Evaluation of the EPI survey methodology for estimating relative risk

World Health Stat Q. 1991;44(3):107-14.


Precision in estimation of relative risks using a standardized sampling method proposed by the WHO Global Programme on AIDS was evaluated using a Monte Carlo model simulating actual populations; the proposed survey design represents a modification of the methodology used by the WHO Expanded Programme on Immunization (EPI) to estimate immunization coverage among children. This study suggests that in actual populations the proposed survey strategy is a reasonable alternative to the use of simple random sampling (SRS) at the second stage of cluster sampling. Although varying such population characteristics as the seroprevalence rate, nonresponse rate, and rate of misclassification of exposure failed to demonstrate a clear advantage of one method over the other, the added cost and difficulty of implementing SRS under field conditions warrant further consideration of the EPI-like methodology for use in estimating relative risks.

PIP: Logistical and managerial constraints have prompted classical probability-proportional-to-size (PPS) cluster sampling to be modified for application to surveys of immunization coverage. This World Health Organization Expanded Program on Immunization (EPI) methodology aims to estimate population rates accurate to within 10% of the true level. Concerned with the accuracy and reliability of the methodology, researchers have employed a Monte Carlo computer simulation model to evaluate the approach's precision in estimating relative risks. Mimicking characteristics of typical African populations, the Monte Carlo model suggests the proposed survey strategy to be a viable alternative to simple random sampling (SRS) at the 2nd stage of cluster sampling. Varying seroprevalence rate, nonresponse rate, and rate of misclassification of exposure failed to prove 1 method advantageous over the other. Given the cost and difficulty of classical sampling techniques, researchers should consider the advantages of using the EPI methodology.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Cluster Analysis
  • Computer Simulation
  • Health Surveys
  • Humans
  • Immunization*
  • Monte Carlo Method
  • Population
  • Risk Factors
  • Risk*
  • Sampling Studies