A comparison of two methods for calculating expected mortality.

A number of statistics based on person-years of survival have use in mortality studies. However, calculation of an expected number of deaths based on person-years of survival and standard death rates cannot be justified. We derive a new method to estimate expected mortality based on a model which assumes that each individual contributes to expected mortality from the time he enters the study until the study ends, regardless of whether he dies during the study. This model shows that what is referred to as expected mortality in the person-years method is a biased estimate of expected mortality. The variance of observed mortality under this new model is substantially less than the variance commonly used with the person-years method.