Illustrating uncertainty in extrapolating evidence for cost-effectiveness modelling

Abstract: To determine cost-effectiveness, an appropriate time horizon must be chosen. In many circumstances, a long term or lifetime horizon is appropriate, particularly when there are mortality effects. Typically the evidence base used to inform cost-effectiveness falls short of this time horizon. Therefore, some form of modelling will need to be employed to extrapolate short term outcomes over the lifetime of the model. This relates to utilities, resource use/costs as well as effectiveness parameters. This extrapolation will inevitably involve assumptions and will be data sparse. This has implications for the decisions that are made on the basis of this extrapolation and the degree of uncertainty surrounding these decisions. Currently there is little guidance on how uncertainty in extrapolation should be represented. This presentation discusses the approaches used to extrapolate all forms of data included in cost-effectiveness models and the limitations of these approaches. Methods to present the assessment of validation and structural uncertainty are considered and case studies used to demonstrate alternative methods of extrapolation.