Maps of disease rates are often used to identify regions with elevated disease rates. The goal of this study was to develop and evaluate a regional measure of spatial autocorrelation for localization of these clusters. A regional spatial autocorrelation coefficient (RSAC) was defined and a theoretical mean and standard deviation was derived for its probability distribution. The RSAC was used to identify spatial units that belong to disease clusters. The sensitivity and specificity of the RSAC method in detecting simulated disease clusters was evaluated. For comparison the simulated data were also used to evaluate methods employed by the National Cancer Institute (NCI) for mapping cancer mortality in the United States. The distribution of pancreatic cancer among the elderly white male population in the United States was also evaluated. Within a simulated disease cluster with a relative risk of 2, the RSAC method detected between 75 per cent and 91 per cent of the units depending on the size of the spatial unit used for the analysis. The corresponding sensitivities of the NCI method ranged from 9 per cent to 68 per cent. The RSAC map of pancreatic cancer demonstrated an area of positive clustering (clustering of high rates) in the south central United States. The RSAC method localized disease clusters with greater sensitivity than the NCI method, particularly when geographic units were small. The RSAC method is an effective tool for the identification of regional disease clusters.