Antibiotic Resistance Threats in the United States, 2019

The CDC’s Antibiotic Resistance Threats in the United States, 2013, was one of the earliest of the recent reports to attempt to quantify the health burden of resistance and to categorize pathogens by the level of threat they pose to public health (CDC, 2013). The CDC revisited this report in 2019, and the more recent publication contains estimates of the burden of resistant pathogens based on laboratory and population surveillance data, complemented with research from electronic medical records, and is weighted to allow for some extrapolation to the national level (CDC, 2019; Kadri, 2020).¹

This analysis indicated there are 2.8 million resistant infections every year in the United States, causing 35,900 deaths; Clostridioides difficile (C.difficile) infection, a problem caused by antimicrobial disruption of the gut flora, kills another 12,800 people a year (CDC, 2019). Despite sudden increases in certain infections, multidrug-resistant Candida auris, for example, was not spreading in the United States until 2015, total deaths from resistant infections declined 18 percent between 2013 and 2019 and deaths in hospitals have declined 28 percent (CDC, 2019). Table 3-1 shows the percentage change for those pathogens for which a longitudinal comparison was possible.

TABLE 3-1

Change in Infections Caused by Some CDC Priority Pathogens Between 2013 and 2019.

The report also explained that the economic costs of resistance can be difficult to estimate with any credibility. Resistant infections undoubtedly cost the health system more in terms of person hours needed to treat them and extended hospital stays. The medicines needed to treat them can be expensive and less well tolerated (CDC, 2019). Nevertheless, there is no consensus methodology to estimate the economic burden of resistant infections.

To put some economic parameters on the problem, the CDC used retrospective cost analysis of patients with six common resistant infections in the Veterans Health Administration medical centers, adjusted for the general population by the Veterans Affairs (VA) Health Economics Resource Center (Nelson et al., 2021). The analysis for C. difficile drew from peer-reviewed literature, and for some pathogens, no reliable cost estimate was available. Table 3-2 shows only the direct medical costs associated with a positive culture for the pathogens of interest, not the downstream costs associated with future disability or the cost to the patient of missed work or even the cost to the health system after discharge. (The long-term asymptomatic nature of resistant gonorrhea infection made it necessary to present a lifetime estimate of costs.) The direct costs of treating six, common multidrug-resistant pathogens was $4.6 billion a year, C. difficile another billion, and drug-resistant gonorrhea another $133.4 million (CDC, 2019; Nelson et al., 2021).

TABLE 3-2

Costs Attributable to Antimicrobial-Resistant Pathogens in the United States, in Constant 2017 Dollars.

OECD Development Work on Antimicrobial Resistance

The OECD, an intergovernmental economic organization, has also published several influential reports on antimicrobial resistance since 2015, often in collaboration with the European Centre for Disease Prevention and Control (ECDC).

The first of these publications, Antimicrobial Resistance in G7 Countries and Beyond, was released shortly after the O’Neill report and drew attention to the fact that only a quarter of the world’s countries had a national antimicrobial resistance plan (Cecchini et al., 2015). The OECD’s 2016 publication drew on data from the ECDC’s Surveillance of Antimicrobial Consumption Network and the Center for Disease Dynamics, Economics, and Policy’s national and subnational resistance data to analyze trends in the emergence of resistance between 2005 and 2014 (OECD, 2016). Using an aggregate measure of resistance based on six, high-priority pathogen–drug combinations, the report concluded that the prevalence of antimicrobial resistance had increased in 23 of 26 OECD countries (see Figure 3-1), though human use of antimicrobials remained largely stable (OECD, 2016).

Figure 3-1

Trends in antimicrobial resistance across OECD countries, 2005 to 2014. NOTE: * Greece missing S. pneumoniae (resistant to penicillin) 2005 and 2014, Slovakia and Belgium missing K. pneumoniae (resistant to third-generation cephalosporins and carbapenem) (more...)

The OECD Health Committee, in collaboration with the European Centre for Disease Prevention and Control, has also given considerable attention to estimating the future health and economic burden of antimicrobial-resistant infections (OECD, 2018). Their analysis was undertaken at the direction of the European Commission and published in Stemming the Superbug Tide. The researchers drew on data from the ECDC European Antimicrobial Resistance Surveillance Network and the laboratory networks that inform the Center for Disease Dynamics, Economics, and Policy’s Resistance Map (OECD, 2018). Their estimates of resistance in pathogen–drug combinations accounted for uncertainty using multiple imputation of missing historical values and estimating correlates of resistance from UN population data and weighted modeling, described in detail in the report (OECD, 2018). The analysis indicated that around 17 percent of bacterial infections in OECD countries overall are resistant to antibacterial medicines, but this prevalence is more than a third in some OECD countries such as Greece, the Republic of Korea, and Turkey (OECD, 2018). Resistance proportions are much higher outside the OECD, over 40 percent in some G20 countries, including China, India, and Russia (OECD, 2018).

The large variation among countries in burden of resistant infections influences projections of mortality. OECD models indicate that there are about 60,000 deaths from resistant infections every year in the United States and Europe (OECD, 2018). By 2050, the OECD model suggests resistant infections will have caused 2.4 million deaths in the same countries (plus Canada, Mexico, and Australia), roughly 30,000 deaths a year in the United States alone (OECD, 2018) (see Figure 3-2). The effects of resistance on quality of life are even stronger. The OECD models of disability-adjusted life years (DALYs), an indicator that accounts for both untimely deaths and time spent in relatively compromised health, suggest 1.75 million years of healthy life are lost every year across 33 study countries (see Figure 3-3). In Italy alone, up to one person out of every 205 could lose a year of life because of infections caused by resistant organisms (OECD, 2018).

Figure 3-2

Projected average annual number of deaths from resistant infections and mortality rate per 100,000, 2015 to 2050. NOTE: AMR = antimicrobial resistance. SOURCE: OECD, 2018.

Figure 3-3

Projected average annual burden of antimicrobial resistance expressed in disability-adjusted life years (DALYs), 2015 to 2050. SOURCE: OECD, 2018.

The OECD’s estimates of mortality and morbidity are the output of published models, the assumptions of which are clearly explained. There are only so many parameters modeling can accommodate, however. All the models presented in Stemming the Superbug Tide were based on resistance in eight common pathogen–drug combinations.² Other resistance patterns will emerge between now and 2050; there are other pathogen–drug combinations that cause serious excess illness and death even today. Accepting the methodological limitations of modeling, the work shows a clear and consistent increasing threat to human health from antimicrobial resistance.

OECD research has also made valuable contributions to understanding the economic consequences of resistant infections. Based on their calculations of morbidity and mortality associated with resistant infections, the report estimated that resistance costs the health system of the 33 countries studied about $3.5 billion a year (adjusted for purchasing power parity), $2 billion a year in the United States alone (OECD, 2018). This finding was consistent with a similar study that estimated the cost to the U.S. health system around $2.2 billion a year (Thorpe et al., 2018).

The OECD also drew attention to the negative externalities (costs to parties other than the patient and prescriber) associated with antimicrobial resistance (OECD, 2018). For example, antimicrobial resistance can undermine confidence in the health system, causing people to avoid in-patient treatment if possible; it can also hurt livelihoods dependent on tourism or agriculture (Thorpe et al., 2018). These kinds of effects are harder to model with any amount of precision, but are useful as a reminder of the potentially devastating downstream effects of a health problem with already devastating short-term consequences.

The O’Neill Report

One of the most influential reports on antimicrobial resistance was the O’Neill report, the 2014 commission from then UK Prime Minister David Cameron to economist Jim O’Neill to analyze the problem of antimicrobial resistance and suggest a mitigating strategy (O’Neill, 2016a). The commission’s final report, Tackling Drug-Resistant Infections Globally, was published in 2016 (O’Neill, 2016b). The report immediately attracted considerable attention in the scientific literature (Matthiessen et al., 2016; O’Neill, 2016a; PLOS Medicine Editors, 2016; Price, 2016; Sugden et al., 2016), from international organizations (FAO and UN, 2021; IACG, 2019; The World Bank, 2016, 2017b), and in the lay media (BBC News, 2016; Boseley, 2016; Roland, 2015; The Economist, 2016). Much of this attention centered around the report’s projection that by 2050 antimicrobial resistance would cause 10 million deaths a year, costing the global economy a cumulative $100 trillion in the same time (O’Neill, 2016b).

These estimates were based on analyses by the RAND Corporation, a nonprofit think tank, and by the management consulting and tax firm KPMG (O’Neill, 2016b). Both models considered resistance to medicines used to treat malaria, HIV, and tuberculosis, as well as hospital-acquired Escherichia coli, Klebsiella pneumoniae, and Staphylococcus aureus (KPMG LLP, 2014; Taylor et al., 2014). Although both models report having made projections for different burden-of-resistance scenarios, only the extreme high-burden scenario were reflected in the O’Neill report (KPMG LLP, 2014; Taylor et al., 2014).

The pathogens causing HIV, malaria, and tuberculosis are not usually central to the discussion of antimicrobial resistance. The burden of disease associated with these infections is vastly greater than those caused by the CDC or the WHO priority pathogens, making it difficult to interpret estimates of their combined affects.

The RAND model assumed that in 15 years none of the medicines licensed to treat these infections will be effective, an assumption that lacks face validity (Friedman, 2020; Taylor et al., 2014). Even in the model appendices, the RAND team cites contemporary estimates of resistance to HIV drugs (~5 percent worldwide) and treatments for multidrug-resistant tuberculosis treatments (~3 percent globally, between 2 and 5 percent in every region except Europe where it is ~16 percent) (Taylor et al., 2014). It is not credible to conclude that total resistance to these medicines in 15 years is in any way likely.

Table 3-3 presents some key results from the RAND and KPMG analyses. These results informed the O’Neill report’s widely publicized estimates of 10 million lives lost to antimicrobial resistance every year by 2050 and $100 trillion cumulative loss in global production (O’Neill, 2016b).

TABLE 3-3

Key Results from the RAND and KPMG Analyses Informing the O’Neill Report and from the O’Neill Report.

None of the analysis informing the estimates in Table 3-3 were formally peer reviewed (de Kraker et al., 2016). The relationship between the commissioned models and the O’Neill report’s conclusions are also somewhat murky (de Kraker et al., 2016; O’Neill, 2014). The O’Neill report’s authors refer to original analyses and information not included in the RAND or KPMG models (e.g., “We estimate that caesarean sections contribute about 2% to world GDP”), but their methods and data are not presented (Friedman, 2020; O’Neill, 2014).

The O’Neill team started from a reasonably credible 2014 base estimate of 700,000 deaths a year from resistant infections (about a third from multidrug-resistant tuberculosis alone) (O’Neill, 2016b; WHO, 2019b). This estimate also has methodological limitations (Schnall et al., 2019). Few if any of the other numbers in the report have such a clear attribution. Though their analytic steps are not clear, one critique concluded, “[T]he scenario that seems to be underlying the most often quoted line [10 million death a year] entails a sharp initial rise of current resistance rates by 40 percentage points, after which rates remain stable until 2050, and doubled infection rates” (de Kraker et al., 2016). A 40 percentage point increase in not consistent with what is presented in the CDC or the OECD analyses discussed previously. One may assume the O’Neill commission believed this to be plausible based on more (rightly) dire predictions in low- and middle-income countries, but it is not clear what scientific research informed their estimates or what their assumptions regarding resistance in different parts of the world might have been. A lack of data from low- and middle-income countries, where the burden of resistant infection is undoubtedly greatest, is a reason to support these countries in routine surveillance for, and prevalence surveys of, resistant infections (Islam et al., 2019). But no estimate of the global burden of resistance can be made in the absence of such data.

The O’Neill report writers may have damaged their credibility by promoting what appears to be only the upper limit of the uncertainty intervals for their conclusions (de Kraker et al., 2016). The writers give no confidence interval for their estimate of 10 million excess deaths a year by 2050 (O’Neill, 2014). (For comparison, 10 million deaths a year is comparable to the global sum of all cancer deaths combined [Sung et al., 2021].) Their estimate that resistance could “cost the world up to 100 trillion USD” contained an uncertainty interval ($60 to $100 trillion) that is usually dropped (O’Neill, 2014).

This is not to say that the O’Neill commission did not produce valuable policy analysis or that its report did not raise the global prominence of the problem. Nor is the modeling of extreme assumptions or worst-case scenarios without value (OECD, 2018). Such models can be useful, especially when they are presented as sensitivity analyses and the caveats on their interpretation made clear. But unfortunately, the O’Neill report’s estimates of both projected mortality and economic consequences of resistance took on a life of their own (Friedman, 2020). When cited, which is often, it is usually without mention of their limitations or the murky analysis that informed them. As a 2016 essay concluded, “Unreliable global estimates like those provided in the [O’Neill report] potentially undermine, rather than support, the fight against a post-antibiotic era” (de Kraker et al., 2016).

The World Bank Report

The World Bank (2017) report, Drug-Resistant Infections: A Threat to Our Economic Future, came out in 2017, building on the momentum of the previous year’s O’Neill report. The report presented estimates of the threat antimicrobial resistance poses to the global economy in terms of lost gross domestic product (GDP) between 2015 and 2050, giving particular attention to costs to international trade, livestock agriculture, and health (The World Bank, 2017).

The World Bank models drew on the RAND estimates that informed the O’Neill report (Ahmed et al., 2017). The model’s low-case scenario was based on the RAND scenario 1, assuming 5 percent antimicrobial resistance from 2015 on; the high-case scenario projected current rates of resistance for 15 years and 100 percent resistance to available treatments after year 15 (Ahmed et al., 2017). Possibly motivated by concerns about the validity of the RAND analyses, the working paper explained, “[W]e avoid the [Rand report’s] extreme cases of absolute resistance”³ (Ahmed et al., 2017). In its final report the World Bank team further clarified that its “simulations are not predictions (rather, a range of outcomes that are possible)” (The World Bank, 2017).

The caveats on the World Bank models are helpful. At the same time, using models driven mostly by data on HIV, tuberculosis, and malaria to inform conclusions about other resistant pathogens in humans and livestock is of questionable validity and should be kept in mind in reviewing the report’s main conclusions.

The World Bank report emphasized that trade and livestock production stand to be seriously affected by antimicrobial resistance, especially livestock production in low-income countries (The World Bank, 2017). Livestock is only a small part (about 2 percent) of the global economy, but its relative value—both in direct terms and as determinant of the health and economic mobility of women and children—is greater in low- and lower-middle-income countries (Ahmed et al., 2017). The World Bank working paper reported that livestock production in low-income countries could fall between 3.1 and 11.1 percent and in lower-middle income countries between 3.1 and 8.9 percent (Ahmed et al., 2017). The effects of reductions in livestock production also influence, but only partially, the projected 1.1 to 3.8 percent deficit in global exports (The World Bank, 2017).

The cost of health services could be perhaps the most directly affected by antimicrobial resistance, given that resistant infections cost more to treat. Increased health expenditures could be felt in cost of medicines, with more expensive antimicrobials being needed, as well as more days spent hospitalized, more consultation time with providers, and increased demand on laboratory diagnostic services. Increasing need for health services puts more pressure on both public and private spending for health, which, coupled with decreasing trade and livestock production, could drive a public deficit. The World Bank models estimate that under a low burden of antimicrobial resistance health costs could increase $330 billion; under a high-burden scenario this increase could be $1.2 trillion (The World Bank, 2017).

Figure 3-4 shows how protracted effects on global economic output might extend for the next 30 years, costing the world between 1.1 and 3.8 percent of annual GDP by 2050 (The World Bank, 2017). These shortfalls could be as serious as during the 2008–2009 global financial crisis (see Figure 3-5) but could extend for much longer. (The shocks of the 2008–2009 crisis lasted only a few years.)

Figure 3-4

Shortfalls in global economic output assuming low and high burden of resistance relative to a baseline scenario, 2019 to 2050. SOURCE: The World Bank, 2017.

Figure 3-5

Costs to the economy of antimicrobial resistance compared to those of the 2008–2009 financial crisis; annual cost expressed as percentage of GDP. SOURCE: The World Bank, 2017.

By taking a disproportionate toll on developing countries, antimicrobial resistance could derail progress on the Sustainable Development Goals, the United Nations’ goals for international development between 2015 and 2030 (UN, 2021; The World Bank, 2017). Through its effects on health costs, trade, and livestock production, the World Bank models indicate that antimicrobial resistance could push between 8 and 28 million people into extreme poverty by 2050 (The World Bank, 2017) (see Figure 3-6).⁴

Figure 3-6

Number of people falling into extreme poverty (living on < $1.90 a day, adjusted for purchasing power parity) by 2050. SOURCE: The World Bank, 2017.

The Health Effects of Resistance in Humans

Chapter 4 discusses the challenges of surveillance for antimicrobial resistance. In short, measuring antimicrobial resistance requires a laboratory capacity and trained clinical microbiologists that are not widely available in low- and middle-income countries (Iskandar et al., 2021). The large, tertiary-care hospitals that can support the required microbiology labs provide a narrow window into the scope of resistance (Gandra et al., 2020; Tadesse et al., 2017; Walia et al., 2019; Wang, 2019). While valuable, this window is not necessarily representative of the national situation. Even in the United States, with its sophisticated laboratory infrastructure, there are challenges in reporting resistant isolates through the regional and national Antibiotic Resistance Laboratory Network, a problem discussed more in Chapter 5.

Regardless of the capacity of the national surveillance systems, there are also challenges in measuring mortality and morbidity from resistant infections. First of all, most resistant infections are seen in patients who have other underlying conditions, making it difficult to know what portion of the clinical outcomes observed can be attributed to the resistant infection (Cassini et al., 2019). For this reason, “scientific debate is ongoing on the appropriate epidemiological study design and statistical inference methods to measure reliable estimates of untoward clinical outcomes attributable to infections with antibiotic-resistant bacteria” (Cassini et al., 2019). For the time being, most research on the clinical outcomes associated with antimicrobial resistance is limited to readily observable, relatively short-term clinical outcomes, including deaths, number of days hospitalized, and risk of developing sequelae (e.g., developing C. difficile infection after treatment for a resistant infection) (Cassini et al., 2019). Risk of death is clearly the most potentially devastating consequence of resistant infection, though only about half of the studies included in a recent systematic review found an increased risk of death in patients infected with a resistant pathogen relative to those infected with a susceptible one (Naylor et al., 2018). It is possible that these differing results are influenced by widely varying methodological approaches, a topic discussed later in the chapter.

Scientists from the ECDC recently published one such analysis of the health outcomes of resistant infections. Drawing on data from the European Antimcrobial Resistance Surveillance Network and health outcome models for specific types of infection (e.g., bloodstream infection, surgical site infection), they estimated between 583,148 and 763,966 infections with resistant bacteria occurred in Europe in 2015, almost two-thirds of them acquired during health care (Cassini et al., 2019). These infections ended in over 33,000 deaths, with the burden of disease (in terms of healthy life years lost) being most severe in infants and older adults; see Figure 3-7 (Cassini et al., 2019).

Figure 3-7

Estimates of the burden of drug-resistant bacterial infections in DALYs by age group, European Union and European Economic Area, 2015 data. SOURCE: Cassini et al., 2019.

Studies such as these help put concrete parameters on the consequences of resistance, parameters that are compelling to policy makers precisely because of their narrow scope and clear boundaries. Nevertheless, some of the most potentially devastating consequences of resistance are the downstream effects that can manifest in increased mortality and complications from seemingly unrelated conditions.

As the previous chapter discussed, antimicrobials are essential for the medical management of surgical care, cancer, and transplant patients, many of whom are immunocompromised. A 2015 model, based on review of randomized and quasi-randomized controlled trials, estimated how loss of effective antibiotic prophylaxis might increase the burden of serious infections and related deaths in the United States (Teillant et al., 2015). The models presented scenarios of a loss of antibiotic efficacy of 10, 30, 70, and 100 percent (Teillant et al., 2015). The authors estimate that a 30 percent reduction in the efficacy of prophylactic antimicrobial treatment for 10 common surgeries and blood cancer chemotherapy would result in an additional 120,000 infections and 6,300 deaths a year (Teillant et al., 2015). Even a relatively minimal 10 percent loss of efficacy would result in 40,000 additional infections and 2,100 additional deaths; the more dire prediction of a 70 percent loss of efficacy would result in 280,000 additional infections and 15,000 additional deaths, as Figures 3-8 and 3-9 show (Teillant et al., 2015).

Figure 3-8

Additional infections per year in the United States under 10 percent, 30 percent, 70 percent, and 100 percent reduction in efficacy of antibiotic prophylaxis. SOURCE: Reprinted from The Lancet, Vol. 15, Aude Teillant, Sumanth Gandra, Devra Barter, Daniel (more...)

Figure 3-9

Number of additional deaths per year in the United States under 10 percent, 30 percent, 70 percent, and 100 percent reduction in efficacy of antibiotic prophylaxis. SOURCE: Reprinted from The Lancet, Vol. 15, Aude Teillant, Sumanth Gandra, Devra Barter, (more...)

Models such as those of Teillant and colleagues draw on an extensive body of research on infection in cancer and surgery patients. The consequences of resistance associated with other common infections, though sometimes more serious, are less amenable to modeling. Drug-resistant infections in the bone and brain, for example, are serious because it is difficult to achieve clinically meaningful concentrations of antimicrobial medicines in these tissues (Nau et al., 1998, 2010; Thabit et al., 2019). Moreover, even small changes in susceptibility of pathogens to medicines can make more surgeries necessary and prompt months-long, or even lifelong, antimicrobial therapy with uncertain results.

Resistant infections can compromise the outcomes of almost every medical treatment. They could also have psychological affects that reduce public confidence in the health system (Foster, 2011; WHO, 2019a). The COVID-19 pandemic has made clear that an infectious disease crisis, accompanied by disruptions in health services and widespread fear or anxiety, can have profound effects on health. English models indicate that cancer mortality may have increased by an estimated 20 percent during the pandemic because of avoided or delayed treatments (Lai et al., 2020). Globally, about 28 million surgeries were cancelled or postponed in the first wave of COVID-19 alone (COVIDSurg Collaborative, 2020). It is not clear how long it will take to clear this backlog or how far-reaching the health consequences could be (Carr et al., 2021). In some ways the most serious risks antimicrobial resistance poses to society are some of the most challenging to quantify.

The overall effects of the COVID-19 pandemic on antimicrobial resistance are hard to predict. On one hand the increased emphasis on hygiene and the decrease in travel and elective medical procedures may have decreased the spread of resistant pathogens both in community and clinical practice (Knight et al., 2021). On the other hand, the vast majority of COVID-19 patients were treated with antimicrobials (Knight et al., 2021). One study found that despite only 7 percent of COVID-19 patients having bacterial infections over 70 percent were treated with antibiotics (Langford et al., 2020). Research in India found that COVID-19 drove over 200 million excess doses of antimicrobials (Sulis et al., 2021). Such extensive exposure to antimicrobials can predispose patients to colonization with resistant organisms. There may also be a cohort of COVID-19 survivors with residual predisposition to lung infections who will need more frequent antimicrobial therapy (Knight et al., 2021).

There has also been considerable presumptive antimicrobial use during the pandemic, both in COVID-19 patients presenting with nonspecific symptoms and “just in case” prescribing to patients with other illnesses who were deflected from care (Knight et al., 2021). As Figure 3-10 shows, the ways COVID-19 has and will continue to influence antimicrobial resistance are varied and warrant further research.

Figure 3-10

Interactions between COVID-19 and the emergence of antimicrobial resistance. NOTE: abx = antibiotics; AMR = antimicrobial resistance; ARO = antibiotic-resistant organism. SOURCE: Knight et al., 2021.

The Economic Effects of Resistance in Humans

Loss of life and disability are devastating health outcomes in their own right. They also affect society indirectly, through the loss of what might have been achieved in years of healthy life. Estimates of how health problems affect the workforce are often of particular interest to policy makers. Most health problems, resistant infections included, hurt the workforce in two ways: the lost productivity among patients suffering from resistant infections and, in some cases, the lost productivity of the workers looking after them (Tillotson and Zinner, 2017). In the same way the health effects of resistant infections can be both direct and indirect, so can the social and economic effects. Table 3-4 shows the many pathways through which health care–associated infections draw a social and economic toll.⁵

TABLE 3-4

The Social Costs of Hospital-Acquired Infections.

In estimating the economic consequences of resistance, researchers must base their analysis on estimates of the health effects. Therefore, the uncertainties and limitations in measuring the health consequences of resistance carry forward to discussion of the costs. For this reason, there are fewer economic studies on antimicrobial resistance (Naylor et al., 2018). The quality of what is published is also lower, a recent systematic review concluded, and held back by a “lack of rigorous, transparent modelling studies which appropriately present or incorporate uncertainty” (Naylor et al., 2018).

The potential for bias in the economic evaluation of antimicrobial resistance is at the root of the many widely variable estimates ($3 billion to $100 trillion) of the economic consequences of antimicrobial resistance published in recent years (Wozniak et al., 2019). In a systematic review on the economic burden of resistant infections, Wozniak and colleagues (2019) commented on this variability: “[E]rroneous or unclear estimates of impact can have alarming effects some of which may contribute to greater action but they also create confusion and potentially undermine the fight against antimicrobial resistance.”

Even a relatively straightforward economic outcome, excess time spent hospitalized, for example, is highly vulnerable to analytic and methodological bias (Naylor et al., 2018; Nelson et al., 2021). Figure 3-11 shows how one recent systematic review identified that different analytic methods, sometimes even reported in the same study, can influence the study’s estimate of excess days hospitalized (Naylor et al., 2018). When measuring excess costs associated with length of hospital stay, for example, studies will often fail to adjust the outcome (i.e., length of stay) to count only those days after the resistant infection started (Wozniak et al., 2019). This time-dependent bias tends to inflate estimates of costs. A recent systematic review found that of 14 studies on the excess costs associated with resistant infections, only two properly accounted for bias in their analyses (Wozniak et al., 2019). After reviewing over 1,000 abstracts the researchers concluded that, while economic valuations of the excess costs associated with resistant infections are sorely needed, especially in low- and middle-income countries, currently the only rigorous and unbiased research available is on health care–associated bloodstream infections with resistant Enterobacterales and MRSA (Wozniak et al., 2019).

Figure 3-11

Estimates of excess length of hospitalization caused by antimicrobial resistance and different analytic methods; (i) through (iii) indicate different methods used in a single study. NOTE: BSI = bloodstream infection; CR = carbapenem-resistant; GN = gram-negative; (more...)

Since Wozniak and colleagues published this meta-analysis, some U.S. papers meeting their criteria for adjusting for bias and confounders have come out. Data from the Department of Veterans Affairs (VA), the largest integrated health system in the United States, with linked records containing cost, microbiological, and clinical information, informed the CDC estimates of costs attributable to resistant pathogens presented earlier in this chapter (Nelson et al., 2021). From a final dataset that included almost 25,000 infections, researchers estimated the costs associated with both community- and hospital-acquired infection for six common resistant pathogens (see Table 3-5). Adjusted over the entire United States, estimates of the direct costs associated with these infections is between $4.1 and $5.1 billion (Nelson et al., 2021).

TABLE 3-5

Adjusted Attributable Cost by Pathogen for Community and Hospital Onset Infections.

Another recent U.S. study compared costs and mortality in MRSA patients to those in patients infected with methicillin-susceptible Staphylococcus aureus (MSSA) (Klein et al., 2019). After adjusting for the many confounders that can influence difference between these groups of patients, the researchers found that costs of hospitalization for MRSA were roughly the same or less than those for hospitalization with MSSA (Klein et al., 2019). This may, ironically, stem from the increasing burden of MRSA infections acquired in the community (as opposed to in a health care setting), as community-acquired MRSA is generally susceptible to second-line medicines, bringing down the overall costs of treatment (Klein et al., 2019). Heightened attention to MRSA in hospitals may have led to the identification and treatment of some minimally invasive MRSA, which in turn influenced these results. It is also possible that aureus infections, even when susceptible to treatment, are simply difficult to manage in clinical practice.

A 2017 study of the burden of carbapenem-resistant Enterobacterales (CRE) found that the cost of one such infection to the hospital was between $22,993 and $35,503, to the insurer or payer between $13,701 and $18,286 (Bartsch et al., 2017). The authors also considered the cost such infections have on society, a cost influenced mainly by assumptions about mortality attributable to such infections. Assuming mortality attributable to the resistant infection was 35 percent, infection with CRE causes between 1,131 and 5,790 deaths a year, costing society between $681 and $3,489 million (Bartsch et al., 2017). Assuming an attributable mortality of 9 percentage points higher would mean 1,422 to 7,279 deaths, costing society between $819 million and $4.2 billion (Bartsch et al., 2017). As a reference, this means the cost of CRE infection alone is higher than many chronic diseases such as high blood pressure (estimated cost to society $672 per patient per year), asthma (estimated direct cost $4,008 per patient per year) and diabetes ($13,015 estimated per patient per year) (Bartsch et al., 2017).

The incurred costs to society from resistant infections is an important point to capture in economic analysis of antimicrobial resistance, partly because of the negative externalities, or the harm associated with antimicrobial use not incurred to the patients or prescriber (Broughton, 2017). The largest part of the negative externality associated with misuse of antimicrobials is the loss of useful antimicrobial medicines in the future. As the previous section explained, the loss of these drugs would influence the risk calculation underlying many basic surgeries as well as more sophisticated treatments such as organ transplantation and cancer chemotherapy. It is difficult to even imagine the potential health consequences of antimicrobial resistance, making the consequent economic burden “at present inestimable” (Smith and Coast, 2013).

The Effects of Antimicrobial Resistance in Food-Producing Animals

As the previous chapter discussed, the contribution of antimicrobial use in food-producing animals to total antimicrobial use and the concentration of resistance genes and drug residues in the environment are a serious concern. Resistance traits that emerge in animals will be found also in manure and water; resistant pathogens from animals can be passed to their handlers, and from them to their family members (Ma et al., 2021). As early as the 1980s researchers have shown an association and plausible pathway through which resistant pathogens emerging in livestock eventually cause human infections (Ma et al., 2021). These include direct contact through consuming food from an infected animal, or indirect routes involving water or a shared environment. Produce can also be a link between resistant bacteria in water or soil and humans (ASM, 2017).

Part of the challenge is that, as in human medicine, it is difficult to know what antimicrobials are being used in livestock and in what doses. The best estimates of use come from U.S. Department of Agriculture (USDA) surveys, which, coupled with the Food and Drug Administration sales data give a rough picture of trends in use (although very little about actual consumption can be inferred from sales data, discussed more in Chapter 5) (Hope et al., 2020). In low- and middle-income countries, the matter is much more difficult to discern. As emerging economies such as Brazil, Russia, India, and China face an increased demand for animal protein they are shifting to more intensive and more efficient systems for raising chickens and pigs (Van Boeckel et al., 2015). Intensive farming of cattle is generally limited to North America, Argentina, and Brazil (Laxminarayan et al., 2015). The shift in production systems will have an effect on global antimicrobial consumption (Laxminarayan et al., 2015).

The projected rise in the use of antimicrobials in livestock, mostly related to an increased demand for animal protein in low- and middle-income countries, has raised a more urgent need for clarity on if and to what extent antimicrobials use in livestock influences human health. Especially in regard to growth promoters, if the gains in efficiency of production are marginal, then it would be easy to justify prohibitions on antimicrobial growth promoters citing only public health concerns (Laxminarayan et al., 2015). But if the gains are larger, then the burden of proof shifts to establishing that the use of antimicrobial growth promoters in agriculture affects human health (Laxminarayan et al., 2015). This question is difficult to answer. Use of antimicrobial growth promoters has been banned in many countries, including the United States (EU, 2005; FDA, 2021). Data on antimicrobial use on farms are often high-level (i.e., sales data), few countries have farm-level data on antimicrobial use (Mesa Varona et al., 2020). What is more, linking antimicrobial use or resistance data from animals to human health outcomes is tenuous.

The Global Research on Antimicrobial Resistance Project

The Global Burden of Disease, Injuries, and Risk Factors program on antimicrobial resistance (the Global Research on Antimicrobial Resistance project) is a welcome addition to the literature on the health consequences of antimicrobial resistance (IHME, 2020a).⁷ This analysis of the health consequences of 23 resistant bacteria (88 microbe–drug combinations) drew on 471 million patient records or isolates from collaborators and public data from around the world (AMR Collaborators, in review). Under a counterfactual assumption of infection with a susceptible pathogen, the authors estimated that antimicrobial resistance killed 1.27 million people in 2019 (95% confidence interval [CI]: 0.91 to 1.71 million) (AMR Collaborators, in review). Such analysis suggests that antimicrobial resistance is the 12th leading cause of death worldwide (AMR Collaborators, in review).⁸ Under a counterfactual assumption of no infection the estimate was 4.95 million deaths associated with antimicrobial resistance (95% CI: 3.62 million to 6.57 million) (AMR Collaborators, in review). Despite limited data from low- and middle-income countries, the models indicated that this is where the burden of resistance is worst, with death rates from antimicrobial resistance highest in sub-Saharan Africa (AMR Collaborators, in review).

Of the 23 pathogens studied, six (E.coli, S. aureus, K. pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa) accounted for a majority (about 72 percent) of deaths (AMR Collaborators, in review). MRSA, a serious burden in high-income countries, caused an estimated 100,000 deaths worldwide in 2019 (AMR Collaborators, in review).

The Global Research on Antimicrobial Resistance study provides a scientifically rigorous framework through which to evaluate antimicrobial resistance and an exhaustive review of the epidemiological data to estimate its burden (IHME, 2020b). Generating comparable estimates of the burden of resistance in key microbe-drug combinations across countries is especially valuable (IHME, 2020c). As Table 3-6 shows, even the estimates of the health consequences of resistant infections vary so widely, including variability in the way they are reported, that it is difficult to identify trends in the literature.

TABLE 3-6

Estimates of the Effect of Antimicrobial Resistance on Mortality from Recent Prominent Reports.

The Economic Component of Antimicrobial Resistance

The problem of wide variability in research is more obvious in reviewing estimates of the economic consequences of resistance. As Table 3-3 showed, even economic researchers working on similar datasets and making ostensibly similar analytic assumptions can arrive at such widely different estimates of the problem as to be unrecognizable. It may be that the most important message regarding the economic fallout of antimicrobial resistance is that it cannot be compartmentalized. As the COVID-19 epidemic has made clear, infectious outbreaks can devastate the global economy and people’s quality of life in far-reaching ways. Even estimates of the cost of the pandemic in the trillions do not account for the long-term, less tangible consequences of disrupted schooling and income (Cutler and Summers, 2020).

Antimicrobial resistance is a One Health problem, so estimating its economic component means untangling the relative contributions of resistance in any one sector and tying them to larger economic indicators. This is not a direct analytic question, and the economic fallout of resistance is not easily reduced to a number. The burden of any one resistant pathogen depends on context; the same resistant infections can have drastically different consequences in humans or animals, if acquired in a hospital or outside of it, in a high-income country or a low-income one. The downstream consequences of resistant infections can be felt on food safety, on livelihoods, on social relationships, and, of course, on health. A lack of communication among the different One Health disciplines may contribute a relatively one-sided body of research on the health and economic effects of resistant infections in humans. This is a major barrier to developing better estimates of the consequences of resistance.