Post #1229: COVID-19 trend to 9/1/2021, back-to-school uptick

Posted on September 2, 2021


Yesterday, there was a little uptick in the numbers that I didn’t expect to see.  Today, the message in the tea leaves appears much clearer.  Enough so that I’m going to jump to a conclusion based on two days’ deviation from trend.

It’s starting to look like we can toss my prediction about the path of the Delta wave into the trash.  But the upside is that we can do that for a very specific reason: School is back in session.

Even if that’s right — if this current little uptick in cases is a back-to-school uptick  — it’s not yet clear the extent to which that reflects more actual infections, or merely more screening, testing, and discovery of infections.  That’s what I’ll try to look at next.

The data

I’ll start with my usual chart.  This last little change doesn’t look like much, but when you’ve stared at these numbers as much as I have, it does stands out.  Particularly given that this was foreshadowed by yesterday’s number.  In other words, it doesn’t look like much, but to me, this looks like something new.  This isn’t the previous predictable dynamic of the pandemic playing out.

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 9/2/2021, from”  The NY Times U.S. tracking page may be found at

In this case, the picture is a lot clearer if I do the same plot in “natural units”, that is, in the count of cases, not the log of the count of cases.  Now this last little uptick really sticks up.

And things were moving along so nicely before this.  The data were trending pretty much as predicted, by eye, a couple of weeks ago.  But in the last two days, the data have decisively broken through that projected trend line.

Still not very impressive.

Here’s the chart of growth rates.  Now the uptick looks a little clearer.  Not a huge deviation from trend, but clear some deviation from trend.

But, separately, there’s another little oddity that I’ve been watching.  I just hadn’t put two and two together until this morning.

The U.S. CDC shows that cases are ticking up, but hospitalizations are ticking down.  I’ve been dismissing that as some sort of data reporting glitch.  But now I have to ask, well, doesn’t this mean that the new cases have a lower likelihood of hospitalization?  Given that neither the virus nor the practice of medicine can turn on a dime, that implies that the cases are shifting toward a population with lower risk of hospitalization once infected.

Source:  CDC COVID data tracker, accessed 9/2/2021.

The hypothesis

To summarize the above, we’ve had not one, but two unexpected things happen this week.  The new case counts ticked up, beyond the trend line that I expected to see.  And the new hospitalization numbers ticked down.

Is that just a coincidence?  Well, FWIW, I don’t think I’ve ever seen that combination of cases up/hospitalizations down before, in this pandemic.  To me, that suggests that this is unlikely to be a coincidence.

First, recall that it takes about two weeks for new infections to work their way into the data.  (Or, at least, it used to take that long, back when the only test was a PCR test (no antigen (instant) tests) and all the COVID-19 infections were in adults.)

That means that we need to start thinking about what might have happened, in mid-August, in the U.S., that would have materially increased risk of infection.  Some event large enough to affect the numbers for the U.S. as a whole?  And because we know that infections outside the home occur almost exclusive in crowded indoor settings, what would have increased the amount of indoor social interaction of the population?  But that seems to be focused on a population that has lower-than-average risk of hospitalization, once infected.  (Which, if have to spell that out, means younger as opposed to older people).

So, what could  increased the total amount of indoor social interactions among (primarily) younger U.S. residents, sometime around mid-August 2021?

When I put that together, my tentative conclusions is that we’re seeing the impact of schools being back in session.

This really shouldn’t be a surprise.  Impact of school re-opening had already been noted as a factor affecting the course of the pandemic in other countries.  Perhaps it just took the vastly-more-infectious Delta variant to make that clear in the U.S. data.

Interestingly, if that’s true, the staggered school start dates across the country provide a sort of “natural experiment” for testing this hypothesis.

And Virginia itself provides an excellent natural experiment, as the most populous county (Fairfax) returns to school earlier than the rest of the Commonwealth.  (It’s a long story for non-Virginians, but has to do with the legislature’s desire to maximize the tourist dollar here in Virginia.)

So that’s the sort of thing I’ll be looking at next.  It’s one thing to leap to a conclusion.  It’s another thing entirely to present a careful analysis of the data.  But for now, I’m betting that the factor that perturbed my nice, neat top to the Delta wave was school openings.

Quick check.  Do the schools really get back in session as early as mid-August?

Let me give you one more chart.  If you read the blog, you’ve seen it before:


Source:  Google, accessed 9/2/2021.

Practically slaps you in the face, if you know to go looking for it.  Ask Google a simple question, and there’s the answer on the first page of results.  Doesn’t prove anything, but it’s consistent with the hypothesis.

I’ll do at little more data-based followup on this in the next few days.  I suspect that a week from now everybody will have reached the same conclusion that I just did.  This recent uptick is attributable to school reopening across the country.