1. What's the infection fatality rate for COVID-19? @MatteoParadisi and I looked at 10 Lombardy towns locked down Feb 21. With a fully functioning hospital system(early days of epidemic), we estimate 0.67% overall, <1% for under 70 e 2-5% for over 70 (1/8)

To get this, we matched the daily deaths data by town up to March 21 to the 2018 demographic data for each of those towns, deaths data is only available for the 7 largest, ~50k total population (2/8)

We computed excess deaths for each day as the differences between 2020 recorded deaths and the 2015-2019 average, can see an obvious spike at the end of Feb @DellAnnaLuca @margheritamvs @matheusagaso (as in https://twitter.com/gian_rinaldi/status/1245466677615022081) (3/8)

Then we simply divided these excess deaths by a proportion of the total population in each age range to obtain the estimates above (4/8)

For the midpoint we assumed 60% of the population was infected: this is high but 66% of a sample of asymptomatic blood donors was positive to antibody testing ( https://www.ilgiornale.it/news/cronache/asintomatici-anticorpi-loro-plasma-cura-virus-1848999.html, in Italian) (5/8)

The error bars allow for uncertainty on the excess deaths (+/2 standard deviation of prior year deaths by age range) and infection rate in the population (30% for upper bound and 80% for the lower bound) (6/8)

If you do not believe this upper and lower bound, we show the full range for any possible infection rate between 15% and 80% of the population. Since we fix the number of deaths, the rates get worse as a smaller fraction of the population is infected (7/8)

The rates might be higher than this because we only have deaths until 21 March. These town were in lockdown already on Feb 21 so hopefully additional excess deaths are not too many (8/8)