voi-package {voi}

Methods to calculate the Expected Value of Information

Description

evppi calculates the expected value of partial perfect information from a decision-analytic model. The default, recommended computation methods are based on nonparametric regression. evpi is also provided for the expected value of perfect information.

evsi calculates the expected value of sample information. Currently this implements the same set of nonparametric regression methods as in evppi, and methods based on moment matching and importance sampling. enbs can then be used to calculate and optimise the expected net benefit of sampling for a simple study with a fixed upfront cost and per-participant costs.

evppi and evsi both require a sample of inputs and outputs from a Monte Carlo probabilistic analysis of a decision-analytic model.

Analogous functions evppivar and evsivar calculate the EVPPI and EVSI for models used for estimation rather than decision-making. The value of information is measured by expected reductions in variance of an uncertain model output of interest.

A pure "brute-force" Monte Carlo method for EVPPI calculation is provided in evppi_mc, though this is usually computationally impractical.

The package overview / Get Started vignette gives worked examples of the use of all of these functions.

References

Heath, A., Manolopoulou, I., & Baio, G. (2017). A review of methods for analysis of the expected value of information. Medical Decision Making, 37(7), 747-758.

Heath, A., Kunst, N., Jackson, C., Strong, M., Alarid-Escudero, F., Goldhaber-Fiebert, J. D., Baio, G. Menzies, N.A, Jalal, H. (2020). Calculating the Expected Value of Sample Information in Practice: Considerations from 3 Case Studies. Medical Decision Making, 40(3), 314-326.

Kunst, N., Wilson, E. C., Glynn, D., Alarid-Escudero, F., Baio, G., Brennan, A., Fairley, M., Glynn, D., Goldhaber-Fiebert, J. D., Jackson, C., Jalal, H., Menzies, N. A., Strong, M., Thom, H., Heath, A. (2020). Computing the Expected Value of Sample Information Efficiently: Practical Guidance and Recommendations for Four Model-Based Methods. Value in Health, 3(6), 734-742.