I am back in Minnesota, the world changed while I was in Florida–the spotlight moved off the epidemic and onto the destruction of a country and the wanton slaughter of its population. All of a sudden CV-19 isn’t that dangerous and we didn’t need all those restrictions. A reader asked me if I felt vindicated. Not really, I mostly feel angry. Angry that people who pointed out the obvious for two years were ignored and pilloried as Grandma killers. Angry that a generation of our children had their eductations and lives ruined. Angry the many people died because they were terrorized into not seeking health care they needed. Angry because we wasted trillions of dollars and jeopardized our financial well-being. So, no, I don’t feel any satisfaction at being right about what the appropriate response would have been. And I feel even worse because I don’t think any lessons will be learned.
Dave Dixon is a tremendous data analyst. I always knew what I wanted to do with the data and I knew what it said, but I lack the technical skills or the time to really dig into the data and to communicate it in an easy to understand fashion. Dave volunteering to help was the best thing that happened to the site. He is dogged in accruing all the data as it is released and in comparing various data sources. He finds and illuminates the most meaningful data. And his charts and tables always tell the story at a glance.
Dave noted to me a couple of weeks ago that some of the per capita rates DOH puts out comparing vaxed and unvaxed groups seemed to be changing dramatically. We had wondered about those rates since they were first published because they appeared very inconsistent with what other data was saying and they implied much greater effectiveness of the vaccines than any other research was showing. Dave, as I noted above, is astoundingly persistent in keeping copies of every data sheet as it is released. He has therefore been able to compare the per capita rates published over time and he has found an extremely large change in those published rates. I will add a chart to this post later, but wanted to get it out now.
I will just quote verbatim from Dave’s emails to me in a second, but first a brief explanation. Per capita rates obviously have a numerator–cases, hospitalizations or deaths; and a denominator–the population in each sub-group, in this case those who are vaxed or those who are unvaxed. There are cumulative rates at any point in time, but there also are rates published as of a specific date. Looking at those rates as of a specific date, we have seen very large revisions that really make no sense. You can understand some numerator changes. All cases are initially reported and basically treated as in the unvaxed, because DOH uses a separate process to match cases to their vax records (which by the way are incomplete). There is always at least a two week lag in that process because of the definition of being fully vaxed. So the numerator as of a certain data can change, but Dave tracks those changes and they aren’t typically large.
What is harder to understand are big changes in the denominator. The state constantly publishes the number of people who are totally vaxed. Pretty simple to subtract that number from the total state population and know the size of the sub-groups of vaxed and unvaxed on any given day. But we have seen massive changes in the size of the denominator, which have dramatically lowered the per capita rates, as detailed by Dave below. Now this could just be typical government incompetence, someone forgets to build an automatic data feed to update numbers or pulls the wrong numbers. But the fact that the state continually harped on these differences in per capita rates makes me suspicious that it was more than that; that perhaps the DOH was holding back on updating or just plain fiddling with the rates to try to make the vax look more effective than they are. We are trying to get an explanation, and DOH needs to give the public one.
Dave is continuing to work on this, and as noted, we will add some graphs and Dave is going to do his own per capita calculations. Dave’s comments, in the order of receipt:
This is interesting. In the graphic, and the underlying data, they have reduced the unvaxed per capita rates for cases, hospitalizations, and deaths by roughly 20% for all weeks going back to the very beginning. There are no corresponding changes to the VBT Counts. This is true for the 50-64 and 65+ age groups. At the same time there were minor increases in the corresponding vaxed rates, but increases did not go back to the first week like the unvaxed rate changes did. No notes on the web page to explain it.
I went back and checked, and after their population size error they had lowered all of the unvaccinated rates starting for the week of 12/27/2021. Those reductions were about 40%. This set of reductions of about 10-20% means that many of their case rates for the unvaccinated are less than half their initial published rates. For example. 65+ unvaccinated, for the week of 11/14/2021. The first published case rate was 3302.8 per 100k on 12/20/21. This was lowered to 1802 on 12/27/21, drifted lower to 1717 on 2/21/22, and then Monday further lowered to 1320 per 100k. For the unvaccinated 65+ age group the week of 11/14/21, hosp. were initially 797 per 100k, and are now 320.8. Similarly, deaths the week of 11/14/21, deaths were initially 237 per 100k, and are now 116 per 100k.
I’ve been looking at the changes in the case rate, hosp, rate, and death rate data again this morning, and nothing about it is consistent or makes sense.
VBTCounts did not change all the way back to 5/2/21, but the cases, hospitalization, and deaths rates mainly did change back to 5/2, especially the older age groups. Bucket shifting from unvaxed to vaxed may be a partial explanation, but there are plenty of weeks where the unvaxed event rate went down without a corresponding rise in the vaxed rate. I’m starting to wonder if what we are seeing is either some errors that got introduced into their charts, or they fixed older errors, superimposed on bucket shifting. The note about including 5-11 year olds last week is still suspicious, but this wouldn’t explain the changes to the older group rates.
Another nagging doubt I have is that it isn’t clear how they are handling reinfections. I think reinfections should be excluded from the vax/unvax rate analysis, and calculated separately, because we don’t know whether the reinfection occurred in a vax or unvax person. This could be another source of error. I also have reinfections by age and date so I can calculate reinfection rates separately, but I don’t have reinfection hospitalizations or deaths by age, only cases.