Thanks to these GIS tools, we can finally make a statistical comparison of maps describing environmental, health and socio-demographic indicators.
An ecological geographic study is a descriptive epidemiological analysis that describes the spatial variability of the incidence of a disease for a given territory and period, and seeks to identify the influence of risk factors on the appearance of cases within groups of individuals (Wakefield, 2008). These retrospective studies are used to formulate etiological hypotheses, and are carried out prior to analytical epidemiological studies (i.e. cohort or case-control studies carried out on an individual level).
During a geographic ecological study, an incidence rate is calculated for each spatial unit in the study area. The study aims to (i) highlight spatial disparities in incidence (incidence mapping: Wakefield & Salway, 2001), (ii) detect statistically atypical areas, called clusters or blackspots, through the use of spatial statistics tools (scan statistics: Kulldorff, 1997), and (iii) establish geographic correlations between this incidence and environmental (proximity to sources of pollution, environmental surveillance and biomonitoring) and socio-economic (indicators of deprivation) factors.
Due to the aggregation of population data, the causality expressed by ecological studies is less robust than that expressed in a study carried out on an individual scale in a cohort or case study. Indeed, ecological studies primarily seek to highlight a health effect at group level in order to develop hypotheses on the potential risk factors for the occurrence of this effect (Richardson, 2000). The causal nature of this effect cannot be clearly defined and extrapolated to the individual due to ecological biases. These biases have been widely described in the literature (Piantadosi et al., 1988; Greenland & Morgenstern, 1989; Morgenstern, 1995; King, 1997; Wakefield, 2007; 2008).
Specification bias is linked to the variability of risk factor exposure levels or that of confounding factors between individuals in the same group (Greenland, 1992).
Aggregation bias indicates that the individuals showing evidence of a health effect are not necessarily those who were exposed to the risk factor studied.
Finally, confounding factors indicate the possible existence of another unmeasured risk factor that may influence the observed effect (Wakefield, 2008). This risk factor can be linked to issues such as daily migrations (commuting to work) or lifestyle habits (food).
A reliable interpretation of the results is therefore dependent upon the greatest possible homogeneity of the population groups compared in terms of confounding factors. An ecological study remains the least costly option in terms of time and financial investment, and makes it possible to formulate hypotheses in order to optimize within the population.
Environmental and social health inequalities are shown by a concentration within a territory of populations that have health problems, are vulnerable in socio-economic terms and live near nuisance or in a contaminated environment. The final objective of our work therefore consists of identifying these territorial blackspots in order to shape public health policies for the prevention of health risks.
Greenland S. 1992. Divergent biases in ecologic and individual level studies. Stat. Med, 11: 1209–23.
Greenland S & Morgenstern H. 1989. Ecological bias, confounding and effect modification. Int. J. Epidemiol, 18: 269–74.
King G. 1997. A Solution to the Ecological Inference Problem. Princeton, NJ: Princeton Univ. Press
Kulldorff M. 1997. A spatial scan statistic. Communications in statistics: theory and methods, 26 (6): 1481–1496.
Morgenstern H. 1995. Ecologic studies in epidemiology: concepts, principles, and methods. Annu. Rev. Public Health, 16: 61–81.
Piantadosi S, Byar DP, Green SB. 1988. The ecological fallacy. Am. J. Epidemiol, 127: 893–904.
Richardson S. 2000. Problèmes méthodologiques dans les études écologiques santé–environnement. Life Sciences, 323: 611–616.
Wakefield JC & Salway R. 2001. A statistical framework for ecological and aggregate studies. Journal of the Royal Statistical Society, series A (164): 119-37.
Wakefield JC. 2007. Disease mapping and spatial regression with count data. Biostatistics, 8: 158–83.
Wakefield JC. 2008. Ecologic studies revisited. Annual Review of Public Health, 29: 75–90.