SAS and R code for probabilistic quantitative bias analysis for misclassified binary variables and binary unmeasured...

Area of Science:

• Epidemiologic research methods
• Quantitative bias analysis (QBA)
• Statistical software implementation

Background:

• Systematic errors like selection bias, confounding, and misclassification are common in epidemiology.
• Quantitative bias analysis (QBA) is infrequently used due to a lack of accessible software tools.
• This limits the accurate quantification of systematic error's impact on research findings.

Purpose of the Study:

• To provide readily modifiable computing code for implementing QBA in epidemiologic studies.
• To enable analysts to tailor QBA methods to their specific datasets.
• To facilitate the quantification of bias from uncontrolled confounding and misclassification.

Main Methods:

• Description of QBA methods for misclassification and uncontrolled confounding.
• Provision of example code for SAS and R to implement QBA.
• Demonstration of QBA using both summary-level and individual record-level data.

Main Results:

• Code examples illustrate bias adjustment for uncontrolled confounding and misclassification.
• Bias-adjusted point estimates are compared with conventional results to show impact.
• 95% simulation intervals are generated and compared to confidence intervals to assess uncertainty.

Conclusions:

• Accessible code can increase the adoption of QBA in epidemiologic research.
• Implementing QBA helps prevent inaccurate inferences caused by unquantified systematic error.
• This approach improves the reliability and validity of epidemiologic study findings.