"An Efficient Gaussian Approximation and Regression Metamodeling Approach to Value of Information Analysis"
Please note: All research in progress seminars are off-the-record. Any information about methodology and/or results are embargoed until publication.
Value of information (VOI) analysis is based on statistical decision theory, and has recently gained increased recognition for its potential application in resource allocation, research prioritization, and future data collection designs. However, VOI remains underutilized due to many conceptual, mathematical and computational challenges of implementing Bayesian decision theoretic approaches in models of sufficient complexity for real-world decision making. In this study, I propose a practical approach for conducting VOI analysis using a combination of probabilistic sensitivity analysis, linear regression metamodeling, and Gaussian approximation as an efficient replacement for traditional Bayesian updating. I leverage the Central Limit Theorem to simplify the Bayesian updating process. I illustrate my approach using a previously published cost-effectiveness analysis of several treatment strategies of gouty arthritis. I present several measures of VOI, including the expected value of sample information for various sample sizes and the optimal sample size that maximizes the benefit of research while addressing many of the challenges of traditional VOI analyses.