BACKGROUND: This paper is about constructing small areas for the analysis of health data with the aims of health service delivery in mind. The areal framework should enable the analyst to link health data and census data and the areas should have large enough populations to ensure that rates are reliable and be homogeneous with respect to important socio-economic attributes. METHODS: An information-based statistic is used for the construction of regions in Sheffield based on the Townsend deprivation index. Enumeration districts are used as the geographical building blocks for the regions. The new regional framework is used for computing Bayes adjusted standardized incidence rates for colorectal cancer (CRC) across Sheffield. The paper then examines the statistical relationship between CRC incidence and deprivation across the set of regions using bivariate regression. RESULTS: The method yields regions that are considerably more homogeneous in terms of deprivation than wards, and using this framework it is shown that there is a (weak) statistical association at the regional scale between deprivation and CRC. CONCLUSION: We conclude that statistical tools can be employed to provide regions that meet the criteria for small area analysis of health data and the analyst does not have to be tied to large administrative units such as wards. There are some benefits to executing this work within a Geographic Informative System. The method should be of interest to those concerned with health service delivery and the identification of 'problem regions'.
BACKGROUND: This paper is about constructing small areas for the analysis of health data with the aims of health service delivery in mind. The areal framework should enable the analyst to link health data and census data and the areas should have large enough populations to ensure that rates are reliable and be homogeneous with respect to important socio-economic attributes. METHODS: An information-based statistic is used for the construction of regions in Sheffield based on the Townsend deprivation index. Enumeration districts are used as the geographical building blocks for the regions. The new regional framework is used for computing Bayes adjusted standardized incidence rates for colorectal cancer (CRC) across Sheffield. The paper then examines the statistical relationship between CRC incidence and deprivation across the set of regions using bivariate regression. RESULTS: The method yields regions that are considerably more homogeneous in terms of deprivation than wards, and using this framework it is shown that there is a (weak) statistical association at the regional scale between deprivation and CRC. CONCLUSION: We conclude that statistical tools can be employed to provide regions that meet the criteria for small area analysis of health data and the analyst does not have to be tied to large administrative units such as wards. There are some benefits to executing this work within a Geographic Informative System. The method should be of interest to those concerned with health service delivery and the identification of 'problem regions'.