Need for Speed single-parameter EVPPI algorithm (Sadatsafavi)
A fast method to calculate the value of information for one parameter in health economic models using existing probabilistic sensitivity analysis results, avoiding slow nested simulations.
At a glance
Use when
Rapidly computing the value of information for individual parameters in health economic models using existing PSA results.
Avoid when
Estimating multi-parameter EVPPI or when PSA samples are too sparse or poorly converged.
Inputs
Probabilistic sensitivity analysis (PSA) sample, including sampled parameter values and corresponding model outcomes (e.g., net benefit, cost-effectiveness ratios).
Outputs
Exact expected value of partial perfect information (EVPPI) for a single parameter.
How it works
An efficient sorting-based algorithm that computes the exact expected value of partial perfect information (EVPPI) for a single parameter directly from a probabilistic sensitivity analysis (PSA) sample. It achieves significant speed improvements—orders of magnitude faster—over traditional nested Monte Carlo simulation methods by leveraging sorting and numerical integration techniques.
- HTA domains
- Costs & Economic Evaluation
- Assumptions
- The PSA sample is sufficiently large and representative of the joint parameter distribution; the parameter of interest is scalar and independent or analytically separable.
- Strengths
- Dramatically faster than nested Monte Carlo methods; provides exact EVPPI for single parameters; leverages existing PSA outputs without requiring model re-evaluation.
- Limitations
- Limited to single-parameter EVPPI; does not scale directly to multi-parameter EVPPI without additional approximations; relies on adequate PSA sample size and quality.
- Also known as
- Need for Speed algorithm, Sadatsafavi's EVPPI algorithm, Single-parameter EVPPI sorting method
Questions this answers
- › What is the value of reducing uncertainty in a specific parameter of a health economic model?
- › Which parameter in the model would benefit most from further research?
- › How much decision uncertainty is attributable to a single parameter?
- › Can we prioritize parameters for future data collection based on their impact on decision uncertainty?
- › Is it worth conducting further research on a particular input parameter?
- › How can EVPPI be computed efficiently without nested simulations?
References & sources
Similar by meaning
Beta record. Based on the original catalogue summary; primary-source enrichment pending.