Last update:
24-Apr-2007

Stratification constitutes the mainstay of epidemiologic analysis. Even with studies that ultimately require more complicated analyses, stratification is an important interim tool. It familiarizes the investigator with distributions of key variables and patterns in the data in ways that other approaches cannot provide. Stratification means that data are separated into categories or strata. For example stratification by sex or age means that data are separated into men and women or into categories by age. Several analytic concerns motivate stratification, the most prominent of which is evaluation and control of confounding. A simple definition of confounding would be the confusion, or mixing, of effects. This definition implies that the effect of the determinant under study is mixed together with the effect of a confounder, leading to a systematic bias. When data are stratified by confounder, for example into men and women, each stratum would be free of confounding. Thus, if the correlation between exposure and disease is analyzed separately, for example in men and women, each category of gender would give an estimation of the result of exposure, independent of gender. Stratification is also used to evaluate and describe effect-measure modification. In addition the study of biologic interactions between two factors is most easily accomplished using stratification methods.

Key words: Confounder, Confounding, Determinant, Effect measure modification, Stratification.