Clustering in Aggregated Health Data

Spatial information plays an important role in the identification of sources of outbreaks for many different health-related conditions. In the public health domain, as in many other domains, the available data is often aggregated into geographical regions, such as zip codes or municipalities. In this paper we study the problem of finding clusters in spatially aggregated data. Given a subdivision of the plane into regions with two values per region, a case count and a population count, we look for a cluster with maximum density. We model the problem as finding a placement of a given shape R such that the ratio of cases contained in R to people living in R is maximized. We propose two models that differ on how to determine the cases in R, together with several variants and extensions, and give algorithms that solve the problems efficiently.