“I roll 1d634.” “That’ll be a critical hit.” “YES!” “…by covid.” “Crap.”
NOTE: This post was originally published at my Substack. The footnote links go there instead of to the bottom of the page.
sigh
This post started as an opinion piece, but then, like too many De Civitate posts, it quickly degenerated into research. Alas, tonight, you will get only facts. Blergh!
This research is inspired by a claim I have seen floating around: “covid has a 99.7% survival rate for those under 50.” (I’ve cleverly put that claim in quote marks to boost my SEO.) Interesting! And plausible! If true, it means that, if you’re infected with covid, your odds of dying are 1 in 333.
So is it true? Let’s take a look.
Doing the Math
First things first, we’ll get the provisional COVID-19 death totals, stratified by age, and we’ll filter to the national totals, both sexes, cumulative death totals only. This data is as of December 1st, 2021.
This data comes from death certificates. It is “provisional” because death certificates take several weeks to arrive at the CDC, so the past few weeks of data will be off by some small factor. Note that the number of both deaths and covid deaths will go up (not down) as that data becomes final.
The only1 interesting column here is COVID-19 Deaths simpliciter. This is the number of death certificates the CDC received where COVID-19 (ICD U07.1) was listed as a cause of death.
Next, we’ll get the national population estimates, also stratified by age group. This data is from 2019, just a few weeks before the pandemic reached our shores2, which makes it our best available baseline.
Now we can calculate what percentage of each age group have died of covid… but that isn’t very useful, because we don’t know what percentage of each age group have been infected by covid so far. We need to get an estimate of how many cases there have been in each age group.
Fortunately, we have some pretty decent data on that. The CDC reports that, as of December 1st, there had been 48,628,497 officially reported cases in the United States.
However, we know that’s an undercount, because some people get sick with covid and do not get tested. No test = no official report. For much of the pandemic, we just had to guess how often these “missed cases” happened. Lots of people came up with very plausible guesses, including me, but they were guesses. Fortunately, earlier this year, there was a really neat study that searched well over a million samples of donor blood for covid antibodies, and managed to come up with pretty solid estimates of how many Americans had actually had covid. As it turned out, we were catching about half of all covid cases (which is a lot more than I had guessed). The study found that there were roughly 2.1 actual cases of covid for each officially reported case of covid, and, having read the study, I think they’re likely pretty close to spot-on—and I assume the number hasn’t changed significantly since May, when their study was conducted.
So that means the total number of covid cases in the United States as of December 1st is 48,628,497 x 2.1 = 102,119,843.7 (estimated). About 30% of Americans have had covid, with many of those cases being entirely asymptomatic. Sounds about right.
We’re getting close, I promise. (Please, some sympathy for your scrivener: you’ve only been reading this for five minutes, but I’ve been beavering away at it since 9:45.)
Now that we know how many cases there have been total, we need to figure out how many cases there have been in each age group, so we can calculate appropriate infection fatality rates for each of them. Unfortunately, that data doesn’t seem to exist. Fortunately, it seems fairly simple to derive!
The CDC tells me that pretty much everyone in the population has an even chance of catching covid. Older people have a higher chance of dying of covid, but everyone can catch it. That means most age groups will have a proportional amount of the total covid cases. 12.5% of Americans are 55-64 years old, so we would expect them to have caught 12.5% of the total covid cases.
The exception, according to the CDC, is children under 5, who are significantly less likely to catch covid. The CDC doesn’t say exactly how much less likely (and they require medical attention so rarely it’s probably quite hard to determine), but I assumed that children under 5 are half as likely to catch covid.3 Then I reweighted my population percentages accordingly.
Now we have everything we need: for each age group, we have good estimates for the total number of covid cases and the total number of covid deaths. Divide deaths by cases to derive the Infection Fatality Rate—your odds of dying if you contract the virus, all else being equal. And that’s what this whole post is about!
The Results
[UPDATE: In a subsequent post, I added odds based on obesity.]
Satisfyingly, this is pretty consistent with the Robert Verity IFR estimates I’ve used for much of the pandemic, as well as with other estimates I’ve seen subsequently.
Substack isn’t giving me a good way to post tabular data4, so I’ll share the whole thing in a Google Sheet. (There’s some neat stuff in there about population structure as well.)
If you are an average 32-year-old, and you catch covid, your odds of dying of it are about 1 in 1800. If you are an average 70-year-old, and you catch covid, your odds of dying of it are around 1 in 60. And so forth. Overall, covid-19 appears to have a survival rate of around 99.2%!
What about the claim at the start, that covid has a 99.7% survival rate for those under 50? That claim appears to be wrong… because it’s too low! I don’t have an age band for 0-49, but the 0-54 age band has a combined total survival rate of 99.88%; odds of death are 1 in 857.
Maybe those sound like small numbers to you.
Here are some other numbers, then, for perspective.
The catastrophic flu pandemic of 1918 had a survival rate of ~98% (estimates vary a lot). If everyone in America catches covid (which will likely happen in the coming months and years), and 99.2% of us survive, we’ll be left with 2.5 million dead to bury. That’s a lotta stiffs. In all American wars ever, combined, only 1.3 million soldiers have died (and only half of those died in combat).
Yes, lots of our covid deaths will be very old people, who didn’t have many years left to live anyway. But lots won’t be:
I really don’t like how many 45-65 year olds die in this chart. That’s an unusual amount of mortality for people who haven’t even reached retirement age.
Those numbers for people my age (25-34) are nothing to sniff at, either.
The National Safety Council shares some interesting factoids with us about some other odds:
Now, do I spend my days worrying about choking to death (odds 1 in 2,535)? Certainly not. If a maitre d’ told me I had to do a Heimlich Maneuver Drill every time he passed by the table, or he wouldn’t serve me the dinner I ordered, I’d tell him to go pound sand. The benefit is not worth the cost.
On the other hand, I do know the Heimlich Maneuver. Learned it in Boy Scouts, never regretted it. My parents also made sure I knew how to swim, reducing my odds of death by drowning (odds 1 in 1,128). And my house has smoke detectors, in order to reduce the risk of death by fire (odds 1 in 1,547). My car has seatbelts. I’ll bet you have all those things, too.
While none of these are common causes of death, none of them are quite rare, either. It’s reasonable to take small precautions against these unlikely events. Relatively small investments of my time, effort, and money are worthwhile if they substantially reduce my chances of death. Of course, a smoke detector doesn’t eliminate my chance of dying in a fire. I could get caught in a fire at someone else’s house, or my smoke alarm might not go off. It happens. But it reduces my chances of death enough that I think every house should have a smoke detector.
Death by covid-19 is not common for my age group, but it’s not quite rare, either (odds 1 in 1,771, assuming everyone gets it). Covid deaths are only truly rare among children under 14.
Existing covid-19 vaccines appear to improve my odds of surviving covid (even Delta) by roughly a factor of ten. Three billion people have taken at least one dose of the vaccines over the past twenty-one months, with no epidemic of vaccine-related permanent injuries, which seems to prove they are reasonably safe. Moreover, the vaccinated continue to be underrepresented in hospitals, which seems to prove that the vaccines are also reasonably effective. That is why I got vaccinated, and why I have recommended my healthy, adult5 friends and families get vaccinated as well.
That last paragraph verges dangerously close on opinion, though, and probably deserves to be fully unpacked in another post.
Until then, please enjoy knowing your Official De Civitate Covid Odds.
All the other columns show how often COVID-19 was comorbid with pneumonia and/or influenza. Those columns are listed only because, in mid-2020, many covid skeptics argued that the CDC death counts were dramatically inflated, because they allegedly included people who died “with covid, not of covid.” These skeptics asked to see how many “covid” deaths were actually just old people who were dying of the flu. (Remember, in mid-2020, many people still believed covid would kill fewer Americans than the annual flu!) The CDC obliged these skeptics and started posting the flu data in June. I think these were reasonable questions to ask and that the CDC’s decision to answer them was wise.
But this data is not very illuminating for our purposes. It is exceptionally unsurprising that many people who die of covid-19, which causes often-fatal pneumonia, end up with death certificates that say “patient died of covid-19 and pneumonia.”
The covid epidemic in the United States began in the second or third week of January 2020. Its initial introduction was from China, although (over the next several months) it would be re-introduced many more times, from all over the world (especially Europe). How do we know this? Phylogenetics!
Covid is constantly mutating very slightly, spinning off thousands and thousands and thousands of tiny variants. Only a handful of these are “variants of interest” like Delta and Omicron, which have actual evolutionary advantages… but all of these variants have their own unique fingerprints. This means you can construct a shockingly detailed “family tree” for the virus. You can see which virus strains were descended from which earlier virus strains… and you can trace the movement of those virus strains around the world, at least in the territories with decent genetic surveillance of disease.
On January 15th, 2020, a traveler to Wuhan returned to Washington state. There was no quarantine in place at the time. CNN was not even talking about the “novel coronavirus,” although we’d had ample warning—and even more warning if we’d listened to Chinese whistleblowers instead of taking the Chinese Communist Party’s word for it. They were out there, on Twitter, quietly getting retweets from right-wing paranoiacs. I know because I was watching them!
Anyway. America’s Patient Zero, aka “WA1/2020” got sick and went to the doctor’s on January 19th. We finished his test on January 21st, confirming his covid case. WA1/2020 wasn’t the only person to bring covid back to the U.S. from overseas in those early days, but, in most cases, we detected and quarantined them before they could spread it. We know this worked because those strains died out; they were never seen again, and no descendants were ever found.
Unfortunately, we discovered on February 29th that we had not actually contained WA1/2020’s case, which entered the general population of Seattle. We might have contained covid at that point if we had closed off all international travel and immediately deployed the military to seal the borders of the State of Washington, but it was probably too late. That outbreak spread across the country and—with help from other, later quarantine failures, which seeded the virus in several other places—the American covid pandemic had begun.
We know that WA1/2020 was the first to bring it to the United States, because we can tie all the strains to specific locations at specific dates, and the strains that later blew up in the US were still evolving elsewhere as of January 15th. If any covid reached the United States before January 15th, it died off before it could start spreading, because, if it had survived, we would see it in this phylogenetic family tree.
China’s claim that covid was circulating in Italy and France as early as 2018 is implausible, because we can trace every strain, from every place in the world, back to a cluster that emerged in Wuhan in late 2019. China’s story would require the novel coronavirus to infect people in Europe in 2018, luckily evade genetic detection, and then completely die off every single place in the world—simultaneously, for no reason—before suddenly re-erupting in Wuhan, China in fall 2019. The simpler explanation is far more compelling: the virus evolved in Wuhan, and infected its first human victim, in Wuhan, in fall 2019.
This is a small educated guess… but it’s a guess, and you are free to cast all kinds of doubt on it.
Nevertheless, five-year-olds are a small enough portion of our population, with such a tiny number of the covid deaths, that it really doesn’t matter very much. Adjust it however you like, with whatever guess makes you feel better. It will change your final estimates of infection fatality by only a tiny fraction in either direction. (Yes, I did check.)
But I will forgive you, Substack, because I am loving your footnote engine.
It’s a way closer call for kids.