Via Nate Silver, most of us are way out of our depth in trying to understand this stuff. I feel more out of my depth each day as I try to reconcile the increasingly hair-raising projections being made by most epidemiologists with the implications of the Oxford model of the disease, which speculates that asymptomatic infections are much more common than anyone realizes. But I’m going to try anyway.
If Oxford is right, it means two things are likely true and they’re both welcome news. One: The case fatality rate is much lower than we’ve assumed. A large number of people are going to die regardless — just turn on the news and feast your eyes on what’s developing in New York City — but deaths are flukier than we currently believe.
Two: This will be over more quickly in some countries than we’ve assumed because those countries are further along the road towards herd immunity than is understood. Once half the population has been infected and acquired immunity (*if* in fact a mild infection confers durable immunity) then the number of cases necessarily will begin to decline. The virus simply runs out of people to infect. That’s the right-hand side of the epidemiological bell curve with which we’ve all become familiar.
The Imperial College, whose projections have been dire, published a new study today that includes this data on estimated infection rates. Check out the numbers for Italy and especially Spain.
The first number is their best guess at the infection (or “attack”) rate in each country and the numbers that follow represent the proverbial margin of error. In Spain the actual rate could be as little as 3.7 percent or as high as 41 percent, which would be approaching herd immunity. In Italy almost 10 percent of the population is expected to be infected right now, although it could be as high as a quarter. What does this mean for the fatality rate?
In Italy there are 11,591 COVID-19 deaths as I write this out of a population of 60.48 million people. If 9.8 percent of the population is infected, that’s 5.9 million cases. That’s a fatality rate of 0.2 percent, lower than the estimates of one to two percent (or higher) that are based on Chinese data.
In Spain there are 7,340 deaths as of today in a population of 46.66 million. Assuming 15 percent are infected, that’s just about seven million infections. Case fatality rate: 0.1 percent.
One more. The Imperial College estimates an infection rate in the United Kingdom of 2.7 percent. Among a population of 66.44 million, that should mean about 1.8 million infections. They’ve had 1,408 deaths so far. That’s a case fatality rate of 0.08 percent.
If you assume that actual infection rates run towards the higher end of the Imperial College’s statistical range, the fatality rate shrinks further. As for our own country, if we assume, say, a two percent infection rate (in line with most of the estimates in the table above), that would mean 6.5 million total infections in the U.S. against 2,945 deaths as of this afternoon. That would be a fatality rate of just .045 percent. If that’s true, then if all 327.2 million Americans were infected, we would expect the total number of deaths to be 147,240 — which, not coincidentally, is smack dab in the middle of the range given by Deborah Birx this morning.
Two things can simultaneously be true, then. One is that the average American’s odds of dying may be very small, something like one in 2,200. (Smaller, actually, for younger age groups.) And two is that a lot of Americans are going to die, whatever the odds may be. Many more than would die in a flu outbreak.
On the other hand, maybe it’s a sucker’s game to compare the size of outbreaks in different countries:
If detection rates vary that much than it is NOT at all reliable to use the number of detected cases to compare countries (or states). Y’all really need to stop doing that until/unless you have a sense for how effective different places are at testing/detection.
— Nate Silver (@NateSilver538) March 30, 2020
As I say, I’m out of my depth. Readers who understand this stuff better are invited to correct me. One additional thought, though: If fatality rates are much lower than we think then China’s death toll of 3,308 people in Wuhan and beyond logically means there were many more infections than the 82,200 known cases that the country is currently reporting. Assuming a fatality rate of, say, .15 percent, that would mean 2.2 million infections, or around 27 times more than the official number. That would fit with the British government’s suspicion that there were 15-40 times as many cases in China than Chinese authorities claim. The Brits, though, seem to think that there are 15-40 times as many confirmed cases known to Chinese authorities than they’re letting on; if there are something like 2.2 million known cases in China, it must be that there’s a gigantic number of unknown/asymptomatic infected population beyond that — maybe something like 50 million people or more — and many, many more deaths than China has revealed, which is what some suspect.
Anyway, just because the Oxford model might be on to something doesn’t mean social distancing isn’t important. Scroll down to page 10 of today’s new Imperial College study and you’ll find an estimate of how many lives social distancing practices in Europe are projected to have saved this past month. Result: 59,000, the bulk of them in Italy.
