1/H
There was a recent effort to champion Nate Silver (@NateSilver538) as a non-expert who speaks uncomfortable truths experts don't want to hear.
That's misguided, as we can see by examining how many SARS-CoV-2-infected get hospitalized.
2/H
Some context:
Infection fatality rate, or IFR, is the proportion of people infected with the virus SARS-CoV-2 who die of the disease COVID-19.
Infection hospitalization rate, or IHR, is like IFR, but with COVID-19 hospitalizations instead of deaths
institutefordiseasemodeling.github.io/nCoV-public/an…
3/H
Seroprevalence studies measure antibody levels to estimate the number of infected people.
Dividing COVID-19 hospitalizations by that number of infected people gives a seroprevalence-based IHR.
IHR is good to know.
4/H
Neil Ferguson's team at Imperial College gave IFR + IHR estimates for Great Britain (GB) and the USA in March 2020.
Their IFR estimates held up well for the mitigated pandemic that actually occurred.
Their IHR was ~4.4%.
spiral.imperial.ac.uk:8443/bitstream/1004…
5/H
This is where Nate Silver objects.
He claims IHR was more like ~2%, and so Ferguson et al.'s ~4.4% value was an over-estimate.
He's been saying this for about a year or more, despite people repeatedly explaining he's wrong.
6/H
Silver recently brought up this up again, after experts correctly criticized his non-expert + uninformed claims on vaccine policy / communication.
So he may have thought pointing out experts being wrong might help him.
7/H
With that background out of the way, it might help to assess how Silver's claims held up in comparison to experts like Ferguson et al.
Well, the CDC's most recent IHR is ~4.9%. So not a good start for Silver.
web.archive.org/web/2021042521…
8/H
IHR is higher for nursing home residents, consistent with higher IFR for nursing home residents + older people due to more severe infections.
So IHR can be higher in older populations + lower elsewhere
ncbi.nlm.nih.gov/pmc/articles/P…
medrxiv.org/content/10.110…
link.springer.com/article/10.100…
9/H
Yet under-estimating IHRs by excluding nursing home residents, leads to IHRs are at or above Silver's value of ~2%.
With the CDC's analysis, that further undermines Silver's IHR claim.
"2.1%"
ingentaconnect.com/content/wk/phh…
"2.7%"
sciencedirect.com/science/articl…
sciencedirect.com/science/articl…
10/H
So where did @NateSilver538 go wrong?
It goes back to the New York study he relied on.
People who've read some of my IFR threads, especially those on Ioannidis, know what I'm about to say:
non-representative sampling. 🙂
ncbi.nlm.nih.gov/pmc/articles/P…
11/H
The study Silver relied on sampled only those in grocery stores. None of the IHR work cited in 7/H to 9/H did that.
So Silver likely over-estimated the number of people infected, + thus under-estimated IHR.
ncbi.nlm.nih.gov/pmc/articles/P…
12/H
Silver messed this up because he's a non-expert.
What he should have done was run this by experts first, + listened when they corrected him.
Instead he stuck to his false claims despite correction, + used this to unfairly criticize experts.
onlinelibrary.wiley.com/doi/full/10.11…
13/H
Silver often does this sort of "epistemic trespassing," where he contradicts experts in a topic, when the problem is that he doesn't understand the information that experts do.
For example, on climate models (after speaking to @ClimateOfGavin):
14/H
To modify @Potholer54T's rule:
If you're a non-expert disagreeing with the evidence-based consensus of scientific experts, then either:
1) experts know less than you
2) experts covered up what they know
3) experts know more than you
Start with #3
15/H
Silver claims that in March 2020 the consensus range for IHR was 5% - 20%.
His citation of the New York Times doesn't make his case, since the range they let people choose is not the same as a best estimate for the model.
16/H
Looking back, several sources either:
- use Ferguson et al.'s IHR value of 4.4%
- use a value of ≤8%
Either option is consistent with the range of IHRs given in parts 7/H to 9/H.
The WHO gives a higher value, when just relaying information on reported cases.
17/H
Examples:
"from Imperial College"
medrxiv.org/content/10.110…
"2.7%"
sites.utexas.edu/dodabalapur/fi…
"1 out of every 20 people"
news.utexas.edu/2020/03/26/a-n…
"5 percent"
montclairlocal.news/2020/03/30/cov…
[predictivehealthcare.pennmedicine.org/2020/03/14/acc…]
"6.2% [...] 6.4%"
uaa.alaska.edu/academics/coll…
who.int/emergencies/di…
18/H
So Silver made unsupported claims on what the "consensus" showed, to make himself look more accurate than experts.
In reality, the evidence-based expert consensus was right, and Silver wrongly downplayed the risk of COVID-19
Twitter isn't showing part 15/H for some reason, so here it is:
Share this Scrolly Tale with your friends.
A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.