Managing risks or generating uncertainties? Ambiguous ontologies of testing in Australian healthcare

Introduction

Testing has become pervasive in modern healthcare, being undertaken for purposes of preventive screening or for clinical diagnosis. While testing is often promoted as an effective tool for risk assessment, diagnosis and health monitoring, recent research suggests that some, if not much, of the optimism for testing may be unfounded. For example, although many people benefit from early diagnosis following testing, some tests may lead to a cascade of further tests, increased uncertainty and, sometimes unnecessary, potentially harmful treatment (Moynihan et al., 2013). Moreover, contrary to the promise of evidence-based medicine (EBM), proven healthcare benefit does not always explain why certain tests are routinely ordered – suggesting a need to consider the sociocultural and epistemological factors that shape testing practices. Conceptually, testing regimes rest on the premise that it is possible to establish certainty and hence individual ‘ontological security’ (Giddens, 1991) by measuring and managing risk which, by definition, is calculable and therefore, knowable. National cancer screening programmes, which operate in Australia, the United Kingdom, the United States and other countries, reflect this striving for calculability, certainty and ultimate control. Similarly, the routine (and growing) use of tests in clinical practice is underpinned by optimistic expectations regarding the prospect of definitive assessments of risk or diagnoses leading to effective treatments and improved care (Bossuyt et al., 2012).

This article examines how a sample of Australian health professionals delineate the boundary between risk and uncertainty in their accounts of medical testing. Health professionals are busy people and it can be difficult to research their views, as we discovered. However, we have been fortunate in securing rich data derived from qualitative interviews with a small sample of professionals (n = 11), along with qualitative data drawn from an online survey (n = 70). The material on which we draw is part of a larger sociological study exploring the factors that shape testing in healthcare with a particular focus on the sociocultural processes underpinning optimism for the use of testing technologies. The focus on health professionals’ views and experiences emerges from the recognition that they play a critical, front-line role in testing regimes: in ordering tests and interpreting the results; in offering risk management advice or deciding on related treatment and care plans; and in subsequent referrals to other specialists who may employ their own distinctive panoply of tests and related interventions. They serve as gatekeepers to the institutions of healthcare and mediators of the many contending truth claims that surround health, risk, treatment and care. Patients and their families depend on their knowledge and advice for personal decisions of great consequence. Consequently, investigating testing practices as forms of ‘evidencing’ on which Australian health professionals rely is crucial for understanding the decision-making processes that contribute to the use of expensive, high-tech and potentially invasive interventions.

Applying concepts from science and technology studies (STS), notably Latour’s influential ethnographies of science-in-action, and drawing on data from the interviews and surveys, we examine how health professionals invoke ‘evidence’ in their accounts of testing. As we argue, for the professionals in our study, the ‘evidence’ produced through testing has ambiguous ontological significance: while it promises diagnostic certainty and a clear direction for treatment and care, it often generates uncertainties, leading to yet further testing. Our analysis also illuminates participants’ views on the multiple functions that testing performs and, in particular, their understanding of the double-edged affective implications of testing, which may either relieve or generate anxiety in the clinical encounter. On the one hand, their responses suggest that by bringing about diagnostic closure, testing may provide reassurance and certainty to patients and clinicians, thereby helping to assuage anxiety; on the other, by reinforcing diagnostic uncertainty and the discomfort of not knowing, it may heighten anxiety for both patients and clinicians. We discuss these findings in light of recent trends in medical testing, including the move towards the ‘mainstreaming’ of genetic testing into healthcare, which promises to reduce uncertainties through the provision of earlier, more personalised tests and treatments. To begin, we elaborate our approach and the foundations on which it rests, and the methods and data that inform our argument.

Medical testing, ‘evidencing’ and the management of uncertainty

Over the past few decades, EBM has become firmly established as the dominant paradigm in medical practice as part of an effort to standardise care (Timmermans and Angell, 2001). Given EBM’s reliance on scientific evidence as the basis for medical decision-making, it is perhaps unsurprising that medical testing, with its purported capacity to produce objective evidence based on standardised measures, has been widely embraced as a reliable, accurate means for diagnosis and risk monitoring (Gøtzsche, 2008). However, work on diagnosis in health sociology and STS has convincingly shown that medical tests do not simply measure an objective reality of health risk or disease. Rather they are shaped and transformed by the contexts in which they are applied, and therefore, careful attention needs to be paid to the clinical practices, health systems and broader socio-technical landscape in which testing is embedded (Engel, 2012; Gardner et al., 2011). The work of Annemarie Mol (2002) is noteworthy in this regard as it tracks how disease is constituted through daily practices across different sites, such as the hospital and the laboratory. Focussing on the diagnosis and treatment of atherosclerosis, Mol argues that diagnostic devices do not simply register the facts of disease; they intervene to enact disease in particular ways, for example, by delimiting the boundaries between what counts as normal and what counts as pathological.

Related research in the sociology of diagnosis has made a similar point, highlighting that diagnostic tools do not exist independently of diseases, practitioners and particular clinical practices, but rather materialise through them (Armstrong and Hilton, 2014; Aronowitz, 2015). Such research has worked to dismantle the common-sense realist view of diagnoses as scientific representations of an underlying biological reality. Instead diagnoses are found to be the outcome of ongoing interactions between health professionals, patients, medical technologies, clinical spaces and social systems. Within this body of scholarship are studies that track the performativity of diagnosis, that is, the ways in which it operates variously as a mode of disease classification, a tool of categorisation and a social process with consequences for those labelled with a particular diagnosis (Jovanovic, 2014; Jutel, 2011; Salter et al., 2011). In a study that brings the sociological literature into dialogue with STS-inspired accounts of diagnosis, Engel et al. (2017) focus on diagnostic processes at point-of-care in India, where a fragmented health system means that patients have to do a lot of coordination work to make testing function and ensure continuity of care. Like Engel et al. (2017), our research draws on the insights of both the STS and sociology of diagnosis literatures to track the work needed to make diagnostic tests function in practice, the assumptions on which they rely and their material effects in the clinical encounter.

We wish to make a critical intervention by exploring how health professionals present the role of testing in relation to managing risk and uncertainty in clinical practice. In doing so, we build on classic and recent research within STS (e.g. Hollin, 2017; Star, 1985) that has considered the diverse forms of uncertainty and ‘uncertainty work’ operating in medical and laboratory settings. As part of a growing literature on the ‘sociology of low expectations’, this research draws attention to the generative power of uncertainty tracing its different forms and effects. In some cases, the presence of uncertainty enables experts to attend to the complexity of the subjects with which they work (Fitzgerald, 2014; Street, 2011), while in others, it can operate more perniciously to allow experts to immunise themselves from liability for technoscientific controversies (McGoey, 2009) or to manage risk and negotiate regulatory contingencies (Tutton, 2011). The dynamics of uncertainty are, of course, not confined to scientific and other experts; they also affect patients, especially those diagnosed with chronic diseases who come to live with these diagnoses by managing the complex interplay of uncertainty, fatalism, hope and the promise of medico-scientific progress (Browner and Preloran, 2010; Petersen et al., 2017).

Our analysis contributes to this literature on the generative power of uncertainty within medical settings. Focussing on professionals’ accounts of medical testing, it considers the processes involved in producing test outcomes as a form of ‘evidence’ to guide diagnosis and treatment. Our analysis interrogates the claims to accuracy and reliability that underpin testing practices, illuminating the epistemological assumptions and the hierarchy of evidence on which they rely. Importantly, we are not implying that these assumptions are somehow incorrect; rather that our analysis suggests that the different processes involved in producing and interpreting tests are often disregarded or ‘bracketed away’ (Law’s (2011) term) when test results are translated between medical sites, for example, between the laboratory and the clinic. This allows the limits and uncertainties of testing to be overlooked when tests are applied in clinical practice. In tracing the processes of evidencing at work in diagnostic testing, our analysis aims to make visible the different ontologies of testing in health professionals’ accounts, the tensions between them and their implications for clinical practice. In this respect, we draw attention to the ways in which testing is paradoxically presented in professionals’ accounts as both a means of achieving diagnostic certainty and a source of uncertainty.

Approach: doing science, making evidence

To theorise the processes of evidencing in relation to testing, our analysis draws on the influential ethnographies of Latour (1987) and Latour and Woolgar (1986) on science-in-action. Challenging the uncritical acceptance of scientific facts as objective truths, this work traces the social and historical processes involved in producing facts. As Latour and Woolgar (1986) explain, the process of generating scientific facts requires ‘slow, practical craftwork’ (p. 236) involving a network of human and non-human actors, including technicians, computers and peer-reviewed publications. Some of the actors in this network function as ‘inscription devices’ which refer to ‘any item of apparatus or particular configuration of such items which can transform a material substance into a figure or diagram which is directly usable’ (Latour and Woolgar, 1986: 51). In the context of medical testing, inscription devices include imaging techniques (mammography, X-ray imaging) and blood tests. The significance of inscription devices is that the image or written document they produce (the ‘inscription’) is assumed to have a direct relationship to the original object of study. This is important because the process of inscription makes it possible to ‘“bracket away” those events and the contingencies involved in [the] production [of scientific facts] and deal, instead, with more mobile and tractable inscriptions’ (Law, 2011: 27). In order for a ‘fact’ to emerge, all evidence of its inscription process must disappear from view, including the inscription devices, the material setting and methods, and the researcher(s). In other words, once knowledge becomes naturalised as ‘fact’, the history of its production, and the negotiations, practices and controversies therein, become hidden from view.

Relevant to this analysis are two key insights. First, facts are constructed through inscription devices and practices that transform the material world into an object of study. Importantly, this does not imply that facts are any less real for being socially constructed, but rather that they are necessarily contingent on particular social processes. This means that different inscription devices and knowledge-making practices will produce different facts. A focus on the construction of scientific facts disrupts the imagined distinction between knowledge and practice, illuminating the ways in which evidence is ‘always inextricably combined with the actions, interactions and relationships of [scientific] practice’ (Wood et al., 1998: 1730). Here, evidence is not a stable, pre-existing body of knowledge that can simply be translated into practice; rather it is a product of particular scientific processes and is therefore emergent and provisional (Rhodes et al., 2019; Upshur, 2001). Second, scientific practice is performative. More than simply generating ‘facts’, scientific practices also produce particular realities and prevent others from materialising (Law, 2004; see Berg, 1996, 1997 for detailed analyses of the performativity of medical practices). Crucially, these realities depend for their existence on certain inscription devices; they are constructed through these devices. Applied to the context of medical testing and the management of uncertainty, these insights prompt us to examine how evidence is assembled to justify particular medical practices and interventions in the face of diagnostic, prognostic and/or treatment uncertainty. A focus on processes of evidencing also makes visible how the objects of testing – such as disease, diagnostic labels and health risks – which are often assumed to be fixed and pre-existing, are in fact products of medico-scientific knowledge-making.

Methods and empirical data

Our empirical data comprise 11 in-depth, qualitative interviews with Australian health professionals (clinicians, microbiologists, specialist physicians) and health policymakers, aged between 30 and 71. Respondents were recruited over a 10-month period by publicising a call for volunteers in the newsletters of peak clinician bodies, on the project’s website and social media, through professional networks and via a short online survey in which respondents were given the option to take part in a follow-up interview. Despite employing a range of recruitment methods over an extended period, we had difficulty recruiting key stakeholders to be interviewed, which may be due to professionals being time-poor. Recognising the limitations of over-reliance on interview accounts (Rapley, 2001) and our small sample of interviewees (n = 11), we supplemented these data with qualitative responses from our online survey on testing in healthcare. Key stakeholder survey respondents numbered 70 and included health professionals, policymakers and consumer advocates. Due to the difficulty of attracting key stakeholders, we did not screen prospective participants and consequently all those who took part are self-selected and cannot be considered representative of the broader population of health professionals in Australia. While our qualitative methods were not aimed at achieving representativeness but rather at capturing diverse views of professionals with different areas of expertise (Barbour, 2001), we acknowledge that the relatively small convenience sample of self-selected participants may have produced a partial picture of the issues and that undertaking further interviews and surveys with a representative sample may have yielded insights other than those identified. However, the responses were broadly similar in revealing the complexities and tensions in health professionals’ accounts of testing and their relationship to managing uncertainty in clinical practice, and we believe that when taken together, the survey and interview data yield valuable insights into experts’ views on some of the benefits, limits and uncertainties of medical testing.

The study was approved by Monash University’s Human Research Ethics Committee (Approval number 12274). All participants provided informed written or oral consent. Interviews were conducted over the course of ten months by the first author. They were undertaken telephonically to maximise geographic variation and for reasons of cost efficiency. While we acknowledge the limitations of telephonic interviews, notably that they may result in data of inferior quality compared to data generated through in-person interviews, we decided that on balance, the advantages of telephonic interviews outweighed the potential disadvantages for the purposes of our study. Interviews ranged in duration from 25 minutes to 2 hours, with an average duration of 57 minutes. They were semi-structured and explored professionals’ experience with medical testing, views on the risks and benefits of testing, how the effectiveness of tests can be measured and whether and how testing should be regulated. Interviews were audio-recorded and professionally transcribed verbatim before being checked for accuracy. To protect participant identities, each was given a pseudonym and all identifying details were removed from the transcripts. Addressing similar topics to the interview questions, the surveys explored professionals’ views on the applications, benefits, risks and regulation of tests; whether the use of tests has increased in recent years; and the use of clinical guidelines in relation to testing. The alignment of the survey instrument with the interview guide meant that the data were comparable, allowing us to supplement the interview data with material from the qualitative survey responses.

The data were coded by the first author using an iterative inductive approach in which a preliminary list of codes was drawn up based on the study’s overarching research questions on the sociocultural factors shaping testing, and the relevant literatures on diagnostic testing and screening. These codes were tested on a sub-set of data by the first author, and the coding frame was then refined in consultation with the project team: supplementary codes were added across the dataset to capture a wider range of themes derived from the interviews themselves. Saturation was reached after this broader coding frame was applied to the interview and survey data. Coding was done in NVivo to enable the authors to collaborate, comment on emerging themes and cross-check for coding consistency, deviant cases and alternative interpretations (Seale and Silverman, 1997). Codes relating to the testing process, the management of diagnostic uncertainty, diagnostic dilemmas and the effectiveness of tests were reviewed in terms of how experts’ accounts of testing and diagnosis depict the evidence produced through testing. We use the following conventions when presenting the empirical data below: interviewer speech or survey questions are presented in italics; square brackets indicate words that have been edited for the purposes of clarity or de-identification in the original verbatim transcript (e.g. [this city]); ellipses [. . .] indicate that some words from the quoted material have been omitted. Accompanying each quotation is demographic information about the participant quoted, including their gender, age (or age range in the case of the survey data), occupation and state of residence. Our analysis focuses on two key themes: (1) the role of testing in the management of uncertainty and bringing about diagnostic closure and (2) the ways in which testing generates uncertainties. In addressing these themes, we highlight the tension between a key promise of testing as being to reduce uncertainty and assist diagnosis, and the fact that it sometimes generates more uncertainties and questions than it resolves. We explore the implications of this tension for professional understandings and expectations of the role of testing in medical practice.

Testing, inscription and the construction of certainty

A key role of evidence-based interventions is to produce certainty, as judged against a pre-determined standard (Timmermans and Berg, 2010). Within the evidence-based medical paradigm testing technologies aim to ‘improve medical practice by bringing it closer to the ideal of rationality that trained physicians are thought to approximate’ (Epstein, 1998: 490). Doing so involves processes of inscription, which generate unequivocal guidance for clinical practice. In other words, testing in healthcare aims to minimise uncertainty and facilitate diagnostic closure (Bossuyt et al., 2012). As specialist clinician David (M, 61, New South Wales) puts it, ‘What tests do is attempt to resolve ambiguity’. While highlighting the role of tests in resolving ambiguity and assisting diagnosis, the majority of experts interviewed were quick to acknowledge that the diagnostic measures provided by tests are themselves the product of clinical judgements and decision-making processes, and thus arguably no more trustworthy and reliable than other sources of medical knowledge:

Doctors] have more faith in a test result than they have in what their physical findings were [. . .] I’ve been fascinated by that [. . .] because when you’re in the lab, you know what decisions people make [. . .] in terms of producing results [. . .] And there’s degrees of uncertainty that pathologists have to deal with all the time. And yet when it’s on a final printed piece of paper, it comes out and it’s like a gospel. I remember the first time, when I was just a junior trainee and I suddenly thought, ‘Oh my goodness, it comes out on a report. They believe what I’ve decided [laughs] but I didn’t have access to all the information when I made those decisions’. (Helena, F, 53, Clinical microbiologist and specialist physician, Western Australia)

Central in Helena’s account is the way in which test results (as putatively objective measures) are prioritised over clinical judgement (as apparently subjective). In the hierarchy of evidence, we might say that test results are presented as impartial and thus elevated as the final arbiter of signs and symptoms (Schubert, 2011), or as Helena puts it, they are treated ‘like a gospel’. In this respect, test results work to objectify aspects of diagnosis, providing apparently impartial evidence to guide diagnosis and treatment, that is, they form the basis for medical action (Armstrong and Hilton, 2014). However, as a clinical microbiologist, ¹ Helena is aware of the contingencies involved in testing and thus appears to be more cautious in assigning them greater value than other diagnostic measures. We suggest that her proximity to the production of clinical knowledge through testing renders her more cautious and sceptical about presenting them as objective evidence to guide clinical practice. Or as Brown and Michael (2003) put it, ‘uncertainty will be more acute for those closely involved in the production of knowledge (where experience of the contingencies of knowledge production in the laboratory make one cautious)’ (p. 12).

Another clinical microbiologist whom we interviewed, whose work is largely laboratory-based, also drew attention to the contingencies of measurement, explaining how it is subject to uncertainty and variation:

[One] thing that we see is that people monitor too frequently [. . .] Within labs, there’s a thing called ‘measurement of uncertainty and variation’. So if I run a test 100 times, that test result will sometimes sit at 100, sometimes sit at 110, sometimes sit at 90. But it’s the same result each time [. . .] So the true value is still the same in the sample but the number that comes out of our machine varies [. . .] But I think it is something that’s difficult for doctors to understand. I think a lot of lab-trained people will get it because it’s part of our day-to-day life [. . .] If you measure something [. . .] 100 times, you are going to get a variation in that measurement [. . .] But when it comes to a lab test result, when you [as a doctor] have an absolute value sitting right in front of you, that extra thought process is kind of gone. (Patrick, M, 33, Clinical Microbiologist and specialist trainee, Victoria)

As Patrick notes, laboratory technicians routinely take into account the margin of variation (the ‘measurement of uncertainty and variation’) associated with a test result, which indicates the degree of confidence in the result. But what does this work of ascribing uncertainty achieve for pathologists, clinical microbiologists and other scientists involved in diagnostic testing? It allows them to openly acknowledge and explain the doubt associated with testing, a move which, somewhat paradoxically, endows them with medical authority and professional capital. In other words, we suggest that admissions of uncertainty, such as those made by microbiologists Helena and Patrick, are significant in the hierarchy of medical professionals: such admissions, and the expert knowledge on which they rely, position those who process and interpret tests as more knowledgeable about the testing process than the medical professionals ordering the tests.

In contrast, some doctors, operating at several removes from knowledge production in the laboratory, are more likely to treat a test result as ‘an absolute value’, forgetting or bracketing out the degree of variation and uncertainty associated with the processes of measurement via testing. The potential for clinicians to ascribe an unequivocal value to a test result is not merely an abstract, epistemological concern about the nature of evidence; it can generate significant, material effects in how medical testing operates, particularly in terms of the frequency of test ordering. In some cases, as Patrick goes on to explain, the tendency to bracket out the contingencies and uncertainties of measurement can lead clinicians to order tests at inappropriate intervals, which renders results ‘uninterpretable’:

If I take a sample on a patient and I test it today and I test it in four days’ time, and it’s gone up, that may actually have gone down within the coefficient of variation, so the variation within the test. So if you test too soon, the result is uninterpretable [. . .] We get a lot of tests [ordered by clinicians] within that seven-day window [. . .] when it’s not interpretable [. . .] so the frequency of testing is one of the things that is most difficult.

While Patrick implies that repeat test ordering is due to clinicians not taking into account uncertainty of test measures, we acknowledge that it may reflect clinicians’ awareness of the potential for errors or their reservations about previous test results. However, continual repeat ordering does suggest a limited understanding of how test values are produced, how they may fluctuate and at what intervals particular tests should be ordered to obtain clinically useful results. As we go on to suggest, it points to a need for improved clinician training on these matters.

To draw these analytic threads together in terms of the processes of evidencing and inscription involved in testing, our participants’ accounts highlight that the apparently objective evidence generated through testing does not stand alone. It depends on the work of the technician, the presence of the technical instruments (microscopes, Petri dishes, test tubes, etc.), the reference standard against which the test’s performance is measured and various processes of inscription that transform it from a test sample into a measure that can guide clinical practice. Yet the emphasis that EBM places on testing over other diagnostic techniques erases the performative aspect of all medical knowledge, that is, the fact that it is done in practice and relies on the careful arrangement of instruments, techniques, machines and human and non-human processes. While other studies have similarly explored the constitutive effects of medical phenomena such as particular diagnostic practices (Gardner et al., 2011) and the patient record (Berg, 1996), to our knowledge, ours is the first to apply these STS-inspired insights to expert accounts of testing in healthcare. Doing so allows us to query the objective status of the evidence produced through testing and highlight the multiple practices, techniques and actors on which regimes of testing depend.

Importantly, our analysis suggests that while some health professionals may readily believe in medical testing’s promises of certainty, others (like clinical microbiologists Helena and Patrick) are more aware of the intricacies involved and the measures of uncertainty associated with testing. We therefore do not wish to overstate the certainty ascribed to test results, nor to oversimplify the complex practices, considerations and different types of evidence involved in establishing a diagnosis. As Greenhalgh’s (1999) study of narrative-based medicine has shown, making a diagnosis is an interpretive act which draws on different types of evidence including the clinician’s accumulated expertise, the illness scripts of patients, physical examination findings and the best available evidence (often derived from the results of tests or other diagnostic technologies). Moreover, health professionals are trained to interpret results in relation to a test’s sensitivity and specificity, the reference standard and the clinical picture before determining the test’s diagnostic or prognostic value (Gøtzsche, 2008). Acknowledging this, and extending the work of Greenhalgh (1999), our findings suggest that far from removing the need for clinical judgement, evidence-based practice relies on an interpretive paradigm in which clinicians weigh up the evidence in order to arrive at a reasoned, integrated judgement that can guide diagnosis and treatment.

Yet notwithstanding the acknowledged measures of uncertainty in clinical practice and the importance of clinical judgement, within the evidence-based paradigm testing practices are a potent means of inscription through which medical promise and certainty is evidenced (Mol, 2002; Rhodes et al., 2019). Like other sources of inscription such as the patient record (Berg, 1996), medical testing is a powerful modality for constituting disease and manufacturing greater certainty in the clinical encounter. As a scientific practice that offers the prospect of closure through fact-finding, it is infused with promise and high expectation for delivering diagnostic and/or prognostic certainty. But the material world is messy and the outcome of testing is not always increased certainty; indeed as we show below, sometimes it is just the opposite: testing can generate uncertainties and ontological insecurity.

Discussion and conclusion

By analysing the accounts of health professionals, we have traced some of the ways in which they depict testing in clinical practice. Rather than offering a stable, predictable object of intervention, testing is variously depicted as an objective tool for measuring health risk and diagnosing disease and as a source of uncertainty that implies new options for action – generally further testing. While existing studies have examined the strength of the evidence produced by testing, we have taken a different approach to questions of evidence as they relate to diagnostic testing. Our aim has been to shift the focus from matters of epistemology – what forms of evidence offer the most objective, certain measures of disease – to matters of ontology – how the evidence generated through testing produces material effects in the clinical encounter, specifically in relation to managing risk and uncertainty. Importantly, such an approach challenges a key assumption underlying much existing research on the uncertainty of medical testing, namely that new, improved testing technologies will lead to more reliable means of diagnosing disease and thus to greater diagnostic certainty. Implicit here is a belief in scientific progress that renders the issue of uncertainty a technical problem (Law and Singleton, 2005) – one that is linked to the inadequacies of existing diagnostic technologies and methods. Against this dominant view of science as advancing ever closer to the certain truth of its object, our analysis treats uncertainty as intrinsically bound up with the contingent and unstable character of scientific evidence. For the respondents in our study, the evidence produced through testing was found to have ambiguous ontological significance: providing both the means for resolving or coping with the anxieties of diagnostic dilemmas and the source of uncertainty in relation to the management of risk. However, within the evidence-based medical paradigm, these ambiguous ontologies are rendered invisible, with the evidence generated through testing being viewed as offering an objective, definitive measure of disease.

Our respondents’ accounts also highlight the multiple roles that testing performs. In addition to its established clinical functions of aiding diagnosis and risk management, testing is also a profoundly affective practice. Situated within a science-based promissory discourse, it operates as a means of managing anxiety and uncertainty in the clinical encounter, yet these affective dimensions tend to be overlooked in the evidence-based medical paradigm. How might we understand these different accounts of medical testing? They illustrate our argument that testing is an outcome of the complexities of medical decision-making: it involves coordinating diagnostic tools and human resources across different clinical sites and managing the relational dynamics of the clinical encounter. Our analysis also draws attention to the work involved in arriving at a diagnosis, highlighting how testing can actually operate to suspend, rather than facilitate, a diagnosis by introducing more diagnostic uncertainties into the clinical picture. These findings are consistent with those of related research in the sociology of diagnosis which similarly highlights the ambiguous ontological significance of diagnostic processes and devices (Mol, 2002; Nettleton et al., 2014), the multiple roles that medical testing performs (Chandler et al., 2012; Daly 1989), how testing is contingent on other medical practices and the dynamics of the clinical encounter (Featherstone et al., 2005; Gardner et al., 2011).

In the era of EBM, the pursuit of objective, reliable evidence is paramount in constituting medical expertise. Test results play an important role in this process, in providing seemingly objective measures of disease. In order to achieve diagnostic certainty, diverging measures must be coordinated through an established scientific hierarchy in which purportedly objective test results are accorded more weight than subjective patient self-reports or clinical findings (Berg, 1997; Mol, 2002). Our analysis underlines the importance of studying how scientific knowledge is constructed, given its significant practical impacts. In the context of medical testing, science serves as an inscription practice for producing certainty and objectivity. One way that testing technologies achieve this is by transforming complex, bodily systems into simple, quantifiable, measurable units of information. In the process, they erase evidence of uncertainty. Yet they also produce uncertainties and hence indecision, or ontological insecurity due to the unstable, contingent nature of scientific knowledge. These findings challenge a key premise of testing, namely that it is possible to establish certainty in medical practice via the measurement of individual health risks and disease markers, which are presumed to exist prior to and separate from efforts to measure them. In other words, within this realist ontology of EBM, the objects of testing are treated as part of an independent, anterior reality that is unaffected by attempts to ‘evidence’ it. In contrast, the approach we have taken treats evidence and its objects as co-constituted; that is, they are made relationally, their ontologies unstable and contingent (Moreira, 2007; Rhodes and Lancaster, 2019). It follows that uncertainty and indeterminacy are built into the very substance of scientific evidence and its objects, rather than being the product of methodological or technological inadequacies.

In tracing the uncertainty of evidence produced through testing, we are not arguing that evidence be dispensed with but rather seeking to draw attention to the limitations of a narrow, objectivist view of evidence. In this respect, we suggest that medical practice might be enhanced by attending to the ways in which evidence is necessarily shaped by particular values, epistemological commitments and knowledge-making practices and, importantly, by considering the performative role of evidence in materialising the objects it purports to describe. Insofar as such a shift in articulating evidence is both conceptual and ontological, it holds political purchase and has the potential to generate far-reaching effects. In the context of medical testing, approaching evidence as emergent and contingent on particular knowledge-making practices has the potential to encourage a greater tolerance of uncertainty (Morgan and Coleman, 2014) and better integration of other forms of evidence and expertise into clinical decision-making such as professional experience, patient preferences and resource availability (Tonelli, 2006). It also invites consideration of the possible unintended effects of particular practices and techniques of evidencing, such as the reliance on increasingly sophisticated testing technologies to generate apparently precise, objective diagnostic evidence.

The rapid growth of medical testing is of concern to many policymakers, who seek to rein in burgeoning healthcare costs, and to clinicians who worry about the associated harms to patients. This growth may be contributed to by many factors, including increasing patient expectations, the influence of powerful vested interests (pharmaceutical companies, medical specialists) and processes of rationalisation and medicalisation. Regardless of the explanation, the underpinning premise and rationale for all testing is that it should be ‘evidence-based’ if it is to provide the certainty that health professionals require to provide timely advice, treatment or care and for policymakers to have confidence that public funds are spent in a cost-efficient way. However, our analysis leads us to question whether ‘evidence’ can provide the envisaged certainty and points to the need to question the fundamental premise that testing – no matter what form it may take – is necessarily desirable and that the benefits generally outweigh the costs. Questioning this premise has become critical with the introduction of increasingly technologically sophisticated means of screening and diagnosis.

Newer forms of testing that are beginning to be used in healthcare, such as ‘personalised’ genetic testing and high-resolution imaging, promise to provide ever more reliable means of identifying diseased and ‘at risk’ populations. It is widely believed that single DNA tests have the potential to provide earlier diagnoses for various conditions for which one may be susceptible and enable treatments and care to be ‘tailored’ to individuals according to their unique genetic profile. Furthermore, big data promises to transform our understanding of disease by using machine learning to identify new patterns and connections, leading to earlier diagnoses for diseases such as melanoma (Osman, 2019). Yet, as our analysis suggests, the promised certainty provided by ever more sophisticated technologies of testing is an illusion and can never be assured since it assumes that ‘truth’ regarding what is tested and the means to its understanding and resolution can be achieved given enough or the ‘right’ kind of ‘evidence’. If instead, following Stengers (2018), we treat uncertainty as productive (rather than defective) and evidence as contingent, it might encourage a more careful reflexive science, one that is driven not by the quest for a single ‘right’ answer, but rather by a willingness to acknowledge and even engage messiness and indeterminacy, and the innovations they yield. As creators of particular realities to the exclusion of others, humans have the potential to narrate and build alternative futures – including ones that are much less reliant on technological interventions and the search for incontrovertible truths about disease, health risks or susceptibility to illness. While testing for purposes of population screening and clinical diagnosis will no doubt continue to play a significant role in healthcare, more questioning of the premises of testing will do much to reduce associated costs and harms.