Talk:Case fatality rate

1
Initially this page redirected to case fatality. After posting to Talk:Case_fatality and receiving one vote in favor I made the change which I think improves clarity. BillpSea (talk) 21:48, 10 August 2009 (UTC)

CFR bias
One might consider adding, after "If some of the cases have not yet resolved (either died or recovered) at the time of analysis, this could lead to bias in estimating the CFR.", something like: "For example, when estimating the CFR as a quotient (number of dead-resolved cases) /(number of identified cases) that bias may lead to an overly optimistic estimate of the final CFR. While an epidemic is still expanding exponentially,  that bias will be larger, in general, - for faster expanding epidemics (with larger expansion factor), - for epidemics with larger mean duration of dead-resolving cases (from identification til resolution)." [Please note that there are further complications, for example if one wants to take into account an unknown number of still unidentified or even unnoticed cases.] Hal19 (talk) 16:37, 30 January 2020 (UTC)

With the definition as the ratio dead/diagnosed it seems unidentified or unnoticed cases are irrelevant for the CFR (provided deaths are correctly diagnosed). However, there ought to be some name for the ratio of deaths to all infected, whether unnoticed or not. Maybe someone knowledgeable could expand on this.Lsutvs (talk) 18:05, 6 March 2020 (UTC)

Thoughts about re-ordering; Fatality metrics; IFR < CFR not generally true
A substantial part of the article seems to be devoted to a discussion about the exact meaning of various rates and ratios, rather than mainly to the case fatality rate in its title. In itself such a discussion is relevant and helpful, in my opinion, although I find it rather difficult to find full clarity from the phrasing and wording in the current text. I also seem to have spotted a flaw in the current discussion, which I will attend to towards the end of this section. Below is my attempt at sorting out various definitions and meanings.

Would it not be helpful to create a page titled "Fatality metrics" which either discusses all the various metrics or - as an umbrella page - refers to "sub"pages discussing the various metrics separately? The following topics could then be addressed:
 * 1) Fatality metrics (general discussion)
 * 2) Causes (e.g. accidents, natural disasters, viral infection)
 * 3) Rates (e.g. number of fatalities per population or group unit per time unit) vs quotients (ratios) (e.g. number of fatalities per population or group unit)
 * 4) Definition of groups (e.g. in case of a viral disease: all affected vs those tested positive for affection by the virus, i.e. the positives)
 * 5) Discussion of fatalities (e.g. in case of a viral disease: all deaths vs the deaths who had/have tested positive for affection by the virus)
 * 6) Discussion of false positives and false negatives
 * 7) Influence of undersampling, both of the infected and of the death

Certain quantities would need to be rigorously defined. I have tried to chose the symbols in an ordered way offering mnemonics that may be helpful.
 * 1) N = number of all population members; generally changes with time, but could under circumstances be approximated as constant
 * 2) S = size of the group of members (surely) having a certain property (i.e. infected with, or having antibodies against, a virus); changes with time
 * 3) T = size of the group of members tested positive for a certain property (i.e. for infection with or antibodies against a virus); in the ideal case where testing would be fully accurate, these members are also known to all have the property; changes with time
 * 4) F = total number of fatalities from a certain condition or kind of incidents up to a certain point in time; changes with time
 * 5) G = number of members gone (i.e. dead) who have/had tested positive for the property under consideration; changes with time
 * 6) P = possibility / chance / risk of ever dying after (or - e.g. in cases where members can be re-infected - while) having a certain property, i.e. after having contracted a virus; while formally changing with time (as a function of the health, the age, and treatment distribution in a population), this quantity / metric seems to be the one that most resembles a constant;
 * 7) Q = fatality quotient (i.e. incidence) per group unit up to a certain point in time
 * 8) R = fatality / mortality rate per group unit and per time unit at a certain point in time (or over a relative short period of time)

Under the condition that there would be neither false positives nor false negatives we certainly have:
 * 1) T ≤ S
 * 2) G ≤ F

If I am not mistaken, we can now define e.g.:
 * 1) IFP = Infection Fatality Possibility = F/S for time to infinity (mathematically speaking), or practically speaking, after e.g. a viral disease has stopped raging
 * 2) CFQ = Case Fatality Quotient = G/T (commonly called CFR)
 * 3) IFQ = Infection Fatality Quotient = F/S (commonly called IFR)

In the article I read: "The term infection fatality rate (IFR) also applies to infectious disease outbreaks, and is closely related to the CFR, but attempts to additionally account for all asymptomatic and undiagnosed infections.[7] The IFR differs from the CFR in that it aims to estimate the fatality rate in all those with infection: the detected disease (cases) and those with an undetected disease (asymptomatic and not tested group).[8] (Individuals who are infected, but always remain asymptomatic, are said to have "inapparent" — or silent, or subclinical, or occult — infections.) The IFR will always be lower than the CFR."

The CFQ (commonly called CFR) may be inaccurate both because the deaths F are undersampled for cause of death and replaced by G with most often G ≤ F if testing is fairly accurate, and because the living members of a population affected by a disease, S, are undersampled and replaced by T, with most often T ≤ S if testing is fairly accurate. There can be no certaintly about whether and by how much G differs from F and T differs from S. So there is no saying either whether IFQ is always less than CFQ: we may have IFQ < CFQ, IFQ = CFQ, or IFQ > CFQ.Redav (talk) 11:23, 5 April 2020 (UTC)

Ratio vs rate
The remark about ratio vs rate seems incorrect. The cited reference (https://wiki.ecdc.europa.eu/fem/Pages/Attack%20rates%20and%20case%20fatality.aspx) explains: "The CFR is a case fatality expressed over time. It is therefore a true rate, since time is included in the denominator. ". The article claims " Case fatality ratio is the comparison between two different case fatality rates, expressed as ratio", whereas it's not true. The cited reference explains "Case fatality ratio" as: "This is simply the comparison of two case fatalities, expressed as a ratio." - so it's a comparison of two case fatalities NOT two case fatality rates. -- — Preceding unsigned comment added by BartoszKP (talk • contribs)
 * What is called "case fatality" is what our article calls CFR. The "rate" in their sense is not discussed here. - Kautilya3 (talk) 18:37, 13 April 2020 (UTC)
 * This is an interesting (but confusing) source I also stumbled across. I never saw that someones defined "case fatality rate" is this way. Their method is stupid because you can never have a rate with case fatility. The term "rate" is exclusively used as a synonym for "ratio". The source tries to "fix" this by coming up with a (in my opinion rather stupid) quantity that is truly a rate. Normally, you have "persons died" in the numerator and "persons diagnosed" in the denominator. So what ECDC is claiming is that you assume a rate for the numerator, but not for the denominator. So you have "100 persons diagnosed" coming out of nowhere (no time dimension) and then you have something like "40 death per 2 week". Only when you set this into comparison, you will get a quantity that is truly a rate. But how should this work. The denominator (persons diagnosed) always also depends on a time frame. (persons diagnosed in so and so many weeks etc.) This is how almost all text books define CFR. If you have a rate both in the numerator and denominator, of course the result is not a rate anymore! Has anyone ever seen the definition from ECDC in practice? I'm interested because in the German Wikipedia, we have the same dispute. :-) --TheRandomIP (talk) 09:55, 1 May 2020 (UTC)


 * I intended to point out two things: 1) the source was cited incorrectly 2) we are discussing "case fatality" here, but call it "CFR" with "R" being expanded as "Rate", which in principle is incorrect, because this is not a rate, strictly speaking. I can see that this part has been edited and problem 1) seems to have been fixed and 2) is at least addressed so I'm happy with the current version, thank you :-). BartoszKP (talk) 17:54, 23 May 2020 (UTC)

CFR definition
A slide shown in the video shows

‘CFR = number of people who died of COVD19 divided by the total number of people who got infected’

but the definition of CFR as given on the main Wki page is:

‘the proportion of deaths from a certain disease compared to the total number of people diagnosed with the disease for a certain period of time'.

The latter definition is genrally agreed to be the standard. — Preceding unsigned comment added by 92.14.172.204 (talk) 11:37, 30 April 2020 (UTC)

The slide show is actually discussing Infectious Fatality Rate as defined in the article. The slide show should be removed from this page, as people who believe it can arrive at - and in some cases publish - erroneous conclusions. Jimosul (talk) 14:29, 24 May 2020 (UTC)


 * I second that. The video uses inaccurate terminology which contradicts this article, and it certainly should not be prominently featured at the top of the article. This is currently an important topic, as Donald Trump has been quoting case fatality rate figures frequently, conflating them with mortality rate. I came here to understand case fatality rate better after watching Trump dissimulate, and this video doesn't help. Fuzzypeg★ 23:53, 22 July 2020 (UTC)

Symptomatic?
An IP (89.206.112.10) wishes to insist that CFR measures deaths as a proportion of symptomatic cases. They defend their position by claiming that the non-symptomatic cases are measured by the IFR. This is to misunderstand the definitions. The difference between those two is that the IFR takes its denominator as the total number of people infected, whether or not there is any confirmation of that. See this discussion for example. In the present pandemic, the number of people tested has always been a small fraction of the total infected population, with the possible exception, latterly, of places like Wuhan.

A person can have the disease without showing symptoms. This is normally confirmed by the presence of antibodies which have frequently been found in such cases although in mild cases these often soon fade away. The IFR for Covid-19 is estimated between 0.1 and 1% whereas the CFR is often as high as 7%. Chris55 (talk) 09:02, 5 September 2020 (UTC)

"case fatality ratio" vs. "case-fatality ratio"
Should it be "case fatality ratio" or "case-fatality ratio"?

In the title and the body, respectively. They are out of sync.

--Mortense (talk) 17:42, 27 April 2021 (UTC)