Marginal Structural Models for Causal Inference Using Observational Healthcare Data: Best Practices and Case Studies


Discussion Leader: Michael Grabner, PhD, Scientific Affairs, Carelon Research, Wilmington, DE, USA
Discussants: Lauren Zalla, PhD, Department of Epidemiology, Johns Hopkins University, Baltimore, MD, USA; Shivani Pandya, BPharm, MS, HEOR/Pharmacy Economics, Carelon Research, Wilmington, DE, USA; Douglas E. Faries, PhD, Global Statistical Sciences, Eli Lilly and Company, Indianapolis, IN, USA

Presentation Documents

PURPOSE: This workshop provides step-by-step guidance on the application of Marginal Structural Models (MSMs) for causal inference using longitudinal real-world healthcare data with time-varying treatments and confounders, illustrated with a case study example. Participants will share their own experiences and take part in a knowledge assessment.

DESCRIPTION: Causal inference in observational research is growing more important, driven by the need for generalizable and rapidly delivered real-world evidence (RWE) to inform regulatory, payer, and patient/provider decision-making. MSMs are a versatile class of causal models frequently (though not exclusively) used for estimating the effects of time-varying treatments in the presence of time-varying confounders and treatment-confounder feedback. Existing literature on this method is rich but can be highly technical and daunting to navigate. Furthermore, the optimal approach to implementing such models can vary by context, and a variety of methodologic questions remain open.

Workshop participants with intermediate-level subject knowledge will be familiarized with key considerations when designing and interpreting MSMs. Dr Grabner will initiate the workshop highlighting the role of MSMs in the context of causal study designs and RWE, and solicit feedback from participants on their experiences with these concepts (~10 minutes). Afterwards, Dr Zalla will discuss different types of MSMs, their applications, and the steps involved in planning and implementation (~20 minutes). Next, Ms Pandya will present a case study example where MSMs were used to estimate the effects of medication adherence among patients with chronic obstructive pulmonary disease. Pitfalls and practical solutions of this administrative claims-based application will be explained (~15 minutes). Dr Faries will conclude by summarizing MSM best practices and ongoing research, and also conduct the second part of the knowledge assessment (~10 minutes).

Besides real-time voting, there will be time for audience questions. In addition, a “skill sheet” for MSMs will be available to all participants.




Methodological & Statistical Research