A framework for investigating geographical variation in diseases, based on a study of Legionnaires' disease

Demonstration of geographical variations in disease can yield powerful insight into the disease pathway, particularly for environmentally acquired conditions, but only if the many problems of data interpretation can be solved. This paper presents the framework, methods and principles guiding a study of the geographical epidemiology of Legionnaires' Disease in Scotland. A case-list was constructed and disease incidence rates were calculated by geographical area; these showed variation. Five categories of explanation for the variation were identified: short-term fluctuations of incidence in time masquerading as differences by place; artefact; and differences in host-susceptibility, agent virulence, or environment. The methods used to study these explanations, excepting agent virulence, are described, with an emphasis on the use of previously existing data to test hypotheses. Examples include the use of mortality, census and hospital morbidity data to assess the artefact and host-susceptibility explanations; and the use of ratios of serology tests to disease to examine the differential testing hypothesis. The reasoning and process by which the environmental focus of the study was narrowed and the technique for relating the geographical pattern of disease to the putative source are outlined. This framework allows the researcher to plan for the parallel collection of the data necessary both to demonstrate geographical variation and to point to the likely explanation.