FINDING TEMPORAL CLUSTERS OF EVENTS IN EPIDEMIOLOGY – COMPARISON OF RESULTS FROM POINT EXPANSION METHOD AND SCANNING STATISTICS

Finding temporal clusters in a sequence of events is often used for epidemiological surveillance. In a previous publication, the authors proposed a new method for searching clusters, the advantage of which is the ability to search for several clusters simultaneously while controlling the probability of false detections. In this paper, this method, which has received the conventional name of the point expansion method, is compared with the most common method in the literature, scanning statistics with a variable window, designed to find a single, most pronounced cluster. It is shown that the power of the new method is usually close to the power of scanning statistics. The problems of monitoring rare events are discussed using the results obtained within the EUROCAT project on registration of cases of congenital anomalies. The use of the Benjamini- Hochberg method here to control the false discovery rate greatly simplifies the task of epidemiological surveillance.