Detection of excess disease near an exposure point: a case study

Many studies have evaluated the likelihood of adverse health effects associated with environmental contamination from point source exposures. Two statistical measures used in these studies are the ratio, R, of the observed to the expected number of cases occurring in the area containing the point and the average distance, D, between the cases and the point. This paper estimates the probability of detecting an association between disease and exposure when one actually exists (power) for R and D for several specific and plausible statistical models. Results are presented in the context of rare diseases such as congenital malformations. The practical implications of using these measures for the evaluation of risk of disease in environmental epidemiologic investigations are discussed.