Discount rates and drug value: A Q&A with Josh Cohen
A RApport Q&A
June 12, 2024
It’s not often we entertain a recurring guest here at Rapport. In fact, this might be a first.
Joshua T. Cohen, PhD, is Deputy Director of the Center for the Evaluation of Value and Risk in Health (CEVR) at the Institute for Clinical Research and Health Policy Studies at Tufts Medical Center, and Research Professor of Medicine at Tufts University School of Medicine. His research focuses on value frameworks, including proper accounting for pharmaceutical life cycle pricing and identification of an appropriate discount rate for health technology assessments.
And now he’s going to be able to add two-time Rapport Q&A participant to his CV, which is undoubtedly a high honor.
We previously interviewed Josh in 2021, to talk about The Right Price: A Value-Based Prescription for Drug Prices, the book he authored with colleagues Peter Neumann and Daniel Ollendorf. When we’re using math to determine what drugs are worth to payers and society, “We should be trying to get the math as right as we can,” Josh reminded us back then.
This time we narrowed the scope and talked to Josh about his research on discount rates and why getting that particular aspect of the math right matters – the subject of his recent review in Value in Health.
Will there be actual math this time around? Yes, there might be a smidge of math, but it’s nothing Rapport readers can’t handle. Just enough to keep you on your toes. “One commentator …described my own review of the discount rate as ‘useful’ although ‘tremendously dull,’” he says. We beg to differ (on the second part, in case that wasn’t obvious).
RApport: Let’s start with the basics. What is a discount rate? Why should we non-economists care about it?
JC: How much something matters to people – whether it’s good or bad – depends on when it happens: all else being equal, people care more about outcomes that happen in the near future than about outcomes that happen later. The discount rate represents how much timing matters. A large discount rate implies that people “discount” the importance of more distant events a lot and hence those future events are relatively unimportant. A small discount rate implies that more distant events remain relatively important.
RApport: Can you give an intuitive example of how everyday folks might use discount rate reasoning?
JC: The discount rate often comes up in the context of borrowing money. A person who takes out a loan for a large purchase – for example, a car or a house – must pay back the original loan plus interest. Implicitly, this willingness means that the borrower places at least as much value on the sum of money they borrow and receive upfront as they place on the greater sum of money they end up paying back in the future (otherwise, they would not willingly take out the loan). For example, say a person borrows $10,000 (e.g., to buy a car) and agrees to pay back the $10,000 plus $2,000 in interest, the implication is that they value the $10,000 received upfront at least as much as they value the $12,000 they pay back in the future. That is, $10,000 now is worth at least as much as $12,000 in the future. In other words, a dollar upfront (of which there are ten thousand in this example) is worth more than each dollar in the future (of which there are twelve thousand). A high discount rate implies that the future dollars paid back are worth substantially less than the upfront dollars received. Equivalently, it means that person is willing to pay more interest on the loan.
RApport: It’s been convention to use 3 percent as the annual discount rate in health economic modeling to value a novel health technology (such as a drug) in the US. What does that mean? Where did that figure come from?
JC: A 3 percent annual discount rate means that each year we delay an outcome into the future, its value decreases by 3 percent. For example, an individual is indifferent between $100 today and $103 in one year ($100 x 1.03). Because the value of money decreases by a further 3 percent during a second year, the individual is indifferent between receiving $103 in one year and receiving $106.09 in two years ($103 x 1.03).
Economists generally agree that the same discount reasoning applies to other types of outcomes, including health outcomes. So a treatment that cures 100 cases of disease today has the same value as a treatment that cures 103 cases in one year. Alternatively, a treatment that extends life by one year today has the same value as a treatment that extends life by 1.03 years a year from now. Equivalently, a treatment that a year from now extends life by 1 year has the same value as a treatment that extends life by around 0.97 years today (because 0.97 years x 1.03 is about 1.0 years). Economists say that the “present value” of a year of added life accrued one year in the future is 0.97 life years.
Throughout this example, I have been using an annual discount rate of 3 percent because that is the discount rate that most health technology assessment guidance documents recommend – including the authoritative Second Panel on Cost Effectiveness in Health and Medicine, which published their recommendations in 2016. The 3 percent value recommendation reflects a series of considerations. First, it corresponds to the interest rate on actual loans (after taking out the effect of inflation and the risk of loan defaults*). Second, it seems to correspond to the reduced value people place on each marginal unit of consumption resulting from the fact that over time, people on the whole have tended to grow wealthier as the economy grows (and the wealthier people are, the less each dollar of consumption matters to them). Third, the Second Panel suggested continuing to use the 3 percent annual discount rate to make economic evaluations more methodologically consistent and hence easier to compare.
RApport: You propose revising the discount rate to the 1.5-2 percent range when considering the net benefits of a novel health technology over time. What would that mean and why do you think the update is necessary?
JC: I have recommended reducing the discount rate because, based on the most recent economic data, a lower rate better represents how people value distant- and near-future outcomes compared to outcomes.
For example, although interest rates have climbed since the end of the COVID pandemic, real interest rates – which represent the tradeoff people make between having money now and having money in the future – have continued a downward trend that has persisted for decades. In addition, economic growth, and hence growth in consumption, has slowed as the population in wealthy countries has aged.
That phenomenon, among other factors, means people are not growing wealthier as quickly, and as a result, the discount rate does not need to be as high to accommodate a diminished value for consumption. The implication is that although people still care about near-term outcomes more than future outcomes, the value they place on future outcomes is greater than it used to be. In short, the future is growing more important.
RApport: To what extent is there consensus around the idea that the discount rate should be lower than 3 percent?
JC: In general, health technology assessment authorities have been reducing the value of the discount rate they recommend. A few decades ago, a recommended annual rate of 5 percent was more common; now, among wealthy countries, a 3 percent annual rate is most common.
Importantly, the US Office of Management and Budget recently concluded (p. 76) that the real interest rate on long-term government debt – the benchmark for informing the discount rate’s value – has averaged 2.0 percent, which is close to my recommendation, and substantially less than the discount rate it had recommended in earlier versions of its guidance. OMB’s update reflected its observation that interest rates have declined since it had issued its earlier guidance.
RApport: Do economists routinely question or update discount rates for other sectors?
JC: Investors update their view of the discount rate almost continuously. If they fail to recognize the market’s “exchange rate” between present and future outcomes, they put themselves at risk of losing money by paying too much for an anticipated future income stream, or failing to pay enough.
Health technology assessment agencies periodically update their discount rate guidance (though less often than Investors adjust theirs) so that their recommended rates correspond to actual preferences for outcomes over time.
RApport: If the discount rate for healthcare valuation changed, who would make that change, and what would be some of the immediate effects?
JC: In some countries, like the UK, the government determines what discount rate health technology assessments use. The government’s involvement in determining the discount rate used reflects the fact that it is the government that is making health care purchases.
In the US, the First Panel and Second Panel on Cost Effectiveness in Health and Medicine issued guidance on health technology practices in 1996 and 2016, respectively. That guidance included recommendations for the value of the discount rate. Adherence to these guidelines is voluntary, as their development is a non-governmental effort. Interestingly, the Second Panel hinted at the need to revisit its three percent recommendation when it noted that this recommendation should be observed for at least ten years. The expiration date for the Second Panel’s recommendation is hence as soon as 2026.
RApport: Is this an esoteric academic point, or does getting the discount rate right in these calculations actually matter to the rest of us?
JC: One commentator, tongue-in-cheek, described my own review of the discount rate as “useful” although “tremendously dull.” Admittedly, discussions of the discount rate do not typically get the blood pumping. The difference between a three percent and a two percent discount rate can seem unimportant. After all, who cares if a health technology assessment assigns a present value of $970 or $980 to a nominal $1,000 outcome projected to take place in a year’s time? But the cumulative effect of the discount rate can become important – especially for health impacts that can last decades.
For example, a hypothetical treatment that cures a fatal childhood disease, saving 70 nominal life years, produces 30 life years discounted at 3 percent annually, but 38 life years discounted at 2 percent annually. That eight-year difference means that use of 2 percent discount rate increases the treatment’s estimated present-value benefit by more than a fourth - a potentially important impact.
RApport: Are there other flaws in current economic models as part of health technology assessment that you see as worth fixing?
JC: Pass.
*We take out inflation because the discount rate represents actual preferences across time and not nominal differences that reflect changes in the value of money due to inflation. We take out default risk because health technology assessments do (or should) account for outcome uncertainty explicitly, rather than building in a generic risk aversion due to uncertainty. Inflating the discount rate to account for default risk would amount to double-counting the impact of uncertainty.
Please click here for important RA Capital disclosures.