B1105
Title: A two-stage-least-square approach for negative control of unmeasured confounding with time-to-event outcomes
Authors: Kendrick Li - University of Michigan (United States) [presenting]
Eric Tchetgen Tchetgen - The Wharton School, University of Pennsylvania (United States)
Abstract: Unmeasured confounding is a universal concern in causal inference. The emerging approach of the double negative control method provides an opportunity to reduce and correct unmeasured confounding bias, leveraging negative control exposure (NCE) and outcome (NCO) variables as proxies to the suspected unmeasured confounders. However, in analysing right-censored time-to-event outcomes, the existing approach does not provide a readily interpretable summary measure of the exposure effect. A simple two-stage-least-square method is described for negative control inference for an additive hazard model with right-censored time-to-event outcomes. Theoretical justification is provided for the proposed approach with different types of NCOs, including continuous, count, and time-to-event data. It is shown in simulation studies that the proposed approach can successfully correct for unmeasured confounding bias. The method is demonstrated to evaluate the effectiveness of right heart catheterization among critically ill patients using the SUPPORT study data.
