Risk functions with outcome measurement error

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Studies linking to vital statistics registries for death ascertainment can suffer from outcome measurement error.
• This error can lead to biased estimates of mortality risk and survival due to uncaptured deaths, false positives, or incorrect timing of recorded deaths.

Purpose of the Study:

• To extend the Rogan-Gladen estimator to address outcome measurement error in risk and survival functions, specifically in the presence of right censoring.
• To apply and evaluate this extended estimator for mortality risk in a cohort of individuals receiving HIV care, accounting for potential linkage errors to death registries.

Main Methods:

• Developed an extension of the Rogan-Gladen estimator to incorporate outcome measurement error.
• Applied the method to data from the University of North Carolina Center for AIDS Research HIV Clinical Cohort (2001-2022).
• Conducted simulation studies to assess the performance of the proposed approach under various conditions, including higher mortality risk in validation subsets.

Main Results:

• The extended Rogan-Gladen estimator effectively accounts for outcome measurement error in mortality risk and survival estimation.
• The method demonstrated robust performance in simulations, even with validation samples having higher baseline mortality risk.
• The approach is flexible, allowing parameterization with internal or external validation data, or use as quantitative bias analysis.

Conclusions:

• The proposed method provides a valuable tool for correcting bias in mortality estimates caused by vital statistics linkage errors.
• This technique enhances the accuracy and reliability of survival analysis in epidemiological and clinical research, particularly in cohorts with potential data quality issues.
• The approach offers a quantitative bias analysis framework for studies relying on registry data for outcome ascertainment.