Maps are generic manner of representation of the results of geographical health analyses. Specific problems are associated with the analysis and interpretation of the regional distribution of health related data in administratively defined areas. The interference from area specific quantities in studies carried out in large areal units tend to be biased i.e. the rate of disease seems to be the same over the whole area and thus information has been lost. On the other hand, the size of population deviates greatly between small areas (for example municipalities) and therefore regional variation in areas with small population tends to present random variation of the rates resulting in extreme incidence rates that dominate the map. There are also complex dependencies between many of the measured variables and response variables, which can be either spatial of temporal or both. Some of the quantities of interest cannot be interpreted as meaningful in the frequentist framework since they are not "repetitions" as required in traditional statistical methods.
Bayesian data analysis together with GIS-techniques provides a flexible framework for assessing spatial data and provides powerful tools for making competent inferences. A Bayesian hierarchical spatial model has been constructed to describe the spatial incidence of non-communicable diseases. This model exploits aggregated pixel-wise location for both the cases and population at risk enabling to use dense grids instead of the conventional areas defined by administrative boundaries. The Bayesian hierarchical model offers an opportunity to study also certain environmental factors using them as covariate variables in the model.
Type 1 diabetes and heart diseases with complex and partly unknown aetiology and a long period of latency have marked geographical variation of incidence in Finland. These spatial differences are independent from changes of incidence and seem to remain rather stable. Our current and future aims are to investigate the association between the spatial distributions of non-communicable diseases and their potential environmental risk factors in order to generate hypotheses of spatial and temporal associations between these diseases and environmental risk.
Leader of the study group:
Academy Researcher Marjatta Karvonen
Publication list of Marjatta Karvonen.
Contact person Marjatta Karvonen