Evaluating bias correction in weighted proportional hazards regression

Area of Science:

• Biostatistics
• Epidemiology
• Health Services Research

Background:

• Observational studies often use biased samples, requiring adjustments for accurate analysis.
• Inverse selection probability weighting (ISPW) is a common method to correct for such biases.
• Existing methods forISPW in Cox models can be computationally intensive and require auxiliary data.

Purpose of the Study:

• To develop and evaluate methods for quantifying the bias corrected by inverse selection probability weighting.
• To assess the impact of bias correction on partial likelihood and Breslow-Aalen estimators.
• To determine the necessity of auxiliary data collection for future studies and evaluate past research.

Main Methods:

• Proposed novel statistical methods to quantify bias correction in survival analysis estimators.
• Derived asymptotic properties for the proposed test statistics.
• Evaluated finite-sample performance through simulations and applied methods to post-kidney transplant survival data.

Main Results:

• The proposed methods successfully quantify the degree of bias corrected by the weighting procedure.
• Asymptotic properties of the test statistics were theoretically established.
• Simulations confirmed the significance level and power of the proposed methods.

Conclusions:

• The developed methods provide a robust way to assess the impact of selection bias correction in observational studies.
• Quantifying bias correction is crucial for validating analytical approaches and informing future study designs.
• Application to organ transplant data demonstrates the practical utility in evaluating survival models.