Post #1343: Two calculations showing much lower illness severity for Omicron

Posted on December 10, 2021


The title of the article pretty much gives you the entire content.

First, I took data on South African COVID-19 cases and deaths from Johns Hopkins, and calculated a “two week lag” mortality rate.  That is, today’s deaths, divided by new cases from two weeks ago.  That’s my best approximation to the true mortality rate.  Near as I can tell, the case mortality rate for Omicron in South Africa is very small.  So small that I can’t distinguish it from zero given the current state of the data.

In addition, I took what could find regarding hospitalizations for Omicron in South Africa.  Most of what you read about that is highly misleading, because something like three-quarters of Omicron hospitalizations in South Africa are people who were actually being treated for something else.  They just happened to uncover an asymptomatic case of Omicron because everybody admitted to the hospital gets tested.  Adjusting for that, and taking a couple of day’s worth of reported data, best guess, the case hospitalization rate for Omicron, in South Africa, is maybe one-tenth of the rate in the U.S., for Delta.  That is once again my calculation, based on what data I could scrounge up.

And so, a bit of good news for a change.  Best guess, based on the data I can get my hands on, the average Omicron case really is a lot less severe than the average Delta case.

Obviously, I’ll revise these estimates as more information becomes available.

Bits and pieces on Omicron.

Rate of spread.

It’s fast.  There’s really nothing more to say, and I’m seeing nothing to dispute that.  Expert opinion is that it’s going to spread as fast in Great Britain as it did in South Africa.  This,  via NBC news, dated 12/8/2021, emphasis mine:

Scientists at the U.K. Health Security Agency said they expected the omicron variant to become the dominant strain in Britain in the next two to four weeks.

The Wall Street Journal, dated 12/10/2021, states the same timetable as a flat three weeks.

The timetable here should be roughly the same.  Three weeks from today is New Year’s Eve.  You should count on it becoming the dominant strain (i.e., account for more than 50% of new cases) in the U.S. sometime around New Year’s. 

At this point, because all the evidence seems to line up and give the same answer, unless I see something to contradict that, I’m just going to assume that.  I would do any personal planning based on the assumption that on or about 1/1/2022, Omicron will be spreading rapidly and will account for about half of new U.S. cases.

Severity of illness:  Case hospitalization rate

The rumor out of South Africa is that Omicron cases are mostly mild.  More importantly, that assessment appears to be nearly uniform within South Africa.  In other words, is not just a hospital or two that appears to have gotten mostly mild cases.  It’s that everybody there is reporting the same thing.

Here’s a South African reporter quoted as saying just that.  This, from Vox, on 12/9/2021

“Of course, that’s anecdotal, but that’s the general feeling,” Malan told me in an interview over Zoom. “That line is pretty consistent across the country. There hasn’t been an institution that’s come out and said, ‘No, it’s completely different in severe cases.’”

I’ve been searching for any hard data on the South African case hospitalization rate and case mortality rate for Omicron. This search is greatly complicated by one fact:  Most of the people hospitalized with Omicron in South Africa were not, in fact, hospitalized for Omicron.  They were hospitalized for something else, and routine testing revealed COVID-19 infection.

This, from the University of Minnesota CIDRAP, dated 12/6/2021, emphasis mine:

In a Dec 4 report from the South Africa Medical Research Council, scientists detailed clinical findings and data from a hospital complex in Tshwane. where the country's Omicron outbreak began. It covers Nov 14 through Nov 29. Since then, South Africa has reported an exponential increase in cases.

For 76% of patients, COVID-19 was an incidental finding, meaning many were admitted for other conditions but were tested because of hospital policy. Admitted patients were younger than in previous waves, with nearly 80% below 50 years. The report authors noted that 57% of people over age 60 have been vaccinated in Gauteng province. Children under age 10 made up 19% of the Omicron cases.

This means that more-or-less every statistic you read about Omicron hospitalizations in South Africa is wrong.  Wrong, in the sense that it mostly represents people who weren’t actually sick from COVID, but instead had some other problem, plus a (presumably) asymptomatic COVID infection.

To be completely clear, all the scary numbers suggesting a doubling of COVID hospitalizations in South Africa are useless.  I was fooled by that, early on.  The fact that the number of persons in the hospital with a positive COVID test may be spiking does not mean that the number of patients admitted for treatment of COVID-19 is spiking.  Most of the count of patient + hospital + COVID is patients who were admitted for something else, and just happened to test positive for COVID.

Because of this, we have to ignore a) the overall rate of admissions, b) the growth rate in reported admissions, c) any statistics such as fraction of patients in ICU or with oxygen, d) admission typical the length of stay, and so on.  That’s because most of what we’re looking at isn’t Omicron patients at all.  It’s people hospitalized for something else, who just happened to have an asymptomatic Omicron infection.

If most of these cases would have ended up in the hospital anyway, what we should see is a spike in “COVID hospitalizations” without a corresponding increase in total hospitalizations.

All I can get my hands on are two pieces of hospitalization data.  So far I have not found any one centralized source of information.

Here’s a reference to a 12/7/2021 figure of about 383 hospital admissions with COVID-19 that day, and 13,147 new cases.  That would be a case hospitalization rate of 2.9%.  Here’s a reference to more recent figure of 374 admissions and 19,842 new cases, for a case hospitalization rate of 1.9%.

But if, as was true in the one hospital system cited above, three-quarters were admitted for other reasons (and merely tested positive for COVID-19), that would be an actual COVID-19 case admission rate of just 0.7% and 0.5%, respectively.

Tentatively, that can be compared to the current 7.5% admission rate for U.S. Delta cases.  (That’s a bit suspect, as admission criteria in South Africa might differ markedly from those in the U.S.)

My conclusion is that Omicron is resulting in about one-tenth as many hospitalizations as Delta, per case.  I’ll revise that if and when I can get better data.  Among other things, this does not account for the reportedly younger age mix of current South African cases.

Severity of illness:  Two-week-lag case mortality rate via Johns Hopkins data.

The only other hard data — other than hospitalization — is the case mortality rate.  And this turns out to be another case where the only way to get a number that would pass the smell test was to gin that up myself.

The problem with mortality rates is that it takes an average of about two weeks to die from COVID-19.  When cases are growing rapidly, the concurrent rate (today’s deaths over today’s new cases) will grossly understate the ultimate  mortality rate.

So, here’s what the concurrent COVID-19 mortality rate looks like for South Africa, based on data from Johns Hopkins:

I’m really only showing this for the cell in yellow.  Historically, the case mortality rate for COVID-19 in South Africa was about 3 percent.  You really can’t make anything out of the week or day cells, because most of the deaths that will eventually occur, for those new cases, haven’t happened yet.

Now let me access the raw data from Johns Hopkins, and use it to calculate a series of two-week-lagged mortality rates.  That is, I’ll take today’s deaths, divided by new cases that appeared two weeks ago.  Absent literal person-by-person data, that’s the best I can do to account for the average lag between diagnosis and death for COVID-19 patients.  In other words, my claim is that these should be a reasonably accurate representation of the actual case mortality rate.  Albeit two weeks out-of-date.  My “12/10” number is actually the mortality rate estimate for new cases appearing on 11/27/2021.

When I do that, it reveals a sharply declining case mortality rate as Omicron displaces Delta in South Africa.  As of about two weeks ago.  In the table above, the older numbers more-or-less match the 3% historical case mortality rate shown above.  But the last number — which would reflect the mix of new cases on 11/27/2021 — shows a vastly lower case mortality rate.

A quick back-of-the-envelope, and a real kicker on this statistic.  Omicron was reported to be 88% of new cases in South Africa on 1`2/3/2021.  For the sake of argument, let me assume that it was 75% of cases as of 11/27/2021. 

If that’s true — if one-quarter of cases on that day were still Delta, and three-quarters were Omicron — that would imply that the mortality rate for Omicron is very close to zero.   If Delta by itself has a 3.2% mortality rate, and a mix of 25% Delta / 75% Omicron has a 0.8% mortality rate, then … well, as a matter of arithmetic, the mortality rate for Omicron must be zero.

By that I really mean that it’s close enough to zero that I can’t peg it down with this method.  All I can say is that it’s quite low.  It would have to be to drag that 3.2% rate down to 0.8%, when one-quarter of cases are still Delta.

Upshot:  Sometimes rumors are true

In the end, it looks like rumors of much lower average case severity for Omicron are true.  Best guess, given what I can get my hands on at the moment, is that it generates one-tenth the case hospitalization rate, and even less than that for case mortality rate, compared to Delta.

I’ll try to revise these estimates as new information comes to light.