Use of directed acyclic graphs (DAGs) to identify confounders in applied health research: review and recommendations

Abstract

Background Directed acyclic graphs (DAGs) are an increasingly popular approach for identifying confounding variables that require conditioning when estimating causal effects. This review examined the use of DAGs in applied health research to inform recommendations for improving their transparency and utility in future research. Methods Original health research articles published during 1999–2017 mentioning ‘directed acyclic graphs’ (or similar) or citing DAGitty were identified from Scopus, Web of Science, Medline and Embase. Data were extracted on the reporting of: estimands, DAGs and adjustment sets, alongside the characteristics of each article’s largest DAG. Results A total of 234 articles were identified that reported using DAGs. A fifth (n = 48, 21%) reported their target estimand(s) and half (n = 115, 48%) reported the adjustment set(s) implied by their DAG(s). Two-thirds of the articles (n = 144, 62%) made at least one DAG available. DAGs varied in size but averaged 12 nodes [interquartile range (IQR): 9–16, range: 3–28] and 29 arcs (IQR: 19–42, range: 3–99). The median saturation (i.e. percentage of total possible arcs) was 46% (IQR: 31–67, range: 12–100). 37% (n = 53) of the DAGs included unobserved variables, 17% (n = 25) included ‘super-nodes’ (i.e. nodes containing more than one variable) and 34% (n = 49) were visually arranged so that the constituent arcs flowed in the same direction (e.g. top-to-bottom). Conclusion There is substantial variation in the use and reporting of DAGs in applied health research. Although this partly reflects their flexibility, it also highlights some potential areas for improvement. This review hence offers several recommendations to improve the reporting and use of DAGs in future research.

Physical Exercise is Associated with Glycemic Control among Women with Gestational Diabetes Mellitus: Findings from a Prospective Cohort in Shanghai, China

“…The potential confounders adjusted in logistic regression were identified by using the directed acyclic graphs (DAGs) method. 19,20 Figures were produced to described the detailed information for the association between total exercise times and percentage of abnormal PG within different groups, as well as the association between the percentage of abnormal PG and total exercise times among GDM women in different gestation weeks. In this study, a p-value of less than 0.05 (two-tailed) was considered as statistically significant.…”

“…In this study, information for pregnancy and childbirth history, morning plasma glucose in fasting state as well as 2-hour after breakfast during routine antenatal checkup were extracted directly from the delivery records and routine antenatal checkup forms in hospital, which contributed to a relatively lower recall bias. Meanwhile, we used directed acyclic graphs (DAGs) to identify potential confounders and then adjusted them in the logistic regression analysis, 19,20 which ensures a relatively unbiased association between physical exercise times and the percentage of abnormal PG, which is another strength of this study.…”

Development and application of an evidence-based directed acyclic graph to evaluate the associations between metal mixtures and cardiometabolic outcomes

“…To the extent that new research is published challenging or adding to the DAG we show here, we encourage researchers to incorporate the new knowledge and update their covariate sets; indeed, a prime advantage of using DAGs is that we can have a scientific conversation about explicit modeling assumptions. 25 For example, others may wish to include hypertension as a cardiometabolic outcome (rather than a node that could be a potential confounder). Nevertheless, given the remarkably similar results observed in our pilot analysis using each of the three minimally sufficient adjustment sets and given the DAG assessment results using the DAGs with randomly permutated nodes, we can suggest that our DAG development process was robust and that the DAG is compatible with real-world data.…”

“…[21][22][23][24] DAGs are also useful for identifying minimally sufficient adjustment sets of variables; when DAGs are used in this way, articles should report the assumed DAG. 25 Few evidence-based DAGs exist in the environmental epidemiology context due to the need to conduct systematic literature reviews and to empirically test the applicability of the DAG for the study context (as one example, see Corlin et al 26 ). No evidence-based DAGs have been previously published describing the structure underlying potential metal mixture-cardiometabolic outcome relationships.…”

Adjustment for energy intake in nutritional research: a causal inference perspective

“…DAGs offer a simple means to identifying the appropriate adjustment set for a particular estimand, and guidelines are now available on how best to report their use. (22) A single model that includes all individual components of the diet may provide the simplest and most accurate approach to estimating both the total causal effect and any relative causal effect of interest.…”