The U.S. now stands at 63 new COVID-19 cases per 100K per day, down 43 percent in the last seven days, and down almost three-quarters from the 1/16/2022 peak of the Omicron wave.
Just eight states remain about 100 cases / 100K population per day, and no states exceed 200.
So far, there is still no clear sign of any letup in the rate of descent. That makes the U.S. experience unusual (possibly unique) by international standards.
Could this lengthy, steady, and extremely rapid descent reported positive cases be a consequence of some sort of problem with the data, such as the rise of in-home testing?
Turns out, the answer to that is yeah, it just might. Hospitalizations and deaths are not falling anywhere near as fast as reported cases. Best guess, that only started to happen in early January.
The trend, FWIW
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 2/11/2022, 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
But is it real?
I keep getting questions about the effect that in-home testing might be having on the numbers. The common assumption is that we’re understating the total number of positives. Now that’s morphing into a concern of a possible overstatement of the decline in the Omicron wave.
I looked at this issue once before (Post #1397, 1/12/2022).
At that time, it was clear that a lot of home tests were being sold. The projection was for 200 million tests to be sold by February 2022. In addition, we now have Federal and local governments supplying those test kits for free.
That said, at that time, nobody knew how many were used, let alone how many gave a positive result. The only evidence I could find, at that time, suggesting that self-testing might be reducing the official count was based on CDC’s seroprevalence data. Between September and October 2021, there was a modest increase in the gap between persons with evidence of prior COVID-19 infection, and the total number of cases that had been officially diagnosed.
Otherwise, there’s already so much slack between the official count and any “true” count that I just couldn’t get too worked up about out. We already know, for example, that there have been at least twice as many actual COVID-19 cases in the U.S. as the official statistics show. (That comes out of the CDC seroprevalence data.)
And, finally, given that there is a large undercount in any case, even without home testing, these new tests only matter to the extent that their use changes over time and diverts positive results that would otherwise have appeared in the official statistics. If that happens, that will perturb the trends. But if it’s just a constant percentage, it’s really no different from what we have now. The official statistics will still show the correct trends, just not the correct total.
Yesterday a sharp-eyed reader forwarded an article with a link to some actual numbers on positive COVID-19 diagnoses from home tests, in one New York county. Based on that, I’m going to revisit this issue.
Seroprevalence data: Nothing through December 2021
Through December 2021, there’s no evidence of an increasing gap between the official count of positives and the estimate of the number of people who have survived a COVID-19 infection. That’s based on a count of individuals who have COVID-19 antibodies in their blood. Near as I can tell, the ratio of “true” cases (persons with antibodies) to officially-tabulated positives has been a more-or-less a constant 2 to 1.
- September 2021: 2.0 to 1
- October 2021: 2.2 to 1
- November 2021: 2.1 to 1
- December 2021: 2.0 to 1
Source: Calculated from the monthly graphics on the CDC COVID data tracker.
Hospitalizations: Rapidly rising case hospitalization rate?
Hospitalizations provide a second benchmark for assessing whether or not there’s been a significant change in the reporting of positive cases. Plausibly, people who need an inpatient level of care for their COVID will more-or-less need that care, regardless of what the initial test was. If there is no change in the underlying variant, there should plausibly be no change in the case hospitalization rate (ratio of hospitalizations to cases).
That’s not strictly true. The case hospitalization rate wanders around a bit randomly from time to time. But in the main, that should be roughly true. Once the case hospitalization rate dropped for Omicron, it should stay down as long as Omicron is the dominant strain.
By contrast, if an increasing share of positive results is being diverted out of the official statistics, the calculated case hospitalization rate should rise. The denominator of the case hospitalization ratio will shrink.
And, hmm. That’s exactly what’s happening. The case hospitalization rate is rising rapidly. Here it is from the start of hospital data collection in August 2020, and then again for just the Delta and Omicron waves. (I’m not sure what happened circa June 2021, but that’s the time when the daily new case counts were in the low single digits per 100K population, so plausibly that extremely high case hospitalization rate is some artifact of the very low prevalence of disease at that point.)
Here’s the same information, just dropping the first year. So this is just the Delta wave and the Omicron wave:
Source: Calculated from data from the CDC COVID data tracker.
The drop in the blue line above is expected. That’s a reflection of the lower average case severity of Omicron, compared to Delta. The transition to Omicron ended some time around the first week of January. For the week ending 1/1/2022, Omicron accounted for 90% of new cases. And, correspondingly, the case hospitalization rate fell from the Delta level of maybe 7 percent, down to the Omicron level of around 2.5 percent, in the first week of January.
The rise in that blue line since roughly the second week of January 2022 is completely unexpected, and as far as I know, has not been remarked upon in the popular press. The apparent case hospitalization rate for Omicron has more-than-doubled in the last month, and is now near the level observed for Delta.
Another way of saying that is that new cases are now at 28% of the Omicron peak, but new COVID-19 hospitalizations are still at 56% of the peak. Hospitalizations are only falling about two-thirds as fast as reported cases.
Source: Calculated from data from the CDC COVID data tracker.
Data on deaths are harder to interpret because there’s about a two-week lag between new cases and deaths. That said, the deaths data seem to reinforce what the hospitalization data are showing. These are the Delta and Omicron waves, below. Note that deaths mirrored cases, with a lag, for the Delta wave. But that deaths are not doing that in the Omicron wave. As with hospitalizations, they are falling more slowly.
Source: Graph from CDC COVID data tracker.
I think I can rule out several plausible alternative explanations.
I can rule out that this apparent increase in the case hospitalization rate is simply an artifact of the rapid decline. (I.e., maybe hospitalizations lag cases by so much that the currently-reported hospitalizations reflect a much larger set of cases from some weeks back.) Reported hospitalizations never lagged cases by more than a day or two in the past, and that still seems to be true. The peak of the new hospital admissions for COVID is within a day or two of the peak in new cases of COVID.
I’m pretty sure I can rule out any mutation of Omicron leading to rapid increase in actual severity of illness of cases. I’m pretty sure that if there had been some new mutation that increased the average severity of illness that much, somebody would have remarked on it. And it’s completely implausible to suggest that such a mutation could have spread that rapidly, in a country that was already clogged with existing Omicron cases of the standard variant.
I don’t think the cross-section of persons infected has changed that much. If anything, it has probably continued to drift toward younger ages, which should reduce the average case hospitalization rate. And in any case, I’ve looked at the age cross-section for Virginia and that hasn’t changed much of late.
I am moseying toward the conclusion that, probably starting in early January, we’re seeing an increasingly large under-count of new COVID-19 cases, compared to the data prior to that.
And the obvious candidate for a reason for that would be ..
Home Testing
Thanks to a sharp-eyed friend of the family, I now have exactly one data point on the extent to which individuals are discovering that they are COVID-positive via home (over-the-counter) tests. The genesis of this is from this reporting via WNYC radio, in New York City.
The point to data reported by Tomkins County, NY (home of Cornell U. and Ithaca College). They have asked residents to report any positive home tests, and they then tabulate and make public the number of positives so reported. After some extensive searching, as far as I can tell, that’s literally the only place in the U.S. reporting that information.
They found that self-reported at-home tests accounted for about 20% of all the positive tests in Tompkins County for the past month or two (which is about as long as they have been tracking it).
They do not show any particular trend, though. Here’s my calculation, for two week intervals, from their data.
Source: Calculated from Tompkins County COVID data website, accessed 2/10/2022.
Now, I need to add many caveats here.
- There may actually be a lot more positives found via home testing that are not being reported.
- There’s no guarantee that these positives would actually have been reported in the official statistics. Tompkins County directs residents to see a professional test if they have symptoms or have been exposed. If residents followed those directions, these home-test results would all be from asymptomatic individuals.
- There’s no guarantee that people aren’t double-reporting their lab (non-home) tests here, despite explicit instructions not to do that.
And so on.
That said, taken at face value, this one data point suggest that home testing might be significantly skewing the official numbers. It doesn’t prove it, but it shows that it’s not entirely implausible.
Conclusion
The rise in the apparent case hospitalization rate, since early January, slipped right past me. And, near as I can tell, everybody else as well.
Best explanation I have, for the present, is that rapid growth in home testing began removing significant numbers of COVID positives that would otherwise have appeared in the official data. And so, it’s not that Omicron has suddenly gotten more virulent, it’s that we’re suddenly missing a lot of cases that would otherwise have appeared in the denominator.
But that also means that we have significantly overstated the rate of decline in daily new COVID-19 cases.
If true, this means that what has appeared to be fantastically good luck for the U.S. — this continued steady rapid decline in reported new cases — may not be a matter of luck at all. Plausibly, we’ve had the same sort of slowdown in the true decline in new cases that every other country has had. It has just been masked by the growth of home testing, which has been diverting an increasingly large number of positives out of the official statistics.
This is not proven by any means, but I’d say the whole story hangs together pretty well.
For me, this means de-emphasizing the case counts from now on, and focusing on hospitalizations and deaths.
