A Generalised Scan Statistic Test for the Detection of Clusters

Abstract

The scan statistic can be used for the testing of clusters of disease in time. However, it is not valid in situations in which there is some change in the population at risk or in the detection rate of the disorder. Furthermore, it cannot distinguish between a cluster due to the changing pattern of known risk factors for the disease and one due to other causes. A generalised scan statistic is proposed which overcomes these limitations. Its use is illustrated with previously published birth defects data.