[M T]

 If the CFR was 1%, then that means that ~350,000 people have been infected (114,000 have been reported)

I have a question on terminology. In earlier discussions of the CFR, I was using the term as you do above (I was speculating that, due to what I thought would be a substantial number of asymptomatic or minimally symptomatic infections, that the CFR would be end up being put at less than 1%--which I still think will be true). But since then I've read elsewhere that this is more properly called the Infection Fatality Rate, and that Case Fatality Rate is a more ambiguous term that depends on how a case is defined (e.g., positive test; hospitalization; ICU transfer, etc.). Can anyone help straighten me out? Thanks!

I have been using the terminology from this report:
Estimates of the severity of COVID-19 disease

Here we attempt to adjust for these biases in data sources to obtain estimates of the CFR(proportion of all cases that will eventually die) and infection fatality ratio (IFR, the proportion of all infections that will eventually die)

I appreciated separating out what could be measured early in this epidemic (known cases) from the unknown (number of unknown cases).  Even today, if I get symptoms and the PCR test returns positive, we are seeing that I would have about a 20% chance of developing pneumonia and a 5% chance of needing to be in the ICU as a result of the disease.   ( If I don't have symptoms at the time of the test, there is some not-well-documented chance I'll get symptoms.  If I never get symptoms, the likelihood of the disease causing pneumonia or the ICU is 0%.)

There are several things left open to interpretation:

Cases:  positive test? clinically diagnosed, what about those that died before testing?

Infections:  How do you count the asymptomatic?   I'm pretty sure you have to estimate it based on some sampling.   
And, for COVID-19, do you treat all cases equally?  Or do you need to account for age, gender, handedness, or whatever parameters you care to differentiate on.   I've been assuming there are some unknown factors that will determine whether someone who gets infected actually shows symptoms or not.  Are those factors something that can be found from the data?  It might be that 90% of 90yo have symptoms and 20% of 2yo.  Or maybe it isn't age dependent.

Deaths: Do you only count current deaths.  H ow do you count future deaths?  (you can just wait)   How do you count deaths that weren't known whether they were from the disease or not?  This was the number that (IMO) was not properly accounted for by the various reporting agencies (WHO, CDC), leading to very low CFRs.


So, (in my non-professional opinion) CFR should start with a known population (for instance, those with symptoms that got a positive test) and then estimate the deaths from that group only (don't count deaths of people that never had the test).  Again, you can pull out interesting subgroups of that population, by age, gender, handedness,...    Post-infection medical care will impact the CFR. 

IFR is squishier.  IFR is a function of the disease, a population, the medical care, the deaths, and a way to estimate the number of infections.  It is an attempt to say how many people in a population that get the disease, die from the disease.  The question is how do you count the asymptomatic, especially in a new disease.   Even recently, we've seen researchers hypothesize vastly different asymptomatic to symptomatic ratios.

Then there's also what is called the attack rate, basically what percent of the population get an infection.  For COVID-19, this would relatively large, if unchecked.  Many places, by doing social distancing, are trying to lower this rate.    Except by avoiding overrunning the medical system, the CFR and IFR would not be affected by social distancing.

(No, I am unaware of any relation of COVID-19 to handedness.  That is just a stand-in for any of the other characteristics that one might use, but without actually being any specific one.) 


Evidence from the Diamond Princess and from the Wuhan evacuation flights puts the number of asymptomatic approximately equal to the symptomatic,  but were those typical of the general population? (I doubt the ages of the cruise customers is typical).  There are some epidemiology studies of the spread that likewise concluded that there were about as many unreported cases as reported cases in China.  (Frankly, I didn't dig into those so I don't know how they got that result.  Uncharacteristically, I've just accepted that.)

[M T]

 If the CFR was 1%, then that means that ~350,000 people have been infected (114,000 have been reported)

I have a question on terminology. In earlier discussions of the CFR, I was using the term as you do above (I was speculating that, due to what I thought would be a substantial number of asymptomatic or minimally symptomatic infections, that the CFR would be end up being put at less than 1%--which I still think will be true). But since then I've read elsewhere that this is more properly called the Infection Fatality Rate, and that Case Fatality Rate is a more ambiguous term that depends on how a case is defined (e.g., positive test; hospitalization; ICU transfer, etc.). Can anyone help straighten me out? Thanks!

I have been using the terminology from this report:
Estimates of the severity of COVID-19 disease

Here we attempt to adjust for these biases in data sources to obtain estimates of the CFR(proportion of all cases that will eventually die) and infection fatality ratio (IFR, the proportion of all infections that will eventually die)

I appreciated separating out what could be measured early in this epidemic (known cases) from the unknown (number of unknown cases).  Even today, if I get symptoms and the PCR test returns positive, we are seeing that I would have about a 20% chance of developing pneumonia and a 5% chance of needing to be in the ICU as a result of the disease.   ( If I don't have symptoms at the time of the test, there is some not-well-documented chance I'll get symptoms.  If I never get symptoms, the likelihood of the disease causing pneumonia or the ICU is 0%.)

There are several things left open to interpretation:

Cases:  positive test? clinically diagnosed, what about those that died before testing?

Infections:  How do you count the asymptomatic?   I'm pretty sure you have to estimate it based on some sampling.   
And, for COVID-19, do you treat all cases equally?  Or do you need to account for age, gender, handedness, or whatever parameters you care to differentiate on.   I've been assuming there are some unknown factors that will determine whether someone who gets infected actually shows symptoms or not.  Are those factors something that can be found from the data?  It might be that 90% of 90yo have symptoms and 20% of 2yo.  Or maybe it isn't age dependent.

Deaths: Do you only count current deaths.  H ow do you count future deaths?  (you can just wait)   How do you count deaths that weren't known whether they were from the disease or not?  This was the number that (IMO) was not properly accounted for by the various reporting agencies (WHO, CDC), leading to very low CFRs.


So, (in my non-professional opinion) CFR should start with a known population (for instance, those with symptoms that got a positive test) and then estimate the deaths from that group only (don't count deaths of people that never had the test).  Again, you can pull out interesting subgroups of that population, by age, gender, handedness,...    Post-infection medical care will impact the CFR. 

IFR is squishier.  IFR is a function of the disease, a population, the medical care, the deaths, and a way to estimate the number of infections.  It is an attempt to say how many people in a population that get the disease, die from the disease.  The question is how do you count the asymptomatic, especially in a new disease.   Even recently, we've seen researchers hypothesize vastly different asymptomatic to symptomatic ratios.

Then there's also what is called the attack rate, basically what percent of the population get an infection.  For COVID-19, this would relatively large, if unchecked.  Many places, by doing social distancing, are trying to lower this rate.    Except by avoiding overrunning the medical system, the CFR and IFR would not be affected by social distancing.

(No, I am unaware of any relation of COVID-19 to handedness.  That is just a stand-in for any of the other characteristics that one might use, but without actually being any specific one.) 


Evidence from the Diamond Princess and from the Wuhan evacuation flights puts the number of asymptomatic approximately equal to the symptomatic,  but were those typical of the general population? (I doubt the ages of the cruise customers is typical).  There are some epidemiology studies of the spread that likewise concluded that there were about as many unreported cases as reported cases in China.  (Frankly, I didn't dig into those so I don't know how they got that result.  Uncharacteristically, I've just accepted that.)