Thread by @GidMK

More estimates of the IFR of COVID-19 out recently

Might start a thread collating the new ones I see

1 - NYC

We already had the basic data, but the testing implies 22.7% infected from tests between 19/04-28/04. Taking the 28/04 date to crudely account for right-censoring, this gives an IFR:

17682/1906573 = ~0.92%

Study is here: medrxiv.org/content/10.110…

IFR for the whole state is a bit lower - using the same date cutoff, it is ~0.84% with 14% infected

2 - Brazil

An impressive effort from Brazil, nationwide seroprevalence sampling. The authors report a corrected IFR of 1%

medrxiv.org/content/10.110…

What's particularly interesting about the Brazilian study is that the IFR range is given, with places reporting between 0% and 2.4%

3 - London

While the testing hasn't been published formally, the estimates appear to be that 17% of the city had antibodies mid-May

So, cumulative incidence of ~1.5million people by 13th

thelancet.com/journals/lanre…

Now, extracting an IFR from these figures is a bit of a headache, because England only publishes specific information on COVID-19 deaths in hospitals, but the deaths appear to be 5,644 on this date from that source

Given that ~75% of London COVID-19 deaths occur in hospitals, that means ~7,500 deaths and ~1.5mil cases, so an IFR of ~0.5% for London

We can also look at the whole UK IFR from this data

~5% of the country infected gives ~3,332,500 infections

24,000 deaths in-hospital gives ~32,000 deaths

Therefore overall IFR is 32k/3.3mil = 0.96% IFR

(Apologies, above tweet should say IFR for the whole of ENGLAND, not the UK. This data is from the ONS testing in England, and the death reports from England as well)