Analysis of Data from Successive Complex Sample Surveys, with an Example of Hypertension Prevalence from the United States Health Examination Survey

Abstract

There are a number of ongoing large complex sample surveys focused on health data being undertaken at present. Because of time and cost constraints, these surveys are usually highly stratified multistage cluster samples. The complexity of this design has made the usual simple random sampling assumptions untenable and therefore analyses are rarely more than descriptive in nature with some pairwise comparison using t-tests. Generalized regression procedures using weighted least squares are now available for solving this difficulty by incorporating the sample design into the data analysis through the variances and covariances of the sample estimates. These procedures are extended to the analysis of data from complex surveys conducted at two or more different points in time (successive surveys). The method is illustrated with an example taken from published data of hypertension prevalence estimates from the Health Examination Survey-Cycle I (1960–1962) and the Health and Nutrition Examination Survey (1971–1975).