Medicines as Subjects: A Corpus-Based Study of Subjectification in Antimicrobial Resistance (AMR) Policy

1 Antimicrobial Resistance (AMR) and its Subjects

Antimicrobial resistance threatens the very core of modern medicine and the sustainability of an effective, global public health response to the enduring threat from infectious diseases. Effective antimicrobial drugs are prerequisites for both preventive and curative measures, protecting patients from potentially fatal diseases and ensuring that complex procedures, such as surgery and chemotherapy, can be provided at low risk. Yet systematic misuse and overuse of these drugs in human medicine and food production have put every nation at risk. Few replacement products are in the pipeline. Without harmonized and immediate action on a global scale, the world is heading towards a post-antibiotic era in which common infections could once again kill (WHO 2015, 1).

In The Subject and Power, Foucault (1982, p. 780) notes that recent struggles such as “opposition to the power of men over women, of parents over children, of psychiatry over the mentally ill, of medicine over the population, of administration over the ways people live” share something in common that is different to previous anti-authority struggles: they do not confront a powerful institution but a technique of power, a way of exercising power by allocating subject positions that predetermine answers to the question of “who are we?” (1982, p. 781). This technique is what Foucault calls subjectification. He goes on to state that the goal of his work in the previous twenty years had “not been to analyse the phenomena of power… [but] to create a history of the different modes by which, in our culture, human beings are made subjects” (1982, 777). This is the understanding of subjectification and subject adopted and expanded here.

Foucault argues that power is exercised in two, complementary, ways: (1) over individual lives, bodies, in the form of discipline and regulation, and (2) over life as a whole, the population, in the form of biopolitics (1982). Biopolitics takes the administration of life and populations as its subject: “to ensure, sustain, and multiply life, to put this life in order” (Foucault and Hurley 1998, 138). Drug resistance – or antimicrobial resistance (AMR), as it is often called – is an issue that concerns biopolitics, the governance of global regimes of production and consumption which have been enabled by antimicrobial drugs and depend on them for continued survival. The World Health Organisation (WHO) defines AMR as microbes becoming resistant to medicine (WHO 2015, 1). This is not a problem but a fact: simply using antibiotics create resistance. Penicillin was first used in 1941, penicillin resistant bacteria were detected a year after, and the circle has repeated ever since, although it had started long before (Landecker 2016, 28). Nowadays “[t]he scale of production is also the scale of resistance” (Landecker 2016, 20). The use of antimicrobials creates resistance, but without antimicrobials current standards of health and economic growth cannot be maintained. It is this dependency on antibiotics as part and parcel of the infrastructure for human life that makes resistance a matter of biopolitics. However, policy, the genre analysed here, is a disciplinary form of power exercised over the actions of human beings, as individual citizens or as institutions, such as governments or hospitals.

A policy is a particular type of document produced by governments and international organisations which sets out programmes for actions to reach goals or solve problems. The key document at the centre of AMR policy is the WHO’s Global Action Plan (GAP). The first paragraph of the GAP (WHO 2015), quoted at the beginning of the paper, explains the importance of AMR to the survival of the population in no uncertain terms. However, as Bacchi (2009), following Foucault, notes, the way a problem is represented in policies contributes to defining the problem. In determining when or why a particular practice or behaviour becomes problematic, policies have real and meaningful effects, both in terms of how we come to collectively think and speak about that problem, and what options are considered possible or, even, imaginable, in order to solve it (Bacchi 2009). The influence of policy documents such as the WHO’s GAP should not be underestimated.

Bacchi’s methodology, designed to reveal the workings of Foucault’s theory of governmentality, has been highly influential in policy analysis. Rather than taking the explicit aims and logic of the policy for granted, Bacchi proposes a list of six questions to ask. The first one gives the methodology its name: ‘What is the Problem Represented to be?’ (WPR). The second question is about underlying assumptions and gaps. The third asks how the problem was constituted, from a genealogical perspective. Subjectification forms part of the fourth question: what effects are produced by this representation of the problem? The final question is how the representation has been disseminated and how it could be disrupted. As Bacchi acknowledges, the questions are interdependent and it is impossible to completely separate one from another. In this study, I am concerned specifically with subjectification processes, although in attempting to answer how actors are represented in health policy, I will inevitably reflect on the assumptions that underlie the representations of subjects and how they impact the representation of the problem as a whole.

Bacchi explains that subjectification results from the way in which different kinds of subject positions are made available in policy discourse (2009, 16). To illustrate, Bacchi uses drugs and alcohol policy and refers to what she calls, after Foucault, dividing practices, which divide people into responsible drug consumers/drinkers, who are in control, and problem drug consumers/drinkers, who are not (Bacchi 2009, 93). A limitation of Bacchi’s methodology is that she seems to assume that the subjects of a particular policy are self-evident. In many cases, they are; not only in policies that target specific human practices, such as drinking alcohol, but also in policies destined to promote inclusivity and equality, which tend to target certain populations explicitly, such as LGBTQ+ or people from underrepresented ethnic backgrounds. However, in AMR policy and others that do not target specific populations, the identification of subjects is less straightforward.

The nouns employed to refer to the population in policy discourse – such as patient(s), consumer(s), citizen(s), – may refer to the same human beings in the real world but they nevertheless have distinct meanings. Each noun indexes different positions available to human beings; for example, their need for medical treatment, in the case of patients, or their financial capacity to access medical treatment, in the case of consumers. Naming is one way in which discursive subjects are represented in texts and, as the examples above show, co-existing and arguably competing representations may interact in the same text. This is why, I argue, a systematic account of patterns of subjectification in texts also needs a more fine-grained model for the identification of subjective agency than the one sketched by Bacchi (2009). Here, I employ van Leeuwen’s “sociosemantic inventory” of the ways in which social actors can be represented in texts (2008, 23).

Like Bacchi’s, van Leeuwen’s framework is grounded on a Foucaldian understanding of discourse. Van Leeuwen’s point of departure is that texts are representations of social practices, and he aims to demonstrate how these are realized linguistically. Social practices, according to van Leeuwen, need “first of all a set of participants in certain roles” (2008, 8). Arguably, the process of allocating people to certain roles is what subjectification is all about. However, linguistic models of social practice such as van Leeuwen’s also have their limitations. An abstract concept such as subject cannot be easily mapped to grammatical functions. The grammatical subject of a clause is not necessarily a discourse subject in Bacchi’s terminology and, vice versa, a discourse subject may take many different grammatical roles. Nevertheless, as I show below, identifying social actors provides a good starting point to reveal how different groups are represented in relation to the problem under construction. What is more, a linguistic model is more easily compatible with corpus-based analysis, and without offering evidence of consistent patterns across collections of texts it is difficult to speak with confidence about patterns of representation in texts.

In brief, this study follows Bacchi’s methodology but makes important innovations to her model. The first one is adopting a more nuanced framework for the analysis of subjectification which is grounded in linguistic categories, but not – it should be noted – limited by them. The second is the use of corpus analysis to provide empirical evidence of patterns of representation across texts. A third way in which this study makes a contribution to the analysis of subjectification in discourse is by extending the concept of subject and agency beyond human subjects. This is a limitation of both Bacchi’s and van Leeuwen’s methodologies. Social agency, as illustrated below with the example of AMR discourse, is not only attributed to people but also to medicines and microbial activity, and only by taking into account the agency of non-human actors can we obtain an accurate picture of the processes of subjectification within an area of policy.

2 Data and Methods: Social actors in the SHE-AMR subcorpus

The analysis reported here was conducted on 36 texts (430,635 tokens) included in the Sustainability and HEalth (SHE) corpus. The SHE corpus, currently under development at the University of Oslo, contains health and health-related texts and aims to support research and teaching in the fields of healthcare and sustainability. The corpus is accompanied by an open access software interface, ModNLP, developed specifically for use with the SHE and the Geneaologies of Knowledge corpora (Luz and Sheehan 2020). One of the advantages of the SHE corpus is that it records the metadata associated with each text and it is easy to select a sub-corpus restricted to a specific genre, date, topic, author and a range of other categories that are annotated alongside the text. The analysis presented here employs a sub-corpus, SHE-AMR, collected with the aim to compare global and national (Indian and Sri Lankan) policies on AMR, in the context of the EquityAMR project. As well as the GAP and other publications by WHO, it includes documents from the Wellcome Trust, and national action plans and related policy documents for India and Sri Lanka. Initial results suggest that Indian and Sri Lankan policies follow the GAP and other global policy documents very closely.

The GAP replaced a previous policy document published in 2001: the Global Strategy for the Containment of Antimicrobial Resistance (GS-CAR). Both documents are included in SHE-AMR. Since my focus is on processes of subjectification, it is important to note that the GS-CAR adopted a “people-centred” approach “with interventions directed towards the groups of people who are involved in the problem and need to be part of the solution, i.e. prescribers and dispensers, veterinarians, consumers, policy makers in hospitals, public health and agriculture, professional societies and the pharmaceutical industry” (WHO 2001, 2). In contrast, the 2015 GAP adopted a “one-health approach”, which acknowledges the interdependence between human, animal and planetary health (WHO 2017, vii). The One Health concept was developed in 2004 and its main novelty was the recognition that the health of the ecosystem, including the health of, for example, wild fauna, was inseparable from human health (Destoumieux-Garzón et al. 2018). The GAP thus appears to offer a less anthropocentric approach than the GS-CAR, although both identify similar group of actors.

Comparing the actors listed in the two documents is useful to illustrate an overarching distinction made by van Leeuwen’s (2008) sociosemantic inventory, that between personalization and impersonalization. Participants are personalized when represented as human beings, that is by “personal and possessive pronouns, proper names or nouns … whose meaning includes the feature ‘human’” (2008, 46).^[1] Policy and other institutional text genres rarely refer to identifiable individuals. More often than not, actors are personalized by means of common nouns, such as prescribers, dispensers, consumers, veterinarians and policy makers in GS-CAR.

Van Leeuwen distinguishes two types of impersonalization: abstraction and objectivation (2008, 47). Abstracting social agency involves representing actors by means of a quality assigned to them; professional societies and the pharmaceutical industry in the GS-CAR can be considered abstractions. Objectivation involves representing actors by means of reference to a place or thing with which they are closely associated (2008, 46); hospital(s) is commonly used with this purpose in AMR policy, as discussed below.

In the GAP, the agency of all actors, with the exception of consumers, is represented in the abstract, as medicine, agriculture, finance and environment. The exception of consumers is important because, it has been argued, health policy tends to locate responsibility at the level of citizens (Bacchi 2009, 138; Bunton and Petersen 1997; Chandler 2019; Mol 2008; Petersen 2003). Petersen, for example, describes a stakeholder society where “[c]itizens are increasingly expected, as a condition of access to health care services, to play their role in minimising their contribution to health care costs by becoming more responsible health care ‘consumers,’ and adopting appropriate practices of prevention” (2003, 194). According to Petersen, in this framework, more and more, the concept of the active consumer has come to replace that of the passive patient (Petersen 2003, 193). However, in SHE-AMR, people are represented as both active consumers, with responsibility for their use of antibiotics (Example 1), and passive patients who require antibiotics (Example 2).

Representing agency through nouns, rather than verbs, as is the case in SHE-AMR, is a well-known linguistic strategy known as nominalisation in systemic functional grammar. As van Leeuwen explains, nominalisations represent processes as facts rather than as the actions of particular groups of people, lending a certain impersonal authority to the text. To illustrate, the word use can, in theory, be used as a noun or a verb. In SHE-AMR, use occurs 1730 times in total; however, only 154 of those instances are verbs. The following examples illustrate how the actor that carries out the action of using can be hidden, as in Example 3, where it is unclear who is the actor that needs to intervene or reduce the use of antibiotics, or explicitly declared, as in Example 4, where it is employed as a verb and responsibility is clearly attributed to ‘poultry farmers’.

The SHE corpus is not tagged with linguistic information so, for the purposes of this study, a frequency list was filtered manually for types that could potentially identify participants and ways of naming them. This could be done in relatively little time given the relatively small size of the sub-corpus and allows a rigorous selection. Table 1 shows the eleven nouns appearing amongst the fifty most common types. It was noted that use appears 154 times as a verb; adjusting the frequency accordingly to reflect only instances of use as a noun would not change its rank on Table 1.

Outside frequency lists and concordances, the main tool employed to identify and display collocation patterns was the concordance mosaic, or mosaic for short (See Figure 1.). A mosaic offers a concise summary of the collocates to be found in the immediate context of a keyword, that is, the string being searched for, which is often a word but can also be a regular expression (‘patient*’ in Figure 1). The results are presented in a key-word-in-context (KWIC) style, with the keyword aligned in the central column and surrounded by the words that co-occur to the right and left of the keyword. While a concordance shows all the words as they appear in the text, the mosaic shows only those that are most frequent at each position in the concordance’s context with respect to the keyword. Thus, a mosaic represents positions relative to the keyword as ordered columns of tiles.

Figure 1:

Mosaic of ‘patient(s)’ in SHE-AMR.

In Figure 1, the results are also filtered by using a list of stopwords, that is, very common function words that are excluded. The horizontal bar to the left of the stopwords button, the slider, was also employed to select the frequency ranges of the collocates to display; this is to ensure readability when the collocates are too many to be presented legibly within the mosaic window.

3 Results: Subjectification in AMR Policy

The frequency list for SHE-AMR shows that there are several nouns amongst the first hundred words, which is unusual for an English corpus. This seems to be a characteristic of the SHE corpus and not only the selection of documents for this study. The frequency lists for both SHE and SHE-AMR include over thirty nouns among the first hundred most frequent words. For comparison, among the most frequent hundred words in the British National Corpus, only one noun, people, appears at rank 85. People also appears in SHE-AMR, at rank 78, but as shown in Table 1, it is preceded by many other nouns. There are a total of 11 nouns amongst the 50 most common types in SHE-AMR.^[2]

The majority of nouns in Table 1 do not denote human participants but actions: resistance, use, action; events: infection(s), and commodities: drug, antibiotics, data. As noted above, nominalisations are characteristic of policy discourse and convey an authoritative and impersonal tone. The only nouns that contain the semantic feature human among the first hundred most common words in SHE-AMR are patients (rank 30), and people. A mosaic of people (Figure 2) shows that they are divided into groups according to age (younger, older) and ethnicity (white, BAME). People (Figure 2) and patients (Figure 1) are often quantified as the majority, many or a large number of, a strategy that van Leeuwen describes as aggregation, allowing participants to be treated as statistics (2008, 37).

Figure 2:

Mosaic of ‘people’ in SHE-AMR.

When referred to as patients, the population is represented as passive. In Figure 1, patient(s) is used as the grammatical subject of verbs such as undergoing, receiving and admitted. Patients are further qualified as infected, as the object of care, and categorized in terms of what increases their risk of infections (organ-transplant, cancer, ICU, HIV, trauma, diabetes, etc.).

A common strategy involves referring to people by their professional role. As noted above, the GS-CAR clearly targeted “prescribers and dispensers” (WHO 2001, 2), but the GAP does not target specific groups of professionals. In SHE-AMR, there are 98 occurrences of prescriber(s), 78 of which appear in GS-CAR. Terms such as doctor(s) and nurse(s) are rarer (72 and 69 instances respectively), and tend to be aggregated in lists, as shown in the mosaic for nurse(s) (Figure 4). One of the features of mosaics in ModNLP is the possibility of clicking on one cell to find what other cells are co-selected with it; in other words, to highlight collocates of collocates (co-collocates). Figure 3 shows that doctor is the most frequent collocate to the left of nurse and, when doctor is selected, pharmacist, staff and professionals are co-selected, indicating that these terms are usually listed together. The mosaic also shows training as a collocate to the left of nurse, suggesting that nurses are often portrayed as the object of, or in need of, training. Had training appeared to the right of nurse, we would expect nurse to be the grammatical subject of training and therefore in charge of delivering rather than receiving training. Although the mosaics cannot be shown for lack of space, training also features as a collocate to the left of doctor(s), pharmacist(s), and staff.

Figure 3:

Mosaic of ‘nurse(s)’ in SHE-AMR, showing co-selections.

Figure 4:

Mosaic of antibiotic* in SHE-AMR.

It seems clear from the analysis above that the agency of most actors, in van Leeuwen’s terminology, is impersonalised and abstracted. However, there are also examples of what van Leeuwen calls instrumentalization: “a form of objectivation in which social actors are represented by means of reference to the instrument with which they carry out the action in which they are represented as being engaged” (2008, 47). The example used by van Leeuwen is that of a “mortal shell” reported in a newspaper as having “slammed into Sarajevo’s marketplace” (2008, 47). Arguably, Example 5 shows an instrumentalization of human agency, whereby antibiotics are left to stand in for the agency of the humans who prescribe or administer them. Example 6 can be found in a report by the Wellcome Trust entitled Framing resistance: How to communicate about antimicrobial resistance effectively, which suggests that “antimicrobial resistance as undermining modern medicine” is the “most compelling frame” to help the public understand the breadth of the issue (Wellcome Trust 2019, 14). Example 6 was one of a series of messages tested for impact, alongside “drugs don’t work” and “medicines are losing the war against bacteria” (Wellcome Trust 2019, 14).

The objectification of human agency is the furthest van Leeuwen will go in terms of blurring the boundaries between the world of humans and that of objects. In his model, as in Bacchi’s, the assumption is that subjects, actors and agency are human. Theoretically, a focus on how power circulates among humans, rather than how it is exerted on other lives, may be an understandable restriction. Methodologically, however, we soon encounter difficulties because, in texts as in life, the distinction between human and non-human agency is not neatly realised. Contemporary sociological theories of agency – Actor Network Theory being one of the best known in the field of health (Cresswell, Worth, and Sheikh 2010; Latour 2005) – and anthropological approaches to agency such as Appadurai’s (1986), adopt a broader understanding of agency, where material objects and technologies, such as medicines, have social agency. Once we dismiss the quality of being or representing human agency as a requirement for social agency and consider the possibility that subjects can be material objects and non-human species such as microbes, the picture painted by the analysis widens considerably. As well as antibiotics, other important actors come to prominence. Reconsidering the list in Table 1 from this broader perspective, it is impossible to ignore that amr (rank 11), health (17), resistance (rank 21) and infections (rank 29) rank higher than patients (rank 30). Resistance embodies the agency of microbes and, arguably, so does infections, as results of their actions. A mosaic created using resist* as a keyword shows that compounds such as antimicrobial resistance, drug-resistant infections and resistant bacteria are often preceded by verbs such as combat, contain, or address. Thus, bacteria and infections are presented as agents, more precisely enemies, to be fought.

On the other hand, a mosaic of antibiotic* (Figure 4) indicates that these drugs are not often represented as active participants, curing or healing patients, but as objects of consumption. The most common verbs to the left are take, taking, taken, using, prescribed, use, prescribe. To the right, use, used and consumption are among the most frequent collocates. A common pattern in the texts seems to be: ‘people take antibiotics without prescription’. Interestingly, however, the mosaic of drugs presents a different pattern. The most common verbs to the left include use but also develop, which, together with the frequent use of development and sale suggest that, although potentially referring to the same substance as antibiotics in the real world, drug* represents it from a different point of view, that of developers rather than users. A mosaic of the 296 concordances of medicines shows that it shares with drugs the three most common collocates to the right – antimicrobial, new, essential. However, it is the use, access and investment in medicines, rather than their development, that is discussed. I come back to these examples in the discussion.

In the same way that the population as a discourse subject is represented alternatively as consumers and patients, medicines as a discourse subject and, more specifically, as a commodity with social agency, can be represented as antibiotics or drugs. What is interesting is that antibiotics and drugs are employed to represent the same commodity from two different perspectives. In the following section, I discuss in more depth the implications of considering medicines as a subject in AMR policy to illustrate the insights that a corpus-based analysis of subjectification in policy discourse may offer when applied within a broader theory of social agency.

4 Discussion: Medicines as Subjects

According to Bacchi (2009, 17) a particularly important subjectification effect consists in the allocation of responsibility: “representations of problems usually have built into them implications about who is responsible for the ‘problem’”. One of tasks of the analyst is to bring these implied attributions of responsibility out in the open “so that we can consider whether or not we believe that responsibility is being sheeted home appropriately” (2009, 17). It was suggested above that the GAP presents an ostensibly less anthropocentric approach to policy. However, as Chandler notes, despite emphasizing AMR as a one health problem involving not only human but also animal health and agriculture, the proposed solution, ironically, focuses on individuals as targets of interventions (Chandler 2019, 7). Locating risk and responsibility at the level of citizens is a characteristic of the neoliberal context within which the policy is developed, which involves reduced government spending on health care and increasing reliance on technologies of self-governance, to make individuals and communities more self-reliant and self-determining (Petersen 2003).

As this study has shown, however, AMR policy represents the population as both active consumers and passive patients. People are represented as at risk and responsible for the problem that makes them vulnerable, by taking antibiotics without prescription. Alongside the production of patients as consumers, suggests Bacchi, “we see the production of doctors as service providers. Potentially then doctors become subject to the kind of surveillance that accompany other forms of ‘client-provider’ relationships” (2009, 199). This is also clearly the case in SHE-AMR, where healthcare professionals are represented as passive and in need of training as well as surveillance. Another important process of subjectification revealed by the analysis is one whereby drugs, infections and resistance are subjectivised, becoming social actors. Infections and resistance represent the agency of microbes, who are portrayed as the common enemy.

The traditional representation of microbes and humans at war has been examined and questioned by authors such as Sariola and Gilbert (2020) and alternative ways of framing that relationship have been proposed (Ironstone 2019). Less attention has been paid to the agency of drugs, with the exception of Whyte et al.’s Social lives of medicines, inspired by Appadurai’s The Social Lives of Things. Appadurai (1986) proposes that following an object’s biography – their social life – can provide insights into how they acquire and exchange value. Whyte et al. (2002, p. 5) define medicines as “substances with powers to transform bodies” and argue that they are social phenomena because they “are used intentionally to achieve an effect in some body” (2003, p. 6, emphasis in original).

Value, notes Appadurai, “is never an inherent property of objects but is a judgement made about them by subjects” (Appadurai 1986, 3). In SHE-AMR, collocates such as essential, effective, new, quality, investment, and development, show that drugs and medicines are highly valued and in demand. What is less clear is whose agency they represent. According to Whyte et al. (2002, p. 17) “[p]hysicians are the prototypical providers of medicines in professional medical traditions”. In SHE-AMR, however, medicines are not represented as embodying the power of physicians but as a commodity, and health professionals are generally represented as in need of training and surveillance because of the risk of overusing or misusing antibiotics.

In declaring drugs essential, the policy is ostensibly concerned with human and planetary health; however, patients, consumers, and healthcare staff are also portrayed explicitly as responsible for optimising the use of medicines and implicitly at fault. The agency behind the development and promotion of medicines, which was clearly, if indirectly, targeted in GS-CAR, has all but disappear in policy documents since the publication of the GAP. These findings align with those presented by Collins, Jaspal, and Nerlich (2018), who focused on UK news discourse. Collins et al. carried out a transitivity analysis, focusing on overt ways in which particular agents are implicated. They conclude that it is antibiotics and the infections they are designed to treat that are instilled with agency, marginalising pharmaceutical industry responsibilities in the process. It is this industry, however, that would benefit first from the investment in the development of new drugs.

The GAP (WHO 2015, 1) explicitly declares that its “goal … is to ensure, for as long as possible, continuity of successful treatment and prevention of infectious diseases with effective and safe medicines that are quality-assured, used in a responsible way, and accessible to all who need them”. A literal reading of this goal suggest that it is medicines, and not patients nor the general community, never mind planetary health, who are the utmost concern of the policy. Read against the background of a health emergency, as in the quote at the beginning of this article, and supported by objectives that prioritise information, prevention and access, the meaning of the goal as formulated above may go unnoticed. However, when read against the analysis presented here and in Collins, Jaspal, and Nerlich (2018), it suggests a coherent principle underlying the problematisation of AMR resistance in health policy.

5 Conclusion – Medicines before Humans

In this paper, I have explored a corpus of AMR policy discourse to demonstrate how corpus analysis can contribute to the study of subjectification in policy analysis, along the lines proposed by Bacchi (2009). I have demonstrated how van Leeuwen’s (2008) framework for analysing the linguistic representation of agency in texts can be usefully combined with Bacchi’s methodology to offer a more nuanced and detailed picture of competing subjectification processes in policy. As well as offering insights into the processes of subjectification in AMR policy, this paper has made a contribution to the concept of subjectification as a technique of power.

Foucault’s concern was with how human agency was taken away from individuals in the interests of biopolitics, the regulation of life at the level of the species. Foucault’s argument is that power is taken away through subjectification; this power is exerted in and through discourse, by creating subject positions that offer predetermined answers to the question of “who we are?” (1982, 781). In this paper, I have argued that one of the ways in which this field is structured is through making others into things. Medicines are the key subjects in AMR policy, where they are represented as commodities. Commodities are generally understood to obey the law of the market, and this state of affairs is not challenged by AMR policy; therefore, the agency of those in charge of developing and promoting medicines, the pharmaceutical industry, is left outside the disciplinary scope of the policy. This is, therefore, a reversal of the process of instrumentalization described by van Leeuwen, where objects carry out the action of humans. In SHE-AMR, medicines are the primary agent and healthcare staff are the instruments through which they work. Medicines are the precondition for the actions of humans.

The value of medicines in SHE-AMR is attributed to their power to control infections. This means medicines protect human and animal lives, and their positive effect on planetary health is debatable. However, medicines also have financial value, not only for pharmaceutical companies but for all the actors involved. In this study, I have focused on the discourse of policy makers, for whom medicines are a valuable and expensive commodity. Had I focused on the discourse of patients and consumers, the value attached to medicines may have been different. Whyte, van der Geest, and Hardon (2002, 15) note that, for consumers, medicines also have social efficacy, and can express issues of “identity, control and power” (1992, 16, emphasis added).

The relationship between medicines and the concept of One Health needs to be further explored and I hope this initially exploration will pave the way for new questions to be asked. This study has established that the goal of policy makers is to preserve the power of medicines, as a precondition for the health of the population and that of the planet. It seems reasonable to ask, for example, what a policy that had at its goal the health of the population and the planet, independently of the fate of medicines, would look like.