This site provides information on community-based two-stage cluster surveys that features first stage selection with probability proportionate to size (PPS) sampling, and second stage selection with simple random sampling (SRS) or probability sampling with quotas.
Such surveys (see figure) typically identify clusters (geographic, social or census groups) and provide a summary tally of people or households in each. Then using a cumulative list of all clusters, at the first stage 30 or more are selected with PPS sampling. At the second stage, a constant number of people or HHs is selected in each of the 30 or more clusters, either by SRS, or by random start HH and next nearest neighbor HHs.
The site, created and maintained by Dr. Ralph R. Frerichs, serves as a compendium for his:
• rapid survey workshops (if interested in sponsoring, contact Dr. Frerichs)
As such, it provides reference material and software for planning, conducting and analyzing rapid surveys.
HISTORY OF RAPID SURVEYS
Seven publications comprise the history of rapid survey methodology.
Serfling RE, Sherman lL, Attribute Sampling Methods for Local Health Departments, Publication No. 1230, U.S. Department of Health and Human Services, Public Health Service, Washington, DC, 1965, 178 pp.
The notion of rapid surveys as a means to help local health departments learn of the immunization status of 1-4 year old children was first defined and explained in a publication by Serfling and Sherman in 1965.
Several years later, Serfling and Sherman's methodology was modified by Henderson and colleagues to provide community-based information for guiding the Smallpox Eradication Program in West Africa.
Later the Expanded Program of Immunization (EPI) of the World Health Organization took over the method. Their strategy was to sample 30 clusters with probability proportionate to size and select 7 children within each cluster, yielding a sample size of 210 children. The first of the 7 children was selected by a random process and the following 6 were drawn from the next nearest neighboring households. Such small surveys, soon done throughout the world, became know as "30 x 7 " surveys. The rationale for the sampling procedure, however was not fully explained.
Lemeshow S, and Robinson D: Surveys to measure programme coverage and impact: a review of the methodology used by the Expanded Programme on Immunization. World Health Statistics Quarterly 38, 65-75, 1985.
In 1985, Lemeshow and Robinson were commissioned by the World Health Organization to statistically describe the EPI 30 x 7 survey methodology. This was the first published article by a statistician that justified the technique which later would become incorporated in rapid survey methodology (RSM).
While working in Burma (now Myanmar) with a consortium of faculty from UCLA, UC Berkeley and the University of Hawaii on a large primary health care project, Frerichs sensed the need for a more timely measure of community health problems. He took the EPI sampling methodology, incorporated a more general variance formula based on ratio estimators rather than persons, added the use of portable computers, and developed rapid survey methodology (RSM), an approach allowing community-based surveys to be planned, carried out and reported in less than a month.
While the intention of the EPI program was met with the "30 x 7" surveys, others began adopting the survey method for problems of a more general nature. Such surveys violated the narrow statistical assumptions of the EPI program as described by Lemeshow in 1985 and thus were inappropriate uses of the method. In 1991 Bennett and colleagues published a seminal article on two-stage cluster surveys which expanded on the 1989 publications by Frerichs, and formally summarized the statistical basis for rapid surveys with ratio estimators.
While used extensively over the years in developing countries, rapid surveys were infrequent in the developed world. In 1991 Frerichs and Shaheen published an updated article on rapid surveys for use in the United States. Their revised method featured community volunteers, incorporation of Census data, and with the help of community enumerators, random sampling of second stage participants rather than random start and next nearest neighbor as used by EPI. Such sampling resulted in smaller variance estimates (and as a result, narrower confidence intervals) than the classical EPI method.
WHO IMMUNIZATION COVERAGE (EPI) SURVEYS
The Expanded Programme on Immunization (EPI) of the World Health Organization has been the major proponent of simple two-stage cluster surveys (i.e., 30 x 7) for immunization coverage. The survey method was developed by the EPI program, and has been used worldwide.