CHDS Faculty Recommend Directed Acyclic Graphs – Center for Health Decision Science

CHDS Faculty Ankur Pandya, Uwe Siebert, and Myriam Hunink joined colleagues in recommending directed acyclic graphs (DAGs) in Decision-Analytic Modeling in a recent Medical Decision Making commentary.

The more complicated a decision-analytic model, the more likely it is for the causal relationships between variables to become unclear and therefore, the more likely that bias unconsciously affects the model design. The authors argue that using DAGs will help visualize the assumptions behind variable selection, leading to greater clarity, transparency, and model accuracy. They demonstrate how DAGs can show not only the effect of the model structure on the outcome but also the potential biases in the parameter choices. To further demonstrate how DAGs might aid in decision-analytic model design, they discuss their use in two scenarios: statin use for cardiovascular disease and mindfulness-based interventions for students’ stress. They also discuss the challenges to wider adoption of DAGs and propose considerations for future debate.

Learn more: Read the full commentary, Directed Acyclic Graphs in Decision-Analytic Modeling: Bridging Causal Inference and Effective Model Design in Medical Decision Making
Learn more: Explore the CHDS approaches to Models and Tools and Cost-Effectiveness Analysis