Why estimating population-based case fatality rates during epidemics may be misleading

Lucas Böttcher; Mingtao Xia; Tom Chou

Besides the mathematical differences between M 1 (t) and M 0 p (t), CFR, estimating M 0 p (t) and CFR(t) from aggregate populations implicitly incorporate a number of confounding factors that lead to variability in these estimates. In Fig. 4 , we plot the population-level mortality-ratio estimates M 0 p against the CFR for different regions and observe large variations and very little correlation between countries [25] . As of March 26, 2020, the value of M 0 p in Italy is almost 45% and can increase further if the current conditions (e.g., treatment methods, age group proportion of infecteds, etc.) do not change. Differences between the mortality ratios in China and Italy (see Figs. 1(b) and (c) ) might be a result of varying medical treatment strategies, different practices in data collecting (e.g., post-mortem testing), and differences in the age demographics between the countries.