[DATA] Case fatality rate formula is misleading

[Olivyf]

Which Dataset

Case Fatality Rate is currently calculated as
(Total Deaths) / (Total Positive Cases)

Error Description

The faster the infection rate accelerates the lower the calculated fatality rate, because it takes weeks before new cases results in 1. Fatality or 2. Recovered.
Hence fatality rate is understated because it is compared against undetermined cases. (Open cases)
Thus the public is led to believe chance of fatality once infected is lower than the true chances of "fatality vs recovery".

Suggested fixes

True Fatality Rate must be expressed as a percentage of closed cases.
i.e. (Total Deaths) / (Total Deaths + Total Recovered)

Crosscheck: Fatality Rate + Recovery Rate = 100%

[vukosim]

Can you provide any literature to assist in this change @Olivyf?

Currently, the rate used is so that you can compare across countries even with the shortcomings. Making the change, we will then have to list as something else, not as case fatality rate. We can rename to Observed Case Fatality Rate if that makes it more clearer.

See

• John Hopkins https://coronavirus.jhu.edu/data/mortality
• Our world in data https://ourworldindata.org/grapher/coronavirus-cfr
• Lancet

[shaze]

For my 2c it may be worth showing both. Looking at the daily data it seems to me like the recoveries data is the noisiest and without knowing the local processes of reporting this we should not over interpret these rates at a daily level while the epidemic is going on. The formula @Olivyf suggests may be an overestimate because as deaths are likely to be reported faster than recoveries. It would also be useful to show both to address the concern of public perceptions -- rather than reporting one number to show two numbers which show a range may be better in showing the uncertainty in the data.

[vukosim]

The reporting of recoveries is slower and erratic so that's why I am concerned. Thanks for your thoughts @shaze

[shaze]

I think keeping consistency with what other people do is paramount and we shouldn't relabel something that other people have used. But I think there is value in having both. I would suggest as the new one "Reported Resolved Case Fatality Rate". A bit klunky I know. And then explaining in the README