Analysis of the components of a linear trend in proportions

The standard epidemiologic methods for evaluating trends in the prevalence or incidence of disease are reviewed, and a method is presented for assessing the contribution of a specific disease subgroup to the overall trend, based on the slope obtained by a grouped weighted linear regression of the proportions of persons experiencing the outcome of interest in each exposure level. The slopes for specific disease subgroups contribute to the overall slope in an additive manner, and measures based on the individual slopes can be used to assess the relative strengths of the trends for specific disease subgroups as well as their relative contributions to the overall trend. This approach is illustrated with data on social class patterns of mortality in New Zealand males aged 15-64 years during 1974-1978; it is shown that the strongest mortality gradients were for mental disorders, infectious diseases, respiratory diseases, and accidents. The latter two categories together accounted for approximately two thirds of the overall social class mortality gradient.