Bayesian methods for prevalence estimation from incomplete administrative lists.

Many studies aimed at estimating prevalence use several administrative lists from different sources in an attempt to enumerate all persons affected with the health condition of interest. Each list is 'incomplete' in the sense that none of them enumerates all persons affected with the health condition. Further, because the lists are drawn from different administrative sources the probability of enumeration varies from list to list. The goal is to use information from the lists to estimate the total number of affected persons in the population, but with some accounting for the different but unknown probabilities of enumeration on each list. This paper presents a Bayesian method to estimate prevalence when the probability of enumeration varies from list to list. Data from a survey of children with spina bifida illustrate the methodology.