[burger]

  Now, IFRs and CFRs seem to be dropping (a good thing if true), but that's giving people like this med school guy false confidence that we don't need to worry about current growth in cases. 

I'd like to clarify something.  One way to drop IFR and CFR is to add more cases that weren't there before.  For IFR, you can find about a factor of 10 between some lower estimates to higher estimates of uncounted asymptomatic cases that are then used in the denominator.   (Of course there are measures of entire populations, some with  0% infections and some  with 94%(IIRC) infections, but those are generally non-representative samples.)
With CFR, wider testing has increased the denominator, adding in very mild or fully asymptomatic cases.  The early CFR numbers were totally bogus because they usually had a denominator full of cases that weren't yet known to be fatalities or not.  AND, in the US at least, we had suspected cases that weren't counted because they weren't tested.

So, I don't think either IFR or CFR are really good numbers for comparing now to earlier (or now to future).

But, I did see one good number at the CDC site a couple of days ago re Mortality.

Based on death certificate data, the percentage of deaths attributed to pneumonia, influenza or COVID-19 (PIC) decreased from 9.0% during week 25 to 5.9% during week 26, representing the tenth week of a declining percentage of deaths due to PIC. The percentage is currently at the epidemic threshold but will likely change as more death certificates are processed, particularly for recent weeks.

In other words, 5.9% is not complete, and looking at the previous week's numbers now versus this week's numbers is not right either.  But if the new, incomplete numbers each week are dropping, or the 1-week old numbers are always dropping, that's a good sign.
(I suppose it could also be that a backlog of deaths is being cleared out, and there is less to clear out.  Bummer if that's it.)
PIC = pneumonia, influenza, covid-19

Also I'll note that excess deaths are those in excess of the normal deaths.  However, since April 1, I'd guess that the number of flu deaths are far below normal, so our "normal death rate" for April (etc) is higher than it would be for non-COVID deaths.  I take it flu deaths are the cause of the wave in the line above.

I agree with what you said about IFR and CFR.  I probably should have restricted my comment to IFR--I was thinking of a comment I saw from a doctor who said that the survival rate for covid patients on ventilators had increased from 25% at the start of all this to 75% now (I have no idea where I saw this--I might even be misremembering, so take with extra salt).

Speaking of change in denominators due to more testing, some modelers think that the actual numbe of infections right now is basically at the same level as it was in March/April despite the reported cases being far higher now:

[tweet]https://twitter.com/NateSilver538/status/1281372870589939712[/tweet]

As for the mortality data, there is always a lag of like 6-8 weeks before the deaths are fully reported.  So any decrease recently could be an artifact of incomplete reporting.  We were seeing this back in April and May; people would look at the latest CDC numbers and declare things were almost over when it was just missing data.

[burger]

  Now, IFRs and CFRs seem to be dropping (a good thing if true), but that's giving people like this med school guy false confidence that we don't need to worry about current growth in cases. 

I'd like to clarify something.  One way to drop IFR and CFR is to add more cases that weren't there before.  For IFR, you can find about a factor of 10 between some lower estimates to higher estimates of uncounted asymptomatic cases that are then used in the denominator.   (Of course there are measures of entire populations, some with  0% infections and some  with 94%(IIRC) infections, but those are generally non-representative samples.)
With CFR, wider testing has increased the denominator, adding in very mild or fully asymptomatic cases.  The early CFR numbers were totally bogus because they usually had a denominator full of cases that weren't yet known to be fatalities or not.  AND, in the US at least, we had suspected cases that weren't counted because they weren't tested.

So, I don't think either IFR or CFR are really good numbers for comparing now to earlier (or now to future).

But, I did see one good number at the CDC site a couple of days ago re Mortality.

Based on death certificate data, the percentage of deaths attributed to pneumonia, influenza or COVID-19 (PIC) decreased from 9.0% during week 25 to 5.9% during week 26, representing the tenth week of a declining percentage of deaths due to PIC. The percentage is currently at the epidemic threshold but will likely change as more death certificates are processed, particularly for recent weeks.

In other words, 5.9% is not complete, and looking at the previous week's numbers now versus this week's numbers is not right either.  But if the new, incomplete numbers each week are dropping, or the 1-week old numbers are always dropping, that's a good sign.
(I suppose it could also be that a backlog of deaths is being cleared out, and there is less to clear out.  Bummer if that's it.)
PIC = pneumonia, influenza, covid-19

Also I'll note that excess deaths are those in excess of the normal deaths.  However, since April 1, I'd guess that the number of flu deaths are far below normal, so our "normal death rate" for April (etc) is higher than it would be for non-COVID deaths.  I take it flu deaths are the cause of the wave in the line above.

I agree with what you said about IFR and CFR.  I probably should have restricted my comment to IFR--I was thinking of a comment I saw from a doctor who said that the survival rate for covid patients on ventilators had increased from 25% at the start of all this to 75% now (I have no idea where I saw this--I might even be misremembering, so take with extra salt).

Speaking of change in denominators due to more testing, some modelers think that the actual numbe of infections right now is basically at the same level as it was in March/April despite the reported cases being far higher now:

[tweet]https://twitter.com/NateSilver538/status/1281372870589939712[/tweet]

As for the mortality data, there is always a lag of like 6-8 weeks before the deaths are fully reported.  So any decrease recently could be an artifact of incomplete reporting.  We were seeing this back in April and May; people would look at the latest CDC numbers and declare things were almost over when it was just missing data.