G-methods in practice: an example from occupational epidemiology

Healthy worker survivor bias arises in occupational studies due to time-varying confounding by causal intermediates, differential susceptibility to exposure, left truncation, and right truncation. Although recognized for centuries, this bias was poorly defined and controlled until recently, when advances in causal diagrams and g-methods clarified the problem and provided adequate statistical methods to address it. This symposium aims to 1) illustrate the utility of g-methods to epidemiologists who have limited knowledge of these approaches, and 2) describe some recent developments in their application. After a brief introduction, Dr. Jonathan Chevrier will discuss the components of healthy worker survivor bias and suggest methods for bias correction under these causal structures. Next, Dr. Alexander Keil will describe his recent methodological research accounting for competing risks using the parametric g-formula in an occupational cohort. Dr. Sally Picciotto will then discuss the benefits and challenges of using structural accelerated failure time models to inform public policy, including guidelines for application, implementation, and interpretation. Finally, Dr. Ashley Naimi will summarize and lead a discussion.