Calculating the Expected Value of Sample Information Using Efficient Nested Monte Carlo: A Tutorial.

Area of Science:

• Health Economics
• Decision Science
• Computational Statistics

Background:

• Expected value of sample information (EVSI) quantifies the economic benefit of reducing uncertainty in health economic models.
• Practical EVSI evaluations are often limited by the high computational cost of traditional nested simulation methods.
• Efficient EVSI estimation can improve research budget allocation.

Purpose of the Study:

• To present the practical application and implementation of a recently developed, computationally efficient EVSI estimation method.
• To illustrate the key steps of the EVSI estimation procedure using a worked example.
• To discuss the optimal implementation of this method in a practical health economic model.

Main Methods:

• A worked example based on a three-parameter linear health economic model.
• A Markov model structure to evaluate the cost-effectiveness of a new chemotherapy treatment and its side effects.
• Application of a novel estimation procedure to calculate EVSI, reducing simulation requirements.

Main Results:

• The new EVSI estimation method provides accurate results within feasible computation times (seconds vs. days).
• The method is effective even for complex health economic model structures.
• Increased number of nested samples improves EVSI accuracy, even at a fixed computational cost.

Conclusions:

• The novel method substantially reduces the computational cost of estimating EVSI using nested simulation.
• This advancement facilitates practical EVSI evaluations, aiding informed decision-making in health economics.
• The method offers a more accessible approach to quantifying the value of additional information in research.