Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 360 - Contributed Poster Presentations: Section on Bayesian Statistical Science
Type: Contributed
Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #312862
Title: Analysis of Ongoing Intervention Adaptations: The Intersection of Platform Trials, Causal Inference, and Bayesian Techniques
Author(s): Teresa Dianne Bufford* and Hilary Aralis and Catherine Crespi
Companies: UCLA Department of Biostatistics, Semel Institute for Neuroscience & Human Behavior and UCLA Department of Biostatistics, Semel Institute for Neuroscience & Human Behavior and UCLA Department of Biostatistics, Jonsson Comprehensive Cancer Center
Keywords: intervention; implementation science; platform trial; Bayesian; real-world evidence; causal inference
Abstract:

An intervention deemed effective in an initial randomized trial is often broadly implemented in real-world settings where adaptations naturally arise. Prevention scientists have recognized the need for ongoing evaluation of intervention adaptations. Existing statistical methods fall short when using continuously-generated real-world evidence to compare concurrent intervention versions. If we make a simplifying assumption that individuals are randomly assigned to an intervention version, our situation resembles a platform clinical trial with no established end point. Using simulations, we have developed a Bayesian analysis framework for interim decision making throughout the endless platform trial. Since type I error rate and power are not well defined without an end point, we use new metrics to evaluate the analysis framework. As the assumption of randomization is relaxed, we need to account for covariate imbalances across subpopulations which may affect treatment effectiveness. Causal inference methods for observational data can be extended and combined with our Bayesian framework to quantitatively determine which intervention version is most beneficial for each subpopulation.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program