Post #1373: Omicron average case severity in the U.S.

Posted on December 29, 2021


The point of this post is to try to get some handle on the actual, as-observed average severity of Omicron, compared to Delta, in the U.S.A.  The bottom line is that the U.S. results look more like research from Scotland and England — the case hospitalization rate for Omicron is about one-third that of Delta.  They don’t look as good as the results from South Africa, which suggested that the case hospitalization rate was one-fifth that of Delta.

The same analysis verifies that there has been an uptick in pediatric hospital admissions with COVID-19, as a fraction of the total.  Not a huge increase, but a marked increase that coincides with the Omicron wave.  Given that we have no national database that shows current new cases broken out by age, there’s no way to tell whether that’s just an artifact of higher infection rates among the pediatric population.

Finally, I think it’s too soon to say much about U.S. mortality rates from Omicron compared to Delta.  The well-aged South African data seem to show a case mortality rate for Omicron that is about one-fifth that of Delta.  But data from European countries — which may or may not be well-aged enough to provide a valid estimate — show a much smaller difference.

All told, this means that Omicron is not quite as benign as the flu.  Best guess, with current data, case hospitalization rate is about twice that of flu, and my guess for a case mortality rate would be about four times that of typical seasonal flu in the U.S.

You can see my original comparison of Omicron and flu in Post #1364.  If our results had matched those of South Africa, then Omicron would have had a flu-like case hospitalization rate.  Now that it’s looking like our results aren’t that good, the best estimate is that Omicron has not-quite-double the case hospitalization rate of typical seasonal flu.

U.S. case hospitalization rate

Let me start with the case hospitalization rate.   That is, the fraction of newly diagnosed COVID-19 cases that are admitted to the hospital.

Below is the crude case hospitalization rate.  The plot is the seven-day moving average of new adult and pediatric admissions with confirmed COVID-19 at time of admission, divided by the seven-day moving average of new COVID-19 cases.

Source:  Calculated from U.S. DHHS unified hospitalization data set, and new case counts from the CDC COVID data tracker.

You can clearly see what appears to be the Omicron impact circled in red.  I date the Omicron wave to 12/17/2021.  Before that, the case hospitalization rate was about 6.5%.  As of the last day show, it has plummeted to 4.0%.  That fairly abrupt change certainly has all the earmarks of cause-and-effect, given the relatively stable rate prior to that point.

That’s the easy part.

The hard part is guessing what the Delta/Omicron mix is, corresponding to that 4% case hospitalization rate.  And here, I’m not going to get all fancy.  That’s a seven-day moving average ending 12/27.  So the median day is somewhere around 12/23 or 12/24.  Which is roughly contemporaneous with the last CDC estimate of Omicron’s share of cases (week ending 12/25).  So I’m just going to take the CDC’s number.  I’m just going to assume that 59% of those cases were Omicron.

The rest is just a bit of trial-and-error with Excel.

  • If the case hospitalization rate under Delta was 6.5%,
  • and the case hospitalization under a 31%/59% Delta/Omicron mix is 4%,
  • then the case hospitalization rate for Omicron must be 2.3%

To sum that up, it looks like Omicron’s case hospitalization rate is about one-third of Delta’s rate.  That’s in the ballpark of more careful estimates from Scotland and England.  But much higher than the South African estimate that Omicron had just one-fifth the case hospitalization rate.  Plausibly, that would depend, in part, on differences in propensity to hospitalize persons in those countries.

A little sensitivity analysis is called for.  By far, the most uncertain aspect of this is the fraction of newly-diagnoses cases that are Omicron.  The CDC didn’t actually say that Omicron accounted for 59% of cases.  The CDC actually said that there’s a 95% chance that Omicron accounted for between 42% and 74% of new cases.  Like so:

Source:  CDC COVID data tracker, accessed 12/29/2021.

We can use the range of the CDC’s confidence interval for the percent of  cases that are Omicron to show how this calculation of the estimated case hospitalization rate would change.  Same algebra as above, just a different set of assumptions.

I think the important lesson is from the bottom line of the table.  At worst — using the upper bound of the CDC’s confidence interval for its estimate of Omicron as a share of cases — the case hospitalization rate for Omicron is less than half that of Delta.

Some caveats.

Is it too soon to say anything meaningful about Omicron’s case hospitalization rate?  No.  I already dealt with the objection that hospitalizations lag new cases in prior posts.  There are plenty of real-life lags between (e.g.) date of infection, symptom onset, and hospitalization.  But as-reported, with all reporting lags included, the new cases and new hospitalizations series are very nearly exactly contemporaneous. The average lag between changes in reported new cases and changes in reported hospitalizations is one day.

Is this estimate going to mistake changes in vaccination rates and such with differences between Delta and Omicron?  No.  Because Omicron has grown so quickly, I can compare the U.S. today to the U.S. just a couple of weeks ago.  More-or-less the same population, more-or-less the same vaccination rate.   Not much changed in the past couple of weeks.

It’s possible that there may be significant demographic differences between the Omicron-infected and the Delta-infected.  But a) I doubt that, and b) there’s no timely data source in the U.S. that would tell us that.  So that remains a caveat.

Footnote:  Pediatric admissions as fraction of total.

There have been a lot of stories about increased pediatric admissions for COVID.  But, you know, anything having to do with dying children always gets a lot of press coverage.  So the question is, for the U.S. as a whole, is it true that pediatric admissions with COVID-19 have increased disproportionately?

The short answer is yes.   Below you see that pediatric admissions have gone from about 2.5% of admissions with COVID-19 to about 3.5%, over the course of the Omicron wave so far. As with the graph above, that short, sharp increase has all the earmarks of being cause-and-effect.

Source:  Calculated from U.S. DHHS unified hospital admission data. accessed 12/29/2021.

In terms of the absolute number of cases, this amounts to about 350 children per day admitted to the hospital with COVID.

There are two things to keep in mind.

First, we don’t know what fraction of those are admitted for treatment for COVID, versus admissions where the COVID is an incidental finding.  Near as I can tell, the U.S. data track something akin to “COVID present on admission”.  They are not tracking cases with COVID as the principal diagnosis (the reason for admission, see Post #1351).

Second, we don’t know what fraction of new cases are pediatric cases.  This might just be tracking the trend in new cases, for all we know.  There is no timely national data set that breaks out the new case counts by age.

Case mortality rate.

I can’t even hazard a good guess here.

In addition to the case hospitalization rate, I’d like to have some idea of the case mortality rate. What fraction of diagnosed Omicron cases go on to die from it?  In the U.S, for Delta, for the most recent period, that number has been just over 1 percent.

It’s too early to see Omicron’s impact on the U.S. case mortality rate.  There’s roughly a two week lag between reported new cases and reported deaths, and two weeks ago Omicron was barely on the horizon in the U.S.

Instead, I looked at three countries where Omicron hit earlier than it did in the U.S.  It’s not clear that any of these provides a good comparison.

In South Africa, the most recent data are all Omicron.  Between they had a material increase in vaccination rate between their prior (Delta) and current (Omicron) waves.  The apparent five-fold reduction in case mortality might be due, in part, to concentrating vaccinations among the highest-risk (e.g., elderly) population.

For two other countries where we can plausibly compare the mortality rates (Great Britain,  Norway), you really can’t see much of an impact of Omicron on the case mortality rate.  I don’t really know why.  I can only guess that this remains an artifact of the long time lags required to gather the mortality data.  In both cases, we’re looking at new cases that appeared in late November/early December.  Even with those countries, Omicron probably wasn’t a big enough factor to have generated large changes in the mortality rate yet.

At the end of the day, in the U.S., all I can do is take a guess.

I’m going to guess that we’ll see about the same effect on case mortality as we see on case hospitalization.   Probably, the ultimate case mortality rate for Omicron is going to end up being somewhere around a third that of Delta.   (I say that mostly because that’s how the Virginia numbers look, for vaccinated versus unvaccinated persons, under Delta.  The reduction in the mortality rate due to vaccination is always similar to the reduction in the hospitalization rate.)

If so, that will make Omicron considerably deadlier than seasonal flu.  A one-third reduction relative to Delta would result in a roughly 0.4% case mortality rate for Delta in the U.S., compared to a 0.1% case mortality rate for flu in the U.S.

My conclusion is much the same as for hospitalizations.  Omicron appears much less virulent than Delta.  But it’s an exaggeration to say that it’s no worse than flu.  Best guess, given what I can get my hands on today, it’ll result in about twice the hospitalization rate and four times the death rate of flu.

(And, of course, those who claim COVID is no worse than “a cold” are just ill-informed.  First, “a cold” is not a single well-defined disease, but refers to the symptoms of any number of upper respiratory viral infections.  Second, near as I can tell, you can’t even find a case mortality estimate for “a cold”, presumably because deaths attributable to complications of a cold are so rare.  Third, the same goes for a case hospitalization rate for “a cold”.  People who claim COVID is no worse than a cold have probably never even tried to look up the actual case hospitalization and mortality data.)

In short, if you’re in one of the higher-risk populations, I think it’s still well worth your while to avoid this strain of COVID.  It’s as prevalent in the population now as it was at the peak of last year’s winter wave.  And while your odds of hospitalization and death, if infected, are lower now than they were then, they are still non-negligible.  Best guess, as of today, if you get infected, you’ve got a better than 1-in-50 chance of ending up in the hospital.