Accumulating evidence for the case of differential discounting

Keywords::

• discounting
• health gains
• time preference

Discounting of monetary and health outcomes is nowadays common in health economics and specified in all country-specific guidelines on ‘good health-economic practice’ [1,101]. It is the specific approach taken and the exact rates utilized that differ among countries. In particular, regarding the differing approaches, much debate has been spurred concerning whether to adopt equal or differential discounting. Equal discounting is mostly used in practice and guidelines sometimes refer to pragmatic motivations, including consistency arguments or the ease to use one and the same rate for money and health. Differential discounting involves the application of a generally lower discount rate for health outcomes – for example, life years – motivated by ethical arguments on valuing life years for future generations or the potential increase in the monetary value of health-over-time related to differing growth rates in the economy and life expectancy [2–4].

More specifically, equal discounting is motivated by the main argument that the use of differential discounting would lead to infinite postponement of healthcare programs [2]. This argument has recently been illustrated by O’Mahony et al. by the inclusion of additional cohorts in a cost–effectiveness analysis for HPV vaccination [5]. In particular, they showed that the cost–effectiveness of subsequent cohorts becomes more favorable if differential discounting is applied. How to exactly judge and whether this is actually a problem is again topic for discussion. One could argue that this decline in cost–effectiveness can be justified exactly owing to this increasing value of health [6].

Currently, only two countries prescribe the use of differential discounting in their guidelines: The Netherlands at 4% for money and 1.5% for health effects and Belgium at 3.5 and 1.5%, respectively [101]. The methodology underlying differential discounting has first been specified by Gravelle and Smith in the UK that, not surprisingly, is the first country to embark on differential discounting [7]. However, the UK reversed this decision to discount differentially back to equal discounting approximately half a decade ago [8–10]. Of note, recently the Appraisal Committee of NICE recommended applying differential discounting (1.5% health effects and 3.5% costs) for long-term benefits in sensitivity analysis [102]. We previously strongly argued against this reversal, and The Netherlands changed to differential discounting based on this argumentation [8].

Next to an overall lower discount rate for health, an initially similar but decreasing discount rate in time has been proposed as well. This illustrates the general uncomfortable feeling with equal discounting of money and life years. Notably, it is well known from empirical research on time preference that this preference is relatively stronger for the near future than for the distant future [11]. Two approaches have focused on a methodology combining a discount rate for health that is essentially similar to that for money in the short term and a relatively low discount rate farther into the future [12–15]. Building on these methods, other authors have extended the approach from specifying two discount rates for the near and distant futures to a trajectory of discount rates for multiple periods into the future. For example, more than two differentiating periods with decreasing discount rates for health have been proposed by the UK Treasury [103], with the discount rate decreasing from an initial 3.5 to 3.0% after 30 years, on to 2.5% after 75 years, and so on. Obviously, the approach of decreasing discounting for health often results in a net lower discount rate for health than for money.

Discounting certainly has a rich history. Fundamental to the approach is Ramsey’s formula already specified in 1928 [16], explaining the discount rate for money (r_m) into its constituting elements:

with ρ reflecting the rate of pure time preference, ϵ_m the elasticity of marginal utility of money and g_m the growth in the amount of money (in particular, economic growth). A similar formula can be designed for the discount rate of health, for example, concerning life years (r_l, ϵ_l, g_l). Both formulas, having only the pure time preference component in common, imply similar discount rates, which is obviously highly unlikely, again illustrating the need for differential discounting from a purely mathematical point of view since it is not always the case that ε_m = ε_l or g_m = g_l. With the growth in life expectancy lagging far behind economic growth, a lower r_l than r_m is plausible, rather than the other way around. Notably, the pure time preference component has been further subdivided into parts reflecting impatience (wanting to have as soon as possible) and uncertainty (potential occurrence of a catastrophic event avoiding us to reap any future benefits). For example, the recently recommended 3.5% discount rate in the UK [101] may be considered to comprise 0.5% for impatience, 1% for catastrophic risk and 2% for the component reflecting elasticity and economic growth.

From a more philosophical point of view, the specific nature of health gains as life years may give rise to specific considerations. This specific nature refers to indivisibility of a life year within a sequence of life years gained and the dependence of life years within the sequence. In particular, one cannot live an individual life year without living the previous one, that is, the subsequent life year is dependent on the previously lived ones and as such life lived is indivisible (one cannot live half of the previous life year, for example). Two inferences can be drawn from this. First, it might be questioned whether the pure time preference component in the mathematical approach would indeed be exactly similar for both money and life years. In particular, since discounting future lives implicitly assumes their existence, rationality may dictate that pure time preference should be omitted or at least exclude the catastrophic risk component. Second, if life years are dependent and indivisible, it would not be adequate to discount a sequence of life years gained in one individual year one by one, rather, one aught to discount the whole composite of accumulated years once from the start of the sequence back to the moment of implementation of the intervention being investigated. An overall lower discount rate for life years would validly address the first aspect, whereas the second once again motivates the use of a decreasing or even zero discount rate within the sequence of dependent indivisible life years.

We have outlined various motivations and illustrations for differential discounting of money and health. Given the predominant current stream in the guidelines of equal discounting, we would argue that it is time for change. From ethical, economic, mathematical and philosophical points of view, differential discounting is strongly supported and guidelines should start reflecting this better. We would argue that, rather than by sound evidence and reasoning, equal discounting is primarily supported by pragmatic and historic motivations. Fortunately, the expertise in the area of discounting in health economics is rapidly accumulating, involving thorough discussions, complex analyses and strengthening alignment with empirical situations. This expertise now increasingly comprises the specificities in discounting, enhancing fair and valid discount rates and albeit fair and valid assessments of the costs and benefits with different timings for both public health and pharmaceutical strategies. Our arguments have been derived from analysis of the health outcome of life years. Notably, for quality of life, other arguments might exist, ultimately even warranting different discount rates for life years and quality of life. This might illustrate the need to go from a 2D analysis to a 3D one including money, life years and quality separately, designing a framework justifying the specificities of all these three elements and in the end estimate the respective discount rates separately and differentially.

Financial & competing interests disclosure

MJ Postma and M Parouty are supported by an unrestricted grant from GlaxoSmithKline Biologicals (Wavre, Belgium) on the topic of discounting. TA Westra is an employee of GlaxoSmithKline (Zeist, The Netherlands), next to his guest researchship at the unit PE2. MJ Postma received grants, honoraria and stipends from various pharmaceutical companies, including some interested in the subject matter. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.