INLA-SPDE EVPPI method (Heath, Manolopoulou, Baio)
This method helps estimate how much value could be gained by reducing uncertainty in specific parts of a health economic model, using advanced statistical techniques to make calculations faster and more efficient.
At a glance
Use when
You need to compute EVPPI efficiently in high-dimensional models where Monte Carlo methods are too slow
Avoid when
The model has highly non-Gaussian or discontinuous input-output relationships that GP regression cannot capture well
Inputs
Health economic model parameters and their uncertainty distributions, probabilistic sensitivity analysis samples
Outputs
Expected Value of Partial Perfect Information (EVPPI) estimates for individual or subsets of parameters
How it works
The method estimates the Expected Value of Partial Perfect Information (EVPPI) in health economic evaluations by combining non-parametric regression with Integrated Nested Laplace Approximation (INLA) and the Stochastic Partial Differential Equation (SPDE) approach. It projects high-dimensional Gaussian Process regression problems into a lower-dimensional space, enabling fast and accurate approximation of EVPPI without requiring extensive Monte Carlo simulations.
- HTA domains
- Costs & Economic Evaluation
- Assumptions
- The relationship between model inputs and outcomes can be approximated using a Gaussian Process; the uncertainty structure is well-characterized probabilistically; INLA-SPDE projection preserves relevant variance information
- Strengths
- Dramatically reduces computation time compared to Monte Carlo methods; enables practical EVPPI calculation in complex models; implemented in accessible R package (BCEA)
- Limitations
- Relies on accurate specification of the GP structure; performance depends on dimensionality and correlation structure of inputs; requires familiarity with Bayesian spatial statistics methods
- Also known as
- INLA-SPDE EVPPI, EVPPI via INLA-SPDE
Questions this answers
References & sources
Similar by meaning
Beta record. Based on the original catalogue summary; primary-source enrichment pending.