Transmillennial

Excess death data through 6/13/20 were released yesterday. I've established, comparing a series of CDC data releases, that the 2020 data up to May 30 are more-or-less complete. Recorded COVID cases through May 30 are 66.2% of excess deaths over the same period. So we're still counting about two COVID deaths out of every three. This is the total death count by week (blue) versus a typical year (orange). The death rate was still dropping on May 30; it was about 925/day. The undercount rate seemed to have steadied by 5/30; it's possibly slightly declining.

I've updated the data to May 23, using the June 17 release of data from CDC. This also adds some counts to earlier data. But the basic picture remains the same. The calculated undercount (Johns Hopkins deaths to May 23, divided by excess deaths to May 23) × 100% is 65.3%, slightly lower than the previous estimate. The curves took a little smoother, ad we seem to be settling down to an undercount of slightly under 75%. The fraction of over 85s (blue) dying has risen somewhat. Still, steady state seems to be that 60% of the deaths are over 75

A few things I noticed about the undercount data ( see first post here ). Firstly, the undercount was severe at the very beginning, but by May 5 (week 18 of year 2020) or so we were counting all or almost all the deaths. Then the fraction counted seems to start dropping again. And the deaths by age are interesting. About 30% of deaths fall into each of the over-85 (blue) and the 75-84 cohorts (orange). Of course, the older cohort has a higher death rate, but the cohort is smaller. And the very old (over-85) seem to have a spike right at the beginning. They die quickly once the surge starts. That makes sense, but it's interesting it's verified in these data.

Every morning, if you like, you can go to the Johns Hopkins dashboard and get the current count of COVID-19 cases and deaths in the United States. Neither is particularly reliable; we don't test everyone, so we don't know the real number of cases. And COVID-19 deaths are not reliable either. Here's why: At the beginning of the epidemic, in good faith we were failing to code deaths as COVID-19, because the disease was still mysterious at the time Very old people often don't show the canonical symptoms of COVID-19, and simply die unexpectedly. Unless you make an effort to test, you don't know COVID-19 was the cause And finally, some states' governors, all or mostly Republicans, are for political reasons miscoding COVID deaths as pneumonia So these day-by-day numbers, while useful and fairly immediate, don't catch all the deaths. Fortunately there's a more reliable but slower alternative. All deaths in the US are recorded by the CDC as death registration