PURPOSE: To discuss how to utilize the potential of directed acyclic graphs to estimate causal effects to inform health technology assessment (HTA), and how to detect and address challenges.

DESCRIPTION: Directed Acyclic Graphs (DAGs) are a great tool to define and visualize research questions and to understand potential confounding and selection bias. They can be used to help design analyses of observational (or “real world”) data, and of randomized studies when post-randomization confounding is suspected. DAGs may be used to help define appropriate analysis methods and data that are necessary for drawing causal conclusions, and may help identify the direction of potential biases. We will illustrate (1) the potential of DAGS for assisting in the analysis of both randomized and observational studies, (2) discuss the challenges typically encountered in practice, and (3) consider how DAGs might be used by HTA decision makers. We will promote potential practical solutions, taking into account the perspectives of academic methodologists as well as the needs of decision-makers. The workshop will consist of three short presentations, interactive tests and polls, and a concluding discussion. The first speaker (Felicitas Kuehne) will present key concepts covered in DAGs in 10 minutes (conditional independencies, confounders, colliders, mediators), followed by a short interactive test (5 minutes). Application of DAGs in both randomized studies and observational (real world) data will be illustrated by the second speaker (Michael Schomaker) by means of realistic examples, both for clear and challenging contexts (15 minutes). Next, a case study will be presented, including an interactive element to ensure participation of the audience (5-10min). Then, the third speaker (Nicholas Latimer) will offer thoughts on how DAGs might be used within the HTA decision-making process (10 minutes), giving the perspective of a NICE Appraisal Committee member. The workshop will conclude with a moderated discussion involving the audience.