Methods for locating spatial clusters of diseases are typically variations of the circular scan statistic method. They restrict the number of potential clusters by considering all circular, rectangular, or elliptical regions, and then apply a likelihood ratio test to evaluate the statistical significance of each potential cluster. Because disease outbreaks may have variable shapes, there has been recent interest in developing methods to detect irregularly-shaped clusters. Starting with a neighborhood graph of the administrative regions in the study area, certain sub-graphs are evaluated. These include all connected subgraphs within a circular window and sub-graphs of the minimum spanning tree of a weighted neighborhood graph formed by deleting one edge. These methods restrict the maximum cluster size or identify large clusters having greater likelihood ratios than true clusters in the data, suggesting a limitation of using the likelihood ratio to detect arbitrarily-shaped clusters.
Objective
A method for detecting spatial clusters of diseases of any shape based on the Euclidean minimum spanning tree is described and compared to the circular scan statistic.