Regionalizing mortality data: ischaemic heart disease in Norway.

The way regions are delimited has a bearing on the geographical patterns and time trends which emerge from cause specific mortality analysis. Whenever possible, alternative regionalizations should be used to explore the full information potential of the data. For statistical reasons, the size of the regional units (populations-at-risk) should be selected according to the frequency of the cause of death, number of years in the time period, etc. A geographical mortality information system for Norway, based on individual death records and with quick and flexible retrieval options is described. As a demonstration, geographical time trends in ischaemic heart disease from 1970 to 1985 are studied, using different schemes of regionalization. A clear tendency towards regional convergence appears in the rural-urban dimension, but there is no convergence between the five subnational regions of the country. There is no evidence that counties which have received heart disease intervention projects fare any better than those which have not, but here a more thorough analysis is recommended. Within the intervention counties, there are large variations both in mortality levels and trends.