That is the concept of IFR vs CFR. If the denominator contains the total number of infected, instead of the total number of confirmed cases, than this ratio is the IFR. But the CFR is defined exactly as I reported, total number of confirmed deaths divided total number of confirmed cases.
Then we should discuss how many people are infected and not confirmed: I have no reliable data for US, but I see CDC estimates this ratio less than 2. Hence in their model a CFR = 0.4% transforms into an IFR = 0.26%.
Instead here in Italy we start having some reliable data (as known, we are almost one month ahead US in managing the epidemics), and the ratio is around 8-10 for the whole Italy, peaking at 35 for Bergamo province (the one with higher number of deaths per million inhabitants).
One of the sources of this info is here (from Berkeley university):
https://www.medrxiv.org/content/10.1101/2020.04.15.20067074v3.full.pdf
This means that here in Italy we have a number of confirmed cases actually equal to 229,858, and a number of deaths equal to 32785, making the value of CFR = 14.26%. But we know that the total number of infected is around 2 millions, hence the IFR is approximately best estimated around 1.58%.
I would be very, very happy to find data more favourable...
I understand that IFR and CFR are different values, but to use a CFR while testing is low doesn't tell a complete story. There's a reason that the CDC is attempting to calculated a "symptomatic case fatality ratio" rather than just counting cases.
It's also important to keep in mind that Italy's numbers may not be representative of the world as a whole due to differences in demographics, lifestyle, and how deaths are counted.