Commentary: Fine-tuning Life-espectancy Calculations Using Markov Processes

In constructing their decision model, Bruinvels et al. assume that the table contains yearly mortality probabilities rather than mortality rates. It is legitimate to use these age-specific probabilities of death in a decision model if the cycle length is one year. However, if the cycle length is shorter than one year, the probabilities must be transformed by converting them to rates and using the rates to calculate the probability for a shorter length of time. Probabilities cannot be transformed directly. For example, the probability of dying in one month is not equal to 1/12 of the probability of dying in one year. Mortality tables published by the National Center for Health Statistics (NCHS),1 on which age-specific mortality rates are based, provide the number of survivors at a given age among 100,000 people born alive. The probability of death from age n to age n + 1 is: