PEER REVIEWED: Computations of Confidence Intervals for Estimates in the United States National Hospital Discharge Survey, 1979–2000

Introduction The National Hospital Discharge Survey is a primary data source for epidemiology research in the United States. To ensure that estimates are reliable, confidence intervals need to be calculated. The original survey data source is not available to the public, and the usual statistical methods are unsuitable for calculating confidence intervals. Instead, calculating confidence intervals requires using the statistical methods and relative standard errors that the U.S. National Center for Health Statistics has provided. However, the relative standard error parameters differ by hospital, patient category, and group. They also change yearly with sampling and are expressed differently before and during or after 1988. Consequently, manual computations of confidence intervals with multiple groups, diseases, and years are inefficient and prone to error. We developed a SAS program to compute confidence intervals for National Hospital Discharge Survey data from 1979 through 2000, newborns excluded.