Effect Measure/Modification
Definition: Effect modification is present when an effect measure such as sex, age or geographic location is different at several levels in an exposure-disease relationship. In other words, sometimes the exposure-disease relationship is different within different levels of another variable. For example, a specific disease may be associated with an exposure in males but not with the same exposure in females. This relationship is also known as statistical interaction.
Effect measure modification is something that one should be aware of and explain in the data. If there is an effect measure modifier, separate measures of effect for each level of the variable should be presented (i.e., present stratified results – e.g., show separate associations between exposure and disease for males and for females).
Example: When all age groups are considered together, females have two to three times the risk for hip fracture as males (Buhr and Cooke, 1959). However, this overall ratio disguises the fact that at young ages males have a higher risk of hip fracture than females, whereas females at older ages have considerably higher risk than males. The association between sex and hip fracture is modified by age, which in this instance is a surrogate measure of the high prevalence of osteoporosis in older women and of the propensity for severe trauma in young males. (Methods in Observational Epidemiology, Kelsey, et al second edition)
