Constructing causal pathways for premature cardiovascular disease mortality using directed acyclic graphs with integrating evidence synthesis and expert knowledge

Cardiovascular disease (CVD) is a major global cause of premature mortality. While multiple studies propose CVD mortality prediction models based on regression frameworks, incorporating causal understanding through causal inference approaches can enhance accuracy. This paper demonstrates a methodology combining evidence synthesis and expert knowledge to construct a causal model for premature CVD mortality using Directed Acyclic Graphs (DAGs). The process involves three phases: (1) initial DAG development based on the Evidence Synthesis for Constructing Directed Acyclic Graphs (ESC-DAGs) framework, (2) validation and consensus-building with 12 experts using the Fuzzy Delphi method (FDM), and (3) application to data analysis using population-based survey data linked with death records. Expert input refined the initial DAG model, achieving consensus on 45 causal paths. The revised model guided selection of confounding variables for adjustment. For example, to estimate the total effect of diabetes on premature CVD mortality, the suggested adjustment set included age, dietary pattern, genetic/family history, sex hormones, and physical activity. Testing different DAG models showed agreement between expert ratings and data accuracy from regression models. This systematic approach contributes to DAG methodology, offering a transparent process for constructing causal pathways for premature CVD mortality.