Geographical variations in mortality : an exploratory approach
This thesis aims to provide a geographical contribution to the understanding of disease causation, primarily through the development of causal models of chronic disease mortality incorporating both the physical and social environments. The overwhelming impression of previous research in this field is one of conflicting findings. For example, studies examining the relationship between disease and water hardness have found positive relationships, negative relationships and no relationship whatsoever. It is contended that this failure to replicate results is a direct consequence of applying an unsuitable 'confirmatory' approach to the quantitative analysis of geographical data. It is argued also that it is necessary to adopt a more appropriate statistical methodology, that of 'exploratory' statistics, before progress can be made. After an exegesis of the exploratory approach, the commonly used technique of multiple regression is given an exploratory interpretation. Each of the assumptions of this technique is discussed, and attention focuses on the effects of breaking the assumptions and on methods of detecting and overcoming the resultant problems. This exposition is illustrated by the re-analysis of previous studies, and it is demonstrated that inappropriate methods have led some researchers to inferential error. Finally in this methodological part of the research, an examination of the analysis of ratios is undertaken; here too it is suggested that the inappropriate analysis of death rates has resulted in some researchers making incorrect inferences. The empirical aspects of the thesis centre on the analysis of mortality variations in England and Wales. A critical appraisal of previous studies of the relationship between disease and water hardness is undertaken, and it is concluded that quantitative techniques have been poorly applied. Exploratory data analysis is then employed to develop models accounting for geographical variations in mortality experienced by the County Boroughs of England and Wales. In contrast to previous studies that have analysed these variations, no strong relationship is found between disease and water hardness. Moreover, an examination of the mortality experiences of Boroughs whose water supply has changed substantially over time also results in the conclusion that the effects of water hardness have been overestimated. Finally, the study examines the difficult problem of drawing inferences from aggregate data. Although it is concluded that much work remains to be undertaken, it is again argued that the exploratory approach may allow progress to towards the solution of this problem and, consequently, some guidelines for further research are outlined.