Investigating heterogeneity of effects and associations using interaction terms

Abstract

Effect heterogeneity, the variability of an association or exposure across subgroups, usually warrants further investigation. The aim of this deeper analysis is to identify effect modifiers (or moderators) and quantify their relationship with the exposure. We explain why it is better to harness interaction effects within a single analytic model than to use separate models to analyze each subgroup. Using examples, we demonstrate a practical approach to modeling and interpretation with interaction terms from various measurement scales (categorical by categorical; categorical by continuous; and continuous by continuous).