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

Lucas Böttcher; Mingtao Xia; Tom Chou

An unambiguous, physiologically-based definition of mortality ratio is the probability that a single, newly infected individual will eventually die of the disease. If there are sufficient individual-level or cohort data, these probabilities can be further stratified according to patient age, gender, health condition, etc. [10] . The mortality ratio or probability of death should be an intrinsic property of the virus and the infected individual, depending on age, health, access to health care, etc. This intrinsic probability ought not to be directly dependent on the population-level dynamics of infected and recovered individuals. Thus, it can in principle be framed by a model for the survival probability of a single infected individual. Whether this individual infects others does not directly affect his probability of eventually dying. In section II A, we derive a model describing the probability M 1 (t) that an infected individual dies or recovers before time t. Importantly, these models incorporate the duration of infection (including an incubation period) before a patient tests positive at time t = 0.