Classification methods for denominators in small areas

Conventional methods fail to provide adequate assignments of individuals into appropriate subgroups for small area analyses, especially when studies cover only a portion of a county or use ecologic, descriptive variables. An unconventional aid to epidemiological investigations, such as a geographic information system (GIS), can aid public health investigators in determining denominators for studies of disease where overlays of environmental exposure and public health data define the geographic extent of the population being studied or where data must be analysed using a map. The basic functions of a GIS--to store, retrieve, transform and display data which have real earth-based co-ordinates--can facilitate this type of ecologic approach. In this paper, each denominator is chosen using the aid of a GIS to allow overlays of environmental elements with residential areas. A variety of methodological issues are associated with such studies, specifically the issue of thin cells (numerically). One solution is to avoid the use of denominators. Another involves careful classification rules for choosing the 'exposed' group. Each of these methods have relevance for environmental equity evaluations associated with environmental hazards.