Using population attributable risk to understand geographic disease clusters.

This paper describes the use of population attributable risk percent (PAR%) in the study of morbidity and mortality clusters, and in particular, shows how this method of risk characterization can usefully distinguish between multiple geographic clusters of potential interest. Incident lung cancer data in persons 60 years and over from the province of Ontario, Canada, are analyzed for spatial clusters, and each cluster is characterized in terms of statistical significance, relative risk and PAR%. We observe that although relative risk is probably highest in Northern Ontario, highest PAR% is in Eastern Ontario, and in particular, the Ottawa area. These results illustrate the usefulness of attributable risk as a metric to help characterize and understand spatial clusters, which could be important for place-based public health interventions.