Review

# The challenge of antimicrobial resistance: What economics can contribute

See allHide authors and affiliations

Science  05 Apr 2019:
Vol. 364, Issue 6435, eaau4679
DOI: 10.1126/science.aau4679

## Incentivizing restraint in drug use

The accelerating tide of antimicrobial resistance (AMR) is a major worldwide policy concern. Like climate change, the incentives for individual decision-makers do not take into account the costs to society at large. AMR represents an impending “tragedy of the commons,” and there is an immediate need for collective action to prevent future harm. Roope et al. review the issues associated with AMR from an economics perspective and draw parallels with climate change. A major stumbling block for both challenges is to build consensus about the best way forward when faced with many uncertainties and inequities.

Science, this issue p. eaau4679

## Structured Abstract

### BACKGROUND

Antimicrobial resistance (AMR) is increasing, driven by widespread antibiotic use. The wide availability of effective antibiotics is under threat, jeopardizing modern health care. Forecasts of the economic costs are similar to those of a 2°C rise in global average surface temperature, above preindustrial levels. AMR is becoming an urgent priority for policy-makers, and pressure is mounting to secure international commitments to tackle the problem.

Estimating the value of interventions to reduce antibiotic use requires predictions of future levels of antibiotic resistance. However, modeling the trajectory of antibiotic resistance, and how marginal changes in antibiotic consumption contribute to resistance, is complex. The challenge of estimating the resulting impact on health and the economy is similarly daunting. As with the cost of climate change, estimates of total AMR costs are fraught with uncertainty and may be far too low. Much of the uncertainty arises from the complexity of estimating the cost of changes in overall resistance levels. This cost depends on various factors: which drug and pathogen are involved, the mechanism of antibiotic resistance, the prevalence of that pathogen, the types of infections it causes and their level of transmissibility, the health burden of those infections, and whether alternative treatments are available.

Effective new antibiotics are urgently needed. However, without government intervention, R&D for antibiotics is rarely profitable, and most major pharmaceutical companies have left the field. New ways are needed to make antibiotic development profitable, decoupling profits from volumes sold.

### OUTLOOK

Analogies can be drawn between climate change and AMR, both of which have been described as a global “tragedy of the commons.” There is some consensus that we should treat carbon emissions reduction as an insurance policy against the possibility of a catastrophic climate outcome—and avoid waiting for a definitive optimum-abatement policy. A similar paradigm shift is needed to incentivize both the introduction and valuation of interventions to reduce antibiotic use and R&D of new antibiotics.

Rather than taxing the price and letting the market dictate the quantity of antibiotics supplied, an alternative may be to establish a regulatory body that issues prescribers tradable permits and to allow the market to determine the price. Such an approach could create a predictable revenue stream through more-foreseeable licensing fees for important antibiotics by decoupling the return on investment from the volume used. Approaches such as this could incentivize industry to develop new antibiotics for which there would otherwise be too small a market to provide a sufficient return on investment.

Reducing inappropriate antibiotic use while expanding essential access is a difficult challenge, especially in low- and middle-income countries. However, policy-makers and philanthropists are alert to the importance of AMR and increasingly are making substantial research funds available, with much of these funds devoted to the social sciences. We need economists, across many different fields, to engage with this pressing global problem.

## Abstract

As antibiotic consumption grows, bacteria are becoming increasingly resistant to treatment. Antibiotic resistance undermines much of modern health care, which relies on access to effective antibiotics to prevent and treat infections associated with routine medical procedures. The resulting challenges have much in common with those posed by climate change, which economists have responded to with research that has informed and shaped public policy. Drawing on economic concepts such as externalities and the principal–agent relationship, we suggest how economics can help to solve the challenges arising from increasing resistance to antibiotics. We discuss solutions to the key economic issues, from incentivizing the development of effective new antibiotics to improving antibiotic stewardship through financial mechanisms and regulation.

Before the discovery of penicillin by Alexander Fleming in 1928 and its subsequent development as a medicine by Howard Florey and Ernst Chain, death from infection could follow from something as minor as a scratch. Today, in high-income countries people take for granted the relative safety of procedures such as caesarean sections, joint replacements, and chemotherapy—all of which would be much more dangerous, perhaps prohibitively so, without effective antibiotics (13).

These gains in health care are now under threat from antimicrobial resistance (AMR). Here, we exclusively consider resistance against antibiotics used to treat bacterial infections. Resistance is largely driven by the ubiquitous use of antibiotics, often with little or no therapeutic benefit. For instance, antibiotics are commonly used for viral respiratory ailments (4) and are used nontherapeutically in agriculture to promote livestock growth (5). Additionally, widespread use of broad-spectrum antibiotics, which are effective against a variety of bacteria, has greater potential for selecting for extensive antibiotic resistance than drugs that target specific bacteria. However, it is important to acknowledge that even so-called “appropriate” use contributes to the development of AMR.

The potential costs to human life and to the economy are sobering. Recent U.K. government–commissioned reports (6, 7) estimate that if no action is taken, by 2050, AMR will cause up to 10 million annual deaths globally, reduce gross domestic product (GDP) by 2 to 3.5%, and cost US$100 trillion. The World Bank (8) has estimated that the impact on global GDP may be even greater, with an annual cost of up to US$3.4 trillion by 2030. Current estimates such as these are necessarily based on limited data and contestable assumptions, and some have questioned their utility (9). Notwithstanding their limitations, these projections exceed the current burden due to cancer, both in terms of mortality and cost (10). They also place the costs of AMR on par with the best current estimates of GDP loss associated with a 2°C rise above preindustrial levels in global average surface temperature (just under 3%) (11). In September 2016, AMR became only the fourth health issue—after HIV, noncommunicable diseases, and Ebola—to be discussed by the United Nations General Assembly. Pressure is mounting for an international commitment to tackle AMR (12), one that is analogous to the United Nations 2015 Paris Agreement to limit global temperature rise to 1.5°C.

Within economics, there has been some discussion of how AMR may be conceptualized and the difficulties of incorporating the large and uncertain future costs of AMR into assessments of health technologies (13, 14). Beyond this, however, discussion of AMR within economics has been limited. At the time of writing, only 55 of more than 1 million peer-reviewed articles in the EconLit database relate broadly to AMR. By contrast, 16,306 articles relate to climate change. We discuss what might be learned from this extensive economic literature on climate change to inform both current policy and future economic research on AMR. A key feature of these two global challenges is that antibiotic consumption and carbon use will bring about adverse future consequences, but there is substantial uncertainty over the timing and extent of these consequences. AMR also poses unique challenges that do not have a climate change equivalent. Here, we examine some of the concepts underlying these challenges and highlight prospects for future research on the economics of AMR.

## The scientific background

When Fleming, Chain, and Florey were awarded the Nobel Prize in 1945, they sounded an early warning that, owing to natural selection, bacteria could become resistant to antibiotics. They worried that subtherapeutic antibiotic doses would teach bacteria to resist antibiotics. Although this worry may still be relevant (15), most concerns currently center on common commensal bacteria being unnecessarily exposed to antibiotics in the treatment of bacterial (or even viral) infections. The more these commensal bacteria are exposed to antibiotics, the greater the risk of antibiotic resistance in subsequent infections. Nevertheless, the warning of future antibiotic ineffectiveness was prophetic and AMR is now associated with substantial morbidity and costs (16), which will continue to rise if resistance increases. Estimating the potential economically justified levels of investment in the development of new antibiotics and interventions that reduce unnecessary antibiotic use requires predictions of future levels of antibiotic resistance at the population level. However, modeling the trajectory of antibiotic resistance is complex (17).

### Uncertain future extent of AMR

The use of an antibiotic may not only select for resistance against itself but may also coselect for resistance against other antibiotics (18) or even result in susceptibility to other antibiotics (19). Thus, without detailed understanding of the underlying mechanisms and interactions, it is difficult to estimate the overall effect of changes in antibiotic consumption on resistance levels. Although the emergence of mutations conferring resistance to certain antibiotics is predictable, the spread of new resistance genes into mobile DNA or widespread bacterial strains is a so-called black swan event; such event will happen, but their timing, magnitude, and impact are difficult to predict (17, 20).

### Reversibility of AMR and renewability of antibiotics

Although increased selection pressure from rising antibiotic use makes greater resistance inevitable, it is not clear whether decreasing antibiotic use will necessarily be followed by a reduction in resistance (21). Most antibiotic resistance mechanisms are associated with a fitness cost, which reduces competitiveness in the absence of a drug and may allow susceptible bacteria to regrow. However, compensatory evolution, which reduces the fitness disadvantage of the resistance mechanism, is widespread among bacteria and limits the effectiveness of reducing antibiotic use (2123). Whether fitness costs or genetic compensation prevails in natural infections is poorly understood (24).

Another important parameter determining the potential impact of changing antibiotic prescribing is coselection of resistance to more than one antibiotic (21). Some empirical evidence suggests that the effect of reducing an antibiotic’s use depends on what, if any, antibiotic it is replaced by. In an effort to reduce resistance in pathogenic strains of Escherichia coli, use of trimethoprim and sulfonamide antibiotics was reduced in the United Kingdom and Sweden. Disappointingly, this particular policy was not followed by substantial reductions in antibiotic resistance (25, 26). However, the lack of effect was largely explained by the replacement of trimethoprim and sulfonamide with antibiotics that shared genetically linked resistance mechanisms and were therefore subject to coselection (18, 25, 26), resulting in AMR being maintained in the population.

This Review emphasizes the importance of understanding the precise mechanisms of resistance for each antibiotic before rolling out policies to reduce use of one in tandem with increased use of another (18, 27). It can be beneficial to have several types of antibiotics, which disable bacteria in different ways, available to treat a given infection. In this respect, antibiotics differ from most other drugs, for which there is generally little to no clinical benefit associated with producing a new drug that is only equally as effective as an existing one. A new antibiotic that matches the effectiveness of existing antibiotics but disables bacteria in a different way is valuable if it will remain effective in the future, once existing antibiotics become clinically useless. In this way, having access to a diverse set of antibiotics reduces the likelihood that infections will become untreatable as a result of universal resistance (17).

### Obstacles to the development of new antibiotics

As resistance to antibiotics accelerates, so too does the urgency of developing effective new antibiotics. Developing any new drug is expensive, risky, and highly likely to be unsuccessful. Unfortunately, the pace of development of antibiotics is slow, in part because of the challenging logistics and high costs of large clinical trials (28, 29). Even if successful, the need to conserve effective new antibiotics means that they will have restricted use, and their clinical utility will decline as resistance to them inevitably develops. All these factors will limit the volume of antibiotic sales and therefore the profits that pharmaceutical companies reap from their development. Without government intervention, the high cost of drug development and poor sales prospects mean that, on average, antibiotic R&D projects take a substantial loss, and few companies continue to pursue them (30). In July 2018, Novartis joined a growing list of major pharmaceutical companies that have abandoned antibiotic development owing to lack of financial incentives.

It is now widely recognized that better incentives are needed to encourage pharmaceutical companies to reengage with antibiotic development. Incentivization will require “push” incentives, such as research grants and tax credits, to bring down R&D costs. In addition, a variety of “pull” mechanisms aimed at providing sufficiently attractive returns on investment for developers have been proposed. A critical aspect is that the development of new antibiotics needs to be profitable regardless of prices and sales volume, as the existing model provides little incentive to produce a product that must be conserved (3134).

## Impact of AMR on health and the economy

Current worldwide deaths attributable to AMR, including antimalarial and antiviral resistance, have been estimated at ~700,000 per year, rising to 10 million per year by 2050 if present trends continue. Antibiotic resistance levels vary across the world (Fig. 1) (35). These estimates have many limitations that need to be addressed in future work to gain a more accurate picture of the extent of global AMR. For example, estimates are sensitive to frequency of blood culture sampling. In resource-poor settings, where cultures are often only obtained from patients who do not respond to empirical treatment, resistance estimates will be inflated (9).

### Direct and indirect consequences

AMR can result in treatment failures as well as normally uncomplicated infections becoming more complicated and severe (36). Increasing complication leads to increased risk of chronic conditions and death; longer hospital stays; even greater risks for patients with frequent infections, such as cystic fibrosis; and a need for more-toxic antibiotics if others fail owing to resistance. A review of papers that have tried to estimate the costs of AMR per patient episode (2) showed that additional costs varied from less than US$5 to more than US$55,000. Yet these papers estimated only the direct health care costs, such as additional hospitalizations.

In a world where antibiotic prophylaxis is no longer effective, there will be severe indirect consequences and costs, resulting from additional infections after invasive surgical procedures and immunosuppressive chemotherapy (3). It has recently been estimated that a 30% reduction in the efficacy of antibiotic prophylaxis for surgical procedures and chemotherapy would result in 120,000 additional infections and 6300 infection-related deaths per year in the United States (3).

### Wider economic impact of AMR beyond health care

In a recent review of estimates of the burden of AMR, only 11 of 214 studies considered the wider economic burden beyond the health care sector and only two did so at a global level (16). Measuring the economic impact of AMR solely with respect to health care misses broader social costs and benefits of interventions to stem AMR (37). The loss of effective antibiotics may have a substantial effect on the health and productivity of the workforce (3840). It has been estimated that caesarean sections, joint surgery, chemotherapy, and organ transplants contribute almost 4% to global GDP (6). This gives a rough indication of the scale of economic impact if AMR rendered these procedures so dangerous that they were abandoned.

Estimates of the total cost of AMR are fraught with uncertainty, and there is a risk of the cost being far higher than current best estimates. Much of the uncertainty arises from the complexity of estimating the cost of changes in overall resistance levels. For example, if there is a 1% increase in resistance to drug X, then it matters what pathogen is exhibiting that resistance, what the prevalence of that pathogen is, what type of infection the pathogen causes, the health burden of that infection, how transmissible the resistance gene and infection are (both between people and between other organisms), and whether there are alternative treatments available. Because the cost of increased resistance and the threat that it poses are uncertain, the investment that can be justified to control the threat and, ultimately, the value of antibiotics are also uncertain.

## Incentivizing socially desirable antibiotic consumption

To reduce antibiotic consumption to socially desirable levels, we must first consider the reasons why many people take antibiotics unnecessarily. One major factor is imperfect information, which might encourage use in cases where antibiotics are neither essential nor even beneficial. Second, how are people and organizations incentivized to prescribe, consume, and produce antibiotics? We must not lose sight of the fact that many people in low- and middle-income countries (LMICs) still struggle to access life-saving antibiotics. Economics has a number of concepts that can help us understand antibiotic consumption.

### The principal–agent problem, asymmetric information, and moral hazard

In health care, a doctor usually acts on behalf of the patient, who benefits from the doctor’s greater knowledge and skill. In economics, this scenario is known as the principal–agent relationship and applies to situations where a person or institution (the agent) makes decisions or takes action on behalf of (or that affect) another person or institution (the principal).

However, principal–agent relationships can give rise to a moral hazard if the agent’s own best interests do not align with, or if they conflict with, those of the principal. The principal–agent relationship between doctors and patients and, more broadly, between doctors and society can be problematic (41, 42). Patients are less likely than doctors to know if antibiotics are inappropriate, such as when a condition is caused by a virus rather than a bacterium or is a self-limiting illness. Patients are also less likely to realize that antibiotic use causes adverse consequences, for themselves and society, through increased resistance, i.e., a negative externality (43). When prescribing and dispensing are not separated, doctors may have financial incentives to prescribe. For example, in both China and Switzerland, countries with otherwise contrasting health care and social contexts, a lack of separation of prescribing and dispensing increases antibiotic prescribing (44, 45). Even without financial incentives, physicians are more likely to prescribe antibiotics if they think that their patients want them (46). If doctor and patient preferences for antibiotic use do align, they may not reflect the best interests of society more broadly (43).

The imbalance in the doctor–patient relationship could be relieved by altering incentive structures and through better education and public information campaigns for patients (47, 48). Alternatively, mirroring the system proposed for containing carbon emissions, a tradable permit system could be organized in which doctors or their institution are incentivized to reduce antibiotic prescriptions to a regulated optimum level (49).

Another major reason for unnecessary antibiotic use is the fact that physicians often work with imperfect information. Without reliable rapid diagnostic tests, physicians can sometimes only make educated guesses as to whether infections need antibiotics and, if so, whether they will be susceptible to first-line antibiotic treatment. As with antibiotic development, a combination of push and pull incentives could encourage the development of novel diagnostic tests (31). In addition to reducing use of antibiotics for viral or self-limiting bacterial infections, better diagnostic tests could improve the feasibility of R&D for innovative narrow-spectrum antibiotics. Empirical clinical use of such antibiotics might even be feasible for life-threatening infections if suitable rapid diagnostic tests become available (31, 50).

### Access versus excess

Brazil, Russia, India, China, and South Africa accounted for three-quarters of the increase in antibiotic consumption from 2000 to 2010 (51). Although there is substantial variation in consumption, even among countries in similar income groups, the overall consumption pattern has shifted toward broader-spectrum antibiotics (51, 52). Yet lack of, and delays in, access to any antibiotics still results in more deaths worldwide than does antibiotic resistance (53). More than 1 million children with untreated pneumonia and sepsis die annually (54). Across a sample of 101 countries (almost exclusively LMICs), it is estimated that universal provision of antibiotics could avert 75% of 590,000 annual deaths from pneumonia in children under the age of 5 (51). However, among neonates, increased resistance of bacterial pathogens to antibiotics jeopardizes improvements to child survival and causes an estimated 214,000 annual deaths globally (51).

Balancing the need to reduce overall antibiotic use with expanding essential access is challenging. Improving global access to antibiotics will increase the selective pressure for AMR. Conversely, reducing antibiotic use is necessary to stem resistance, but, perversely, this may reduce the incentives for developing new antibiotics, which then would impede access (55). This potential unintended consequence from antibiotic conservation emphasizes the importance of finding new ways for drug developers to profit from innovation, other than the current model based on drug prices and volumes.

## Learning from the economics of climate change

The scale and multifaceted nature of the challenges presented by AMR may seem overwhelming, yet AMR is not the first such challenge the international community has attempted to tackle. It is useful to reflect on what might be learned from other global challenges. There are some notable similarities between climate change and AMR (56, 57). Indeed, there may also be a direct link between the two phenomena; recent evidence shows that higher temperatures are associated with higher resistance levels in common pathogens (58). Both challenges have been described as a global “tragedy of the commons,” in which individuals, acting rationally and according to their self-interest, collectively damage public goods (5963). AMR and climate change are each driven by consumption of goods (carbon and antibiotics) that can provide people with valuable short-term benefits but that also impose long-term costs, such as existential threat from extreme weather or from life-threatening infections. Individuals may feel little incentive to change their course of action and forego short-term benefits because the costs are highly uncertain and harmful events may happen far in the future, which people typically discount (64). Moreover, the costs are unlikely to be avoided unless many other people also decide to reduce their carbon and antibiotic consumption.

The worst effects of both problems are likely to be distributed unequally across the world. In the most severe climate change scenarios, some of the small-island developing states, such as the Maldives and Tuvalu, could become uninhabitable (65). Likewise, increases in health care expenditures resulting from AMR are expected to be most severe in LMICs (66). Inadequate infection control systems within LMIC hospitals can lead to the spread of health care–associated infections and outbreaks caused by resistant pathogens, accompanied by marked increases in treatment costs, morbidity, and mortality (67). Within individual countries, the poor and vulnerable are also likely to be the most severely affected by climate shocks, for example, because of reliance on agricultural productivity for subsistence or employment (68). Similarly, because of its deleterious effect on labor and agricultural productivity, as well as on health care cost, AMR is likely to make it harder to reduce extreme poverty and may even increase the number of people living in extreme poverty (66).

Because of the time scale of cause and effect in climate change (69) and AMR (43), both raise questions of intergenerational equity, as future generations will not gain the benefits of carbon and antibiotic consumption but will face the brunt of the costs. Predicting the trajectory and extent of antibiotic resistance and temperature change is extremely difficult. On the one hand, increased temperatures resulting from carbon emissions are largely irreversible for 1000 years after emissions stop (70), but on the other hand, with the right approach, antibiotic resistance may be reversible within months to years. As LMIC economies grow, both carbon emissions and antibiotic consumption will rise (Fig. 2) (71), likely causing both temperatures and AMR to increase. The trajectory of both problems is strongly dependent on population levels and, in the case of AMR, population densities (72).

Despite similarities, AMR is arguably a more tractable problem than climate change. In many countries, doctors play a gatekeeping role with antibiotics, acting as agents for society. Though imperfect, this principal–agent relationship can be harnessed to steer antibiotic consumption toward optimum levels by providing doctors with appropriate incentives. Unfortunately, there is no comparable agent to act as a gatekeeper for carbon consumption. Reducing antibiotic use could potentially lower AMR in a much shorter time frame than the time frame over which reducing carbon consumption would lower temperatures. Because results from policy interventions may be faster for AMR than for climate change, it might be easier for policy-makers to agree to implement them. Successful cooperation over AMR could conceivably develop goodwill, or even infrastructures, for forging cooperation on climate policy.

Given the many parallels between AMR and climate change, it is instructive to consider the main economic approaches used to inform climate policy and the points of consensus and controversy regarding these approaches.

### Economic approaches to climate policy

Economists agree that the full cost to society of burning carbon is greater than its current market cost (typically measured in U.S. dollars per ton of CO2) (7375). The gap between the full cost and market cost is known as the social cost of carbon (SCC). In theory, if the market price could be increased by an amount equal to the SCC, the laws of supply and demand would result in the quantity of carbon burned falling to the socially optimum level, thus mitigating climate change. This forms the theoretical economic basis for introducing a carbon tax equal in size to the SCC. Economists broadly agree that global imposition of such a tax would be a sensible approach to tackle climate change. The leading alternative, of reducing emissions by issuing tradable emissions quotas, is similar in principle, even if mechanics differ. Instead of intervening in the market price of carbon to achieve the socially optimum quantity, “cap and trade” would instead enforce the socially optimum quantity directly. However, in theory, the price should then rise to the same level that a tax equal to the SCC would achieve. To date, such global approaches have eluded the international community for a variety of reasons, discussed below. More-local initiatives have shown mixed results: From 2003 to 2015, gasoline taxes rose in 83 countries, although they were reduced in another 46, while several other countries continued to subsidize gasoline prices (76).

There are major difficulties and controversies over the implementation of tax and quota schemes. In large part, the controversies arise over calculating the SCC. Estimates from leading experts differ by an order of magnitude (11), leading to profoundly different implications for public policy, ranging from quite modest (77) to a requirement for immediate and substantial cuts in emissions (69).

The standard approach to estimating the SCC has been to attempt a highly complex form of cost-benefit analysis using integrated assessment models (IAMs). An IAM attempts to monetize all the costs and benefits associated with carbon emissions to estimate the SCC. These costs and benefits are then discounted over some extended timeline to give the net present value (NPV) of burning a marginal ton of carbon. The marginal SCC is equal to this value minus its market price.

There are three major uncertainties in the inputs to an IAM. The first is climate sensitivity: the extent to which global temperatures would eventually rise if the atmospheric concentration of CO2 were to double. In AMR research, an equivalent uncertainty is the proportion of a bacterial population that will ultimately become resistant. The second uncertainty is the economic impact resulting from whatever climate change might occur, sometimes called the damage function. Both these sources of uncertainty have been described as not only unknown but unknowable (11, 78, 79). In AMR, the damage function reflects the health costs of future resistant infections, as well as the consequent economic costs. The third uncertainty is the choice of discount rate with which to perform NPV calculations, which requires a judgment of intergenerational equity. This applies to cost-benefit analyses of both climate change and AMR. It is important to remember that over long time frames, small differences in discount rates can make big differences to the NPV of distant outcomes. These uncertainties have led many economists, such as Stern (80), to caution against an overly narrow focus on cost-benefit approaches to climate change mitigation and to call for more drastic and imminent emission reductions as insurance against the non-negligible possibility of catastrophic outcomes (81, 82).

Theoretically, economists may largely agree that a global harmonized carbon tax, equal to the SCC, is an efficient response to climate change. However, lack of consensus regarding the size of the SCC is a major practical impediment. Instead, climate policy for reducing emissions has focused mainly on negotiating a limit for temperature increase on a country-by-country basis (11). The uncertainties are such that the choice of temperature target is highly contentious, on both environmental and economic grounds, but does offer a starting point for international negotiations on country-level emissions reductions.

### What can we learn from the economics of climate change?

Just as the full cost to society of burning a ton of carbon is much greater than its market cost, so too there is a social cost of antibiotics (SCA) that is not captured in market prices. The SCA could be estimated via detailed analyses akin to IAMs. The costs and benefits of antibiotic use can be monetized and discounted over time to obtain the NPV of consuming some unit of antibiotics. In essence, the many difficulties identified by Coast et al. (13, 14) result from a lack of knowledge of the SCA. Like the SCC, estimating the SCA presents analogous difficulties: How do increases or decreases in antibiotic use translate into changes in resistance levels? What is the impact of increased resistance levels on health and GDP (i.e., damage function)? How should the discount rate be chosen and what implications will this rate have for intergenerational equity? An important additional difficulty is that not all antibiotics, nor all resistances, are the same. Consequently, there is no fully satisfactory common unit of antibiotic quantity or, by extension, price in the way that there is for 1 ton of carbon (83). It may be more fruitful to model the costs and benefits, and resistance trajectories, associated with use of specific antibiotics for specific infections. Also, it might be valuable to gain a better understanding of the resistance trajectories of commensal bacteria, which are frequently exposed to antibiotics and may facilitate the spread of AMR (18).

Supposing that the social cost of using specific antibiotics for specific infections can be estimated, who carries the tax burden? If something analogous to the “polluter pays” principle is applied to antibiotic use, in the sense that those who cause externalities must pick up the bill (13, 84), is this the consumer, the prescriber, the national health care system, or the government? It has been argued that taxing patients for antibiotic consumption is unlikely to be effective, at least in high-income settings, due to lack of price sensitivity (7). However, antibiotics could also be taxed at a more aggregated level (e.g., general practice, local or national authority). A tax on each antibiotic prescribed might provide an effective incentive for reducing prescriptions, and the revenue raised could be invested in antibiotic development.

Rather than taxing the price and allowing the market to dictate the quantity of antibiotics, as we have discussed, an alternative may be to establish a regulatory body that gives prescribers permits (quotas) for prescribing, then lets the market determine the price (49). The quantity of permits available would be updated over time, according to resistance levels, and ideally the system would allow the permits to be traded (49). No such schemes have been implemented in practice, although since April 2015, the United Kingdom’s Quality Premium scheme has provided financial incentives to local Clinical Commissioning Groups (CCGs) (up to £140,000 for average-sized CCGs, which serve populations of around 280,000). Promisingly, this scheme has been credited with having reduced antibiotic prescribing in primary care by about 7% in 2015–2016 (85).

Similar approaches could be tailored to other health care systems. The U.S. Food and Drug Administration recently proposed a purchasable permit scheme for antimicrobials of last resort, where acute care institutions would pay a fixed licensing fee for the right to use a certain number of annual doses of a drug (86). The number of licensed doses could be tied to the number of beds or likelihood of encountering certain microbes. An attractive feature of this proposal is that it would create a predictable revenue stream, through more-foreseeable licensing fees, for important antibiotics, in a way that could potentially decouple the return on investment from the volume used. Assigning antibiotics an “option” rather than a “use” value could incentivize industry to develop important new antibiotics for which there would otherwise be too small a market to provide a sufficient return on investment. Another recent proposal for decoupling profits involves developers being paid an insurance premium to provide access to antibiotics, which could be renegotiated at regular intervals (32).

Taxes and quotas have also been considered as potential tools for discouraging unnecessary use of antibiotics in animals. Eighty percent of all antibiotics used in the United States (87) are employed in agriculture and aquaculture for routine nontherapeutic reasons, such as to promote growth, or as low-cost substitutes for hygiene measures to prevent infections (5). There is evidence that restricting antibiotic use in livestock is associated with a reduction in antibiotic-resistant bacteria in animals (88). The World Health Organization recently recommended a complete restriction on all antibiotic use for growth promotion and disease prevention in healthy livestock (89), and some European countries, including Denmark and Sweden, have imposed regulatory restrictions that have achieved substantial reductions in agricultural antibiotic use while maintaining productive livestock sectors (90).

The global average quantity of antibiotic administered per kilogram of animal (50 mg/kg) has been proposed as a possible regulatory target (7). This could reduce total consumption by 64% (5). Alternatively, a user fee of 50% of the current price on veterinary antimicrobials could reduce global consumption by 31% (5). User fees have several advantages (5, 87): They can be easily administered either at the manufacturing or importing stage, they would deter low-value antibiotic practices, and they would generate substantial annual revenues. User fees could be reinvested in R&D for new antibiotics or antibiotic stewardship programs and be used to compensate LMICs, which may be disproportionately affected by a user fee, by investing in strategies that reduce the spread of infection and improve veterinary services (5). In contrast, quantity restrictions would be difficult to enforce without adequate surveillance systems, which could be prohibitively expensive in LMICs (5). Whether through taxes or quotas, reducing predominantly nontherapeutic antibiotic use in agriculture is relatively uncomplicated compared with reducing human consumption.

Despite the shortcomings associated with the approach, combining biophysical and economic systems in climate IAMs is widely considered a worthy endeavor (80), even if the approach remains far from delivering convincing estimates of the SCC. Developing analogous models for antibiotics will be similarly valuable but will also have similar shortcomings. As for climate change, the magnitude of the uncertainties and potential risks has led to calls to try to reduce antibiotic use as a type of insurance policy (2). An analogy with fire prevention has recently been used to illustrate the insurance value of antibiotics (91, 92), in that the microbiology laboratory acts as the smoke detector and medical staff as the firefighters. Like climate change, rather than waiting for analyses of optimum policies to curb AMR, precaution necessitates setting ambitious but pragmatic targets for reducing antibiotic use on a country-by-country basis without delay.

## Outlook

Economics, as a discipline, has made valuable contributions to the debate on climate change mitigation. Economic analyses have successfully informed recommendations by the Intergovernmental Panel on Climate Change and decisions made by the international community, such as the Paris Agreement. We would like to see economists rise to the challenge of creating similar initiatives for AMR (table S1).

AMR presents particular challenges for LMICs, many of which already experience a high prevalence of AMR and its serious consequences. In LMICs, there is inadequate access to lifesaving antibiotics among the poor, coupled with often-substantial overconsumption of antibiotics among the middle classes. A lack of separation of prescribing and dispensing leads to supplier-induced demand, where patients consume more antibiotics than they would if they were better informed (32). The recent demonstration of substantial reductions in mortality from the mass administration of 6-monthly single doses of the antibiotic azithromycin in children (93) highlights this tension between access and excess. Thus, future research should consider mechanisms that reduce overall antibiotic consumption without restricting essential access (94). If the cost of antibiotics is to increase, via taxation or quotas, it will be vital to develop mechanisms that reduce the risk that they will only be taken by those who can afford them.

The challenge of dealing with AMR is sandwiched between two classic market failures. On the demand side is the tragedy of the commons, represented by the misuse of antibiotics as a public good, and on the supply side is the lack of incentive for firms to develop new antibiotics. A key challenge on the supply side is the need to hold back the distribution of new antibiotics until AMR renders existing ones ineffective. The question here is how to design incentives for production of a good that we must avoid using for as long as possible? It is now widely recognized that new product development requires the profits from innovation to be decoupled from prices and volumes (3234). One insight from game theory is that none of the players, neither individually nor collectively, should be able to gain by leaving a coalition. Developing an international agreement to limit antibiotic use is challenging because countries vary widely in their incomes, capacities, and objectives, and therefore in the costs they incur and benefits they derive from membership (95). Enshrined in the Paris Agreement is a principle of “common but differentiated responsibility and respective capabilities.” Similar consideration should be given to the different challenges AMR presents in different countries.

Like climate change, the potential costs of AMR are incredibly uncertain and potentially catastrophic. There is some consensus in the economics of climate change that we should treat the cutting of carbon emissions as an insurance policy. We suggest that a similar paradigm shift is needed for tackling AMR.

There is an urgent need to improve the supply of new antibiotics. At the 2019 World Economic Forum meeting in Davos, there were encouraging signs of recognition among policy-makers of the need to work with industry to change the way antibiotic innovation is rewarded (96). Policy-makers and philanthropists have recognized the importance of AMR and are increasingly making substantial research funds available, including to social scientists. A recent analysis of funding-organization databases, in 19 countries and at the EU level, found that from 2007 to 2013, a total of 1243 antibacterial resistance research projects were funded, at a total public investment of €1.3 billion (97). There is a great opportunity for economists, across many different fields, to engage with this pressing global problem.

## References and Notes

Acknowledgments: Funding: This research was funded by the National Institute for Health Research Health Protection Research Unit (NIHR HPRU) in Healthcare Associated Infections and Antimicrobial Resistance at the University of Oxford (grant HPRU-2012-10041) and in partnership with Public Health England (PHE). The support of the Economic and Social Research Council (UK) (grant ES/P008232/1) is acknowledged. L.S.J.R., L.A., A.S.W., and S.W. are supported by funding from the NIHR Oxford Biomedical Research Centre (BRC). L.A. and C.C.B. are supported by funding from the NIHR Community Healthcare MedTech and In-vitro Diagnostics Co-operative (MIC). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care. Competing interests: The authors declare no competing interests.
View Abstract