DAGDraw: A Responsive and Modern Directed Acyclic Graph Drawing Tool for Causal Inference and Transportability Analysis
Author(s)
Rebecca Metcalfe, BA, MA, PhD1, Quang Vuong, MSc2, Steve Kittredge, DWD2, Jay J. Park, PhD3;
1Centre for Advancing Health Outcomes, Vancouver, BC, Canada, 2Core Clinical Sciences, Vancouver, BC, Canada, 3McMaster University, Department of Health Research Methods, Evidence, and Impact, Hamilton, ON, Canada
1Centre for Advancing Health Outcomes, Vancouver, BC, Canada, 2Core Clinical Sciences, Vancouver, BC, Canada, 3McMaster University, Department of Health Research Methods, Evidence, and Impact, Hamilton, ON, Canada
OBJECTIVES: The validity of causal inference methods depends on the ability of statisticians to identify an appropriate set of variables that can control for confounding. In practice, analysts should consult subject-matter experts and review the available literature to construct causal diagrams, such as directed acyclic graphs (DAGs), that describe the causal relationships between variables of interest. However, it can be difficult to construct and update DAGs when the number of relevant variables is large. We sought to develop an open-source tool to facilitate collaborative DAG development.
METHODS: We reviewed existing digital DAG tools to identify gaps in current implementation. Following review, we consulted potential end-users with experience in causal inference and DAG development to understand software requirements.
RESULTS: Review of existing software found several limitations including: absence of a graphic user interface (GUI) or an unintuitive and outdated GUI; installation requirements; poor compatibility with statistical software; and restricted options for customization. End-users reported similar challenges with existing software and shared additional features of interest including the ability to identify confounders using backdoor criteria; and the ability to identify effect modifiers in the context of novel causal inference methods, such as transportability analysis. Based on these findings, we created “DAGDraw” as an interactive, open-source web application hosted through Shiny. “DAGDraw” pulls from existing R packages “daggity” and “ggdag” to identify backdoor paths, and leverages the visualization functions of “DiagrammeR” to dynamically construct DAGs, thereby integrating existing capabilities into an intuitive GUI. In addition, “DAGDraw” can identify effect modifiers in the context of transportability analyses. DAGs can be output graphically or as R code to allow further customization.
CONCLUSIONS: We developed an open-source web application to support causal reasoning and the collaborative development of DAGs. DAGDraw responds to end-user needs. Further testing is needed to quantitatively assess tool usability.
METHODS: We reviewed existing digital DAG tools to identify gaps in current implementation. Following review, we consulted potential end-users with experience in causal inference and DAG development to understand software requirements.
RESULTS: Review of existing software found several limitations including: absence of a graphic user interface (GUI) or an unintuitive and outdated GUI; installation requirements; poor compatibility with statistical software; and restricted options for customization. End-users reported similar challenges with existing software and shared additional features of interest including the ability to identify confounders using backdoor criteria; and the ability to identify effect modifiers in the context of novel causal inference methods, such as transportability analysis. Based on these findings, we created “DAGDraw” as an interactive, open-source web application hosted through Shiny. “DAGDraw” pulls from existing R packages “daggity” and “ggdag” to identify backdoor paths, and leverages the visualization functions of “DiagrammeR” to dynamically construct DAGs, thereby integrating existing capabilities into an intuitive GUI. In addition, “DAGDraw” can identify effect modifiers in the context of transportability analyses. DAGs can be output graphically or as R code to allow further customization.
CONCLUSIONS: We developed an open-source web application to support causal reasoning and the collaborative development of DAGs. DAGDraw responds to end-user needs. Further testing is needed to quantitatively assess tool usability.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
MSR93
Topic
Methodological & Statistical Research
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference
Disease
No Additional Disease & Conditions/Specialized Treatment Areas