Estimating Cumulative Failure Risk Under Hypothetical Interventions on Time-Varying Treatments in Complex Observational Studies

• Activity Number: 416
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jessica G. Young*+
• Companies: Harvard School of Public Health
• Keywords: g-estimation ; causal inference ; survival analysis ; g-formula ; marginal structural models ; structural nested models

Abstract

Given untestable assumptions, cumulative failure risk under hypothetical interventions on a time-varying treatment may be identified in observational studies with complex time-varying confounding using Robins' g-formula (1986). Various approaches have been proposed for estimating this complex function. These include g-computation, inverse probability weighting of marginal structural models and g-estimation of structural nested cumulative failure time models. In this talk, we review assumptions under which these three estimators will converge to the same desired intervention risk and discuss relative advantages and disadvantages of each method in practical settings. The material presented represents an overview of ongoing work in our group, involving many collaborators, to compare methods for estimating long-term disease risk under hypothetical interventions on time-varying lifestyle factors in the Nurses' Health Study.