Value of Information in Health Technology Regulation and Reimbursement

Abstract

The objective of medical research from a societal or health-care perspective should be to improve the health of a population. In the preceding chapters it is apparent that we need to think about more than simply the effectiveness of the intervention undergoing evaluation if we are to meet this objective. Research is expensive and subjects patients to the risk of experimentation. Both of these factors, set in the context of uncertainty, have the potential to incur opportunity costs within a population. This concluding chapter builds on Chap. 12 setting the context for the use of Value of Information in research prioritisation and research design.

Appendix 13.1: General Monte Carlo Sampling Algorithm for Calculation of Population ENPVSI

Adapted from Ades et al. ( 2004 )

θ _I  = parameters of interest (here assumed independent of θ _I ^c)

First record the net benefit of an optimal decision based on current information. Then define a proposed piece of research from which data X _θI will be collected to inform θ _I.

• A1. For i = 1,2… N simulations

  • B1. Draw a sample θ _I ⁽ⁱ⁾ from the prior (baseline) distribution of θ _I .

  • B2. Draw a sample X _θI ⁽ⁱ⁾ from the distribution of the sufficient statistic X _θI |θ _I ⁽ⁱ⁾ arising from a new study of defined size.

  • B3. Calculate posterior (updated) expected net benefits for each strategy j, using an inner Monte Carlo simulation loop using the posterior distribution θ _I ⁽ⁱ⁾|X _θI ⁽ⁱ⁾.

  • B4. Calculate expected net benefits for each strategy j given the likelihood X _θI ⁽ⁱ⁾, evaluated at its mean, using an inner Monte Carlo simulation loop.

  • B5. Find the strategy j maximising expected net benefit for simulation i based on B3.

  • B6. Draw a sample from the distribution of time to trial reporting (τ) using X _θI ⁽ⁱ⁾._.

  • B7. Using the expected net benefit given the mean of the likelihood X _θI ⁽ⁱ⁾ (B4.), allocate net benefit to patients allocated to trial arms for each strategy j for each time interval up to τ, discounted.

  • B8. Using the posterior expected net benefits (B3.), record the population net benefit for patients not in trial for time intervals prior to time τ who receive the optimal strategy j given a decision based on the prior expected net benefits up to time τ, discounted.

  • B9.  Record the population net benefit for the optimal strategy j given a decision based on the posterior expected net benefits using the discounted population for each time interval after the trial has reported.

  • B10. Record the sum of the expected net benefits over all groups in B7, B8 and B9.

• A2. Find the average of the population expected net benefits (B10), over the N simulations. This is the population expected value of a decision based on sample information.

• A3. Subtract from this the population expected value of a decision based on current information to give the ENPVSI.