Here we are, in the middle of this very serious epidemic, trying to figure out exactly what case, hospitalization, death, etc. trends and relationships are, and you can’t get data that is worth crap from the state. At first, I was inclined to be understanding and just assume they had a million things to deal with and were having trouble figuring out was relevant and what wasn’t and what the best formats were to provide information. Now, at least in Minnesota, because I and others have suggested and asked for certain data or formats, I think it is an intentional obfuscation. The Dictator and his staff know they screwed up big time, relying, and they did rely completely on it, go back and look at the videos of the press conferences, on a model that was worthless to make decisions that destroyed the state unnecessarily. Now they are trying to keep people scared and justify their past and continuing actions. It is beyond despicable. Here is a perfect example.
You might assume than when the states report deaths every day from coronavirus, those deaths occurred on that day or maybe the day before. Well, you would be wrong and the charts produced from the deaths by date reported data are misleading about when an epidemic might have peaked, at least in terms of deaths, in a particular area, and the extent to which it is petering out. Date of the death is the only useful metric. Date reported is subject to lumpiness and other issues and makes it difficult to estimate lags from infection to death and hospitalization to death and to calculate certain ratios. This is critical to understanding trends. (While we are add it, the other data the state won’t give us, but has, is antibody test results, hospitalizations by date of admission, date of discharge, average length of stay, how many people were hospitalized who already had a confirmed diagnosis by test, and how many people were admitted and then tested positive.)
Minnesota, like most states that I have looked at, is only giving us date death is reported data, which is pathetic because they know date of death and they should give us that. And so I kept rooting around trying to figure out where to get it when all of sudden I realized it was right under my nose at one part of the CDC site with data on deaths by date of death by state. To see this, go to this link and click on the area that says download data by state by date of death. (CDC Data) So I made the handy chart at the bottom of the post, which I will update regularly, which compares Minnesota’s date of report versus the CDC’s date of death data by week.
Note that while we don’t know the exact data lags, the state and CDC do and could give them to us. (If I knew how to find archived web pages you could figure out those lags by looking at the daily CDC updates of this spreadsheet and identifying the date various deaths were attributed to a set week, so if any of you know how to do that, let me know.) But the CDC notes say it can take at least 8 weeks later for a particular week to be pretty complete. So what happens is, a person dies, a doctor or someone fills out the death certificate, hopefully accurately, the doctor or medical examiner signs, the certificate is shipped to the state, and then by the state to the CDC. So you are dealing with communication and shipping and processing lags, which probably aren’t consistent for every death. Almost certainly weekends look like they have lower deaths on date reported methods because people are slower to get the certificates in. The state is probably also getting indirect death counts in other ways, which may show up in its death by report date data, but does not go to CDC. But the CDC data is from death certificates.
Now a couple of oddities, the CDC leaves blank a cell with a count of under 9, only the first and last weeks in Minnesota with a death have this problem and the last one is so incomplete it won’t matter anyway, but it does tell you that a lot of the deaths reported last week must have ended up in earlier weeks. The first week we know had only one death by cross-reference to the Minnesota data, which has its first death reported on March 21, which is the end of the first full week I used from the CDC data.
What you see in the table is exactly what you would expect to see given the circumstances described above. Deaths were higher in the earlier weeks of the epidemic than the state reported and they are almost certainly lower now than is being reported. Weeks in the CDC data probably aren’t complete past the end of April. So the epidemic in terms of deaths, peaked pretty early, kind of plateaued and now has likely declined sharply and will have a long, low right hand tail. Among other things, if you already thought deaths were low in the last two weeks, they are likely actually running a few deaths lower. And actual cases (detected and undetected) had likely plateaued even before the Dictator issued stay-at-home and business shutdown orders, indicating how unnecessary they were. A quick update, as I wrote this the difference between total deaths reported by the state and the CDC was 121. Those 121 deaths will eventually be received by the CDC and put in the weeks in which they actually occurred, so the pattern of deaths actually occurring earlier than the state data indicates will continue. And as soon as I get my rusty graphing skills going, I will turn the table into a graph.
|Week||MN Reported Deaths||CDC Date of Death|