Several epidemiological parameters have been introduced for quantifying the population impact of a certain exposure on morbidity on a population level, termed 'attributable risk' (AR). Of these definitions, the AR as suggested by Levin in 1953 or some algebraic transformations of it are most commonly used. A structured literature search, based on the Medline database, identified 334 original epidemiological studies dealing with this AR published between 1966 and 1996 (mostly case-control studies). A considerable increase in the number of published studies incorporating some quantification of the exposure impact on the population level in terms of the AR was observed in the last decade. However, in 64.5% of these studies no exact definition of the AR used was given. Adjustment procedures necessary in multifactorial situations were performed by only 37.5%, confidence intervals for the AR were given by only 19.3% of authors. Thus, although the increasing popularity of this important epidemiological measure is encouraging, its correct application and comprehensive reporting in medical practice should be promoted further.