Exploratory pattern-analysis study of a multifocal measles epidemic

As a preliminary to forecasting and to the design of epidemiological control measures, the features of weekly case occurrence data in 232 towns during a multifocal epidemic of measles in the state of Rio Grande do Sul (Brazil) have been investigated using a signal-processing approach. Data from 153 outbreaks in the period 1971-79 were examined. Evidence was found of a consistent microstructure in the form of a common, recurrent transient pattern of case occurrences in all the outbreaks. This microstructure of the outbreaks was investigated with the aid of an average profile of the 'microepidemics', applied in an inverse filter approach to identify their occurrences in an objective way. The average profile was formed by coherent averaging of many examples aligned at their maxima. However, it was necessary to remove the unavoidable background of associated microepidemic occurrences: estimated by convoluting the observed renewal density of occurrences with the coherent average, iterating the process as needed. The inverse filter, designed with a knowledge of this average profile, was used to refine the process of locating microepidemic occurrences and to study their time interval and magnitude sequences. It is shown that the inferred microstructure offers a good basis for modelling the different epidemic outbreaks, and that it suggests hypotheses, forecasting procedures and possible control strategies in relation to future outbreaks.