Applying causal inference and Bayesian statistics to understanding vaccine safety signals using a simulation study

Area of Science:

• Pharmacovigilance and Vaccine Safety
• Biostatistics and Epidemiological Methods
• Public Health and Community Medicine

Background:

• Community perception of vaccine safety is a critical determinant of vaccine uptake.
• Effective vaccine safety monitoring relies on accurate signal detection from post-vaccination data.
• Understanding factors influencing data availability is crucial for refining monitoring systems.

Purpose of the Study:

• To assess current vaccine safety monitoring systems.
• To examine factors influencing post-vaccination survey data availability.
• To evaluate the impact on signal detection specificity and sensitivity.

Main Methods:

• Utilized causal directed acyclic graphs (DAGs) to model relationships between factors.
• Employed Bayesian posterior predictive analysis (PPA) for signal detection.
• Simulated data based on varied biological and behavioral factors.

Main Results:

• Biological factors (adverse reaction severity) and behavioral factors (healthcare-seeking behavior) significantly influence survey participation and signal detection.
• False signals were detected when reaction severity strongly influenced both survey participation and medical attention seeking, especially with low prevalence of moderate to severe reactions.
• These findings highlight potential biases in vaccine safety signal detection.

Conclusions:

• Vaccine safety monitoring systems are susceptible to biases introduced by biological and behavioral factors.
• Future monitoring strategies should account for the influence of reaction severity and healthcare-seeking behavior on data collection.
• Refined methods are needed to improve the accuracy and reliability of vaccine safety signal detection.