Instrumental variables in studies of health and health inequalities

One way to approach causality in health research is the instrumental variable approach, which is used to avoid confounding and measurement error in observational studies (as opposed to experiments). Beginning with Angrist and colleagues’ studies of mortality differentials using the U.S. draft lotteries during the Vietnam War, instrumental variables have been widely applied in the health literature. Examples of instrumental variables range from birth month, which was used as early as the 1930s to study effects of early life on health outcomes later in life, to famines, wars, German reunification, or - more recently - genes. This chapter introduces the idea of instrumental variables and reviews their definition. It provides an overview of the study questions and instruments used in health research. It discusses what makes a good instrument, the advantages and disadvantages of the approach, and outlines how this approach can be combined with other study designs used to examine causality.