There’s a video going viral of two doctors from Bakersfield, CA — Dan Erickson and Artin Massihi — who argue the testing results they’re seeing at the urgent care businesses they own makes COVID-19 more like the flu and is evidence that isolating the healthy doesn’t make sense and social distancing restrictions can be safely eased, leading to the opening of the economy. It’s a powerful message in these times, but, unfortunately there is a “solid counter argument to be made from the other side” that’s not getting shared:
Here is an example of great journalism. Talks about what the Kern County Doctors said, then presents a solid counter-argument from the other side. https://t.co/pDeTpn19n3
— Dillon Casey (@DillonCasey) April 25, 2020
First up, here’s the video, which was shared by Elon Musk:
Docs make good points https://t.co/WeXuZpMghY
— Elon Musk (@elonmusk) April 26, 2020
Now, the counterpoint from the article: The two are being called out on their use of the testing data:
Kern County Public Health Services Spokeswoman Michelle Corson and an epidemiologist contacted by The Californian said they didn’t agree with the doctors’ recommendation to end social distancing and immediately start reopening society.
“This is a many-headed hydra. It’s really unfortunate to boil this all down to it’s just flu,” said Andrew Noymer, associate professor of public health at UC Irvine. “There’s no flu season that looks anything like New York does right now.”
And:
Noymer of UC Irvine disagreed with the doctors’ premise that COVID-19 is as widespread as Erickson and Massihi think, saying the idea that nearly 5 million Californians have had the virus is a gross overestimate. The people tested in California were not a random sample; they were mostly people who were symptomatic, Noymer said. Therefore, extrapolating the positive test rate across the entire population of the state is not an accurate way to arrive at how widespread the virus is.
And even if 12 percent of the state has had the virus, that still leaves 88 percent vulnerable to it, Noymer said.
“They’re advancing factual inaccuracies and playing off the esoteric nature of the mortality stats to make a case that the economy should be reopened,” Noymer said. “I agree it should be reopened, but it should be opened deliberately, bit by bit, and informed by science. Not informed by a misreading of the mortality.”
More from the University of Washington ‘s Carl Bergstrom where he explains how the sample they’re using as the base for all their calculations is not random at all, which is throwing off the entire estimate:
Unfortunately the misleading claims of those two doctors in Bakersfield keep making the rounds, so I want to very briefly address the problem with what they are saying.
I won't get into their possible motives, past political activity, etc.https://t.co/IGzeamfqce
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
What they did was simple: they looked at the fraction of patients who tested positive for #COVID19 at the clinics they own. They found 340 out of 5213 tests were postive, about 6.6%
Then they assume the same fraction of the whole population are infected. pic.twitter.com/5DNl0EdTit
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
From there, they scale up to the state level and claim 12% incidence statewide. The news story says it is using the same calculation, but it can't be—how did they get from 6.6% to 12%? Perhaps they estimating infected *ever* versus infected *currently*. It's not clear. pic.twitter.com/Xj8kX7ZmZP
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
Using that 12% infected figure, and a known 1400 deaths in California, they assume 1400 out of 4.7 million have died. That gives them an infection fatality rate of 0.03%. That is, they think that if 10,000 are infected, 3 will die on average.
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
The problem with this approach is that during a pandemic, the people who come into an urgent care clinic are not a random sample of the population.
A large fraction of them are coming in precisely because they suspect that they have the disease.
This generates sampling bias.
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
Estimating that fraction infected from patients at an urgent care facility is a bit like estimating the average height of Americans from the players on an NBA court.
It's not a random sample, and it gives a highly biased estimate.
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
Moreover the estimate does not pass even a basic plausibility check.
In New York City, 12,067 people are known to have died from the virus, out of a population of 8.4 million.
This is a rate of 0.14% of all people. Not just infected people. All people.
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
That gives us a lower bound on the death rate in New York. Not an estimate, a lower bound.
The death rate for infected people is obviously higher than 0.14%, because not everyone in New York has been infected.
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
And yet that 0.14% lower bound is nearly *five times as high* as the 0.03% that the Bakerfield duo are claiming. They've used absurd methodology to arrive at an implausible number.
If the pandemic were not so severely politicized, this would be a non-issue from the start.
/fin
— Carl T. Bergstrom (@CT_Bergstrom) April 26, 2020
***