Modeling of under-detection of cases in disease surveillance.

PURPOSE: Accurate epidemiological surveillance of leprosy is a matter of international public health concern. It often suffers, however, from potential problems of under-registration of reported cases, particularly in poorer and more socially deprived areas. Such problems also apply in the surveillance of many other communicable or transmissible diseases. We develop a Bayesian model for small-area disease rates that allows for censoring of case detection in suspect districts and can therefore be used to estimate under-reporting of cases in a given study region.

METHODS: Such methods are applied to leprosy incidence in a municipality of Pernambuco State in North Eastern Brazil, using a social deprivation indicator as the basis for considering data from certain districts to be censored. The time period we consider was immediately prior to an extension of the coverage and efficacy of the control program and model predictions concerning under reporting can therefore be compared with more reliable data subsequently collected from the same region.

RESULTS: The proposed method produces informative estimates of under detection of leprosy cases in the defined study region and these estimates compare well, both in size and in geographical location, with the numbers of cases subsequently detected.

CONCLUSIONS: As illustrated by the application discussed in this article, the proposed model provides a general tool that may be used in spatial epidemiological surveillance situations where the available data is suspected to contain significant under-registrations of cases in certain geographical areas.