A Not Quite as Quick but Much Cleaner Alternative to the Expanded Programme on Immunization (EPI) Cluster Survey Design

Abstract

Although the Expanded Programme on Immunization (EPI) cluster survey methodology has been successfully used for assessing levels of immunization programme coverage in developing country settings, certain features of the methodology, as it is usually carried out, make it a less-than-optimal choice for large, national surveys and/or surveys with multiple measurement objectives. What is needed is a ‘middle ground’ between rigorous cluster sampling methods, which are seen as unfeasible for routine use in many developing country settings, and the EPI cluster survey approach. This article suggests some fairly straightforward modifications to the basic EPI cluster survey design that put it on a solid probability footing and render it easily adaptable to differing and/or multiple measurement objectives, without incurring prohibitive costs or adding appreciably to the complexity of survey operations. The proposed modifications concern primarily the manner in which households are chosen at the second stage of sample selection. Because the modified sampling strategy maintains the scientific rigor of conventional cluster sampling methods while retaining many of the desirable features of the EPI survey methodology, the methodology is likely to be a preferred ‘middle ground’ survey design, relevant for many applications, particularly surveys designed to monitor multiple health indicators over time. The fieldwork burden in the modified design is only marginally higher than in EPI cluster surveys, and considerably lower than in conventional cluster surveys.