The Evidence That A Million Americans Died Of COVID

...

I.

Many commenters responded to yesterday’s post by challenging the claim that 1.2 million Americans died of COVID:

I'm still not convinced that number can be directly linked to Covid. Yes, I'm talking about the old cliché "died with Covid" instead of "died of Covid". Shouldn't we talk about excess deaths only?

Nobody talks about "1.2 million COVID deaths" because the medical establishment played obvious and ridiculous games with the way it counted "COVID deaths." Any patient with a positive PSR was required to be reported as such, even patients who were in vehicular accidents or suffering from late-stage cancer. This has been known since the beginning.

I'm sure some people did die from covid, but given how much we now know the statistics were faked, it seems to me most of these covid cases were other conditions that got labeled as it for variety of reasons, or died from the ventilators that were incorrectly prescribed in the early stage of the pandemic.

As many of the comments note, the attribution of 1.2M deaths to COVID is as questionable as the zoonotic origin story, the value of the vaccines to healthy young people, and the rest of the COVID rhetoric. Some, like Scott Alexander, still “trust the experts”. But many don’t. There doesn’t seem to be much objectively verified consensus ground truth here, just conflicting narratives and numbers.

I didn’t know this was still such a topic of debate, but since it seems the comment section is split, let me present my case for the 1.2 million number and see if people still disagree.

If people died “with” COVID, ie of normal causes like flu, cancer, or car accidents while only incidentally having a positive COVID test, then total all-cause mortality during the COVID pandemic would be the same as always. If people died “of” COVID, then total all-cause mortality during the COVID pandemic would be higher than usual.

It was higher. This is via census.gov, from the National Center For Health Statistics:

It’s hard to read the exact numbers, but it looks like about 500,000 - 700,000 excess deaths in each of 2020 and 2021¹, which adds up to most of the 1.2 million (although I think the full number might include some residual deaths during 2022+).

From the CDC, via the White House, correlation between reported COVID-19 deaths and excess deaths throughout the pandemic:

They’re pretty much the same, maybe margin of error of 10% or so but not consistently in any direction.

(does that mean that the doctors and coroners charged with determining whether a given death was “with COVID” vs. “from COVID” did an amazing job? Most of the doctors I know are smart and hard-working and I’m sure they did their best on this - but these statistics don’t necessarily mean that. Suppose that the average elderly person has COVID for two weeks before either dying or recovering, and that they get one case of COVID during the pandemic. The yearly mortality rate for 70 year olds is 2%, so the risk-during-the-time-they-incidentally-have-COVID is 2%/(52/2) =~ 0.08%. There are 55 million seniors in the US, so if 0.08% of them incidentally died with COVID, that’s only 44,000 deaths - less than 5% of total excess mortality deaths. So even if doctors had been maximally lazy and dishonest and recorded every single case with incidental COVID as a COVID death, even car accidents - something I’m confident they didn’t do - this couldn’t have produced the observed death numbers.)

Is it possible that the CDC and National Center For Health Statistics are lying about all-cause mortality patterns? Seems unlikely, because individual states reporting separately found similar patterns, and so did the other countries that reported data. This would take a truly global conspiracy.

II.

I think this disproves the claim that it was just normal deaths being reported as COVID deaths. What about the alternative claim - that it wasn’t COVID that caused the extra deaths, but various treatments - ventilators, remdesevir, vaccines?

We know it wasn’t a specific single one of those treatments, because the treatments were only used during certain subsets of the pandemic, but the excess mortality was a constant function of COVID cases. So for example, it can’t just be vaccines, because people only started getting vaccinated in December 2020, but there was the same amount of excess all-cause mortality before that time. It can’t just be ventilators, because doctors significantly cut down on ventilator use by mid-to-late-2020, but there was the same amount of excess all-cause mortality after that time - also, the effect of ventilators on mortality was exhaustively studied, and is low.

(Also, the reason people were getting put on ventilators was that they were having trouble breathing, their oxygen saturation had gone down to critical levels, and they were in the ICU. To claim that there was minimal direct COVID mortality and it was all ventilator-related would require that this was some completely new type of respiratory distress that could bring patients’ oxygen saturation down to levels usually considered fatal, but which in this case wouldn’t kill them in the same way respiratory distress usually kills patients, for unclear reasons, unless doctors tried to treat it.)

You could still potentially stitch together several treatments, saying that all of them increased mortality (the same amount?) and by the time ventilators were falling out of fashion, vaccines and antivirals were revving up. But this would require every single COVID treatment to be unprecedentedly dangerous, at exactly the same level, and in a way invisible to studies.

Why would you do this? People get mad when I overuse the term “priors”, so let’s talk about burden of proof. There’s a new virus. It looks exactly like the kind of virus that people predicted would be a deadly pandemic - so much so that when the Wuhan Institute of Virology set out to create and study viruses that might cause deadly pandemics, they made ones so similar to this virus that we’re still not sure whether it was actually one of theirs. It’s known to cause thromboembolism, cardiac damage, and kidney damage in susceptible individuals, all complications that can potentially result in death. It infected hundreds of millions of weak old people of exactly the sort who die from viruses like this all the time. Why are we even trying to come up with a weird Rube Goldberg collection of precisely-calibrated-yet-undetectable iatrogenic injuries that would mimic the mortality pattern of a pandemic respiratory virus, rather than just acknowledging that the pandemic respiratory virus killed a number of people about in the middle of the range that pandemic respiratory viruses usually kill (more than seasonal flu, less than the Spanish Flu)?²

III.

One commenter gives a plausible argument for doing this:

1.2m dead is just not believable. I only know 1 person who died from catching covid (over 80, in a hospital for other reasons, caught it there and got lung damage, fairly late dueing the official pandemic). I have substantive network in Hawaii where lots of people got sick a few weeks before covid officially arrived in the USA. Not a single person I know there died out of hundreds, and that's early infections when it was worse and includes elderly people.

This is my experience too - the only person I “know” who died from COVID was a relative of one of my patients, who I had never met and only heard about secondhand. I can’t deny COVID deaths seemed weirdly thin on the ground in my and many other people’s personal experiences - in a sense, that’s exactly what my earlier post was about.

But is this really so mysterious? There are 340 million Americans, so if 1.2 million died of COVID, that’s about 1/300. This number - 1/300 - is also the prevalence of multiple sclerosis³. Do you know someone with multiple sclerosis? No? Then it’s not surprising that you also don’t know someone who died of COVID.⁴

I actually asked about this on the 2022 ACX survey (as part of the research for this post). 6.5% of respondents said a family member had died of COVID (with “family member” described as “first and second degree relatives - ie self, brother, sister, mother, father, child, aunt, uncle, grandparent, grandchild, niece, or nephew”). I think this number is compatible with both “it killed a million people” and “it’s not surprising that most people don’t know anyone who it killed”.

If you still disagree, tell me why I’m wrong!

1. The graph is slightly confusing, but remember, it’s per year, not total. So the large jump in 2020 and the very small jump in 2021 both mean that about 500-600K people died during each of those years.

2. Could the excess mortality have been caused by the negative effects of lockdowns, like suicide or drug overdose? No. Remember this graph?

It shows that whatever the cause of the excess mortality, it was closely correlated with reported COVID-19 deaths. Although you can almost imagine a doctor not being able to tell the difference between a COVID death in the ICU and a respirator-injury death in the ICU, they could probably tell the difference between a COVID death in the ICU and a bullet through the brain in a young person who probably didn’t even have a positive COVID test.

(also, suicides didn’t rise during the strictest part of lockdown, and may even have paradoxically fallen. I agree this is mysterious, but it’s only one of the many mysteries of suicides, which often fall during bad weather or social disasters - see my article here. Drug overdoses rose, but by a tiny fraction of the COVID death toll.)

3. Other categories containing ~1/300 Americans: police officers, Jehovah’s Witnesses, doctors, prisoners, Rhode Islanders. I like these less because they’re class- and location- stratified, so your chance of knowing them goes up or down a lot depending on your own characteristics.

4. More formally: if there’s a group containing 1/300 people, you need to have 208 acquaintances before you have a 50-50 chance of having one acquaintance in the group. This isn’t quite the right number, because COVID deaths were concentrated among elderly people, and your acquaintances might not be elderly, but I’m not sure how to apply the correction for the mix of elderly and non-elderly people that most people know.