Variants of the Attributable Risk in a Multifactorial Situation: Theory and Computational Realization

Statistical methods for quantifying the population impact of exposure factors by means of estimating some variant of the attributable risk parameter for a binary disease variable based on stratification methods or multivariable models have been extensively developed during the last decade. In epidemiologic studies, these risk parameters enjoy growing popularityas they allow a practice-oriented interpretation for public health purposes. Despite this demand by epidemiologic researchers, none of the leading statistical software packages including SASTM has incorporated any attributable risk parameter in the catalogue of measures of association to be automatically calculated in some procedure. In the paper, an introduction to the theory and application of attributable risk estimation from epidemiologic data is provided. Then, a general method and its corresponding practical realization are described how to calculate different attributable risk estimators using the SAS software. Data from the German GRIPS-study, a cohort study on cardiovascular risk factors for myocardial infarction, are used to illustrate the methodologic concept and its computational realization.