Power and sample size considerations for detecting deviations from secular trends in surveillance surveys

Abstract

SUMMARY On-going public health surveillance is essential to the detection and monitoring of epidemics. We present statistical methods for determining the sample size that is required to detect unacceptable deviations from existing secular trends in prevalence with specified power. Between 1958 and 1969, a large study was conducted for providing surveillance of the prevalence of infection for a well-known disease. Data from this study indicated that the prevalence of infection in the mid-1960s increased significantly over previous secular trends among important demographic groups tested. These data are used to illustrate the statistical methods that we propose for detecting departures from existing temporal trends, estimating the year in which the changepoint occurred and specifying sample sizes for on-going active surveillance studies.