Marginal Structural Modeling in Comparative Effectiveness Research: Illustration in Diabetes Research

• Activity Number: 517
• Type: Topic Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Committee on Statistics and Disability
• Author(s): Romain Neugebauer*+
• Companies: Kaiser Permanente
• Keywords: causal inference ; survival analysis ; dynamic treatment interventions ; inverse probability weighting ; targeted learning ; diabetes

Abstract

A goal of comparative effectiveness research (CER) is often to contrast survival outcomes between exposure groups defined by time-varying interventions. For instance, the progressive nature of type 2 diabetes mellitus (T2DM) results in frequent revisiting of treatment decisions for many patients as glycemic control deteriorates but the effects of intensive pharmacological treatment remain uncertain. For this reason, using the electronic health records from patients of seven sites of the HMO Research Network, a large retrospective cohort study of adults with T2DM was assembled to evaluate the impact of progressively more aggressive glucose-lowering strategies on several clinical outcomes. With this practical CER example, I review the motivation for and principles of marginal structural modeling and discuss alternate estimation approaches based on inverse probability weighting and targeted learning with large healthcare databases.