Causal diagrams for encoding and evaluation of information bias

Area of Science:

• Epidemiology
• Biostatistics
• Medical Research

Background:

• Bias is a critical concern in epidemiological and clinical research.
• Common bias categories include confounding, selection bias, and information bias.
• Directed acyclic graphs (DAGs) are increasingly used to model confounding and selection bias.

Purpose of the Study:

• To examine common types of information bias, specifically disease-related and exposure-related bias.
• To explore the utility of causal diagrams (DAGs) in representing information bias.

Main Methods:

• Analysis of information bias through the lens of causal diagrams.
• Focus on disease-related and exposure-related information bias.

Main Results:

• Information bias involves effects of the true variable of interest (disease or exposure status).
• Bias often arises from uninteresting causal or associational paths.
• Strategies may exist to prevent or reduce certain types of information bias.

Conclusions:

• Information bias can be clearly and helpfully represented using causal diagrams (DAGs).
• The DAG framework provides a unified approach to understanding major bias types in research.
• Causal diagrams offer a valuable tool for addressing information bias in epidemiological studies.