Associate Professor Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill University of North Carolina at Chapel Hill Chapel Hill, United States
1. Course Aim This course will introduce students to multiple approaches to addressing unmeasured confounders through study design and analysis, and the requirements and assumptions of each.
2. Requisites Statement This is an intermediate level course for researchers using healthcare data for pharmacoepidemiology. There are no precourse materials to be reviewed.
3. Course Objectives · Identify common unmeasured covariates in pharmacoepidemiology and the types of bias that can result. · Describe multiple approaches to addressing unmeasured covariates in pharmacoepidemiology through study design and analysis. · Understand the assumptions of each approach, ways of evaluating these assumptions, and when they are likely to be violated.
4. Syllabus Outline Introductions: Course faculty lead will provide an overview of the course objectives and introduce the presenting Faculty. Section 1: Use of study design to control unmeasured confounding: During the first set of lectures, Drs Lund, Raman, and Gokhale will introduce three ways in which study design can be used to limit bias due to unmeasured confounding: a) the active comparator new user (ACNU) design, b) self-controlled designs, and c) instrumental variable designs. Faculty will provide foundational knowledge about each approach illustrated through real-world examples. · Module 1: Active comparator / new user study design with Dr. Jennifer Lund · Module 2: Self-controlled study designs with Dr. Sudha Raman · Module 3: Instrumental variable study designs with Dr. Mugdha Gokhale Section 2: Analytic approaches to unmeasured confounding: During the second set of modules, Drs Lund, Stürmer, and Young will introduce three analytic approaches to reduce bias due to unmeasured confounding: a) proxy measures, b) external control for confounding, and c) quantitative bias analysis. Faculty will provide foundational knowledge about each approach illustrated through real-world examples. · Module 4: Proxy measures with Dr. Jennifer Lund · Module 5: External control for confounding with Dr. Til Stürmer · Module 6: Quantitative bias analysis with Dr. Jessica Young Section 3: Choosing among approaches: A case study Dr Michele Jonsson Funk will present a case study in which we pose a hypothetical research question and model thinking about the likely unmeasured covariates, considering how each of the approaches might be used in that setting, and weighing the pros/cons of each. This case-study will provide integration across the approaches and give students an opportunity to consider the selection of an appropriate strategy for addressing important unmeasured covariates in light of real-world constraints. Format: Live, in person lectures followed immediately by a Q&A with the course faculty.