PRM92 - EVPI Curves in Practice

Abstract

When making decisions about the allocation of scarce health-care resources it’s not only important to consider the estimated cost-effectiveness (CE) with current evidence, but also the value of additional research. Expected value of perfect information (EVPI) is the amount a decision maker should be willing to pay to eliminate uncertainty surrounding the decision about which option is optimal for different levels of the CE threshold (λ). Although EVPI analysis is being requested increasingly by reimbursement agencies there is still limited literature on the interpretation of EVPI curves. The typical ‘textbook’ example represents just one of the possible shapes that the curve can take. To explore and explain different shapes of EVPI curves based on the position of alternative treatment choices on the incremental cost-effectiveness plane. A hypothetical probabilistic decision model was developed in which two interventions were compared. Key input parameters were varied to force the model outcomes into different quadrants on the incremental CE plane. The population EVPI, based on a hypothetical number of future patients and the estimated lifetime of the new intervention, was plotted. The result of this study demonstrates a number of scenarios where the EVPI curve takes a different form compared to the one illustrated in the typical ‘textbook’ example. For example, when the majority of the plotted outcomes are spread over the northern quadrants the traditional EVPI peak is absent, and this could be explained by the fact that the reduction in decision uncertainty does not outweigh the increased value of opportunity loss. Further, plots spread over the eastern quadrants present a maximum EVPI at zero λ which then gradually decreases. This study may inform the interpretation of EVPI curves, and add value to the analysis of the value of additional research.