Increasing efficiency in evaluation research: the use of sequential analysis.

This paper describes Wald's sequential analysis and briefly reviews the history of its applied use. Two public health applications are presented as examples of how the method helped overcome two common problems associated with evaluation research. In one case, the sequential technique reduced dramatically the workload in an evaluation project where hospital records were being reviewed. In the second case, prompt feedback of data to a breast screening program was facilitated by the sequential method allowing program administrators to refocus their efforts on problem areas before the end of the funding period. On the average, Wald's sequential method results in a savings of 50 per cent in observations as compared to classical sampling procedures. The sample sizes will always terminate with a finite number and will rarely exceed three times the average sample size for any single sample.