1/U
You may recently have heard that COVID-19 has a fatality rate of ~0.15%, making it akin to a bad flu.
In reality, a more accurate fatality rate would be closer to ~0.6%, as per the WHO.
That's ≥10X worse than seasonal flu, and ~100X worse than the 2009 swine flu pandemic.
2/U
Background:
Infection fatality rate (IFR) is the proportion of people infected with the virus SARS-CoV-2 who die of the disease COVID-19.
IFR for seasonal flu is <0.1%, as per the WHO, among others:
who.int/emergencies/di…
institutefordiseasemodeling.github.io/nCoV-public/an…
3/U
So, following @BallouxFrancois, who in their right mind would claim COVID-19 has an IFR comparable with that of seasonal flu?
Well, I can think of at least 2 people.
You likely know who one of them is. 😉
from 2:27 :
4/U
Make that 3 people
"Ioannidis [said it] has an “infection fatality rate that is in the same ballpark as seasonal influenza.”"
buzzfeednews.com/article/stepha…
Gøtzsche:
"[IFR] seems to be about the same as for influenza"
bmj.com/content/371/bm…
Tegnell:
unherd.com/2020/07/sweden…
5/U
Oh wait, there were at least dozens; see the thread below.
The list includes *all* the authors of the Great Barrington Declaration, the organization behind it, and a lot of people from Stanford.
Probably a coincidence.🤔
archive.is/QLmJt#selectio…
6/U
The "0.15%" IFR figure seemingly first appeared in an October 2020 paper from Ioannidis.
There he lists at least 2 methods for getting that figure.
As we'll see, there are at least 3 methods overall. And none of them work.
onlinelibrary.wiley.com/doi/pdf/10.111…
7/U
Method #1:
Cite the World Health Organization (WHO) as saying 10% of people were infected.
Problem with that is WHO officials said *less* than 10% of the population was infected. That includes a WHO expert on this subject.
web.archive.org/web/2020111415…
8/U
Also, from February through October 2020, WHO officials kept saying SARS-CoV-2 had a higher IFR than seasonal influenza.
"Mortality for COVID-19 appears higher than for influenza"
who.int/emergencies/di…
October 12:
who.int/publications/m…
9/U
And the WHO continues citing papers that show a higher IFR than Ioannidis claims; that's an IFR much higher than that of seasonal influenza.
They're not even citing his IFR work. That continues into 2021.
January 8, 2021:
apps.who.int/iris/handle/10…
10/U
So method #1 dies.
Ioannidis moves to a March 2021 paper, i.e. method #2. He conveniently removes mention of the WHO (probably because they contradict him).
That paper under-estimated IFR by using non-representative sampling, among other issues:
11/U
That leaves method #3, which Ioannidis tried in his October 2021 paper:
Decrease IFR from another one of Ioannidis' IFR studies, by claiming that study focused on places with abnormally large IFR.
That method doesn't work:
onlinelibrary.wiley.com/doi/pdf/10.111…
11/U
Unsurprisingly, the study Ioannidis adjusts under-estimated IFR by using non-representative samples that over-estimated the number of infections:
But @sschinke also points out under-estimating IFR by under-estimating deaths:
12/U
For example, Ioannidis gives an IFR of 0.06% for Scotland, which is impossible since >0.13% of their total population died of COVID-19.
But he gets that using 47 deaths by April 1, which is too low by at least a factor of 4
web.archive.org/web/2020111809…
13/U
Now this might not be Ioannidis' fault, since the death information could have been updated after he checked. He notes such changes in another paper.
But other times the error lies with him.
academic.oup.com/ije/advance-ar…
14/U
So I tried to address these death issues in the randomized/representative sampling studies that Ioannidis collected + adjusted for method #3.
After I got a median IFR of 0.58% (~0.6%).
A bit higher than before:
15/U
But that ~0.6% IFR matches what WHO officials said for months before *and after* they were aware of Ioannidis' work (see part 8/U), including his work that was submitted to the Bulletin of the WHO.
That fits with the following hypothesis.
16/U
WHO experts (😉) knew how to recognize representative sampling. So they removed studies with non-representative sampling from Ioannidis' analysis + addressed his errors on deaths.
That led to their 0.6% IFR
And they said not to cite his flawed work
apps.who.int/iris/bitstream…
17/U
That was just a hypothesis/conjecture. But it does neatly explain the WHO's IFR + their response to Ioannidis' work.
In any event, the '0.15% IFR' claim is nonsense.
It's likely closer to ~0.6%; much worse than seasonal flu
sciencedirect.com/science/articl…
18/U
Some people may want to check the median IFR of ~0.6% from part 14/U (I miswrote 0.61% as 0.58%).
So below is a numbered list of studies from Ioannidis' paper, along with which studies I used:
web.archive.org/web/2020111809…
19/U
Study #49 for Los Angeles County should be left out.
Ioannidis' IFR of 0.18% is impossible since >0.22% of the county died of COVID-19. Also had an accelerating outbreak.
dashboard.publichealth.lacounty.gov/covid19_survei…
coronavirus.jhu.edu/us-map
link.springer.com/article/10.100…
20/U
~0.6% IFR is consistent with other non-Ioannidis studies.
0.68% (0.53–0.82%)
0.76% (0.37–1.15%) for higher-quality studies
sciencedirect.com/science/articl…
0.79% (0.68–0.92%)
median range: 0.24–1.49%
nature.com/articles/s4158…
imperial.ac.uk/media/imperial…
21/U
Part 5/U included the wrong link be mistake. The thread it's referring to is below
22/U
Thread with more details on the Scotland deaths mentioned in parts 11/U and 13/U:
23/U
IFR increases with age
link.springer.com/article/10.100…
cdc.gov/coronavirus/20…
medrxiv.org/content/10.110…
So we can check if countries with ages close to the global median have IFRs near Ioannidis' global 0.15%
web.archive.org/web/2021040513…
24/U
On how well countries catch COVID-19 deaths:
medrxiv.org/content/10.110…
github.com/akarlinsky/wor…
Colombia's death reporting is pretty good, with an IFR of ~0.4% or more.
blogs.worldbank.org/developmenttal…
25/U
Peru under-estimates COVID-19 deaths, so its IFR should be >>0.5%.
seroprevalence:
dge.gob.pe/portal/docs/to…
web.archive.org/web/2020122009…
thelancet.com/journals/eclin…
dge.gob.pe/portalnuevo/wp…
28 - 35min: facebook.com/minsaperu/vide…
deaths:
covid19.minsa.gob.pe/sala_situacion…
26/U
Iran under-estimates deaths, so its IFR should be >0.4%
onlinelibrary.wiley.com/doi/full/10.10…
medrxiv.org/content/10.110…
seroprevalence:
- nationwide:
medrxiv.org/content/10.110…
(regions:
thelancet.com/journals/lanin…
wwwnc.cdc.gov/eid/article/27…
)
deaths:
covid19.who.int/region/emro/co…
27/U
Brazil slightly under-estimates deaths, with an IFR >0.6%.
seroprevalence:
thelancet.com/journals/langl…
ncbi.nlm.nih.gov/pmc/articles/P… (excludes nursing home deaths, lowering IFR)
estado.rs.gov.br/estudo-epidemi…
deaths:
transparencia.registrocivil.org.br/especial-covid
web.archive.org/web/2021040513…
28/U
So for 4 countries with randomized seroprevalence studies + median ages near the global median:
- IFR is larger than Ioannidis' global 0.15%
- IFR is compatible with the WHO's ~0.6%
Why are people still peddling Ioannidis' shoddy estimate?
🤔
publichealthontario.ca/-/media/docume…
29/U
Also, excess deaths in 24/U + 25/U are a good indicator of under-estimated COVID-19 deaths, regardless of Ioannidis' falsehoods on that.
Other indicators also point to under-estimated COVID-19 deaths.
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.