This article, titled “Confounder Selection in Observational Studies in High-Impact Medical and Epidemiological Journals,” was authored by Luis C. L. Correia, Rafael F. Mascarenhas, Felipe S. C. De Menezes, Jeronimo S. Oliveira Junior, Viola Vaccarino, Joseph S. Ross, and Joshua D. Wallach. It was published in JAMA Network Open on July 25, 2025.
The research addresses the increasing interest in using observational data to evaluate the causal effects of exposures on outcomes, highlighting that such studies rely on strong assumptions, including the absence of uncontrolled confounding. Traditional statistical methods, like the change-in-estimate strategy, are now considered inadequate for selecting confounders, with current recommendations advocating for an informed approach based on theoretical models, such as directed acyclic graphs (DAGs). Following a 2004 study that pointed out the lack of sufficient rationale for confounder selection in observational studies, the STROBE reporting guideline emphasized the need for clear definition of potential confounders.
This study’s objective was to evaluate whether there have been improvements over time in the reported methods for selecting confounders to control for in observational studies published in the highest impact factor medical and epidemiological journals. The researchers conducted a cross-sectional study, reviewing PubMed-indexed articles from 2003, 2013, and 2023 in the 10 highest impact factor medical and epidemiological journals. They classified the reported methods for confounder selection into categories such as no justification, selection based on established association, statistical criteria, or a causal model.
Key findings indicate that among 623 eligible observational studies, 281 (45.1%) reported selecting confounders without justification. While the use of a causal model to identify confounders increased from 0% in 2003 to 22.6% in 2023, only 42 (6.7%) studies overall reported using a causal model (with 35 employing DAGs and 7 providing textual explanations). The proportion of studies selecting confounders without justification remained relatively constant between 2003 and 2023, from 48.7% to 41.5%. These findings raise concerns about how confounders are selected and justified in observational studies and highlight the need for journals and the STROBE guideline to provide more explicit guidance on confounder selection requirements to improve reporting practices.
Correia, L. C. L., Mascarenhas, R. F., De Menezes, F. S. C., Oliveira, J. S., Jr., Vaccarino, V., Ross, J. S., & Wallach, J. D. (2025). Confounder selection in observational studies in high-impact medical and epidemiological journals. JAMA Network Open, 8(7), e2524176. https://doi.org/10.1001/jamanetworkopen.2025.24176
