Performances of some common confidence interval construction procedures for direct standardized incidence or mortality rate were compared. The asymptotic method, the method based on matching moments, the method based on gamma distribution and the method based on beta distribution were investigated through Monte Carlo simulations based on data from an epidemiologic study of cardiovascular disease and randomly generated scenarios. These interval estimation procedures, as well as some mixed procedures developed for the present study, were evaluated in terms of their coverage probability and expected length.