The economics of alternative payment models for pharmaceuticals

Abstract

Pharmaceuticals are priced uniformly by convention, but vary in their degree of effectiveness for different disease indications. As more high-cost therapies have launched, the demand for alternative payment models (APMs) has been increasing in many advanced markets, despite their well-documented limitations and challenges to implementation. Among policy justifications for such contracts is the maximization of value given scarce resources. We show that while uniform pricing rules can handle variable effectiveness in efficient markets, market inefficiencies of other kinds create a role for different value-based pricing structures. We first present a stylized theoretical model of efficient interaction among drug manufacturers, payers, and beneficiaries. In this stylized setting, uniform pricing works well, even when treatment effects are variable. We then use this framework to define market failures that result in obstacles to uniform pricing. The market failures we identify include: (1) uncertainty of patient distribution, (2) asymmetric beliefs, (3) agency imperfection by payer, (4) agency imperfection by provider, and (5) patient behavior and treatment adherence. We then apply our insights to real-world examples of alternative payment models, and highlight challenges related to contract implementation.

Introduction

Concerns about the prices of pharmaceutical products have dominated health policy debates for many years. Earlier research points out the many difficulties of pricing drugs per unit sold instead of value [1]. In response, alternative payment models (APMs), also described as ‘performance-based risk-sharing arrangements’ or PBRSAs, have gained popularity as more payers and manufacturers elect to tie reimbursement to specific outcomes [2]. Some of the reasons cited for their implementation include ‘uncertainty about future clinical and economic outcomes’ [3, 4], varying ‘safety and cost-effectiveness [profiles] for different indications’ [2] as well as ‘the “hidden information problem,” which gives rise to adverse selection; and the “hidden action problem,” which is the root cause of moral hazard in incentive contracting’ [5, 6]. In Europe, such contracts are justified in efforts to ‘maximize the value of health-technology investments’ [7] or to manage budget impact of a medicine [8].

While at least 500 such contracts have been implemented around the world [9], limited evidence has been presented on the net benefits they have produced for the healthcare system, especially once implementation costs are taken into account [10, 11]. In addition, while value-based pricing appears initially to present obvious benefits, it remains a niche method for pricing pharmaceuticals. Some have argued this is due to legal and regulatory barriers, rather than economic incentives [12]. However, these arguments are typically specific to the US and cannot explain the lack of widespread adoption overseas. Moreover, even in the US context, regulatory barriers apply only to certain types of novel pricing schemes [13].

Moreover, only a handful of studies employ price-theoretic analysis to reveal the efficiency rationale and likely consequences of value-based pricing. One such study, by Chandra and Garthwaite [14], considers the economics of indication-based pricing, noting that enabling one more dimension of price discrimination for innovators is likely to raise their total profits, not lower them as some proponents of indication-based pricing contend. Garber, Jones, and Romer caution that boosting innovator profits might lead to inefficiency in contexts where innovation spending is already excessive [15]. Another study, by Pauly [16], notes that the value of health improvement is patient-specific in ways that alternative payment models cannot easily accommodate. For example, there are few if any proposals that raise the value-based price for patients who place a higher value on a given health improvement. Therefore, adoption of a single value-based price effectively trades away one binding constraint for another.

Our paper continues in this tradition of theoretical analysis. Our aim is to study the underlying economic rationale for value-based pricing, and to determine how and when it is most likely to improve efficiency. We begin by presenting a stylized framework of fully efficient market interaction between payers and drug manufacturers. We then apply our insights to real-world examples of alternative payment models, and highlight challenges related to contract implementation.

We aim to illustrate how market efficiency largely erases the need for value-based pricing, but varied types of inefficiency create gains from alternative value-based pricing schemes. Conversely, efficient markets act as a nearly perfect substitute for value-based pricing arrangements, whose value is circumscribed by the degree of inefficiency present in pharmaceutical pricing and utilization. This perspective departs from the notion that uniform pricing is fundamentally inefficient in itself. We argue instead that uniform pricing can support efficient outcomes, if markets are functioning well in other respects. Unrelated sources of market failure might justify the use of value-based pricing as a corrective.

Theoretical framework

Most pharmaceuticals are priced uniformly but vary in their effectiveness.¹ Some have argued this variable effectiveness provides a rationale for more sophisticated payment models that go beyond uniform pricing [17]. We develop a short theoretical argument to illustrate that uniform pricing in itself can support efficient outcomes [18].² However, other sources of market incompleteness and misaligned beliefs can provide an efficiency rationale for novel payment models. In other words, while uniform pricing rules in efficient markets can handle variable effectiveness quite easily, market inefficiencies of other kinds create a role for different value-based pricing structures.

We begin with a stylized theoretical model of efficient interaction among drug manufacturers, payers, and beneficiaries. In this stylized full-efficiency setting, uniform pricing works reasonably well, even when treatment effects are variable. We then use this framework to precisely define market failures that create problems for uniform pricing and a corresponding typology of alternative payment models that address these failures.

Economic environment

Suppose that one or more third-party payers and a single drug manufacturer negotiate over the price of a drug therapy and over utilization rules that will govern beneficiary access to the therapy. We have in mind a patent-protected drug manufacturer that can earn rents from its product. Without loss of generality, assume the drug is costless to manufacture.³ Similarly, it sacrifices little generality to confine our attention to the case of a single-payer; even with multiple payers, each payer would perceive itself to be the marginal payer in a Nash-bargaining framework [19].⁴ The single payer covers i = 1, …, I types of beneficiaries, and there are n(i) beneficiaries of each type. Each type i beneficiary’s baseline untreated health status is H_i. If type i beneficiary is granted access to the therapy, she experiences the benefit Δ_i +∊_i, where Δ_i is the average benefit and ∊_i is a white noise error term—for simplicity, suppose ∊_i is normally distributed, with mean zero and variance ${\sigma }_i^2$. Both the mean and variance of the treatment effect can vary across beneficiary types. Rank-order beneficiary types according to their mean treatment effects, such that Δ_i ≥ Δ_i+1 for all i. If type i beneficiary pays the premium π_i and consumes c_i, her expected utility is EU(c_i −π_i, H_i + Δ_i + ∊_i); we assume the consumer is strictly risk-averse in consumption and weakly risk-averse in health. Define the per capita gross consumer surplus—before paying the premium—accruing to type i beneficiaries as $G_i\left({\Delta }_i,{\sigma }_i^2\right)$, where the partial derivatives have the signs, G_Δ > 0, and $G_{\Delta }>0,\text{ and }{G}_{{\sigma }^2}\le 0$.⁵ The latter is a strict inequality if consumer utility is strictly concave in health [20].⁶

We assume that therapies are more likely to be covered for patients that derive greater benefit from them. In addition, we assume that the insurer can charge higher premiums when it provides beneficiaries with more gross surplus—i.e., surplus prior to premium payments. Both these assumptions will hold in a marketplace where insurers have market power, or one where competitive insurers are negotiating against drug manufacturers with market power [21]. We embed these relationships using the stylized assumption that total premium revenue is equal to total gross consumer surplus. This assumption leads to a first-best outcome, even under uniform drug pricing, as we will show.

Efficient markets equilibrium

We first consider an equilibrium in which the payer and the drug manufacturer negotiate over a uniform price for the therapy, P, and over the number of patient groups, i*, that will receive the drug. For instance, the payer and manufacturer might bargain over a price and over formulary rules that determine what diagnoses or other clinical characteristics will permit access to the drug. If the payer and manufacturer possess Nash-bargaining weights α and 1−α, respectively, for some 0≤ α ≤1, the equilibrium price-utilization agreement solves:

$$\underset{P,i^{*}}{\text{max}}{\left[P\sum _{i=1}^{i^{*}}n(i)\right]}^{\alpha }{\left[\sum _{i=1}^{i^{*}}n(i)G_i\left({\Delta }_i,{\sigma }_i^2\right)-P\sum _{i=1}^{i^{*}}n(i)\right]}^{1-\alpha }$$

Defining overall consumer surplus for type i beneficiaries as $C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)=n(i)G_i\left({\Delta }_i,{\sigma }_i^2\right)$, this problem has the following two first-order conditions:

$$[P]:\frac{\alpha }{P}-\frac{(1-\alpha ){\sum }_{i=1}^{i^{*}}n(i)}{{\sum }_{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)-P{\sum }_{i=1}^{i^{*}}n(i)}=0$$ (1)

$$\left[i^{*}\right]:\frac{\alpha }{{\sum }_{i=1}^{i^{*}}n(i)}+\frac{(1-\alpha )\left(C{S}_{i^{*}}-P\right)}{{\sum }_{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)-P{\sum }_{i=1}^{i^{*}}n(i)}=0$$ (2)

These two conditions imply that:

$$\begin{gathered} \alpha =\frac{(1-\alpha )P{\sum }_{i=1}^{i^{*}}n(i)}{{\sum }_{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)-P{\sum }_{i=1}^{i^{*}}n(i)} \\ =-\frac{(1-\alpha )\left(C{S}_{i^{*}}-P\right){\sum }_{i=1}^{i^{*}}n(i)}{{\sum }_{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)-P{\sum }_{i=1}^{i^{*}}n(i)} \end{gathered}$$ (3)

It is straightforward to show that this expression implies CS_i* =0. This reflects first-best efficiency, because the payer allows access for every type of beneficiary whose gross surplus exceeds the zero marginal cost to manufacture the drug.

In addition, the first-order condition for i*, coupled with the result that CS_i* =0, implies:

$$\alpha \left(\sum _{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)\right)=P\sum _{i=1}^{i^{*}}n(i)$$ (4)

This expression indicates that the drug company’s total profits (on the left-hand side) equals the share α of total consumer surplus.⁷ The equilibrium price is given by the manufacturer’s share of the total surplus, $P=\frac{\alpha \left({\sum }_{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)\right)}{{\sum }_{i=1}^{i^{*}}n(i)}$.

Even though we have assumed uniform pricing for a drug with variable treatment effects, this equilibrium is first-best efficient. Payers and manufacturers share the first-best surplus between them. Thus, in the absence of market failures, there is no need for variable pricing strategies (or alternative payment models), even when the effectiveness of the drug varies.

Our analysis presumes that manufacturers can bargain with payers rather than setting a single monopoly price. This seems realistic, since manufacturers routinely bargain with public and private payers over prices [22]. Both manufacturers and payers have good reason to prefer this approach, since price-bargaining leads to more joint surplus for both compared to simple monopoly pricing.⁸ Indeed, price-bargaining—with or without variable pricing strategies—produces greater efficiency than simple monopoly pricing. The latter leads to underutilization, where CS_i* > 0 and some patient groups are denied access to a drug that would benefit them.

In contrast, price regulation leads to the opposite problem—overutilization—and can interfere with the efficiency of bargaining. Our framework helps illustrate this, if we consider a public or private payer operating in the context of a price ceiling, $\overline{P}$. Focus on the substantive case where this price ceiling binds, in the sense that the optimal price computed above exceeds the ceiling. In this case, raising the price ceiling increases total surplus, as in:

$$\frac{\alpha }{\overline{P}}-\frac{(1-\alpha ){\sum }_{i=1}^{i^{*}}n(i)}{{\sum }_{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)-\overline{P}{\sum }_{i=1}^{i^{*}}n(i)}>0$$

This inequality leads to a condition analogous to Eq. (3) above:

$$-\frac{(1-\alpha )\left(C{S}_{i^{*}}-P\right){\sum }_{i=1}^{i^{*}}n(i)}{{\sum }_{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)-P{\sum }_{i=1}^{i^{*}}n(i)}=\alpha >\frac{(1-\alpha )P{\sum }_{i=1}^{i^{*}}n(i)}{{\sum }_{i=1}^{i^{*}}C{S}_i\left({\Delta }_i,{\sigma }_i^2\right)-P{\sum }_{i=1}^{i^{*}}n(i)}$$

This expression in turn implies that CS_i* < 0. Intuitively, price ceilings lead to overuse of the drug by some groups of patients that do not derive sufficient surplus from it. This type of inefficiency would be difficult to correct via value-based pricing, which typically results in lower prices for low-value groups. Such an arrangement would only exacerbate overuse, and it highlights the limitations of VBP as a tool for mitigating inefficient pricing policies.

Related issues might arise if insurance is incomplete due to patient deductibles. Deductibles might discourage wealth-constrained patients from purchasing a drug, even if they belong to a group that derives positive consumer surplus from it. Variable pricing does little to mitigate these inefficiencies, which must be addressed by altering cost-sharing rules.

Market failures

Variable treatment effects in themselves—whether across indications, patients, or other characteristics—do not provide an opening for value-based contracts, provided markets are efficient. Intuitively, a risk-neutral payer finds it costless to bear variable treatment effects, so long as it can negotiate a single lump-sum payment that matches the aggregate expected value of the medical technology.

Instead, the rationale for value-based contracting flows more directly from market inefficiencies, not variable treatment effects. Each of the failures we will enumerate below creates an opportunity for welfare improvements from novel payment strategies.

For this discussion, it is convenient to differentiate between “observable heterogeneity” and “unobservable heterogeneity.” “Observable heterogeneity” means that the benefit of the therapy varies in ways that the payer can readily predict, or that Δ_i ≠Δ_j for at least one pair (i, j). “Unobservable heterogeneity” means that the benefit of the therapy varies in ways that the payer cannot predict, or that ${\sigma }_i^2>0$ for some value(s) of i. Above, we showed that observable and unobservable heterogeneity on their own do not justify departures from uniform pricing rules. However, when they intersect with alternative market failures, which we enumerate below, uniform pricing may result in inefficiency.

Uncertainty of patient distribution

“Uncertainty of patient distribution” means that, in the presence of observable heterogeneity, the payer cannot form an unbiased estimate of n(i). This means a payer cannot accurately estimate how many patients of each type i will receive treatment and will thus overestimate or underestimate the aggregate consumer surplus that can be attained. As a result, the payer cannot achieve the first-best, because she cannot solve the bargaining problem described above, and the equilibrium uniform price cannot be calculated. However, the first-best can be achieved by paying variable prices, P_i = α(CS_i), which do not require unbiased estimates of n(i). This is a type of outcomes-based price, in which value to the consumer is the relevant outcome. Under this solution, the payer earns positive surplus from treating any patient type for whom CS_i ≥ 0. This results in the first-best efficient outcome.

However, the value-based pricing solution here requires detailed information that might extend beyond what is practical. The payer must understand not only variation in health outcomes, but also variation in the way patients value health improvements. Pauly (2017) has emphasized the importance of this variation in a market. This poses a particular problem in cases where the patients deriving the greatest health benefit also place the lowest value on health improvement—e.g., socioeconomically disadvantaged patients. In this case, health outcomes might negatively covary with consumer surplus, and tying prices to health outcomes may be a poor substitute for “consumer-surplus-based” pricing. An example might be intensive treatments for diabetes, which tend to disproportionately benefit the poor [23], or cardiovascular treatments that might substitute for behavioral changes like diet and exercise [24]. Thus, this solution works best when consumers place reasonably uniform values on health improvement, so that variation in consumer surplus is well captured by observable variation in health outcomes.

Asymmetric beliefs

“Asymmetric beliefs” imply that the payer and drug manufacturer hold different expectations about the value of Δ_i or of ${\sigma }_i^2$, for at least one patient type i. Or, equivalently, asymmetric beliefs may arise because Δ_i or ${\sigma }_i^2$ are unknown exante (due to structural uncertainty over treatment effects or unknowable patient types). Define $C{S}_i^p$ as the payer’s expectation of consumer surplus, and define $C{S}_i^d$ as the drug company’s corresponding expectation. Our efficient markets case presumed that $C{S}_i^p=C{S}_i^d=C{S}_i$, but this need not be the case. If these beliefs differ, the first-best can still be achieved if P_i = αCS_i, for the ex post realization of CS_i. Such a payment scheme addresses the asymmetry of beliefs (or information) by contracting on the actual (consumer surplus) outcome, rather than on expected outcomes. Such outcome-based payment would be determined after a patient has received therapy. It preserves the first-best efficient outcome, even when payers and manufacturers differ in their beliefs. In a hypothetical case where manufactures believe a treatment is worth x and payers believe its value is x − δ, the price they agree upon may be tied to the realized consumer surplus of the treatment so that the identity P_i = αCS_i holds. In theory, the realized consumer surplus may exceed x or be lower than x − δ, hence any value-based contract should allow for price to reflect the realized consumer surplus (and not be arbitrarily pegged to either x or x − δ.

The same caveats regarding consumer surplus valuation apply to this case as well: if patients differ in how they value health, and especially if health valuation varies negatively with treatment effects, payers might not be able to identify contracts that link prices to consumer surplus. On the other hand, this pricing mechanism works most efficiently if consumers are reasonably homogeneous in their value of health improvement.

A special case of asymmetric beliefs

The solution above breaks down if payers and manufacturers disagree on which therapy produces the greatest ex ante expected consumer surplus. For example, if payers believe in a cheaper, older drug, while manufacturers believe in a newer, more expensive drug, contracting on the price of a single drug will not bridge that divide. That is, outcomes-based contracts solve disagreements about pricing [33], but do not always solve disagreements about which therapies are ex ante optimal. Efficiency may be unattainable when there are competing clinical options with uncertain and controversial benefits [34]. For this reason, the utility of value-based pricing could be somewhat limited in therapeutic areas where treatment selection is controversial. However, if agreement exists about the benefits based on varying treatment staging (e.g., first-line, second-line, and adjuvant), reimbursement could differ based on the stage-specific consumer surplus a therapy provides.

Payer agency imperfection

“Imperfect agency” means that the payer fails to act as a perfect agent for its beneficiaries. For example, if the payer is deciding how to pay for a therapy with long-term benefits, but beneficiaries stay with the payer for a short period of time, the payer may internalize only a portion of these benefits, so that $C{S}_i^p<C{S}_i$ for some type i. Agency problems are even harder to solve for than asymmetric beliefs, but there are cases in which solutions might present. Continue with the example of a therapy with long-term benefits. In this case, the surplus from a treatment is shared across current and future payers, where each payer j perceives surplus $C{S}_i^{p_j}$. One solution, in principle, is for these payers to share the cost proportionally: each payer j pays $\alpha C{S}_i^{{\rho }_j}$. If ${\sum }_{j=1}^{J}C{S}_i^{p_j}=C{S}_i$ and if $S_i^{p_j}\ge 0\forall j$, this pricing strategy results in the first-best equilibrium—assuming, of course, that all sides agree on treatment benefits.

To see why, note that if CS_i > 0, then there must be some payer j that perceives positive surplus from the technology; moreover, the set of all payers j ∈ J_B that derive positive surplus must be willing to pay a total of αCS_i for the technology, implying that $P_i^j=\alpha \left(C{S}_i^j\right)$. In theory, a payer whose share of the consumer surplus is negative may be compensated by payers who derive a positive surplus.

While there is at least a theoretical solution, the practical difficulties remain considerable. As mentioned above, asymmetric beliefs about treatment benefits pose problems, because multiple payers have to share similar beliefs among themselves, and with a drug manufacturer. In addition, there may be no way to contractually hold a future payer responsible for a technology used today; in other words, future payers have strong incentives to renege on an agreement, since the patient has already been treated with the drug. A regulatory requirement for payers to comply would likely be required, since treatment cannot typically be reversed ex post. Potential solutions proposed include installment payments shared across payers [44]. However, handing off liabilities to subsequent payers may be complicated by legal obstacles to contracts between payers [45].

Payer agency imperfections can also arise in single-payer settings without interactions among multiple insurers. There is no guarantee that single-payer decision making perfectly internalizes the values of consumers. Instead, single-payer systems may tie value-based pricing to the explicit or implicit measurement of value via cost-effectiveness or other methods [46]. In practice, such measurement approaches might lack transparency [46] or fail to align with value to society. These challenges are more difficult to address with a single solution but instead require a robust and accurate health technology assessment regime.

Provider agency imperfection

Another type of agency problem arises with providers. Efficiency requires that payers and providers are perfectly aligned with each other and with the interests of patients. The earlier section discussed misalignment between payers and patients, but misalignment between physicians and patients is also possible—indeed, often likely [6]. Providers may care about patient health or well-being [50], but they also care about their own profits, which could be at odds with patient utility. There may be overuse of profitable therapies or underuse of therapies with limited financial returns, such as preventive therapies.

The agency problem here results in financial incentives being misaligned with consumer surplus. One solution to this problem is to tie provider reimbursement to patient outcomes or other outcomes—e.g., total cost—that may not directly relate to provider profits. The pricing of treatment, P_i, may then involve multiple services and has resulted in payment models such as bundled payment, capitation, and shared savings programs, which may address inefficient allocations of resources due to the provider incentive problem. There is a large literature discussing the strengths and weaknesses of these various payment models. Bundled payment, for example puts providers “at risk for efficient care” during an episode of care [6]. Among the leading downsides of such models is the diagnostic and treatment uncertainty, which creates a performance risk faced most acutely by smaller providers who cannot sufficiently diversify such risk [6], as well as practical challenges related to defining a bundle, agreeing on a payment method that results in an acceptable risk allocation between the payer and provider, and implementing measures that prevent sicker patients from receiving suboptimal care [51].

Patient behavior and treatment adherence

Our model assumes that treatments impose only financial costs. In practice, patients often have to invest time and effort into treatment interventions. Examples include prescription drug adherence, along with required diet or exercise regimens. Thus, the realized Δ_i may be lower than expected if the patient does not comply with the necessary behavioral requirements. This behavior is unknowable to the payer and manufacturer ex ante, producing information asymmetry and, similarly to earlier cases, inequality of beliefs about consumer surplus: $C{S}_i^p\ne C{S}_i^d\ne C{S}_i$. Here, the first-best can be achieved if P_i = α(CS_i), for the ex-post realization of CS_i, and can be achieved by a value-based contract conditional on adherence. This is a type of outcomes-based payment contract that faces the same challenges described above.

Implications for healthcare markets

The emergence of targeted therapies and increase in healthcare spending as a proportion of overall US national spending will continue to elevate the prominence of healthcare market failures [60]. As a result, public and private payers have debated value-based pricing agreements for pharmaceuticals and have in some cases implemented them [2, 61]. Earlier research emphasizes the need to address variable benefits by making prices variable as well [62], and the potential for cost-savings from value-based pricing in single-payer systems like the NHS in the United Kingdom [63].

However, we show that variable treatment effects on their own do not imply the superiority of variable pricing over uniform pricing. Variable pricing improves efficiency only in the presence of other market failures, including uncertainty about patient distribution, asymmetric beliefs about efficacy, agency imperfection by provider and payer, and patient adherence. Without these failures, payers and manufacturers would bargain efficiently over uniform pricing and formulary rules, even if treatment effects vary across patient subgroups. Moreover, we follow earlier economic research that more efficient pricing strategies may not end up reducing costs for consumers [14].

The market failures we have identified suggest different types of treatments may threaten efficiency absent appropriate value-based pricing strategies. First, therapies that target heterogeneous patient populations with uncertain distribution may risk underutilization. Payers may opt to cover these therapies for narrower subsets of eligible patient populations, such as by requiring step-therapy or delaying access before a patient meets other eligibility criteria. As we show, this may constrain use to inefficiently low levels. Examples of such therapies may include gene therapies for hereditary and oncology conditions with variable response rates, therapies for neurodegenerative indications such as Alzheimer’s disease, which may target multiple non-discrete disease states, and treatments of patients in multiple indications with the same drug, such as due to off-label use, provided that the variance in patient outcomes is unknowable at the time of treatment.

Second, an asymmetric beliefs-related market failure will directly affect therapies with imperfect information about a treatment’s long-term efficacy (such as due to relatively short trial duration), including many emerging cell and gene therapies as well as drugs in primary and secondary prevention, such as treatments of hypercholesterolemia.

Third, payer and provider imperfection-related market failures will impact therapies with long-term benefit accrual trajectory, including cures (such as treatments of viral infections like Hepatitis B and C) and therapies for diseases that affect patients near Medicare eligibility age, such as in Alzheimer’s disease.

Finally, market failures related to adherence and patient behavior will remain dominant in treatments that rely on long-term patient compliance or have high patient cost-sharing, resulting in lower adherence due to behavioral and economic factors. Examples of such therapies include insulin products and other chronic therapies in primary and secondary prevention.

Although our model does not estimate transaction costs, these are acutely present in all value-based pricing agreements and may often impede their implementation. In addition, regulatory constraints (or absence of regulatory solutions to the collective action problem) prevent individual payers from implementing value-based pricing agreements, even if economically rational. Among the key obstacles are Medicare’s average sales price calculation, Medicaid best price rebate program, the federal anti-kickback statute, the design of the 340B drug pricing program, and regulations pertinent to off-label promotion of drugs [13, 64]. These are not present in ex-U.S. settings, but other constraints (such as equity concerns about variable pricing [65]) may affect the use of alternative payment models in these markets. Finally, collection of data about monetary, health and societal benefits of healthcare interventions is fraught with challenges which often impede the determination of a treatment’s benefits and costs, and thus the implementation of value-based pricing agreements.

Conclusions

Uniform pricing does not prevent first-best efficiency in efficient health care markets. Payment for medical treatments faces complex challenges related to the role of individual stakeholders and their agreement on payment that results in Pareto efficiency. In principle, value-based pricing agreements may overcome market failures existent in healthcare markets, and their recent uptake around the world suggests that they are expected to resolve some of the market failures identified above. However, the implementation of such payment models continues to face numerous challenges, including transaction costs, regulatory constraints and data collection issues. We may expect more value-based agreements for pharmaceuticals to emerge as solutions to these challenges are developed.

Box 1: Handling varying patient population sizes in the real world.

An example of value-based contracts addressing uncertainty in patient distribution is indication-based pricing. Here, different indications are used as proxies for variation in benefits to different patients from the same treatment (although within-indication variation exists) [14]. For example, indication-based prices have been negotiated for drugs like Herceptin^® (trastuzumab) and anti-inflammatory drugs [25]. Similarly, drug manufacturers have pursued regulatory approval of the same compound under different brand names as a strategy for receiving indication-based prices. For instance, Pfizer’s sildenafil was first approved in 1998 for male erectile dysfunction under the brand name Viagra^® [26] and in 2005, sildenafil received a separate approval indicated for pulmonary arterial hypertension under the brand name Revatio^® [27]. The prices for the two branded versions of sildenafil have differed [28]. A different strategy was chosen for two anti-vascular endothelial growth factor (VEGF) antibodies, marketed as two products: bevacizumab (Avastin) for cancer indications and ranibizumab (Lucentis) for age-related macular degeneration. Both have demonstrated roughly comparable therapeutic utility in blocking VEGF-induced angiogenesis in these indications [29]. Avastin is considerably cheaper per use for this purpose [30].

While regulators have been slow to catch up to these market trends [13], the Centers for Medicare and Medicaid Services have recently announced that beginning in 2020, Part D plans will be able to cover prescription drugs for certain indications that they are approved for, and deny coverage for other indications [31]. Under the United Kingdom’s 2009 Pharmaceutical Price Regular Scheme (PPRS), manufacturers are permitted a one-time price increase for a new indication [32]. The price of the original indication remains unaltered. This price request is conditional on an approval by NICE (National Institute for Health and Care Excellence) approval. If NICE does not review the request within 12 months, then the price increase goes into effect.

Box 2: Asymmetric beliefs in the real world.

Outcomes-based contracting ties payments to ex post outcomes, and has been used or proposed in treatments with uncertain real-world outcomes such as PCSK9-inhibitors [4, 35]. In mid-2017, Amgen and Harvard Pilgrim signed an outcomes-based refund contract for the PCSK9-inhibitor evolocumab, which is indicated for certain adults with heterozygous familial hypercholesterolemia (HeFH) or clinical atherosclerotic cardiovascular disease, or homozygous familial hypercholesterolemia [36]. Under this contract, Amgen provides Harvard Pilgrim a rebate for the cost of evolocumab for patients who have a heart attack or stroke while receiving evolocumab. Following this, Amgen entered into at least four other outcomes-based contracts with payers (Cigna, CVS Health, Prime Therapeutics, and Abarca) [37–39]. Prime’s contract included an adherence component [40], while Cigna’s contract focused on meeting clinical trial results in clinical practice—implying Cigna was skeptical about the replicability of trial results in non-supervised real-world conditions [37]. While outcomes-based contracts have been gaining traction, their precursor—coverage with evidence development—accounts for around half of performance-based risk-sharing agreements in the biggest-available database [25]. Italy reportedly collects hundreds of millions of euros in rebates every year, amounting to about 1% of total pharmaceutical spending [41], and has developed staggered, outcome-based payment schemes for high-cost cell and gene therapies [42]. Other OECD countries have similar schemes in place [43].

Box 3: Benefit non‑internalization by payer in the real world.

Some healthcare interventions such as high-cost cures produce public goods that may not be internalized by payers. Benefit mandates and patent buyouts have been proposed as potential solutions to these constraints but never seriously discussed [47]. In states like Louisiana and Washington, value-based payment models for Hepatitis C have been unveiled, using a subscription payment approach dubbed the ‘Netflix model’ [48]. Generally, spending is capped at a certain level with either free or “extremely low” pricing for any additional patients [49]. As a result, government budgets are better positioned to absorb the costs of an infectious disease cure, which results in positive externalities, and patients receive access sooner than under existing contracts, thus increasing consumer surplus. Similarly, future treatments of progressive diseases like Alzheimer’s or Parkinson’s disease will face challenges related to benefit non-internalization by private payers, who may be asked to pay for the cost of therapies in patients months or years before they become eligible for Medicare. It is unlikely that any individual payer and manufacturer can resolve this market failure on their own without some government intervention given the high likelihood of free-riding.

Box 4: Agency imperfection provider problem in the real world.

Bundled payments encourage physicians to choose treatments that offer value for money while simultaneously achieving specific health outcome objectives [52]. For example, bundled payments can encourage providers to account for total costs or patient health outcomes, depending on how incentives are structured. Bundled payment should reflect patient mix so as to avoid market failure concerning the uncertainty of patient distribution [53].

For instance, UnitedHealthcare operates the UnitedHealthcare Care Bundles Program in fee-for-service Medicare and Medicare Advantage plans, and gives providers the option to receive bundled payment for specific procedures like hip and knee replacements, spinal fusions and coronary bypasses [54]. This model rewards providers for exceeding specific standards, implicitly resulting in higher benefit internalization by providers, and aims to result in better care coordination and lower costs (ibid). Bundling of payment for prescription drugs, however, faces unique challenges related to the risk this creates for providers given increasing drug costs [55]. UnitedHealthcare does not bundle drug and non-drug expenditures, but includes drugs in the calculation of spending targets in its shared savings opportunities (ibid).

In the Netherlands, for example, bundled payment models for type 2 diabetes care, chronic obstructive pulmonary disease and vascular-risk management have been implemented, with an explicit goal of improving primary care and avoiding high-cost specialty and hospital care [56].

Box 5: Patient behavior and treatment adherence contracting in the real world.

Measuring adherence and linking it to reimbursement has been a challenge. Yet, it has been shown that adherence declines with patient cost-sharing and greater adherence improves outcomes [57]. An adherence-linked contract for evolocumab to treat hypercholesterolemia between Amgen and Prime Therapeutics, a pharmacy benefit manager, has an adherence component [40]: prime receives payments in case of poor performance.

Other value-based, adherence-linked contracts have been publicly disclosed, such as an adherence contract between Harvard Pilgrim and Eli Lilly for the treatment of osteoporosis, between Harvard Pilgrim and Amgen for evolucumab, and between Express Scripts and AbbVie for a hepatitis C treatment [58]. Typically, these contracts focus on persistence (continuing treatment for a certain duration, such as by tracking medication refills). New technologies may provide better instruments to measure true adherence at the patient-level for payment purposes [59].