查看文章

The potential outcomes framework has gained traction and is now, arguably, the prevailing methodological perspective in epidemiology. While it has not yet fully permeated the applied literature, the current generation of epidemiology students is trained in this approach and it dominates the methodological literature. This perspective requires a reframing of epidemiologic questions and the retooling of many epidemiologic methods. The traditional hallmarks of study validity—no confounding, selection bias or information bias—are being supplanted by exchangeability, positivity and well-defined interventions (Hernán, 2005; Hernán and Taubman, 2008). Here we examine how the requirement to study well-defined interventions impacts the kinds of questions asked by epidemiologists.

To meet the “well-defined intervention” requirement, exposures must be factors that can be conceptualized as treatments in a …