Potential Applications of Directed Acyclic Graphs in the Design and Interpretation of Biomedical Research

This article presents an algorithm for constructing and applying directed acyclic graphs (DAGs) in the planning of epidemiological and biomedical studies. DAGs are graphical tools for modeling complex relationships between variables, which is particularly relevant in biomedical science, where accurate assessment of causal relationships requires accounting for potential confounding factors. The importance of DAGs is emphasized for conceptualizing scientific hypotheses and understanding the structure of relationships between factors based on scientific data review and findings from previous studies. The use of DAGs enhances the quality of both study design and data analysis, providing a more grounded approach to selecting variables for inclusion in statistical models. DAGs make it possible to determine the minimal and sufficient set of factors for adjustment, with consideration of the roles of variables (confounders, mediators, colliders) in relation to the exposure (a probable risk factor) and the outcome (a disease or condition), thus reducing the likelihood of analytical errors. The article highlights the practical application of DAGs using available software and provides specific examples of their use in biomedical research. Finally, recommendations are offered for integrating DAGs into biomedical research practice, which may contribute to the broader adoption of modern multivariate statistical methods, improved interpretability, and enhanced reproducibility of scientific findings

Algorithm · Data science · Directed acyclic graph · Epistemology · Management science · Programming language · Bioinformatics and Genomic Networks · Computational Drug Discovery Methods · Computer Science · Engineering · Health, Environment, Cognitive Aging · Philosophy · Theoretical Computer Science

• Reducing bias through directed acyclic graphs

  Open Access•Ian Shrier, Robert W Platt•BMC Medical Research Methodology•2008

  Incluir item

• Are patients with hypertension and diabetes mellitus at increased risk for Covid-19 infection?

  Open Access•Lei Fang, George Karakiulakis et al.•The Lancet Respiratory Medicine•2020

  Incluir item

• Variable selection – A review and recommendations for the practicing statistician

  Open Access•Georg Heinze, Christine Wallisch et al.•Biometrical Journal•2018

  Incluir item

• Robust causal inference using directed acyclic graphs

  Johannes Textor, Benito van der Zander et al.•International Journal of…•2017

  Incluir item

• Purposeful selection of variables in logistic regression

  Open Access•Zoran Bursac, C Heath Gauss et al.•Source Code for Biology and…•2008

  Incluir item

• Causal Diagrams for Epidemiologic Research

  Sander Greenland, Judea Pearl et al.•Epidemiology•1999

  Incluir item

• Principles of confounder selection

  Open Access•Tyler J Vanderweele•European Journal of Epidemiology•2019

  Incluir item

• Factors associated with Covid-19-related death using OpenSafely

  Open Access•Elizabeth J Williamson, Alex J Walker et al.•Nature•2020

  Incluir item

• Sample size calculation for cross-sectional studies

  Open Access•Nikita A Mitkin, Sergei N Drachev et al.•Ekologiya Cheloveka (Human Ecology•2023

  Incluir item