Calculating expected deaths: a comparison of two methods.

A comparison is presented between the traditional "person-years" and more recently described "prospective model" methods for calculating mortality expectations. Problems arising from the fact that expectations under the person-years method are calculated on the basis that a null hypothesis is true, which results in artificial figures that, at least theoretically, are meaningless if the hypothesis is rejected, are discussed. Data are presented from two studies in which expectations have been calculated both ways, showing important differences between the two methods with an exaggeration of the expectation when the study group has an above normal mortality experience. It is suggested that the person-years methodology should be replaced by that of the prospective model.