Estimating expected value of sample information for incomplete data models using Bayesian approximation.

Area of Science:

• Decision Analysis
• Bayesian Statistics
• Health Economics

Background:

• Expected value of sample information (EVSI) calculation is crucial for decision-making under uncertainty.
• Bayesian updating in models with incomplete data often relies on computationally intensive Markov chain Monte Carlo (MCMC) methods.
• Existing EVSI computation methods can be inefficient for complex health economic models.

Purpose of the Study:

• To develop and present a revised Bayesian Laplace approximation for EVSI computation.
• To support decision-making processes in the presence of incomplete data models.
• To improve the computational efficiency and accuracy of EVSI estimation.

Main Methods:

• Developed a Bayesian Laplace approximation for EVSI, including mathematical formulations for likelihood and log posterior density.
• Compared the accuracy of the proposed approximation (first- and second-order) against traditional Monte Carlo methods.
• Evaluated performance in a cost-effectiveness model with incomplete data.

Main Results:

• The revised approximation offers a new approach for EVSI computation.
• Accuracy of EVSI estimates was compared between the approximation methods and traditional Monte Carlo.
• Computational efficiency is influenced by model complexity, sample size, and MCMC requirements.

Conclusions:

• The proposed Bayesian Laplace approximation provides a valuable and potentially more efficient method for EVSI computation in health economic decision models.
• This methodology has potential applications beyond health economics in fields requiring Bayesian approximation.
• The approach facilitates better-informed decisions by improving the estimation of the value of additional information.