Estimating the Expected Value of Sample Information Using the Probabilistic Sensitivity Analysis Sample: A Fast,...

Area of Science:

• Health Economics
• Decision Analysis
• Biostatistics

Background:

• Health economic models inform decisions by estimating net benefits.
• Model input parameters often have uncertainty, necessitating methods to quantify the value of reducing this uncertainty.
• Expected value of sample information (EVSI) quantifies the value of new data for research design.

Purpose of the Study:

• To introduce a computationally efficient, nonparametric regression-based method for estimating per-patient EVSI.
• To provide an alternative to the computationally intensive 2-level Monte Carlo procedure for EVSI estimation.
• To demonstrate the superior efficiency of the proposed regression method compared to the traditional Monte Carlo approach.

Main Methods:

• A novel nonparametric regression technique is employed to estimate EVSI.
• The method utilizes probabilistic sensitivity analysis samples, avoiding direct posterior distribution sampling.
• It bypasses the need to rerun complex decision models for each data set.

Main Results:

• The regression-based method significantly improves computational efficiency for EVSI estimation.
• It successfully estimates EVSI without requiring sampling from posterior parameter distributions.
• A case study confirmed the superior performance of the regression method over the 2-level Monte Carlo procedure.

Conclusions:

• The developed nonparametric regression method offers a fast and efficient approach to estimating EVSI in health economic modeling.
• This method simplifies the quantification of the value of information, aiding research prioritization.
• The technique is broadly applicable to various model complexities and parameter distributions.