Construction Specification of Directed Acyclic Graphs: An Introduction to the ESC-DAGs Method and Case Interpretation in Cerebrovascular Disease Research

Abstract

Abstract: Directed acyclic graph (DAG) is a visualization tool to construct causal hypotheses in causal inference, which can be used to control confounding bias in observational studies by screening the minimal adjustment subsets that need to be adjusted through the backdoor criterion. However, existing studies lack standardization in drawing DAGs and the quality of DAGs used in the literature varies, which causes application dilemmas. The emergence of the ESC-DAGs method provides a standardized process for constructing DAGs. Based on the basic idea of systematic review, ESC-DAGs extensively searches relevant literature and divides the construction process into three phases: mapping, translation, and integration. Firstly, a saturated DAG is constructed based on each retrieved study. The directed edges are screened in the graph according to the Hill criterion and the potential outcome framework, and finally, all the retained directed edges are integrated into one DAG. The ESC-DAGs method makes full use of the literature information and the expertise of researchers. It provides a reasonable and feasible criterion for applying DAGs, which has important guiding significance. This article presents the content of the ESC-DAGs method and elucidates the specific process of constructing a DAG using an example from a clinical study on stroke.

Keywords

• directed acyclic graph
• confounding bias
• causal inference
• ischemic stroke