Beyond observational data: understanding anaesthesia research better with causal diagrams

Research that assesses causal relations based on observational data remains challenging because of the well known tension between natural causal thinking and classic statistical association methods. This tension, over a period of decades, has generated the development of a statistical framework, with specific methods and reasoning, to allow the drawing of causal inference. A part of this framework is the directed acyclic graph (DAG), a specific type of causal diagram that is based on graph theory. This intuitive representation of the mechanistic processes of a specific problem, and the logical consequences that come with it, closes the gap between observed associations and causality. The advantage of integrating DAGs into observational research has been emphasised in several reviews. This review is an anaesthetist-friendly (re-)introduction to this graphical reasoning. We provide some examples of how this framework can be used beyond observational data and studies. We explain how important aspects of randomised controlled trials like covariate adjustments, handling of missing data and protocol violations can be both understood and taught by drawing and interpreting a DAG. We consider the case of the titration paradox and show how combining knowledge of how a specific dataset was built, with pharmacology, into a more advanced DAG, can help solve and understand these seemingly paradoxical findings. In all of this, we use anaesthesia-oriented examples to illustrate how DAGs can be a valuable scientific language that helps us to understand, organise and communicate study results and research questions.