Estimating the case fatality rate (CFR) is a high priority in response to COVID-19 pandemic. This article details issues with interpreting the CFR during the middle of outbreak and cautions against drawing causal inferences from observational data.

To date, there appear to be wide variations in the CFR of COVID-19 among countries. The current CFR in Korea is 1.3% (120 deaths out of 9,037 confirmed cases), as of March 25, 2020,2 while the CFRs are 9.5% in Italy, 7.9% in Iran, 6.6% in Spain, and 4.0% in China.

The authors caution against making comparisons between countries for several reasons:

1. Current CFRs place ‘reported’ number of deaths in the numerator and ‘reported’ number of confirmed cases in the denominator in each country. They may not reflect the true number of deaths (numerator) and COVID-19 cases (denominator) occurred in each country.
2. There are differences in testing capabilities and public health response policy for detection of COVID-19 cases between countries, so it is imperative to check the completeness of death reports following COVID-19 before making a comparison between the countries.
3. CFRs are also dependent on each country’s surveillance performance to detect cases of COVID-19. Most of asymptomatic or presymptomatic cases would not be tested for COVID-19 in many places, resulting in higher CFR by smaller denominator.
4. Differences in age structure of affected population between countries may guide to the true burden of COVID-19 in terms of disease severity and fatality proportions. Age-specific or age-stratified
   CFRs will be useful in comparing between the countries with different affected age groups.
5. There are time differences between disease onset and death. The denominator and the numerator of CFR should be composed of patients infected at the same time as those who died in order to accurately represent the actual CFR. Cross-sectional comparison of these figures may be biased since the time span for illness is different among countries during the course of the ongoing epidemic. To avoid this bias, time-delay adjusted estimates between symptom onset and death need to be calculated for a fair comparison of CFRs among different countries.