Multi-level analysis in epidemiologic research on health behaviors and outcomes

Abstract

Individual-level health behaviors and outcomes have multi-level determinants (individual and environmental). Multi-level analysis seeks to explain individual outcomes in terms of both individual and environmental or aggregate variables. Ecologic fallacy (improper inference about individual-level associations based on associations measured only at the aggregate level) can result from confusion about the level of inference that is of ultimate interest. The perspective of multi-level analysis acknowledges the importance of both individual and environmental variables in determining health behaviors and outcomes at the level of the indivisible unit--the individual. The authors review concepts and methods of multi-level analysis and their application to epidemiologic research on health behavior and health outcomes.