Cluster detection of diseases in heterogeneous populations: an alternative to scan methods

Abstract

Cluster detection has become an important part of the agenda of epidemiologists and public health authorities, the identification of high- and low-risk areas is fundamental in the definition of public health strategies and in the suggestion of potential risks factors. Currently, there are different cluster detection techniques available, the most popular being those using windows to scan the areas within the studied region. However, when these areas are heterogeneous in populations' sizes, scan window methods can lead to inaccurate conclusions. In order to perform cluster detection over heterogeneously populated areas, we developed a method not based on scanning windows but instead on standard mortality ratios (SMR) using irregular spatial aggregation (ISA). Its extension, i.e. irregular spatial aggregation with covariates (ISAC), includes covariates with residuals from Poisson regression. We compared the performance of the method with the flexible shaped spatial scan statistic (FlexScan) using mortality data for stomach and bladder cancer for 8,098 Spanish towns. The results show a collection of clusters for stomach and bladder cancer similar to that detected by ISA and FlexScan. However, in general, clusters detected by FlexScan were bigger and include towns with SMR, which were not statistically significant. For bladder cancer, clusters detected by ISAC differed from those detected by ISA and FlexScan in shape and location. The ISA and ISAC methods could be an alternative to the traditional scan window methods for cluster detection over aggregated data when the areas under study are heterogeneous in terms of population. The simplicity and flexibility of the methods make them more attractive than methods based on more complicated algorithms.