Using directed acyclic graphs to illustrate common sources of bias in diagnostic test accuracy studies.

BACKGROUND: Diagnostic test accuracy (DTA) studies, like aetiological studies, are susceptible to various sources of bias, including imperfect reference standard, partial verification, spectrum effect, confounding, and misassumption of conditional independence. While directed acyclic graphs (DAGs) are widely used in aetiological research to identify and illustrate biasing structures, they have not been systematically applied to DTA studies. METHODS: We developed DAGs to illustrate causal mechanisms underlying common sources of bias in DTA studies. For each source, we also illustrate its equivalent in aetiological studies. Real-world examples are used for illustration. RESULTS: We present DAGs for five major sources of bias in DTA studies with structural parallels to aetiological studies: imperfect reference standard corresponds to exposure misclassification, misassumption of conditional independence creates spurious correlations similar to ignoring unmeasured confounding, spectrum effect parallels effect modification, confounding operates through backdoor paths in both settings, and partial verification acts through selection mechanisms that can induce selection bias. These DAG representations help reveal causal mechanisms underlying each source of bias and suggest appropriate adjustment strategies. CONCLUSION: DAGs provide a valuable framework for understanding sources of bias in DTA studies and should complement existing reporting and quality-assessment tools such as the Standards for Reporting of Diagnostic Accuracy (STARD) and Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2). We recommend incorporating DAGs in the conduct and reporting of primary DTA studies, systematic reviews, and meta-analyses to identify potential sources of bias and enhance transparency. DAG construction requires interdisciplinary collaboration and sensitivity analyses examining alternative causal structures.