A Bayesian approach to estimating causal vaccine effects on binary post-infection outcomes

Area of Science:

• Biostatistics
• Epidemiology
• Vaccinology

Background:

• Estimating causal effects of vaccines on post-infection outcomes is crucial for public health.
• The principal stratification framework and maximum likelihood estimators have been used previously.

Purpose of the Study:

• To propose and evaluate a Bayesian approach for estimating causal vaccine effects on binary post-infection outcomes.
• To assess the identifiability of the causal vaccine effect under various selection bias assumptions.

Main Methods:

• Developed a Bayesian framework for causal inference in vaccination studies.
• Investigated identifiability of the vaccine efficacy estimand (VEI) under different selection bias scenarios.
• Compared Bayesian and maximum likelihood methods via simulations and real-world case studies.

Main Results:

• The Bayesian approach yielded similar inferences to frequentist methods in case studies.
• Simulation studies indicated that the Bayesian method offers reduced bias and shorter confidence intervals, particularly with small sample sizes.
• Identifiability of the causal vaccine effect was analyzed under varying assumptions.

Conclusions:

• The proposed Bayesian method provides a robust alternative for estimating causal vaccine effects on post-infection outcomes.
• The Bayesian approach demonstrates improved performance in bias reduction and precision for smaller datasets.
• This work extends causal inference methods in vaccine effectiveness research.