To answer this question, one could naively divide the number of deaths with the total number of cases for a given country. For instance, one could say that the mortality rate in France is 0.018 (4/212). I have seen this logic online but there are several issues: Should we, for instance, expect ZERO people to die in Germany (0/196)? Moreover, for countries with only a few cases, a single death could change our estimates dramatically. For instance, Denmark with 10 reported cases could jump from a rate of zero to one of 0.1 by a single observation.
A much better approach would be to take all the observations across different countries into account so that we learn from all global observations while still allowing countries to differ from each other. Aiming to do just that, I made a logistic generalized linear mixed model with the publicly available data as of 2020-03-04.
We see that the model is generally in line with WHO’s estimate of 0.02 but that there is some estimated variance across countries. The USA immediately pops out with an estimate of 0.058. Note that this estimate is lower than the naively estimated number of 0.076 (9/118). On the other end, we have Germany with an estimated rate of 0.005 compared to the naively estimated 0 (0/196). We can also see big differences in the two countries with the most reported cases. China with a mortality rate of 0.037 and S. Korea with a rate of 0.007.
The analysis is a snapshot that poses more questions than it answers. Why do we see for instance see a ~ 5 times higher mortality in China than S. Korea and will the estimate for the USA stay high or is it simply from 118 cases that just happened to be very unfortunate?
This model is obviously only as good as the data it has been fed with and as data is still collected, there are plenty of potential pitfalls: We cannot tell how many people who are currently infected could pass away from the virus which could bring the mortality rate up. Conversely, there could be many undetected (mild) cases which would bring the true mortality rate down. Moreover, we can suspect that there are differences in how different countries detect and report cases.
It is important to remember that the mortality rate is nothing like a law of nature and that it can change depending on how well a country responds to it. Hopefully, all the countries will do their utmost and keep getting better as research and experiences are shared.
Take care out there. I recommend not to trust random strangers found online, such as myself, completely and to always verify from official sources such as
https://www.who.int/emergencies/diseases/novel-coronavirus-2019