Most observational studies measure data at individual level. In an ecological study, the unit of observation is at population or community (ecological) level. The disease or outcome and the exposure of interest are measured in a number of populations and their relationship is examined.

A common use of ecological studies is to look at geographical correlations. Figure 8a illustrates an example where coronary heart disease mortality correlates with infant mortality rates 60 years previously. This finding led to the hypothesis that impaired foetal development may lead to coronary heart disease and its risk factors in later life. Figure 8b shows the relationship between gross domestic product (GDP) and life expectancy at country level. Ecological studies may be used more where the key outcome of interest, such as equity of access, or an explanatory variable, such as GDP, can be measured better at an area or community level. Although ecological studies are useful for generating hypotheses, findings should be confirmed in more rigorous studies on individuals.

As in any other study, ecological studies are subject to confounding and bias. If the group level data come from very different populations – for example, in terms of age, risk factor prevalence or ethnicity – then comparisons between them are difficult, as information on the confounders may be confined to the average population level.

Information bias