Interpretation of the steps in the construction and application of directed acyclic graphs_Chinese Journal of Evidence-Based Medicine

Authors：

WEI Baian ¹ , CHEN Genghang ² , LIU Shaonan ² , CHEN Xueyin ² , HUANG Qian ² , YANG Lihong ² ,  GUO Xinfeng ²

1. The Second Clinical Medical College, Guangzhou University of Chinese Medicine, Guangzhou 510405, P. R. China;
2. The Second Affiliated Hospital of Guangzhou University of Chinese Medicine (Guangdong Provincial Hospital of Chinese Medicine), Guangzhou 510120, P. R. China;

Corresponding?author：

GUO Xinfeng, Email: guoxinfeng@gzucm.edu.cn

Keywords：

Directed acyclic graphs; Causal inference; Confounding bias; Guideline interpretation

DOI：

10.7507/1672-2531.202506078

Video：

Export PDF Favorites Scan Get Citation

Abstract Full text Figures/Tables Video References Cited by

Directed acyclic graphs (DAGs) are essential tools based on prior assumptions, capable of clearly depicting causal relationships between variables, aiding in the identification of confounders and assessment of bias. They have been widely applied in the field of epidemiology. However, the construction of DAGs is prone to the subjectivity of researchers. This paper interprets the DAGs construction and application guidelines from the BMJ and elaborates on the seven steps for building DAGs: clarifying the research question and target population, identifying relevant variables, reaching consensus and pre-registering, determining data collection or datasets based on the consensus graph, selecting analysis methods and variable measurement, conducting sensitivity analysis, and reporting the DAGs. By following these steps, researchers can construct DAGs more scientifically, identify the minimal sufficient adjustment set, and enhance the scientific validity and reliability of their studies. Despite its advantages, DAGs have limitations. For instance, failure to consider all relevant factors may lead to unexplained disturbances in the analysis results or incorrect causal inferences, thus failing to verify other key assumptions in causal reasoning. Therefore, it is recommended to integrate DAGs with data-driven methods to maximize their strengths and compensate for their shortcomings.

1.	Chowdhury MZI, Turin TC. Variable selection strategies and its importance in clinical prediction modelling. Fam Med Community Health, 2020, 8(1): e000262.
2.	聶朦, 鄔娜, 袁志權, 等. 臨床流行病學研究中多因素分析時的變量選擇方法. 中華內科雜志, 2024, 63(5): 462-467.
3.	Xie J, Lin Y, Yan X, et al. Category-adaptive variable screening for ultra-high dimensional heterogeneous categorical data. J Am Stat Assoc, 2020, 115(530): 747-760.
4.	Tennant PWG, Murray EJ, Arnold KF, et al. Use of directed acyclic graphs (DAGs) to identify confounders in applied health research: review and recommendations. Int J Epidemiol, 2021, 50(2): 620-632.
5.	Digitale JC, Martin JN, Glymour MM. Tutorial on directed acyclic graphs. J Clin Epidemiol, 2022, 142: 264-267.
6.	Dekkers OM, Laugesen K, Groenwold RHH. Directed acyclic graphs in clinical research. Eur J Endocrinol, 2024, 190(4): E5-E7.
7.	Staerk C, Byrd A, Mayr A. Recent methodological trends in epidemiology: no need for data-driven variable selection. Am J Epidemiol, 2024, 193(2): 370-376.
8.	Talbot D, Massamba VK. A descriptive review of variable selection methods in four epidemiologic journals: there is still room for improvement. Eur J Epidemiol, 2019, 34(8): 725-730.
9.	Corazza GR, Formagnana P, Lenti MV. Bringing complexity into clinical practice: an internistic approach. Eur J Intern Med, 2019, 61: 9-14.
10.	Gr?tz M. When less conditioning provides better estimates: overcontrol and endogenous selection biases in research on intergenerational mobility. Qual Quant, 2022, 56(5): 3769-3793.
11.	Feeney T, Hartwig FP, Davies NM. How to use directed acyclic graphs: guide for clinical researchers. BMJ, 2025, 388: e078226.
12.	Suzuki E, Shinozaki T, Yamamoto E. Causal diagrams: pitfalls and tips. J Epidemiol, 2020, 30(4): 153-162.
13.	Westreich D, Greenland S. The table 2 fallacy: presenting and interpreting confounder and modifier coefficients. Am J Epidemiol, 2013, 177(4): 292-298.
14.	Byeon S, Lee W. Directed acyclic graphs for clinical research: a tutorial. J Minim Invasive Surg, 2023, 26(3): 97-107.
15.	Matsouaka RA, Liu Y, Zhou Y. Variance estimation for the average treatment effects on the treated and on the controls. Stat Methods Med Res, 2023, 32(2): 389-403.
16.	Ferguson KD, Mccann M, Katikireddi SV, et al. Evidence synthesis for constructing directed acyclic graphs (ESC-DAGs): a novel and systematic method for building directed acyclic graphs. Inter J Epidemiol, 2020, 49(1): 322-329.
17.	郝允逸, 夏雪, 胥芹, 等. 有向無環圖的構建規范: ESC-DAGs方法介紹及腦血管病研究案例解讀. 中國卒中雜志, 2024, 19(10): 1221-1229.
18.	Hasani WSR, Musa KI, Chen XW, et al. Constructing causal pathways for premature cardiovascular disease mortality using directed acyclic graphs with integrating evidence synthesis and expert knowledge. Sci Rep, 2024, 14(1): 28849.
19.	Rodrigues D, Kreif N, Lawrence-Jones A, et al. Reflection on modern methods: constructing directed acyclic graphs (DAGs) with domain experts for health services research. Int J Epidemiol, 2022, 51(4): 1339-1348.
20.	Xu J, Wang Y, He Q, et al. Evaluating the agreement between sensitivity and primary analyses in observational studies using routinely collected healthcare data: a meta-epidemiology study. BMC Med, 2025, 23(1): 393.
21.	Inoue K, Sakamaki K, Komukai S, et al. Methodological tutorial series for epidemiological studies: confounder selection and sensitivity analyses to unmeasured confounding from epidemiological and statistical perspectives. J Epidemiol, 2025, 35(1): 3-10.
22.	Vanderweele TJ, Ding P. Sensitivity analysis in observational research: introducing the e-value. Ann Inter Med, 2017, 167(4): 268-274.
23.	Faries D, Gao C, Zhang X, et al. Real effect or bias? Best practices for evaluating the robustness of real-world evidence through quantitative sensitivity analysis for unmeasured confounding. arXiv, 2023.
24.	盛松, 郭曼萍, 趙陽, 等. 效應改變法和有向無環圖在線性回歸變量篩選中的應用及實現. 中國循證心血管醫學雜志, 2024, 16(10): 1175-1178.
25.	Lee PH, Burstyn I. Identification of confounder in epidemiologic data contaminated by measurement error in covariates. BMC Med Res Methodol, 2016, 16: 54.
26.	Parida PK, Marwala T, Chakraverty S. A multivariate additive noise model for complete causal discovery. Neural Netw, 2018, 103: 44-54.
27.	Viinikka J, Hyttinen A, Pensar J, et al. Towards scalable Bayesian learning of causal DAGs. arXiv, 2020.
28.	Piccininni M, Konigorski S, Rohmann JL, et al. Directed acyclic graphs and causal thinking in clinical risk prediction modeling. BMC Med Res Methodol, 2020, 20(1): 1-12.
29.	Zhu S, Ng I, Chen Z. Causal discovery with reinforcement learning. arXiv, 2020.
30.	Sirazitdinov A, Buchwald M, Heuveline V, et al. Graph neural networks for individual treatment effect estimation. IEEE Access, 2024, 12: 106884-106894.

1. Chowdhury MZI, Turin TC. Variable selection strategies and its importance in clinical prediction modelling. Fam Med Community Health, 2020, 8(1): e000262.
2. 聶朦, 鄔娜, 袁志權, 等. 臨床流行病學研究中多因素分析時的變量選擇方法. 中華內科雜志, 2024, 63(5): 462-467.
3. Xie J, Lin Y, Yan X, et al. Category-adaptive variable screening for ultra-high dimensional heterogeneous categorical data. J Am Stat Assoc, 2020, 115(530): 747-760.
4. Tennant PWG, Murray EJ, Arnold KF, et al. Use of directed acyclic graphs (DAGs) to identify confounders in applied health research: review and recommendations. Int J Epidemiol, 2021, 50(2): 620-632.
5. Digitale JC, Martin JN, Glymour MM. Tutorial on directed acyclic graphs. J Clin Epidemiol, 2022, 142: 264-267.
6. Dekkers OM, Laugesen K, Groenwold RHH. Directed acyclic graphs in clinical research. Eur J Endocrinol, 2024, 190(4): E5-E7.
7. Staerk C, Byrd A, Mayr A. Recent methodological trends in epidemiology: no need for data-driven variable selection. Am J Epidemiol, 2024, 193(2): 370-376.
8. Talbot D, Massamba VK. A descriptive review of variable selection methods in four epidemiologic journals: there is still room for improvement. Eur J Epidemiol, 2019, 34(8): 725-730.
9. Corazza GR, Formagnana P, Lenti MV. Bringing complexity into clinical practice: an internistic approach. Eur J Intern Med, 2019, 61: 9-14.
10. Gr?tz M. When less conditioning provides better estimates: overcontrol and endogenous selection biases in research on intergenerational mobility. Qual Quant, 2022, 56(5): 3769-3793.
11. Feeney T, Hartwig FP, Davies NM. How to use directed acyclic graphs: guide for clinical researchers. BMJ, 2025, 388: e078226.
12. Suzuki E, Shinozaki T, Yamamoto E. Causal diagrams: pitfalls and tips. J Epidemiol, 2020, 30(4): 153-162.
13. Westreich D, Greenland S. The table 2 fallacy: presenting and interpreting confounder and modifier coefficients. Am J Epidemiol, 2013, 177(4): 292-298.
14. Byeon S, Lee W. Directed acyclic graphs for clinical research: a tutorial. J Minim Invasive Surg, 2023, 26(3): 97-107.
15. Matsouaka RA, Liu Y, Zhou Y. Variance estimation for the average treatment effects on the treated and on the controls. Stat Methods Med Res, 2023, 32(2): 389-403.
16. Ferguson KD, Mccann M, Katikireddi SV, et al. Evidence synthesis for constructing directed acyclic graphs (ESC-DAGs): a novel and systematic method for building directed acyclic graphs. Inter J Epidemiol, 2020, 49(1): 322-329.
17. 郝允逸, 夏雪, 胥芹, 等. 有向無環圖的構建規范: ESC-DAGs方法介紹及腦血管病研究案例解讀. 中國卒中雜志, 2024, 19(10): 1221-1229.
18. Hasani WSR, Musa KI, Chen XW, et al. Constructing causal pathways for premature cardiovascular disease mortality using directed acyclic graphs with integrating evidence synthesis and expert knowledge. Sci Rep, 2024, 14(1): 28849.
19. Rodrigues D, Kreif N, Lawrence-Jones A, et al. Reflection on modern methods: constructing directed acyclic graphs (DAGs) with domain experts for health services research. Int J Epidemiol, 2022, 51(4): 1339-1348.
20. Xu J, Wang Y, He Q, et al. Evaluating the agreement between sensitivity and primary analyses in observational studies using routinely collected healthcare data: a meta-epidemiology study. BMC Med, 2025, 23(1): 393.
21. Inoue K, Sakamaki K, Komukai S, et al. Methodological tutorial series for epidemiological studies: confounder selection and sensitivity analyses to unmeasured confounding from epidemiological and statistical perspectives. J Epidemiol, 2025, 35(1): 3-10.
22. Vanderweele TJ, Ding P. Sensitivity analysis in observational research: introducing the e-value. Ann Inter Med, 2017, 167(4): 268-274.
23. Faries D, Gao C, Zhang X, et al. Real effect or bias? Best practices for evaluating the robustness of real-world evidence through quantitative sensitivity analysis for unmeasured confounding. arXiv, 2023.
24. 盛松, 郭曼萍, 趙陽, 等. 效應改變法和有向無環圖在線性回歸變量篩選中的應用及實現. 中國循證心血管醫學雜志, 2024, 16(10): 1175-1178.
25. Lee PH, Burstyn I. Identification of confounder in epidemiologic data contaminated by measurement error in covariates. BMC Med Res Methodol, 2016, 16: 54.
26. Parida PK, Marwala T, Chakraverty S. A multivariate additive noise model for complete causal discovery. Neural Netw, 2018, 103: 44-54.
27. Viinikka J, Hyttinen A, Pensar J, et al. Towards scalable Bayesian learning of causal DAGs. arXiv, 2020.
28. Piccininni M, Konigorski S, Rohmann JL, et al. Directed acyclic graphs and causal thinking in clinical risk prediction modeling. BMC Med Res Methodol, 2020, 20(1): 1-12.
29. Zhu S, Ng I, Chen Z. Causal discovery with reinforcement learning. arXiv, 2020.
30. Sirazitdinov A, Buchwald M, Heuveline V, et al. Graph neural networks for individual treatment effect estimation. IEEE Access, 2024, 12: 106884-106894.

Chinese Journal of Evidence-Based Medicine

Latest ArticlesInterpretation of the steps in the construction and application of directed acyclic graphs

Abstract Full text Figures/Tables Video References Cited by

Format

Content