Synthesizing Waterborne Infection Prevalence for Comparative Analysis of Cluster Detection Methods

Abstract

When water is an important direct or indirect facilitator in the transmission of disease it is reasonable to expect that clusters of disease may occur near or along these water sources. As such, searching for water-related disease clusters can be an important part of spatial analysis process, particularly when there may be unknown spatial heterogeneities in the relationship between proximity to water and illness. We illustrate the value of using a new class of disease cluster detection methods in the spatial analysis of diseases suspected to emerge in unusual and irregular spatial patterns. Our experiment uses synthetic Schistosoma mansoni prevalence data created from information on environmental factors known to influence risk of infection. Our simulations suggest that cluster detection methods that assume circular cluster shapes are less precise in the delineation of cluster areas, even when the difference between cluster and non-cluster areas is large. We conclude that methods able to find irregularly shaped disease clusters are particularly well suited to applications in which features of the physical environment are suspected to influence risk of illness or infection.