Handling multiple time-varying exposures in survival analysis using real-world pediatric data from the pedianet...

Area of Science:

• Biostatistics
• Epidemiology
• Pediatric Research

Background:

• Survival analysis traditionally focuses on single time-varying exposures (TVEs).
• Handling multiple TVEs simultaneously presents statistical challenges and is an active research area.
• Real-world data applications are crucial for validating statistical approaches.

Purpose of the Study:

• To apply and compare different statistical models for multiple time-varying exposures.
• To investigate the association between antibiotic use, influenza vaccination, and influenza/influenza-like illness (ILI) onset in children.
• To assess the impact of statistical modeling choices on parameter estimates.

Main Methods:

• Utilized the Italian national pediatric database (Pedianet) for children aged 6 months to 14 years during the 2017-2018 influenza season.
• Modeled influenza vaccine administrations and antibiotic prescriptions using time-fixed and time-varying approaches.
• Employed Cox proportional-hazard models with random intercepts to analyze the association between exposures and ILI onset.

Main Results:

• Estimates for influenza vaccination remained stable across different modeling approaches.
• Estimates for antibiotic use showed significant variation depending on the statistical model employed.
• The choice of statistical handling significantly impacts the interpretation of results for antibiotic exposure.

Conclusions:

• Careful evaluation of exposure characteristics is crucial when dealing with multiple TVEs.
• Statistical methods specifically designed for multiple TVEs are necessary for accurate analysis.
• The findings underscore the need for robust statistical techniques in pediatric epidemiological studies involving complex exposures.