Significant advances have been made in New Zealand in recent years in the application of geographical information systems (GIS) to the health sector. Initially, the principal focus of this effort was directed towards infectious disease surveillance (www.phew.govt.nz). It is not surprising, given the relatively minor impact of such diseases on population health, that attention has recently turned to examining a wider range of health conditions, including chronic diseases (www.nzpho.govt.nz).

It should be noted, however, that attempts to conduct area-based investigations into patterns of health and disease may be hindered not only by the ecological fallacy,1,2 but also by the exposure fallacy.3 The ecological fallacy, wherein assumptions are made on the basis of aggregate data about the relationship between two or more variables without confirmation of an individual-level link, is acknowledged as a perpetual bugbear of spatial analysis.

Unfortunately, the issue of population mobility and migration, the essence of the exposure fallacy, is seldom addressed. Although some elements of the exposure fallacy are occasionally acknowledged,4,5 such concerns are generally ignored and few authors make any significant attempts to incorporate population mobility into their health/disease models (there are exceptions6,7). The exposure fallacy is the unfounded assumption that population mobility occurs at a low and consistent level across areas, and that the populations involved are uniformly similar. However, population movements occur in a spatially inconsistent manner and this factor has the potential to seriously confound results. There are two elements to this spatial inconsistency. First, there is the nature of the people involved, which can differ markedly (areas may, for example, experience gentrification or residualisation); second, the levels of migration and mobility within areas can also vary dramatically.

The exposure fallacy has been identified as a significant issue in ecological analysis of health status in Ireland.3 However, analysis of national population mobility rates has identified Ireland as having a low rate by international standards. This issue, therefore, may be a much more significant problem for countries such as NZ, Australia and the US, which historically have had relatively high rates of population mobility.8 In an analysis by Long, using data from (circa) 1981, NZ had the highest rate of observed population mobility over a one-year period (19.4% had moved compared with 6.1% in Ireland).9 Results from the 2001 Census indicate that half of the NZ population (aged 5+) had changed residence since 1996 (www.stats.govt.nz). An examination of possible cultural differences may prove useful, as well as an examination of subgroups (eg, what is the profile of Maori who move away and then return later in life to their turangawaewae?).

It should be noted that the exposure fallacy will obviously be less of an impediment to the examination of health issues with relatively short lead-in periods. However, NZ health researchers may still have to explore the development of an area-based mobility index (for use in weighted-least-squares regression analysis), or the introduction of a Finnish-style, centralised population and housing register,10 to help overcome these problems if they intend to continue conducting area-based analysis of health. Alternative, partial solutions may include examining larger areal units (most people do not move far) and excluding young adults from analysis (they move the most).3