Post #948: COVID trends through 1/12/2021

Posted on January 13, 2021

Source:  Calculated from NY Times Githib COVID data repository.  All graphs below have the same data source.  Thick blue line is the U.S. average.  Dotted red line is my guess as to trend.  The dips in the blue line are true reductions in reported cases during/following the holidays, as discussed in Post #929.

This post is something of a catch-up.  It’s about current trends, and about the post-Christmas surge.  Which, you might notice, you aren’t reading much about, currently.  And it’s about correcting some recent errors I made.

First, some clean ups. The following posts were wrong:

Post #944: Last of the holiday data anomalies, 1/10/2021.  I thought that little quick uptick was the last vestige of the glitches in data reporting around the holidays.  But in hindsight, that clearly was not the last of it.  And if I’d had the good sense to be able to count to small integers (7-day moving average), I should have known that.

Which means that the companion post, Post #943: The post-Christmas surge is here, is also probably wrong.  Because, as it turns out, a lot of what I was reacting to was the last of the holiday reporting anomalies, working their way through the seven-day-moving average that I use to plot the data (and that every else uses as well.)

Why were those wrong?  No sooner had I published them than more artifacts — in the opposite direction — appeared in the data.  And I could look back and see that this rapid rise and rapid fall was a purely mechanical artifact of the seven-day moving average, moving past the “speed bump” of post-holiday catch-up reporting.

Below you see, in effect, the rear wheels of my seven-day-moving average going over the speed bump that I thought I had already crossed.  Said speed bump being a lot of catch-up reporting and changes in reporting that occurred on 1/3/2021.  The graph is from data report through 1/10/2021.

And, in the spirit of driving an analogy straight into the ground,  as soon as the seven-day moving average cleared that, the rear wheels dropped back down onto the actual roadway, meaning, the true trend.  And that brings us to the picture at the top of the posting.

And that brings me back to the current trend, where I’m now saying, huh, is there really a post-holiday surge?  Or are things just returning to the prior trend, after all the direct effects of the holidays are over?

And so, we’re back where we were with Thanksgiving.  If the current rate of new cases is high, people are going to call that a post-holiday surge.  When, in fact, nothing of the sort has occurred.

Look at the pictures, judge for yourself.


State graphs.

Recall the rules for identifying a post-holiday surge.  We’re talking about an increase in cases caused by the holiday.  That is, new infections above the existing baseline trend, caused by the travel and socializing associated with the holiday.

1:  A post-holiday surge ought to start somewhere between 12 to  21 days following the onset of holiday-related travel and socializing.  That’s based on the lag between infection and onset of symptoms, and then seeking care and having the COVID test reported.  Based on air travel data, the start of holiday travel is the Sunday before Christmas (documented in Post #939).  And so, the “surge” should start no earlier than 1/1/2021, and no later than 1/10/2021.

2:  The hallmark of a holiday surge is simultaneity.  We ought to see it across all or most states, all at the same time, because whatever holidays we celebrate, those are all celebrated at the same time.

3:  And we’re looking for a sharp upward departure from trend.  If new cases are rising before the holidays, and they rise at the same rate after the holidays, you really can’t call that a surge.  I.e., the holidays didn’t seem to change anything.  Look at Post #920 to see what I mean, based on a summary of the past 20 flu seasons.

How are the states stacking up, by those rules?

California is getting a lot of press for its post-holiday surge.  But, in fact, California has been moving sideways for the last three weeks.  There has been a modest upturn in the seven-day moving average.  But I’m pretty sure that this is not what people mean when they say “surge”.  Lot of new cases for sure, but no more than they had before Christmas.  So far.

Separately, Alaska and Washington may have something that appears to be “a surge”, but I don’t think that’s correct.  That’s the decline in new cases that occurs on and during the holiday.  Followed by a return to normal.  Trace through the Washington State “surge” above, and you’ll realize that all you do is return to the level of daily new cases that Washington was reporting prior to the holiday.

Just looking at that one graph, you can pretty much dismiss the idea of a post-holiday surge.  Unless they don’t celebrate the holidays in Oregon.  Because, for sure, what you’re not allowed to do, under the rules, is pick and choose.  If all the states faces the same holidays, you can’t take the ones where rates increased after the holidays and somehow attribute that to a post-holiday surge.

 

The Midwest was and is in lockstep, showing a dip in reported cases around the holidays.  Now, two weeks out, all of them appear to be over that, appear to have returned to prior levels of new case reporting, and appear to be read to resume their prior downward trend.

Now, that sure looks like the lines are starting to form an arch on the right — starting to form the downward leg of that arch.  And if that were just one state, you’d be nuts to make anything of that.  Could be random.  But when it’s a dozen states, spanning maybe 1000 miles east-to-west, and all of them seem to be doing the same thing?  Yeah ,that would be a heck of a coincidence, if that were purely by chance.

We’ll need another week or so to be sure, but that doesn’t look like a post-holiday surge to me.  That looks like a dip in actual new case counts during and following the holiday, followed by a return to a pre-existing downward trend.

Or, if that’s “the surge”, it isn’t much.

Mountain states.  Everything on the bottom of the graph looks just like the Midwest.  No surge there.  But AZ, UT, and arguably NV have increases that match the rest as to timing, but are vastly larger in extent.  AZ and UT hit new high rates of new cases per day.  Plausibly, those three — but only those three — meet the timing and extent criteria for a “post-holiday surge”.  But why it that would only occur for those three is a mystery.

South central states:  Long-standing upward trend.  If I didn’t have the dates on the bottom line, you’d be hard-pressed to tell when or whether a holiday occurred.  This is just an unbroken continuation of the slow upward trend that started around the end of October 2020.  No surge there, just a long-standing trend.

Northeast and South Atlantic:  Slow upward trend, with variations.  This is more-or-less the same story as South Central, with more variation around the trend line.  In the typical state, new case counts continue to rise slowly.  If there’s a post-holiday surge there, I sure can’t spot it.  Maybe DC/MD/VA (bottom of the SA chart).  Maybe not.


Summary.

It took longer than I anticipated for the direct effects of the holidays to work their way through the data.

Mainly, there is a true reduction in reported COVID-19 cases during the holiday season.  This occurred at Thanksgiving (in hindsight), and it occurred, to a larger and longer extent, for the year-end holidays as well.

This is confounded by glitches in the data reporting — states that report zero cases on the holiday, then double cases the next day.  States that change their reporting definition at the start of the year to include (e.g. ) cases identified by antigen tests as positives.  And so on.

Once that all of that truly washes through the data, then the picture starts to get clearer.  As with Thanksgiving, there is no evidence for a nationwide post-holiday surge in cases.  But beyond a doubt, every place where the new case rates are now high, the popular press will refer to that as a post-holiday surge.

But in terms of an actual cause-and-effect surge — a sharp upward departure from prior trend, across all or most states, all at roughly the same time — no such thing occurred.  Or, if it did, it was far too subtle to be seen on a graph of cases, which is close enough for me.

No doubt, the most troubling thing in the entire national picture is that the only region where the new-case rates might appear to be heading back down is the region where a large fraction of the population has already had COVID-19. 

That argues that we may have been slowing down the spread, but we haven’t been stopping it.  Which I guess we’ve been told, all along.

At some point, one way or the other, enough people will be immune to this that it will no longer spread.  Everything between now and that point is just a question of minimizing the damage as we wait.

If this coronavirus follows the seasonality of other human coronaviruses, new cases rates will begin to recede before spring.  All you can do is try to stay safe in the interim.  Truly, this is public health straight out of the pre-World-War-II era.