Test-negative Designs with Various Reasons for Testing: Statistical Bias and Solution

Area of Science:

• Epidemiology
• Biostatistics
• Vaccinology

Background:

• Test-negative designs are crucial for post-market vaccine effectiveness evaluation when randomized trials are infeasible.
• Recent adaptations include individuals with diverse testing reasons, potentially introducing bias.
• Formal statistical examination is needed for these modified designs.

Purpose of the Study:

• To statistically examine potential bias in modified test-negative designs.
• To develop methods for unbiased estimation of vaccine effectiveness.
• To improve precision by incorporating multiple reasons for testing.

Main Methods:

• Statistical derivations and causal graphs were used to analyze bias.
• Reasons for testing were categorized into symptoms, mandatory screening, and contact tracing.
• Stratification was employed for consistent estimation and bias elimination.
• A novel stratified estimator was proposed and evaluated.

Main Results:

• The standard odds ratio estimator can be biased if diverse testing reasons are not considered.
• Stratification effectively eliminates bias and allows for consistent estimation of vaccine effectiveness.
• The proposed stratified estimator can improve precision by incorporating multiple testing reasons.

Conclusions:

• Modified test-negative designs require careful statistical consideration of testing reasons.
• The proposed stratification method provides a robust approach for vaccine effectiveness estimation.
• This work enhances the reliability of post-market vaccine safety and effectiveness surveillance.