2.1 Spatial aspects of causality

The purpose of this object is to explore specifically spatial aspects of the investigation of non-communicable disease causality. This is particularly relevant to the use of GIS for analysis of health. Much theoretical work on disease causation gives little attention to the role of space and time yet spatially referenced data can play a significant role in the investigation of disease causality (Cossman et al. 2003). Further, it is important to recognise that understanding of spatial pattern in disease incidence may have implications for the planning of health service provision even when the precise causal mechanisms are poorly understood. At its simplest level, the ability to detect similarities in the spatial patterning of disease and potential causal factors can draw attention to factors that need further investigation. This situation most commonly applies where both the observed outcome and potential cause are distributed extensively in geographical space, and our aim is to investigate the nature of the relationship. In some cases where the potential cause is spatially distinct (a pollution source, power transmission line or geologically distinct region) the research task is rather different and methods for the detection of spatial clustering are required.

 
The starting point for most causal investigations using GIS is exploratory spatial analysis. At this stage no specific hypothesis is being tested, but it is useful to construct indicators of association or clustering within the data, and to experiment with different visualisations. This analysis, combined with previously published evidence, may lead to the identification of potential causal relationships which are then investigated by statistical modelling.

The construction of causal models is complicated by the fact that observed associations between variables do not necessarily imply causal connections between them. Variables such as smoking and employment in manufacturing industry may both be associated with chronic bronchitis, but there are different causal models that may be appropriate. While it is possible that they both have direct impacts on bronchitis causation, they may interact with one another in such a way that employment in manufacturing industry has no direct effect on bronchitis, but does lead to increased smoking due to the social pressures of the workplace environment. A third alternative is that both factors have an effect but that they also interact. Investigating the relationships between variables in these types of situation is a challenging task and there is a role both for experimental research in a clinical setting and analytical studies using aggregate data. In the latter context, understanding spatial associations has an important role to play.

---

---