This is the (tentative) conclusion of a meta-analysis found on the WHO's website, written by John Ioannidis, which in its abstract says:

Across 51 locations, the median COVID-19 infection fatality rate was 0.27% (corrected 0.23%). [...] In people < 70 years, infection fatality rates ranged from 0.00% to 0.31% with crude and corrected medians of 0.05%.

[... and in its conclusion... ]

Most locations probably have an infection fatality rate less than 0.20%.

Ioannidis also self-cites that in a later paper to say a bit more concisely:

Global infection fatality rate is 0.15-0.20% (0.03-0.04% in those <70 years)

According to Wikipedia, Ioannidis was criticized for some previous (primary) studies on the matter. So is this meta-analysis essentially correct or possibly misleading in some way?

Note 1: according to the mods, this a distinct question from Did WHO publish a bulletin stating that COVID-19 is “equivalent in lethality to seasonal flu”?, although it's obviously related as it was considered in an answer there as being what someone (else) implicitly referred to.

Note 1-bis: That meta-analysis has also been cited in a letter to the editor of the BMJ (by a certain Eshani M King) to say that

A recent peer-reviewed paper by one of the world’s most cited and respected scientist, Professor John Ioannidis of Stanford University, quotes an infection fatality rate (IFR) for Covid of 0.00-0.57% (0.05% for under 70s), far lower than originally feared and no different to severe flu. This paper is published on WHO’s own Bulletin but ignored by UK mainstream media.

(emphasis mine). But Ioannidis himself never seems to make an explicit comparison with flu and there are some other questions here on Skeptics about the appropriateness of the comparison, so let's stick with the original/technical claim in Ioannidis' own paper(s) for this Skeptics question.

(The sub-claim in that letter of the paper being ignored by the press is somewhat inaccurate. The Daily Mail, for instance, did cover the paper in a stand-alone article.)

Note 2: infection fatality rate is not case fatality rate. The former includes in the denominator all those infected, even if asymptomatic.

  • 2
    Not sure if that is the same paper that made rounds rather early in the pandemic, but I vaguely remember lots of criticism of cherry-picking the studies for the meta analysis there. And on the IFR comparison to the flu I've seen a few comparisons recently between typical flu season deaths and Covid, and it's not even close, so that might be one way to answer this by comparing it directly to previous flu seasons as that is more straightforward than the IFR calculations
    – Mad Scientist
    Jan 13, 2021 at 9:06
  • 8
    Ionnadis' work, in particular, has been terrible throughout the Covid epidemic. He's infamous for predicting that there would be fewer than 5,000 deaths in the US, and then quickly revising that to 10,000 as events rapidly overtook his cherry-picked statistics. It's become clear that his papers on the subject reflect his biases rather than scientific reasoning.
    – antlersoft
    Jan 13, 2021 at 18:43
  • 3
    The UK Corona Infection Survey reported in Nov 2020 that 8.7% of the population had antibodies, at which time 0.08% of the population had been killed by it. So that's closer to a 1% infection fatality rate. Jan 13, 2021 at 22:23
  • 3
    Current US fatalities are well over 330k. Even assuming 330k that gives a population fatality rate of 0.1% (divide by your estimate of infection rate to get the IFR so, if you think 10% of the population has been infected the IFR is 1%). This doesn't reinforce the estimates in that paper unless you make absurd assumptions about the infection rate.
    – matt_black
    Jan 15, 2021 at 20:30
  • 3
    @matt_black - US fatalities are verging on 400k rather than a "mere" 330k. Either way, a 0.1% IFR would imply that every American has been either infected at least once, or that people can easily be infected twice, thrice, or even more often. An IFR of 0.2% or lower is either ludicrous or a sign that this disease is even more highly contagious than people thought, and a disease where immunity does not last very long at all. Whatever the case may be, it makes the seasonal flu look like a case of the sniffles. Jan 16, 2021 at 16:05

4 Answers 4


It is implausible if we assume the estimate of 400,000 USA Covid deaths to date is correct. If the IFR is really 0.2%, we would have 200,000,000 infections, i.e., well over half of the USA population. At that level, we should start seeing herd immunity effects. Although I see the original estimate of 60-70% infected for herd immunity is being revised upward, it's hard to see this as compatible with recent significant growth in the number of daily cases and deaths.

The claim becomes even more dubious when we restrict to a hot-spot. New York City has recorded 26,161 deaths, implying over 13 million infections on 0.2% IFR. However, the population of NYC is only 8.3 million. Some of the differential can be explained by better treatment since the beginning of the pandemic, but the worst NYC test positivity rates are about 10% at the moment. The pieces of the puzzle fit together best by rejecting the low IFR.

  • Somewhat similar to this argument, another paper (this one published in Science) estimated that in Manaus, Brazil > 70% of the population had the virus. But the same location is currently experiencing another Covid-19 surge. The situation there is a bit more complicated because a variant of the virus found in that country does have enhanced ability to evade antibodies and there is a confirmed case of reinfection with old-then-the-new variant. OTOH, that (new) Brazilian lineage hasn't been found in the US (but something similar has been in South Africa). Links to follow. Jan 20, 2021 at 6:06
  • Also: jamanetwork.com/journals/jamainternalmedicine/fullarticle/… : "the estimated percentage of persons in a jurisdiction with detectable SARS-CoV-2 antibodies ranged from fewer than 1% to 23%. Over 4 sampling periods in 42 of 49 jurisdictions with calculated estimates, fewer than 10% of people had detectable SARS-CoV-2 antibodies." this is for July to September, you would need to scale up in some appropriate way, or more plausibly use death counts from that time range
    – Ben Bolker
    Jan 21, 2021 at 23:00
  • Can you assert any truth to the claims though? I think you miss half the question here. What's the data for people under 70? So far I've heard that covid has extended the average lifespan of a given human, lending extreme credence to the second figure.
    – tuskiomi
    Feb 3, 2021 at 7:54
  • @tuskiomi Which numbers need further assertion? They are official government population estimates and official government death counts. COVID–19 is causing a huge reduction in USA life expectancy, not an extension. Feb 3, 2021 at 17:16

If you take a closer look at the paper, then you will notice that the methodology to detect infected people was based on studies which tested more or less randomly selected people for antibodies against SARS-CoV-2, looked at the percentage of people with antibodies, extrapolated that number to the whole population of the region from which the tested people were drawn, and then compared that to the reported number of deaths from COVID-19 in that region.

The assumption is that anyone with antibodies already had an infection with SARS-CoV-2, but might not have realized it because they were asymptomatic and there was no reason for them to get tested during their infection.

This is based on a couple assumption which might or might not be true:

The official numbers you can see on various COVID-19 dashboards (like the one from the John Hopkins university) list only confirmed active infections vs. deaths. With that methodology, you come to a fatality rate of ~2% with regional differences ranging from 1% to up to 10%. These regional differences can be explained by prevalence of testing (less testing usually means only obviously sick people get tested and reported, who are usually people who are less likely to survive) and quality of health care.


In interviews, Ioannidis has stated that IFR is not a constant. It varies geographically (which the question implies) and in time. So I think the question also needs to specify a time frame. An example:

Around Boston, USA the current viral peak in wastewater is over twice as high as in the Spring. Yet death rates are less than half as high. An exact geographic match to the wastewater area is unavailable, but state and city numbers confirm this. Apparently, IFR has shrunk considerably since the Spring.

  • He starts by talking about the Infection Fatality Rate, but then goes to describe it as calculated from the factors going into the Case Fatality Rate. So while I think he has an important point about the CFR, OP is specifically asking about the IFR. Jan 25, 2021 at 3:41

The IFR in Denmark, around April-May 2020, was estimated as 0.089% for patients younger than 70.

This was based on analysis of blood donors, who tend to be healthier than the regular public.

Results: The first 20,640 blood donors were tested and a combined adjusted seroprevalence of 1.9% (CI: 0.8-2.3) was calculated. The seroprevalence differed across areas. Using available data on fatalities and population numbers a combined IFR in patients younger than 70 is estimated at 89 per 100,000 (CI: 72-211) infections.

Conclusions: The IFR was estimated to be slightly lower than previously reported from other countries not using seroprevalence data. The IFR is likely several fold lower than the current estimate. We have initiated real-time nationwide anti-SARS-CoV-2 seroprevalence surveying of blood donations as a tool in monitoring the epidemic.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .