Small area health statistics in Europe*

Abstract

The recognition that health may be adversely affected by environmental pollution in the vicinity of industrial plants, together with the advent of new industrial and chemical processes with unknown or poorly recognised health effects, suggest the need to investigate and monitor the health of populations living in quite small and localised areas. Traditionally, such investigations were costly and time-consuming, since health data relevant to a particular location had to be assembled and analysed on an ad hoc basis. However, recent developments in computing, statistical methodology and the use of geographically encoded data sets have enabled the initial investigation of disease in small areas to be largely automated and based on routinely available statistics. Although routine statistics have been used to display and analyse variations in disease patterns over relatively large areas (for example, counties in England and Wales [1] or municipalities in Finland [2]), their use in describing disease in small areas has been limited. This is the result partly of the lack (in most countries) of both a precise geographical location for events (e.g. deaths) and appropriate small area population (denominator) data; but it reflects also the need for appropriate statistical methods to interpret unusual clusters of cases in the context of natural variations in disease across small populations. For health monitoring in small areas to be successful, a number of require­ ments should be satisfied: (1) There should be good quality data on health events to include mortality and some morbidity data (e.g. a national cancer registry). It should be appreciated that whereas very large (national) data sets are required to generate statistics for any small area, and to compare results with other areas, the analysis in anyone area is based on a small subset of the whole data; hence good quality data are essential if spurious results (artefacts) are to be avoided.