a master spreadsheet that for every person has their age; if vaxed, the date of each dose and vaccine received; if ever infected, the date of infection, and for each infection in the person, the date of that infection (to detect reinfections and breakthroughs); whether the person was hospitalized and when for each hospitalization in the person; and whether they died and when. With that spreadsheet you could calculate relative risks on a days of exposure basis and you could do all the case rates and per capita rates. So, we have demanded that DOH provide that.
It is in the interest of sound public policy that full data be released and be available for independent analysis. And it is clearly true that even with 70% of the population, including children, vaccinated in Minnesota we have a high level of cases, hospitalizations and deaths. We had almost as many deaths in Minnesota in September and October this year among the vaccinated, as we had in September and October last year when no one was vaccinated.
So anything you can do to get fulsome data released and to conduct your own analysis of such data would be helpful. Thank you.”
The epidemiologist very kindly replied and said while she doesn’t work in infectious diseases and isn’t familiar with the state data, she was concerned that a misleading impression could be given that vaccines aren’t working. I agree with that, which is why I try to caveat everything and now that we have better data, we can actually do some work on case rates, by age, and on per capita rates.