Partitioning linear trends in age-adjusted rates

Objective: Surveillance of chronic diseases includes monitoring trends in age-adjusted rates in the general population. Statistics that are calculated to describe and compare trends include the annual percent change and the percent change for a specified time period. However, it is also of interest to determine the contribution specific diseases make to an overall trend in order to better understand the impact of interventions and changes in the prevalence of risk factors. The objective here is to provide a method for partitioning a linear trend in age-adjusted rates into disease-specific components. Methods: The method presented is based on linear regression. The decreasing trend in age-adjusted cancer mortality rates for the total United States during the period 1991-96 is analyzed to illustrate the method. Results: Trends in mortality for cancers of the colon/rectum, breast, lung/bronchus, and prostate are found to be responsible for 75% of the decreasing trend in cancer mortality. Conclusions: It is possible to partition an overall trend in age-adjusted rates under the assumption that it and the trends for all mutually exclusive and exhaustive subgroups of interest are linear.