The U.S. is now down to 10.5 new COVID-19 cases per 100K population per day, down 18% over the past seven days. The rate of decline continues to slow, but we haven’t reached bottom yet.
I think I have abandoned all hope of understanding what’s happening for anything beyond the new case counts. Hospitalizations and deaths have not come down anywhere near as much as new cases. There are several plausible explanations, but none of them really seems to explain the size of the discrepancy.
The death rate is particularly puzzling. On paper, the COVID-19 case mortality rate for the U.S. is approaching 2 percent. By contrast, in the U.K., recent reporting pegged it at somewhere around 0.3 percent. That’s a vast difference.
The only explanation I can think of is some sort of data reporting oddity. And, sure enough, what the CDC COVID data tracker says about deaths does not match what some large states say about deaths. Something is off with whatever source the CDC is using to track COVID-19 deaths.
Details follow.
Case trend
Data source for this and other graphs of new case counts: Calculated from The New York Times. (2021). Coronavirus (Covid-19) Data in the United States. Retrieved 3/15/2022, from https://github.com/nytimes/covid-19-data.” The NY Times U.S. tracking page may be found at https://www.nytimes.com/interactive/2020/us/coronavirus-us-cases.html
Short summary: Newly-reported COVID-19 cases continue to fall, but at an ever-slower rate.
Hospitalizations and Deaths
Note: I edited this section a few hours after posting, when I went back to the California data again. Some original text remains as struck-out text.
The odd thing about the downside of the Omicron wave is that hospitalizations and (particularly) deaths have not fallen in proportion to the reduction in cases.
The first graph below plots all three quantities, synchronized so that the peaks are all on Day One. That meant shifting the hospitalizations number by about a day, and shift the deaths number by about two weeks.
Source: CDC COVID data tracker (cases and deaths), calculation from US DHHS uniform hospital dataset (admissions).
As you can see, hospitalizations and (particularly) deaths have not fallen anywhere near as much as reported cases. Possible explanations include a falloff in official case statistics due to widespread use of home testing, or possibly a falloff in immunity among the elderly, who were first to be immunized and who account for the majority of deaths and hospitalizations.
Source: CDC COVID data tracker (cases and deaths), calculation from US DHHS uniform hospital dataset (admissions).
U.S. deaths data show COVID-19 deaths increasingly concentrated in the elderly. But a direction calculation on U.S. hospital admissions shows only a slight shift toward the elderly during the Omicron wave. The chart below shows the elderly’s share of COVID-19 admission starting December 1, 2021.
Source: US DHHS uniform hospital dataset.
Arguably the purely weirdest aspect of this is that the the case mortality rate (deaths per case) continues to climb in the U.S., but has plummeted in the U.K. (as reported in the just prior post). The U.S. COVID-19 case mortality rate is approaching 2 percent. The reported case mortality rate for the U.K. is about 0.3 percent. That’s almost an order of magnitude lower. (Or has it now fallen to 0.03 percent? And so is it almost two orders of magnitude lower. See this reference).
And yet, the U.K. has the same vaccines and roughly the same vaccine timing as the U.S. If we’re seeing declining immunity among the elderly now, they should be as well. And the U.K. was ahead of the U.S. in terms of availability of home testing. So if we’re seeing home testing substitute for officially-recorded tests in some massive way, there’s a good case that the U.K. should be seeing that as well.
Given that I’ve just cast doubt on the only two “real” explanations I can think of, it’s time to take the low road and start casting doubt on the numbers themselves. In the past, states frequently did “catch up” reporting of COVID-19 deaths, sometimes months after the peak. Plausibly, if enough of that is going on, that might keep the reported deaths data artificially high.
But, in fact, there’s an even simpler explanation. Whatever the CDC is using for its deaths data is grossly at odds with with what some large states are reporting.
Pick a number, any number: First, from the CDC, here are the three largest U.S. states, showing new cases and reported deaths. Note that the new cases data is similar, but the deaths data is all over the map. In just those three states, you can see c) slowly declining deaths, b) plummeting deaths, and c) no change in deaths:
Source: CDC COVID data tracker, accessed 3/15/2022
The CDC’s death count for California looks spectacularly odd, because deaths haven’t fallen at all, even though the new case count is negligible compared to where it was two months ago.
So I decided to double check that. Below is what California itself reports as its COVID-19 deaths.
Source: State of California
And so, at least part of the problem is that whatever-it-is that the CDC is reporting for California, that has nothing to do with the COVID-19 deaths as they are occurring in California.
And, to be clear, this is a large difference. Of late, California is reporting on order of 10 deaths per day. By contrast, the CDC reports a seven-day total moving average of 2.6 deaths per 100,000. Which, given California’s population of about 39 million, means that the CDC is counting (39,000,000*(2.6/100,000)/7)) = ) ~150 deaths per day, or more than ten times more that what California is reporting.
And it doesn’t take look to look where to point the finger: At California. The chart above was by date of death. The chart below is by date of report: The date-of-report series looks like the CDC data.
Source: State of California
And so, the issue is that California takes a ludicrously long time to report COVID-19 deaths. In an era of steeply declining new cases, this creates a huge disconnect between cases and deaths.
That one state, by itself, does not explain the apparent high U.S. case mortality rate, but it makes me wonder how much more of that is going on, and what, exactly, the CDC is reporting.
So I then looked up Texas. And, even acknowledging that death certificate data will be incomplete toward the end of the timeseries, it sure looks to me as if deaths in Texas have declined far more than the CDC shows:
Source: Calculated from data downloaded from the State of Texas COVID-19 dashboard
Texas does not, however, provide ready access to deaths by report date. (Or if so, I couldn’t find it.) But I’d bet that if I could find it, deaths by report date would look like the CDC data.
Things are now getting clearer, I think. Whatever the CDC is reporting, in the case of two of the three largest states, it has little to do with the actual deaths, as they are occurring. And, note that for Texas, the CDC shows almost a month gap between the peak of cases and reported deaths, but the data directly from Texas show the standard two-week gap between the peaks (corresponding with the median time-to-death for COVID-19 decedents).
At this point, I can’t even guess what the CDC is reporting in its deaths data. At this point, I think I understand what’s going on. CDC isn’t simply reporting Texas deaths, by death date, with some time lag. If that were true, it wouldn’t have anything to report for the last few weeks of the time series. It’s probably reporting deaths by date of report from Texas, and Texas is another large state that takes its time about reporting COVID-19 deaths.
In any case, the bottom line is that the CDC ends up systematically over-reporting current COVID-19 deaths, for two of the three largest U.S. states, compared to what those states themselves are as the trend in deaths by date of death. And, I repeat, these are not small errors. As with California, the latest CDC-reported death counts for Texas appear to be a large multiple of what Texas appears to report.
How that could possibly happen, I haven’t a clue. But it means that any calculation of case mortality rates for Omicron, based on the CDC data, likely overstates it. We may actually be on the same path as the U.K., in terms of deaths. You just can’t tell that based on whatever-it-is that the CDC is currently reporting. Which, plausibly, is whatever-it-is that the states are reporting to the CDC. As opposed to what the post on their own COVID-19 dashboards.
Either way, the bottom line is that the true, contemporaneous death rate is almost certainly more in-line with the new case rate. And that you can’t use the CDC death rate to assess the COVID-19 case mortality rate, in an era of rapidly declining new case counts.
