Applications of a parametric model for informative censoring

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Informative censoring can bias survival data analysis.
• Assessing the impact of informative censoring is crucial for reliable results.
• Existing methods may not adequately address the nuances of dependent censoring processes.

Purpose of the Study:

• To develop a parametric model for survival data analysis that explicitly accounts for informative censoring.
• To provide methods for sensitivity analysis to quantify the potential bias introduced by informative censoring.
• To derive bounds for key parameters that are robust to the choice of bias function.

Main Methods:

• Utilized a parametric survival model incorporating a dependence parameter (delta) and a bias function B(t, theta).
• Calculated the expectation of potential bias to measure the impact of non-ignorable censoring.
• Derived bounds for distribution parameters and relative risk, independent of specific bias function choices for fixed delta.

Main Results:

• The proposed model allows for quantitative assessment of bias due to informative censoring.
• Calculated bounds provide a range of plausible estimates for parameters of interest.
• Demonstrated through an application to systematic lupus erythematosus data that additional information can reduce uncertainty in location parameter estimates.

Conclusions:

• The developed parametric model offers a flexible framework for sensitivity analysis in the presence of informative censoring.
• The methodology provides valuable insights into the potential impact of censoring on survival analysis results.
• The approach enhances the reliability of statistical inferences in biomedical research, particularly in complex datasets.