Abstract

The investigation of disease risks in small areas is complicated by many issues including data quality, the retrospective nature of the statistical testing, the problems of boundary definitions in time and space around a putative disease cluster, and the lack of generally accepted definitions of the key terminology. Routine data systems have revolutionised the initial investigation of disease risks near sources of environmental pollution, although problems of data analysis and interpretation remain. This is especially true of unmeasured socioeconomic confounding, which could generate apparent positive results near a pollution source.