Methods to Overcome Violations of an Instrumental Variable Assumption: Converting a Confounder into an Instrument

• Activity Number: 425
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Michelle Shardell*
• Companies: National Institute on Aging
• Keywords: causal inference ; generalized method of moments ; instrumental variables ; structural mean models ; two-stage least squares

Abstract

Instrumental variable (IV) methods have potential to consistently estimate the causal effect of an exposure on an outcome in the presence of unmeasured confounding. However, validity of IV methods relies on strong assumptions, some of which cannot be verified. One such assumption is that the effect of the proposed instrument on the outcome is completely mediated by the exposure. We consider the case where this assumption is violated, but a weaker assumption holds where the effect of the proposed instrument on the outcome is completely mediated by measured variables, including the exposure. That is, the proposed instrument is actually a confounder. We review some conventional IV methods and propose easy-to-use adaptations to use when the IV assumption is violated, but the weaker assumption holds. The proposed methods involve analytically `converting' the confounder into an IV, then applying conventional IV methods. Potential applications of the proposed methods to epidemiology include studies where the exposure and outcome exhibit seasonal variation and studies using Mendelian randomization with genetic variants that affect multiple phenotypes that may affect the outcome.