Measurement error and information bias in causal diagrams: mapping epidemiological concepts and graphical structures.

This article focuses on the application of directed acyclic graphs (DAGs) to represent measurement error and information bias in epidemiology, areas traditionally underrepresented in causal diagrams compared to confounding and selection bias. It argues for incorporating both conceptual (true underlying constructs) and empirical (measured variables subject to error) components into DAGs to better depict the data-generating processes and clarify the impact of measurement error on causal inference. Using a clinical epidemiology example involving hearing loss and cognitive impairment, the article illustrates how measurement error and differential misclassification can create information bias pathways distinct from confounding, which are visible only when empirical variables and their errors are included in DAGs. The authors conclude that integrating measurement error and information bias into causal diagrams enhances epidemiological understanding and supports improved study design and interpretation.