Understanding and Managing Confounders, Mediators and Colliders in Research

ABSTRACT Rationale Researchers often make causal inferences about relationships among variables and constructs. However, third‐variable effects may obscure the relationship among studied variables. Third‐variable effects generally include confounders and mediators, but recently there has been an emerging discussion on colliders. Aim To provide a concise introduction of confounders, colliders, and mediators for health researchers and outline strategies for minimising the impact of confounders, colliders and mediators in quantitative research. Methods Methodological literature from biostatistics textbooks, methodology papers, and methodological reviews published in nursing, health, psychological and behavioural sciences. Conclusions Understanding third‐variable effects is crucial to conducting rigorous research and drawing valid causal inferences from research data. Health researchers should embrace both theory and model‐based thinking as a foundational element of their methodology. This involves explicitly theorising the underlying causal structures before data collection and analysis using Directed Acyclic Graphs which are useful for visually representing hypothesised causal pathways and their relationships with potential third variables.