Bounding Bias Due to Selection.

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Selection bias poses a significant threat to the validity of causal inference in epidemiologic studies conducted on population subsets.
• Quantifying selection bias and performing sensitivity analyses are often complex and challenging for researchers.

Purpose of the Study:

• To develop simple, bounded expressions for quantifying the magnitude of selection bias.
• To provide researchers with practical tools for assessing the potential impact of selection bias on study findings.

Main Methods:

• Demonstrated bounding of selection bias magnitude using simple expressions based on unmeasured factors and measured variables.
• Introduced summary measures derived from bounds to calculate the minimum bias magnitude needed to nullify a risk ratio.
• Illustrated the application of these methods with examples of varying selection mechanisms.

Main Results:

• Developed simple, assumption-free expressions to bound the magnitude of selection bias.
• Created summary measures to determine the strength of selection bias required to explain away observed results.
• Showcased the practical implementation of these sensitivity analyses through diverse examples.

Conclusions:

• The proposed bounding methods offer a straightforward approach to quantify and assess selection bias in epidemiologic research.
• These tools empower researchers to better understand the robustness of their findings against potential selection bias.
• The methods are adaptable and can be simplified under specific assumptions or contexts, enhancing their utility.