Post #1395: COVID-19 hospitalizations flattening out?

Posted on January 11, 2022


To answer the question in the title, um, yeah.  That’s how it looks.  We probably need another few days to be completely sure.  Let me now work through that.

The data source for this analysis is what the US CDC refers to as the Unified Hospital Dataset.  The actual file I’m using can be seen and downloaded at this U.S. DHHS link.  More-or-less every acute care hospital in the U.S. is required to submit data on a daily basis.  Data fields include not only COVID case counts, bed and ICU occupancy rates and such, but also include (e.g.) self-reported critical staffing shortages, the amount of PPE is on hand, how many doses of key medicines are in stock, and so on.

The file has a few quirks and takes a bit of getting used to.  It’s a process of sorting out how things work, and which pieces of of data actually provide any information.

For example, you may have recently seen that umpty-percent of hospitals are reporting a critical shortage of staff.  What you don’t see reported is that it’s been that way for a long time, well before Omicron hit the scene.  So, it may well be true and it certainly could be a significant problem for some hospitals.  But it’s not some brand-new crisis caused by Omicron. 

Source:  Calculated from Unified Hospital Dataset, cited above.

In Post #1391, I showed a lot of the oddness of patterns of COVID-19 hospitalization across the states.  In particular, COVID-19 case hospitalization rates varied seven-fold across states.  Among other things, this means that what we observe as “the national rate” is going to depend, to a large degree, on where those cases are showing up.

Source:  Calculated from Unified Hospital Dataset, cited above.

But of all the oddities of the file, the most important one here is a weekly pattern of under-reporting and catch-up.  It appears that many states allow smaller hospitals to skip reporting over weekends, and just book those new cases every Monday.  (Or something — the exact timing on this file still baffles me).  (Separately, the file also tails off in the last two days of reporting, but that doesn’t matter materially for this analysis because I stop before the last day of reported data.)

Anyway, the raw count of new hospitalizations looks like this:

Source:  Calculated from Unified Hospital Dataset, cited above.

I smoothed that out by ( … does anybody really care? … ) estimating and apply a percentage adjustment based on day of the week.  Not rocket science.  I calculated the percentage amount that things seemed to be up or down on each day of the week, historically, on average.  Then applied the reverse of that to the data.  blah blah de blahdity blah blah.  And blah blah blahblah Bob Loblaw’s Law Blog.  And blah.  This seems to remove any obvious weekly pattern, which is exactly what it should do, leaving this:

Source:  Calculated from Unified Hospital Dataset, cited above.

And when you lay that curve up against the count of daily new US COVID-19 cases, you can see that hospitalizations are not keeping pace with new cases.

Source:  Calculated from Unified Hospital Dataset, cited above.

And if you do the long division, you find that the case hospitalization rate continues to fall.  Under Delta, 6.5% (and up) of new COVID-19 cases ended up in the hospital.  Under Omicron, it’s now down to 2.6% and still falling.  Which is good news, for a change, I think.

Source:  Calculated from Unified Hospital Dataset, cited above.

Why the COVID case hospitalization rate continues to fall, I could not tell you.  Maybe cases are simply shifting to lower-hospitalization-rate states, such as Florida.  Maybe there really was some residual of the higher-hospitalization Delta hanging around.  (CDC says no to that, with a most recent estimate that 98.3% of cases were Omicron for the week ending 1/8/2022).  Maybe the age mix is continuing to shift toward younger, less hospitalization-prone populations.

But if I had to guess, I’d guess that physicians are slowly re-calibrating their criteria for hospital admission.  On the one hand, it’s generally frowned upon to send somebody away from the hospital and have them die.  On the other hand, it’s also frowned upon to admit a lot of patients with two- or three-day stays.  At the very least, that will eventually attract the attention of Medicare’s auditors, who will not only question why you have all those short-stay cases, they may very well say that Medicare isn’t going to pay for them as inpatient stays, forcing the hospital to re-file the bills as much-less-well-paid outpatient observation stays.

It’s a balancing act.  For a disease this new and this mutable, surely a lot of it is rule-of-thumb.  It would not surprise me if we started the Omicron wave with Delta-based standards for who should and should not require inpatient care.  And as the fact of lower mean severity of illness gets discovered, those standards might reasonably change.

In any case, for sure, as you can see from the graphic earlier in this post, the hospitalization rates are, in fact, all over the map.  It certainly appears that some practices — either data reporting, or standards for hospital admission, or both — are varying across the states.

The upshot of this is that, as best I can tell, COVID-19 hospitalization rates are not keep up with new cases.  Instead, they appear to be leveling off.  Given the ongoing increase in daily new cases, that’s a very good and very lucky thing.