COVID-19 (artist formerly known as Wuhan strain novel Corona virus)

Thats the definition of case fatality rate. If detected cases go up and deaths don’t it must go down.

Infection fatality rate is another story. But there are signs that has gone down too.

 

If there is no attempt to match the case to the death, by date, then it's worthless.

And as near as I can tell, there is no attempt.

If you look at it by day, one case, one outcome, you'd get:


Oct 3rd:
1 case
0 deaths

Case fatality rate: infinite

October 24:
0 cases
1 death
Case fatality rate: infinite

That's a dumb measure, it's non deterministic. It should be 1/1, 100%, deterministic.

Which isn't infection fatality rate, but a deterministic case death rate.

 

You would typically use a period and aggregate over it.

Point is that if you are detecting a higher percentage of infections as cases then the case fatality rate goes down by definition, assuming no increase in infection fatality rate.

 

I believe fatality rate has gone down, certainly compared to march. it still is higher than the flu and the risks increase if hospitals are stretched

 

That's a dumb measure to me. If it can't, at its base, measure a specific outcome, it's a bad measure.

The bottom is also bad, to me. You can have multiple infections, per case, depending on your time aggregate. So non-deterministic. Bad measure.

It's like measuring a shadow the way you describe it. It's going to depend on more than just the measure. So, bad measure.

 

The lower limit of case fatality rate is the infection fatality rate at 100% case capture. How you calculate case fatality rate can also be an issue during the epidemic, as you point out, due to lag. But my point is that if you are increasing testing and getting better case capture, your final case fatality rate is going to be lower.

 

And my point is that won't, necessarily, but the way it is being done, it will, always. It could be constant. But lag makes it lower.

 

How would it be constant if the fraction of infections you actually measure and declare a case goes up?

 

If those cases then die. That would be a constant ratio, with an increase.

But it isn't tracked case to case, you say, but by some date. Which means it's just measuring shadow. Useless.

 

If they die then they’re added to the fatality numbers.

I don’t get your attack on it. No it’s not real time but a useful gage

 

ya, it is useful, even if not direct. certainly good enough when looking at rolling trends, so long as you dont cherry pick right at a spike like has been done 3 out of 3 times now.

 

Even if tracked case to outcome vs by date (which can absolutely be done but often isn’t how it is reported), the case fatality rate will go down over time as case capture rate goes up. This is because you aren’t going to die of it without being diagnosed. Therefore all deaths were a case but not all infections were a case.

If the likelihood of missing an infection that would ultimately lead to death is equal to missing one that wouldn’t, then increased testing should have no impact. I’ll give you that. But due to asymptomatic infections that absolutely isn’t the case.

 

No, it isn't useful. It's measuring shadow. As TennTradition has pointed out, based on how y'all are defining this thing, it'll always go down. Always. Which means it can't be used to show when things are going up. And if it can't measure two directions, it's useless.

 

You've just demonstrated why it is useless.

 

It shouldn't always go down. But you're telling me it will always go down. So it can't measure anything, because measurement goes both ways.

The way I would define case to death is literally looking at each case individually, and looking at each outcome, individually. And that would allow me to say that I don't know for sure that case to fatality is going down. It might be going up. We're just masking it by measuring lots of cases.

If case to death is a measure that is always going down, I'm not going to use that definition, because that's stupid.

 

i don’t know if it will always go down but with this virus it will due to that it’s just not as deadly as we first feared.

But if it did mutate to become more deadly this would show it by rising

 

No I’m not telling you it will always go down in a vacuum. It will go down if case capture rate goes up - AND (which was in my head bit not written until my last post, but is critical to the point) it goes up because you are capturing more of the less severe or asymptomatic cases.

At one extreme is everyone who dies of a respiratory virus gets tested. High CFR. Next is everyone who is very sick with a respiratory virus gets tested. Lower CFR. Next is some mildly asymptomatic people get tested. Lower. Next is asymptomatic people actually start getting captured. Even lower.

 

It won’t go down because it is less deadly than we thought it was.

 

I'm talking about only cases, and their outcomes.

And only measuring a case by its, and only its, outcome.

Anything else is a useless measure to me.

Reporting on a given day that you have 10,000 cases, and 30 deaths, and those 30 deaths aren't in your 10,000 cases is useless.

And that is what you are telling me that measure is capable of doing.

 

Deaths to cases may be going up right now. We miss it because new cases are dominating deaths.