PCV59 Where Did the Time Go?: Temporal Uncertainty in Cost-Effectiveness Decision Models

Abstract

WHERE DID THE TIME GO?: TEMPORAL UNCERTAINTY IN COSTEFFECTIVENESS DECISION MODELS Mahon R1, Manca A2, Palmer S1 1University of York, York, North Yorkshire, UK, 2University of York, Heslington, York, UK OBJECTIVES: Since the required time horizon in a cost effectiveness decision model often exceeds the evidence time horizon, numerous temporal uncertainties arise regarding model parameters and structures. The objective of this study is to demonstrate, through a motivating example: (i) why temporal uncertainty ought to be addressed more thoroughly than it has been to date; (ii) how this uncertainty might be expressed in decision models; and (iii) the consequences for the costeffectiveness results when temporal uncertainty is incorporated into the analysis. METHODS: Taking the example of a decision model seeking to estimate the costeffectiveness of an early interventional strategy for patients with non-ST-elevation acute coronary syndrome, we firstly highlight the model components that are exposed to temporal uncertainty. Focusing on two key model parameters, we explore the extent to which the existing short-term evidence could reasonably be extrapolated over time. We then suggest a means to quantitatively convey the temporal uncertainty pertaining to these parameters within the model. RESULTS: Temporal uncertainty is shown to have a significant impact on the cost-effectiveness results. Value-of-Information analysis (specifically population EVPPI) suggests that for this example, it may have been more cost-effective to delay adoption recommendation until further evidence on the temporal behaviour of parameters was collected. CONCLUSIONS: Temporal uncertainty, though rarely formally modelled, is a significant characteristic of cost-effectiveness decision models. It is possible and desirable to express temporal uncertainty within a decision model, as the complete model may show that it is more cost-effective to collect further information on the temporal behaviour of model parameters before issuing an adoption recommendation.