Strategies for analysing ecological health data: models of the biological risk of individuals.

Frequently, the analysis of environmental health hazards using ecological data does not involve explicit recognition of the difficulties in translating health effects expressed in the aggregate to the health risks of individuals. We discuss these difficulties and suggest the need for the appropriate conceptualization of risk mechanisms at the individual level and of the population processes that determine the form in which these risk mechanisms are expressed in aggregate data. To illustrate the implications of these concepts we develop a biologically motivated model of lung cancer risk and apply it to both national and county data. In addition, to measure the total health effects of the long term elevation or depression of lung cancer incidence rates, we calculate prevalence distributions from the time series analysis of incidence patterns in county data.