The impact of diagnostic hierarchies on prevalence estimates for psychiatric disorders

This article presents an algebraic treatment of the impact of diagnostic hierarchies on the estimation of prevalence rates of psychiatric disorders. A method for correcting for this "hierarchy effect" is developed and illustrated. Using the terminology of Boyd et al., when the dominant disorder is common and/or the odds ratio for the dominant and the excluded disorders is high, the observed prevalence of an excluded disorder can substantially underestimate its true prevalence. This "hierarchy effect" can be particularly important in genetic-epidemiologic investigations which compare the prevalence of an excluded disorder in two populations which differ in the prevalence of the dominant disorder. The impact of certain kinds of diagnostic hierarchies can be easily understood and corrected for; with others, particularly those based on etiologic assumptions, a straight-forward interpretation is not always possible.