The impact of data quality on spatial analysis of cancer registry data: The example of missing stage at diagnosis and late-stage colorectal cancer

Most disease surveillance systems currently geocode case data. This, coupled with advances in geographic analysis tools, has led to a rise in epidemiologic studies on distribution of disease that rely on analysis of secondary data, e.g. from cancer registries. However, while the data and tools are available for performing geospatial analyses, there are challenges with which methodologies to apply, how to interpret and translate results, and how results are impacted by data quality. The issue of data quality is the subject of this paper. Mapping cancer rates highlights spatial patterns that can help elucidate environmental, clinical, or social causality pathways that drive differences in disease burden by geographic locations. Locating areas with high rates of cancer incidence or variations by stage at diagnoses can help prioritize cancer control efforts. Once the geographic patterns of cancer are mapped, the ideal action is to follow with effective public health interventions for the high risk communities. However, before using results of spatial research to inform public health response, it is important to consider whether the results are spurious due to methodological issues, such as data quality. Missing or incorrect data can distort research conclusions and result in ineffective public health policy. Using colorectal cancer (CRC) as an example, the impact of missing stage at diagnosis on late-stage at diagnosis cluster detection is evaluated. The impact on cluster detection, areabased modeling, and distance from services analysis is described.