Spatial contouring of risk: a tool for environmental epidemiology

The possible elevation of disease rates in the proximity of site-specific environmental hazards is much investigated. Single-site studies are subject to problems of reporting bias and statistical power, and multisite studies to heterogeneity of exposure. Both types of studies usually use concentric circular regions centered on a site as a surrogate for defining the exposed and unexposed populations. This approach does not take into account the actual spatial pattern of toxicant dispersion or the spatial pattern associated with the population, and so much useful information is wasted. We report a kernel density technique to map risk contours for disease, which is not influenced by the coordinates of any putative environmental hazard and which could be married to actual spatial exposure patterns.