Instrumental Variable Estimation of Marginal Structural Mean Models for Time-Varying Treatment

Abstract

Robins introduced Marginal Structural Models (MSMs), a general class of counterfactual models for the joint effects of time-varying treatment regimes in complex longitudinal studies subject to time-varying confounding. In his work, identification of MSM parameters is established under a Sequential Randomization Assumption (SRA), which rules out unmeasured confounding of treatment assignment over time. We consider sufficient conditions for identification of the parameters of a subclass, Marginal Structural Mean Models (MSMMs), when sequential randomization fails to hold due to unmeasured confounding, using instead a time-varying instrumental variable. Our identification conditions require that no unobserved confounder predicts compliance type for the time-varying treatment. We describe a simple weighted estimator and examine its finite-sample properties in a simulation study. We apply the proposed estimator to examine the effect of delivery hospital type on neonatal survival probability. Supplementary materials for this article are available online.

KEYWORDS:

• Causal inference
• Delivery hospital
• Marginal structural model
• Time-varying endogeneity
• Unmeasured confounding

Supplementary Materials

The supplementary materials include proofs, additional examples, and details on the simulation and data analysis.