[Application of directed acyclic graphs in identifying and controlling confounding bias]

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Observational studies are crucial for etiological research but are susceptible to bias from confounding variables.
• Traditional methods for identifying and adjusting confounders can be complex and may lead to incorrect adjustments, introducing new biases.

Purpose of the Study:

• To explore the utility of directed acyclic graphs (DAGs) in improving confounder identification and adjustment in observational etiological studies.
• To provide a more intuitive and accurate approach to causal inference in observational research.

Main Methods:

• Utilizing directed acyclic graphs (DAGs) to visually represent causal relationships between variables.
• Comparing the application of DAGs with traditional confounder definition methods in study design.

Main Results:

• DAGs provide a more intuitive framework for identifying true confounders.
• Employing DAGs helps avoid the over-adjustment or under-adjustment of variables, thereby reducing bias.
• DAGs facilitate the selection of appropriate adjustment strategies for causal inference.

Conclusions:

• Directed acyclic graphs (DAGs) offer a superior method for confounder management in observational etiological studies.
• The use of DAGs enhances the accuracy and interpretability of causal inference from observational data.