Instrumental Variable Additive Hazards Models

• Activity Number: 117
• Type: Topic Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Jialiang Li* and Jason Fine and Alan Brookhart
• Companies: Duke University/NUS/SERI and The University of North Carolina and The University of North Carolina at Chapel Hill
• Keywords: Additive hazards model ; instrumental variable ; 2-stage least square ; survival analysis ; confouding variable ; endogeneity

Abstract

Instrumental variable (IV) methods are popular in non-experimental studies to estimate the causal eects of medical interventions. These approaches allow for the consistent estimation of treatment eects even if important confounding factors are unobserved. Despite the increasing use of these methods, there have been few extensions of IV methods to censored data problems. In this paper, we discuss challenges in applying IV techniques to the proportional hazards model and demonstrate the utility of the additive hazards formulation for IV analyses with censored data. Assuming linear structural equation models for the hazard function, we develop a closed-form, two-stage estimator for the causal eect in the additive hazard model. The methods permit both continuous and discrete exposures, and enable the estimation of causal relative survival measures. The asymptotic properties of the estimators are derived and the resulting inferences are shown to perform well in simulation studies and in an application to a data set on the eectiveness of a novel chemotherapeutic agent for colon cancer.