Impact of spatial aggregation error on the spatial scan analysis: a case study of colorectal cancer.

The paper aims to estimate the level and impact of spatial aggregation error for spatial scan statistics where disaggregated data below the zip code level are not available. Data on colorectal cancer cases in Cook county, Illinois, USA with a 5-year interval were used. An innovative procedure using SAS and Java was designed to make SaTScan auto-run. Characteristics of clusters at each reference level were compared to those at zip code level to observe differences related to spatial aggregation. The comparison reveals that spatial scan statistic at the zip code level can generate reliable clusters in areas with a large number of cases, but fail to detect clusters in areas where there are a sparse number of cases, since the spatial aggregation error is minimised in areas with sizeable numbers of cases. Without localised cancer data, zip code level data can be used effectively to identify dominant clusters. However, smaller clusters located in low-density areas may be missed.