Post #1382: Gap-fill using second-order extrapolation.

Posted on January 3, 2022


If you read this blog, you’re aware the most states don’t bother to release new COVID-19 case counts on holidays or weekends.  Also, you know that we have to wait a day to get the full set of U.S. data.  Together, those two things mean that tomorrow (Tuesday) will be the first time in several days that we’ll have been able to get an accurate fix on the level of new COVID-19 cases in the country.

One of the few benefits of this is that nobody knows what the rate actually is, right now.  So newspapers have to be quiet on that subject, for a couple of days, waiting for the new data to come in.  Same as I do.  Which is something of a relief, in the current situation.

Up to now, my solution to the missing data has been to assume that states carry on at their current level of daily new cases, until such time as they report new data.  If their seven-day average stood at 100 on Friday, I assumed it would be 100 on Saturday and Sunday as well.  Until the state finally reported new data on Monday.

In effect, my old gap-fill method assumed that every curve remains level until new data show otherwise.

In normal times, that’s a pretty good gap fill.

These aren’t normal times.

For today’s estimate, I ginned up a “second order” gap fill.  Instead of assuming the level of cases remains constant, this one assumes that a state’s most recent trend remains constant, over the period where data are missing.  If cases were increasing (e.g.) 10 percent per day just prior to Friday, I’ll now assume they continue to increase 10 percent per day over the weekend.

In effect, the new gap-fill method assumed that every curve remains at its prior slope until new data show otherwise.

Really, this is just the nerd’s way of extending the curve along the existing slope.  It’s a bit nicer than that, as it uses current data where available.  But that’s the gist of it.

So, take this for what it’s worth.  Here’s my best guess as to where the U.S. actually stands on Sunday 1/2/2022:

Based on the same rate of growth, we can plausibly expect the data for Monday 1/3/2022 — the data that will show up tomorrow — to show about 180 new cases / 100K / day, for the U.S. as a whole.

I like to put a prediction down on paper to keep things honest.  But this one serves two more purposes.

First, it’s a good test of whether or not U.S. new case growth is slowing.  If new case growth actually is slowing down — and in South Africa, that happened abruptly — then tomorrow’s number will come in well under 180.

And if it hasn’t, and tomorrow’s number, based on actual data, comes in around 180, then we’re definitely into the territory where we’ll start to see as many daily hospitalizations from Omicron as we saw at the peak of the Delta wave.  (That was, recall, where two states — AK and ID — declared crisis standards of care because they’d run out of ICU beds).

Second, this is just to get people ready for the shock.  Because if new case growth didn’t slow down, it’s a good bet that the new cases numbers will be all over the media tomorrow.

One caveat is that, even assuming I did the arithmetic right, there’s no way to know how good this new extrapolation is.  It’s the first time I’ve tried it.  But the bottom line is that simple extrapolation-of-trend puts us well into the hospital-admissions territory that caused trouble last time.  If we actually end up where my prediction suggests, then only hopeful thing we can point to is the lower ICU use per case, so far, for Omicron compared to Delta..  So maybe even if we fill all the acute-care beds, maybe we won’t fill the ICUs.  Just yet.

I never thought the U.S. would get anywhere close to this number of cases.

But I never thought we’d still be sleepwalking through the pandemic, either.  I see where the U.S. House of Representatives is finally going to require not just a mask, but a good mask, when in the Capitol complex.  Plausibly a NIOSH-certified N95.  All I can say is, what took you so long.  And how about suggesting that for the rest of us.