Multilevel and Quasi Monte Carlo methods for EVPPI
This method helps estimate how much value there would be in collecting more information about certain uncertain inputs in health technology assessment models, especially when those inputs are complex and correlated. It does so more efficiently than traditional simulation methods.
At a glance
Use when
Estimating EVPPI in models with many correlated uncertain parameters, especially when inputs come from MCMC; when regression approximations are infeasible or biased
Avoid when
When computational resources are not a constraint and parameter sets are small; when simpler methods (e.g., standard Monte Carlo) are sufficient
Inputs
Cost-effectiveness model structure, probabilistic input parameters (especially MCMC-sampled), sets of parameters of interest for EVPPI calculation
Outputs
Estimate of the expected value of partial perfect information (EVPPI) with quantified precision and reduced variance
How it works
The method applies quasi Monte Carlo (QMC) and multilevel Monte Carlo (MLMC) techniques to estimate the expected value of partial perfect information (EVPPI) in cost-effectiveness models. It reduces computational cost by lowering variance compared to standard Monte Carlo methods, particularly when input parameters are numerous, correlated, and derived from Markov chain Monte Carlo (MCMC) sampling. The approach preserves accuracy and allows control over bias and precision.
- HTA domains
- Costs & Economic Evaluation
- Categories
- Cost-effectiveness Modelling
- Assumptions
- The underlying cost-effectiveness model is well-specified; input parameters are characterized with sufficient accuracy (e.g., via MCMC); the EVPPI is defined over a fixed subset of parameters
- Strengths
- Reduces computational cost significantly for large and correlated parameter sets; maintains control over bias and precision; applicable to complex models with MCMC-derived inputs; outperforms regression approximations in scalability and accuracy in high-dimensional settings
- Limitations
- Implementation complexity is higher than standard Monte Carlo; performance gains depend on problem structure (e.g., diminishing returns for small parameter sets); requires careful integration with MCMC sampling workflows
- Also known as
- QMC for EVPPI, MLMC for EVPPI, Multilevel Monte Carlo EVPPI, Quasi Monte Carlo EVPPI
Questions this answers
- › What is the value of reducing uncertainty in a subset of parameters in a cost-effectiveness model?
- › How can we efficiently compute EVPPI when parameters are correlated and estimated via MCMC?
- › Can QMC or MLMC reduce computational burden compared to standard Monte Carlo for EVPPI?
- › How do regression-based approximations (e.g., GP, GAM, INLA-GP) compare in accuracy and scalability to QMC and MLMC?
- › When is it more efficient to use QMC or MLMC over traditional Monte Carlo methods for EVPPI?
- › Can these methods handle large sets of correlated parameters in complex models like Markov structures?
References & sources
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.