From an epidemiological and public health perspective there is an interest in quantifying differences in incidence and mortality between either time points, geographical areas or males and females. We propose a method for splitting such a difference in the number of cases/deaths into three components: (1) those due to risk; (2) those due to population structure (i.e. age distribution); and (3) those due to population size. We also propose graphical methods for presenting the results. Three examples are used to illustrate our methodology.

No abstract