COVID: The Dunning-Kruger Effect In Epidemiology

My most recent blog entry was, by my assessment, fairly mild in its assertions: I simply attempted to show two graphics to make two specific points about COVID: namely, that the "official" state counts that only include those who were PCR positive for COVID represents an undercount; and that there is now mounting evidence that people spread the disease before their symptoms begin, which makes containment via contact tracing much more difficult.

Inherent in those two messages is a theme I've written about at length previously, namely, that data in medicine has an essential fuzziness to it, and one shouldn't get too caught up in thinking that a number represents a precise, mathematical reality in the same way a physics equation does. Though maybe physicists might indicate that they have the same issues with interpreting data as we do in medicine. Hopefully mechanical engineers are more precise.

Anyway, I didn't think I was making an especially debatable point, and in the first half of the blog, the part about the relationship between the official count and the true count, I certainly didn't think I made any statements that could be construed as political (the end of the blog tosses in a typically Billy Rubenesque political potshot). Keep in mind I made no confident pronouncement about the exact total: I just pointed out that the real number of deaths from COVID is somewhere between the official count and the total number of deaths in a given county or state. Let me be clear: that's not controversial. Any epidemiologist could tell you this is true. Here, for instance, is a CDC document explaining this very phenomenon for influenza.

So imagine my surprise when a reader writes back to tell me--politely but forcefully--that I was full of shit:

We "know" that's an undercount. 

What you "know" ain't exactly "so". Compare 2019 deaths 

from 2020 deaths for NY. What we actually know 

is that there is a high degree of uncertainty. Sloppy thinking.

As I wrote, we do know that the official COVID mortality total is an undercount--though again, nobody knows by how much. Which is to say, there is a degree of uncertainty, and I don't think I ever said that wasn't the case. But I was informed that what I know ain't exactly so, and was told to compare 2019 from 2020 deaths for NY.

No link was provided so I can't be sure what the reader was referring to, but here is the CDC's rolling estimates of total deaths by month in New York dating to 2017:

Spot the one part of the graphic that's different? So while there is some uncertainty about the precise number of deaths, there's no debating that COVID mortality is quite substantial. You can also break it down by including known COVID cases, and even including that, you still end up with a huge number of deaths above recent historical averages, which strongly suggests missed COVID diagnoses (you want to look at the lime-green bars above the average):

The reader then goes on to imply that epidemiologists must suffer from some kind of rain-is-falling-everywhere bias with what certainly reads like a sneer about the perceived snobbery of coastal elites:

People keep saying that covid is coming to my area 

of flyover country, but our new infections have been flat 

for a month and our random swab sample showed almost 4% infections. 

I figure we're at 16% who have or have had covid. 

Herd immunity by end of June.

Why and how this reader from "flyover country" figures "we're at 16 percent" is not made fully clear, nor is it clear whether he is qualified to pronounce with confidence that his community, wherever it is in flyover country, will have herd immunity by the end of June. (We'll assume for a second that we're dealing with a he, given the whiff of testosterone ambient in the I'm-splainin'-to-you-you-pointy-headed-professor commentary.) The fact that this reader has previously gravitated toward embracing what amounts to folk remedies for COVID, touted by people too ignorant to understand their own limitations as they conduct what they mistakenly believe are "clinical trials," might suggest that they regard assertions made with bravado to have high truth value, regardless of the expertise of the person making those assertions.

I simply don't know what motivates these declarations, although I am confident that his confidence is overblown, given that virtually every qualified epidemiologist thinks we are far from the finish line of this pandemic. What does our reader know that every PhD steeped in epidemiology, virology, and biostatistics does not? Or is this just denial masquerading as swagger? I have an opinion on this; I'll let you guess at it.

Wherever his section of flyover country might be, it is curious that he might assert everything is hunky and dory in rural or less urban America. Take a look at this graphic from WaPo displaying the ten counties in the US with the highest rate of deaths, adjusted for population:

In short, half of the ten counties with the highest rate of deaths are in less populated, non-urban settings: four counties in southwest Georgia and a suburbanish community outside of New Orleans. That would strongly suggest that this virus will spare no area, and supposing that it has a predilection strictly for urban centers is wishful thinking.

Why has our reader's area been spared? Nobody knows. Japan has the oldest population in the world, and so they should be experiencing a level of misery on par with Italy. So far, they're not. Most countries from warm-weather climates aren't seeing much death; not so for Ecuador and Brazil, which are getting hammered. But one reasonable interpretation is that, in the words of Kim Carpenter, we've only just begun, and over time we're going to see these numbers even out if people return to their normal habits. Russia, for instance, was fine until it wasn't. (Though skepticism of official Russian statistics, given its political structure, is probably warranted, especially as news stories of strange "accidents" befall Russian doctors. Given Trump's warm regard for Putin, Anthony Fauci should watch his back.)

As with hydroxychloroquine and clinical trials, on the topic of armchair epidemiology, it might be best to follow the simple, but lifesaving, dictum: don't try this at home.

I grew up in flyover country, by the way. I have never forgotten where I have come from.

--Billy

(Postscript: we note, with embarrassment, that the great Ms. Carpenter's first name was, of course, Karen and not Kim. Anonymous below in the comments has decided we're not worthy as a consequence, and we find it hard to argue the point, as we'd say the same thing. Some things are not worthy of mercy.)