SaTScan is a freely available software that uses the scan statistic to detect clusters in space, time or space-time. SaTScan uses Monte Carlo hypothesis testing in order to produce a p-value for the null hypothesis that no clusters are present. Monte Carlo hypothesis testing can be a powerful tool when asymptotic theoretical distributions are inconvenient or impossible to discover; the main drawback to this approach is that precision for small p-values can only be obtained through greatly increasing the number of Monte Carlo replications, which is both computer-intensive and time consuming. Depending on the type of analysis being done, the number of geographical areas included, the amount of historical data, and the number of Monte Carlo replications, SaTScan can take anywhere from seconds to hours to run. In doing daily surveillance of many syndromes, we need to limit the amount of time it takes to generate each p-value while still retaining enough precision in the p-value to determine how unusual a cluster is. Since the type of analysis done and the geographic regions being used cannot be changed in most cases, we focus here on trying to reduce the number of Monte Carlo replicates needed.
Objective
Our goal was to increase the precision of the p-value produced from SaTScan while reducing the amount of CPU time needed by decreasing the number of Monte Carlo replicates.