Estimation of causal effects with repeatedly measured outcomes in a Bayesian framework

Area of Science:

• Biostatistics
• Epidemiology
• Causal Inference

Background:

• Longitudinal observational studies present complex data structures due to time-dependent treatment assignments.
• Existing Bayesian causal methods have limitations in handling longitudinal data, particularly with repeatedly measured outcomes.

Purpose of the Study:

• To extend Bayesian causal inference methods for analyzing longitudinal data with repeatedly measured outcomes.
• To enable causal estimation of treatment effects at each clinical visit in observational studies.

Main Methods:

• Extended Bayesian estimation of marginal structural models.
• Adapted two-stage Bayesian propensity score analysis for longitudinal data.
• Utilized time-dependent inverse probability of treatment weights derived from Markov chain Monte Carlo samples.

Main Results:

• Developed and validated novel Bayesian approaches for causal inference with repeatedly measured longitudinal outcomes.
• Demonstrated the utility of the proposed methods through a simulation study.
• Applied the methods to analyze intravenous immunoglobulin therapy for juvenile dermatomyositis.

Conclusions:

• The proposed extended Bayesian methods effectively handle complex longitudinal data for causal effect estimation.
• These methods offer a valuable tool for researchers studying time-dependent treatments and outcomes in observational settings.
• The approaches are applicable to various clinical research scenarios involving longitudinal data analysis.