Thursday, January 27, 2022

"3895 people died of COVID in this country yesterday" (why that's misleading, what the numbers really mean, and implications for omicron's severity)

I want to say up front, because if I didn't I think this could be easily misconstrued, that my purpose here is not to question the seriousness of what's been happening with the COVID pandemic. It's a horrific situation. And I'm also not questioning whether all those deaths were really caused by COVID. There's no reason to doubt that they were.

Also, I'm not trying to call out people who make this error, because I find it very understandable why people who haven't directly engaged with the data might make this mistake.

I also want to say up front that the most important thing to know about COVID is that everyone who's eligible should get vaccinated and boosted. People who are unvaccinated have a roughly 68x chance of dying from COVID as people who are vaccinated and boosted according to the latest numbers. We are incredibly lucky to be able to receive, for free, such an effective intervention against a deadly disease.

Anyway, I just want to talk about a thing that I see all the time that isn't accurate and that bugs me, and try to educate people a little. Because it's important to me to try to understand what's going on in the world, and I like trying to share that knowledge.

I guess I've always been someone who is a stickler for accuracy. I have a strong tendency to "well, actually" people. When I see someone make a blatant error, it's like something is triggered in my brain and I have a strong urge to correct the error. Literally, when I was in second grade I frequently corrected my teacher's spelling mistakes in front of the class. So this tendency has been around for a long time!

I very often see people on social media say "[x number of] people died yesterday from COVID" to complain about how other people aren't taking the pandemic seriously enough. (And yes, certainly, people not taking the pandemic seriously enough has been a problem since the very beginning of the pandemic.) The most recent example being "3895 people died of COVID yesterday," talking about the numbers (for the United States) that have come out in the past day as I write this post.

It is not accurate to say 3895 people died of COVID yesterday. In fact, it's impossible to say how many people died of COVID yesterday, because the vast majority of those deaths have not yet been officially recorded as COVID deaths.

Here's a graph, from the New York Times, showing where this number came from.


So, a couple things to immediately notice. One is that it's a graph of "new reported deaths by day," meaning that the deaths were reported on that day, which says nothing about when the deaths occurred. (I wrote a post in June 2020 - yeah, this has really been going on for almost two years, sigh - with a lengthy explainer on this very topic. It's been very frustrating to me seeing people continuing to make this mistake over and over for almost two years, but again, I don't really blame people who haven't directly engaged with the data for not understanding this.) The second thing is that the daily average for the last week is 2466, meaning that 3895 is a big outlier. If you're in a period of time where about 2500 people are dying each day from a disease, it's extraordinarily unlikely that suddenly nearly 4000 would die in one day. That's just not how these things work. And I notice the "[x number of] people died yesterday of COVID" trope usually comes up when there's a big outlier like this.

Why is there such a big outlier? It's because sometimes a big backlog of previously unreported deaths will be cleared in a single day.

But when did the deaths actually occur?

Well, right now I find this particularly interesting, because it relates to the issue of how big of a toll the omicron wave will end up taking.

Lately, when I see this trope about so and so many people died of COVID yesterday, I sometimes see someone add something like, "at least they were mild deaths." A snarky reference to the news stories saying the omicron variant is milder than previous forms of COVID.

The implication here is that 3895 people died yesterday of COVID, and because we're in the omicron wave those 3895 people died from omicron, and therefore omicron can't be that mild. But did those people really die from omicron?

Let's try to figure that out.

I took a look at the data reported by the state of Ohio and identified the dates of death for all the COVID deaths that Ohio reported in the one week period from January 20 to January 26, 2022. Here's a graph of those deaths showing the number that occurred on each day (there's also a handful that occurred on earlier dates that I left off the graph):


What this tells us is that more than half of the deaths that Ohio reported in the last week occurred on or before January 3; in other words, most of the deaths occurred several weeks ago. (Certainly not "yesterday.")

Incidentally, when I analyzed this for that post in June 2020, the median delay from date of death to report date was 5 days, so the delay has become much longer. I can't say what the typical delay is in other states. I'm sure that substantial delays are quite common, but as substantial as those in Ohio? I don't know. Perhaps yes in some states and no in some others.

That graph shows us when the recently reported deaths occurred. The next piece of the puzzle is, when did those people get COVID?

It's been reported that the median duration from symptom onset date to date of death, for people who get COVID and then die from it, is roughly 17 days. (I don't know if it's exactly the same for the delta and/or omicron variants, but I expect it's at least fairly similar.) So for a quick and dirty estimate of the case onset dates (which, it should also be noted, would be a few days after the infection dates), let's send everything back 17 days:


More than half of the deaths reported by Ohio in the last week would be people whose cases started on or before December 17. At that point in time, what was the prevalence of the omicron variant?

It turns out, for the week ending December 18, about 62% of the cases in the US were still the delta variant, and about 38% were the omicron variant (from the CDC's variant tracker):


I will note that, at that point in time, omicron had already reached a higher prevalence in Northeast Ohio, where I live. Still, delta hadn't been wiped out, and the rest of the state probably had numbers comparable to the country as a whole.

So we have several pieces of data from which we can extrapolate.

1. About half of the COVID deaths reported in Ohio in the last week were likely the result of cases that began in mid-December or sooner.

2. In mid-December, the majority of cases were still delta, not omicron.

3. Delta has a higher infection fatality rate than omicron. There is now an abundant amount of evidence showing this, although we don't know exactly how much higher, but it's probably at least several times higher.

Taken together, this implies that the majority, and perhaps even a large majority, of COVID deaths reported in Ohio in the last week were caused by the delta variant, not the omicron variant. This might also be true for the country as a whole, but I'm not sure because I don't have an easy way to see deaths by date of death for all the other states.

We've seen that a lot of other countries have had massive surges of cases from omicron without anywhere near as dramatic of increases in deaths as what we're seeing in the US now. This has been attributed to the poor vaccination rate in the US. And this is indeed true, but an important point that is being missed when interpreting the current death numbers is the poor vaccination rate in the US also means that the omicron wave is happening on top of an already large baseline of delta cases. Therefore, it's hard to say right now how much of the latest wave of deaths is the direct result of omicron.

Really, to be able to tell what the impact of the omicron wave is on the US's death rates, we'll have to wait at least a few more weeks. That's not the most satisfying answer, but it's the most honest answer. Additionally, teasing out the impact of omicron vs. delta on death numbers is likely going to require a more rigorous level of analysis than what is being done by most people who are commenting on COVID death numbers.

We are actually incredibly lucky that omicron does cause considerably less severe disease and death per case than all previous versions of COVID. It's horrific to contemplate how things might look right now if that weren't the case.

I hope this was informative. My purpose here was simply to try to educate people. I'm not trying to tell anyone what to do with this information. At this point, almost two years in and with no real way to know what's going to happen in the future, I hesitate to try to tell anyone what they should do in regard to COVID matters. With one definite exception, which is that everyone who's eligible should get vaccinated and boosted, and that we collectively have to increase our efforts to vaccinate people both in this country and around the world. I don't know how many people are going to end up dying from omicron vs. delta, but it's going to be a lot either way, and it continues to be true that the vast majority of those deaths could have been prevented by vaccination.