Read on Twitter
Thread on how lethal Covid is using available evidence rather than modelling guesses. One place to start is population fatality rates in hard hit city districts eg in New York one district saw 0.6% of entire population (not just those infected) die /1 https://www.nytimes.com/2020/05/18/nyregion/coronavirus-deaths-nyc.html">https://www.nytimes.com/2020/05/1...
Likewise some districts in London saw over 0.1% of the population die, again an underestimate as testing was limited at that point in time. 0.1% is 1000 people per million which some have suggested is a magic upper limit yet Edgemere NYC was 6000 per million /2
In some places Covid19 went so out of control that it may even have got close to so called & #39;herd immunity& #39;. Manaus Brazil is one such case where 0.28% died or 2800 per million. However only 6% of its population are over 65s, if you adjust for age the IFR was around 0.6% /3
Mumbai in India was prematurely thought to have reached herd immunity but now a 2nd surge is sweeping through it. IFR was thought to be as low as 0.2% but this now appears to be based on many deaths being missed with 0.35% now being more likely /4
https://science.thewire.in/the-sciences/covid-19-mumbai-all-cause-mortality-data-ifr-bmc-seroprevalence-survey/">https://science.thewire.in/the-scien...
https://science.thewire.in/the-sciences/covid-19-mumbai-all-cause-mortality-data-ifr-bmc-seroprevalence-survey/">https://science.thewire.in/the-scien...
How can deaths be uncertain? Financial Times has done a very good job of tracking often enormous gaps between official Covid19 deaths & excess deaths in same period. Jakarta was 15 times! Suggest& #39;s very much higher deaths & so much higher IFR& #39;s. /5 https://www.ft.com/content/6bd88b7d-3386-4543-b2e9-0d5c6fac846c">https://www.ft.com/content/6...
Looking at excess deaths suggests 0.58% of the entire population of Bergamo may have died (5800 per million) which would represent an IFR around 0.8% presuming 69% got infected /6 https://www.medrxiv.org/content/10.1101/2020.04.15.20067074v3">https://www.medrxiv.org/content/1...
These real world examples suggest IFRs quite close to each other, in particular once you allow for the age structure of the population. 0.6% seems reasonable after age adjustment which is much lower than many including myself first feared /7
We also get some idea of how many people can be easily infected 60-70% which is consistent with herd immunity thresholds derived from the presumed R value. So you can crudely estimate how many might die in Ireland for instance (5 million * 0.006 * 0.65) = 19,500 /8
Ireland however has an older population, a crude comparison between the 1.7% infection rate detected by our seroprevalence study in July and 1700 deaths at that time would give you an IFR nearer 2.8% but thats because there were disproportionate outbreaks in nursing homes /9
To get a real estimate of likely death toll you really need to calculate mortality by actual age structure of population & IFR for each of those age bands. Here the incidence of who makes up the 65% infected becomes important . Our estimate of deaths might rise to 30,000+ /10
The key take away here is that anyone who speaks with great certainty about what the IFR is or makes exact predictions about possible mortality is at best a bluffer. In most countries we don& #39;t yet have a good idea of how many were really infected or how many really died /11
We can look at emerging evidence and suggest some upper and lower bands, although even those depend on what segment of the population gets infected. For a country like Ireland those guesses might run from 0.3% to close to 2% at an unlikely extreme /12
