OBJECTIVES: Recently, value of information (VOI) analysis is more popular in health economics to investigate the added value of conducting future research to reduce parameter uncertainty in decision-analytic models. Such analysis generally calculate a bounded value (i.e., expected value of (partial) perfect information (EV(P)PI)) in a compute intensive way. We applied in this study an efficient approach to calculate EV(P)PI based on a previously conducted model-based economic evaluation on heart failure disease management.

METHODS: The comparators considered were a conventional nurse-led management program and an alternative strategy using a novel point-of-care testing device. A continuous-time three-state Markov model was developed to describe the disease progression with the uncertainty of the transition rate of the novel device arm being captured probabilistically from two cardiologists. The Markov model was simulated in five years. Generalized additive model (GAM) regression was used to model the incremental net monetary benefit (averaging between cardiologists) with the experts’ elicited transition rate. The EV(P)PI was then calculated based on the predicted value of the GAM. The base-case EV(P)PI was calculated with an €20,000 willingness-to-pay (WTP) threshold. The elicited transition rates were assumed to be independent in base case. We subsequently altered the WTP threshold and relaxed the assumption of independence to see how the results were influenced.

RESULTS: In base case, the EVPI per patient was €401 and the EVPPI for the elicited rate from the second cardiologist was €340. The EVPI value ranges between €337 and €1131 when altering the WTP threshold from €10,000 to €80,000. Both values hardly changed when the independent assumption was relaxed.

CONCLUSIONS: GAM can be successfully applied in our case to efficiently calculate EV(P)PI. There is a profound difference between the two cardiologists regarding the added value of conducting future research to reduce parameter uncertainty existed in the current decision-analytic model.