The official count of U.S. new COVID-19 cases per 100K population per day now stands at 57, down 42% over the past seven days. The U.S., unlike any other country, is seeing an extended, uniform, smooth decline in new case counts, with no slowdown in sight.
See caveat section below before you get too excited about that.
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
Caveat
Part 1: Recap, the data are suspect.
Per yesterday’s post, I believe that the official count of new COVID-19 cases has recently picked up a significant and increasing downward bias. The most plausible explanation is a recent rapid increase in use of over-the-counter COVID-19 test kits. In theory, some people who would have gotten a lab test (and been counted in official statistics) are now discovering their COVID-positive status via an over-the-counter test and are no longer being counted in the official data.
I can’t directly prove that, because there is no reporting requirement for self-testing. But, for sure, a) estimates of OTC test sales are in the hundreds of millions per month, b) Federal and local governments are supply those for free upon request, and c) in the one U.S. county that captures and reports self-test positive, those make up 20% of the total known positives in the county. (That’s all documented in yesterday’s post).
I think the most telling argument is the rise in the apparent case hospitalization and case mortality rates. Either we’re missing an increasing fraction of cases, or somehow Omicron has suddenly gotten a whole lot more virulent in just the past few weeks, and nobody noticed. Hospitalizations per (officially reported) case are almost at the level that existed under Delta, and deaths per case seem to be following the same trend. (Again, see yesterday’s post). I don’t think hospitalizations and deaths are suddenly up. I think the official case count is suddenly down.
Part 2: That’s great news!
First, let me emphasize that if this is true, that’s great news. That’s exactly what a COVID-19 control strategy based on cheap rapid tests should do. The hope is that people who suspect they may be carrying the virus will self-test without hesitation, and if positive, will isolate and so break the chain of infection going forward.
In other words, if an emphasis on frequent self-testing works well, it ought to generate a increasing divergence between the true case count and the official case count. I just don’t think anyone expected it to work that well.
Let me rephrase that: I think that this is the first evidence that a Federal and state strategy centered around self-testing is working. That’s a bit of speculation on my part. But OTC tests have been available for the better part of a year now. Of late, they were selling at the rate of a hundred million a month. But somehow, only in January did this discrepancy in the new case counts become noticeable.
Best guess: This is just another instance of “free is different from every other price”. We see this in health care all the time, and it’s the reason that Medicare (and many other payers) strive to make preventive services free. If you charge anything at all for a preventive service — even a tiny amount — a whole lot of people will just skip it. But if you give it away, people won’t hesitate to take advantage of it. (Off the top of my head, I recall one large HMO-based study where a $10 copay for screening mammograms cut the mammography rate in half. That’s how sensitive people are to having to pay for preventive care.)
And so, if I had to fine-tune my explanation, I’d point to a strategy of providing free tests as the likely source of the January 2022 divergence of the official count of positives from the hospitalization and death data. Prior to January 2022, hundreds of millions of tests had been sold in drug stores and the like. But January was, by my crude reckoning, the start of large-volume government giveaways of test kits. And maybe that’s why we’ve seen this sea-change in the apparent fraction of COVID-19 positives that are not appearing in the official statistics. Free is different from any other price.
(In Virginia, I see this same effect playing out in reverse, in the new rules requiring stores to charge a nickel for every plastic bag. I don’t know about you, but that’s just enough of a nudge to get me to remember to bring a re-usable bag to the grocery store every time. Which was the whole point — any price other than zero will make you think about it. I’m now being forced to ration my dwindling stock of cheap plastic grocery store bags. And that’s a good thing.)
Part 3: Great news, but not for this blog.
To a first approximation, a blog that focuses on tracking the new case count is now toast.
The numbers are what they are. There’s no sense moaning about it.
It was already well-established that the official COVID-19 case count is only loosely related to the total number of cases. Best guess, based on CDC seroprevalence data, over the course of the pandemic, the number of true infections has been about twice the official count.
As long as that was stable, it didn’t much matter. One could still track short-term changes in the U.S. COVID situation by tracking the official count. If the ratio of official count to true count was some fixed number, changes in one mirror changes in the other.
By contrast, if the ratio of officially-counted cases to true cases is rapidly changing, trends in the official count can be misleading. In particular, it certainly appears that the extremely rapid, smooth, and sustained drop in the official count of new cases overstates the actual degree of improvement in the U.S. COVID-19 situation.
Part 4: A little peak-shifting exercise.
Let me use the Delta wave and Omicron wave peaks to show that the behavior of the case counts is now qualitatively different. I’m going to shift the peaks of the hospitalization and death counts so that they coincide with the peak of the official case count, then graph that, norming each series to a value of 1.0 at the peak. The result will show how rapidly cases, hospitalizations, and deaths declined, following the peak. All on the same scale, all in the same timeframe.
Source: Calculated from CDC COVID data tracker data, accessed 2/12/2022
It doesn’t take a genius to notice that these two graphs are qualitatively different. In the first — the Delta wave — the three lines are practically on top of each other. To my recollection, that has been the norm all along. In the second — the Omicron wave — there’s a big spread across the lines. The count of cases has fallen far more rapidly then either hospitalizations or deaths.
So, something ain’t right. And the obvious candidate for the source of the discrepancy is the rise of home testing.
Part 5: What next for me?
I’m still trying to get my thinking straight on this. I’m usually one to two weeks ahead of the popular press. Surely if I have noticed this, it will eventually catch the eye of some acknowledged “expert” in these matters, which will then duly be reported in the popular press.
And then we can squabble about the quality of the case counts, in addition to everything else.
In the meantime, I guess this means emphasizing the hospitalizations and deaths data, over the case counts. I really can’t see any alternative for a blog that purports to track the trends.
That has a lot of downsides. Not only do deaths lag cases by about two weeks, the hospitalization data can’t be down-scaled to state-level estimates. I can get the state-level admissions data, I just can’t make any particular sense of it in a cross-section as the apparent propensity to hospitalize varied considerably across states.
The upshot is that it’s not clear what the best way to track the pandemic is, going forward. At least not to me.
