Potential biases in observational infectious disease studies: examples using Directed Acyclic Graphs

Abstract Background Bias is a systematic error that threatens the (internal) validity of research. Observational studies in infectious diseases are especially prone to bias because they are typically non-randomized and often rely on routinely collected or retrospective data. Objectives To provide an overview of common biases in observational infectious diseases research, introduce the concept of directed acyclic graphs (DAGs), and outline practical approaches for bias identification and mitigation. Source This narrative review draws on biases commonly encountered by the authors as editors, reviewers, and clinicians, supplemented by PubMed searches and recent reviews from other medical fields. Content The paper introduces DAGs as a conceptual tool to identify bias. Biases are grouped into selection, time-related, and information bias. Examples from published infectious diseases studies are given together with strategies to detect and reduce bias. Implications Intended as an introductory guide for clinicians, investigators, and reviewers, this paper aims to improve critical appraisal and study design, while acknowledging it does not cover all possible biases.