Marginal Structural Models for Multi-State Outcomes

• Activity Number: 581
• Type: Topic Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Consulting
• Author(s): Wei Yang*+ and Marshall M. Joffe
• Companies: University of Pennsylvania and University of Pennsylvania
• Keywords: Marginal Structural Models ; Multi-state outcome ; Time-dependent confounding

Abstract

The progression of chronic disease is often described in discrete stages clinically. For example, patients with chronic kidney disease are classified into five stages based on the glomerular filtration rate. Multi-state models are well-established for modeling transition rates between different stages. However, when estimating the effect of time-updated treatments on the transition rates between stages, the model typically gives biased estimates in the presence of time-dependent confounding. Time-dependent confounding occurs when some time-updated covariates confound the association between subsequent treatments and the outcome and are themselves affected by prior treatments. We use inverse probability of treatment weighting to deal with time-dependent confounding. Using the Cox proportional hazards model, the causal hazard ratio for the transition between any two stages can be estimated in the weighted sample without further adjustments for time-updated covariates. We illustrate the method through simulation and data from an ongoing cohort study of patients with chronic kidney diseases.