Ecological and cross-level studies
Ecological studies differ from other study designs by having groups of individuals as their unit of analysis. Studying differences in risk factors between individuals cannot fully explain variations in the health of populations in different regions or over time, and ignores the fact that a population's health impacts on social functioning and collective economic performance (McMichael and Beaglehole 2000). Similarly, there is increasing interest on the impact of social capital, or those aspects of the social environment that promote cohesion and cooperation, which have been associated with improved health and other outcomes. For example, a recent study found higher rates of common mental disorder among British women living in areas of low social capital (McCulloch 2001). Thus ecological associations can be analysed to gain insight into aetiological mechanisms at the level of individuals (cross-level inference), although there is ongoing debate about whether ecological analyses can add to insights obtained from studies of individual persons. Kasl (1979) stated that ‘ecological analyses lead to results which, in themselves, are opaque, unhelpful, potentially misleading’. Others emphasize that population health is more than the sum of the health of individual population members, and that therefore, ecological studies have a separate role alongside individual-level epidemiological research (Rose 1992). This chapter summarizes the principles and the place of ecological studies in the history of epidemiology, and the distinguishing properties of ecological data. After a description of ecological study designs and their analysis, the chapter argues that ecological data have added value even if individual-level information is available on the associations of interest.