- To develop state-of-the-art methods and software based on Bayesian statistics for the analysis of small area variations in health;
- To use the methods and software developed to quantify the extent to which socio-economic, geographic and other factors may explain small area variations in health in different European settings;
- To quantify the impact of errors in data recording and coding on estimates of disease occurrence and small area variations in health needs, and to develop methods which account for these during analysis
Planning the delivery of health care takes many forms in the different member states of the European Union. However, a common requirement is for adequate, timely and accurate information on the health needs of populations to facilitate health service provision in an efficient, effective and equitable manner. By necessity, estimation of such information must be done at local level, to match the delivery of health care. Unfortunately there are formidable technical barriers to overcome in the assembly and correct analysis of such 'small area' data. This Concerted Action aims to make these analyses feasible for people working in the public health field by developing guidelines for the assembly of health data and assessment of data quality, and by developing and disseminating methods and software for their analysis. In particular, the project will involve an assessment of the completeness and quality of various small area datasets available to each of the participating partners, and the implications for analysis and interpretation. The project will also address the technical difficulties associated with geographic inconsistencies in the data (e.g. due to measurements being available at different spatial scales), spatial dependence between observations in nearby areas, socio-economic confounding, ecological bias and measurement error. Bayesian hierarchical models will be used to provide the methodological framework for this purpose. Computer intensive Markov chain Monte Carlo simulation algorithms will be used to carry out the necessary statistical calculations, and will be implemented in the BUGS software for Bayesian modelling (currently under ongoing development by the coordinating partner). A specialist interface between BUGS and market standard relational database programs and Geographical Information Systems will be developed to facilitate automatic data extraction and graphical display (mapping) of results for small area studies. The methods and software will then be applied to a series of substantive studies in an attempt to quantify and explain the sources of small area variations in health and their implications for informing health policy and resource allocation.
Bayesian hierarchical models, BUGS software, ecological bias, ecological regression, health needs, small area methods, socio-economic confounding.