Interaction on an additive scale.

摘要

Many clinical epidemiological studies investigate whether a certain exposure, or risk factor, is associated with the incidence of disease or mortality. It may be of interest to study whether this association is different in different types of patients, or to study joint effects. To investigate whether the effect of one risk factor differs across the strata of another risk factor, the presence of interaction among two risk factors can be examined. In statistics, interaction refers to the inclusion of a product term of two risk factors in a statistical model. Statistical interaction thereby evaluates whether the association deviates from either additivity or multiplicativity, depending on the scale of the model. From a public health perspective, the assessment of interaction on an additive scale may be most relevant. For a transparent presentation of interaction effects, it is recommended to report the separate effect of each exposure as well as the joint effect compared to the unexposed group as a joint reference category to permit evaluation of interaction on both an additive and multiplicative scale.