A simulation study comparing methods for calculating confidence intervals for directly standardized rates

This study provides a supplemental report of the performance of the confidence interval for direct standardized rates obtained by the approximate bootstrap method (ABC) method. The ABC method was not considered by the Ng et al. (2008) paper which compared different methods of interval construction. A graphical comparison of the coverage probability and the ratio of the right to left non-coverage probabilities, as a function of the variance in the weights used for each simulation point, are given for the ABC method, as well as three of the recommended procedures by Ng et al. The expected confidence interval lengths are also reported. The ABC intervals are shown to be good competitors compared with the other three confidence intervals.