SHOULD WE CONSTANTLY BE COUNTING WITH CONSTANT DISCOUNTING?
Author(s)
Redig J1, Klint J2, Åkerborg Ö2
1ICON Plc, Stockholm, AB, Sweden, 2ICON Plc, Stockholm, Sweden
Objectives: In economic evaluations, costs and effects are recommended to be discounted in order to reflect the loss-in-value over time based on preference observations. Guidelines most often recommend use of a social discount rate, constant over time. There are, however, situations where non-constant discount rates are recommended. In HAS’ reference case, a constant discount rate is recommended. For economic evaluations with long time horizons, HAS recommends replacing the time-consistent model with a time-inconsistent, arguing that discounting follows a hyperbolic function with a relatively larger discount rate closer in time. The hyperbolic function is, however, difficult to translate intuitively to discount rates. Therefore, a discount rate of 4% for the first 30 years and 2% thereafter is suggested; typically implemented with the Lebègue formula. Methods: A tool was implemented contrasting constant discount rate of 4%, and time-dependent step function implementation from 4 to 2% after 30 years to the hyperbolic discounting function. In addition, a linearly changing discount rate from 4 to 2% over 30 years was explored as a novel option. Pairwise comparisons among the approaches were performed using squared error estimation. Results: Over different time-horizons (5, 10, 30 and 100 years) the different discounting functions were compared, showing that the step-function performed better in estimating a hyperbolic function compared to the constant discount rate. When implementing a linear change in the discount rate, an even better fit was obtained. Errors at year 100 amount to 2.68, 1.26 and 0.364, respectively. Conclusion: The discrete step function gives discount factors similar to the desired time-dependent hyperbolic function. However, a linear and also continuous implementation results in an intuitive and more accurate representation of the hyperbolic shape of the discount factors. The large impact of discounting methodology - relative to other model parameters – motivates careful handling as well as consistency across jurisdictions.
Conference/Value in Health Info
2019-11, ISPOR Europe 2019, Copenhagen, Denmark
Code
PNS37
Topic
Economic Evaluation, Health Policy & Regulatory, Health Technology Assessment, Methodological & Statistical Research
Topic Subcategory
Decision & Deliberative Processes, Modeling and simulation, Public Spending & National Health Expenditures
Disease
No Specific Disease
Explore Related HEOR by Topic