Rethinking 'resistance' to big IT: A sociological study of why and when healthcare staff do not use nationally mandated information and communication technologies

Article history

The research reported in this issue of the journal was funded by the HS&DR programme or one of its proceeding programmes as project number 10/1011/01. The contractual start date was in September 2011. The final report began editorial review in June 2013 and was accepted for publication in April 2014. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HS&DR editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

none

Disclaimer

this report contains transcripts of interviews conducted in the course of the research and includes language that may offend some readers.

Copyright statement

© Queen’s Printer and Controller of HMSO 2014. This work was produced by Greenhalgh et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

The problem of ‘failed’ information and communication technology programmes in health care

The history of information and communication technologies (ICTs) in health care includes some expensive disasters. For example, England’s National Programme for Information Technology (NPfIT) was established in 2006 with a projected cost of £12.7 billion over the ensuing 8 years,1 but after numerous delays and difficulties, most of its component technologies were abandoned in 2011–12, and expensive penalties were incurred for breach of commercial contracts. 2–5 To accompany the release of a damning summary of the programme’s failures in May 2011, the head of the National Audit Office wrote:

The original vision for the National Programme for IT [information technology] in the NHS will not be realised. The NHS is now getting far fewer systems than planned despite the Department paying contractors almost the same amount of money. This is yet another example of a department fundamentally underestimating the scale and complexity of a major IT-enabled change programme. The Department of Health needs to admit that it is now in damage-limitation mode. I hope that my report today, together with the forthcoming review by the Cabinet Office and Treasury, announced by the Prime Minister, will help to prevent further loss of public value from future expenditure on the Programme.

Amyas Morse, head of the National Audit Office, 18 May 20116

In 2007–10, our team evaluated the introduction of two components of the NPfIT: the Summary Care Record (SCR) and a linked patient portal, HealthSpace (see Chapter 3, The primary data sets). Policy-makers had predicted that these technologies would be widely adopted and would lead to rapid improvements in the quality, safety and efficiency of care as well as empowering patients, promoting choice and reducing health inequalities. 7 Lack of realisation of their hoped-for benefits was attributed mainly to the fact that they were not being used. 8–11 Similar findings were documented by the evaluators of the implementation of hospital records within the NPfIT. 3,12–14 At the time of writing, most of the technologies in the NPfIT (including hospital-based records and HealthSpace) had been withdrawn because of delays in supply and/or non-uptake by clinicians, and a new policy emphasising local development and/or procurement introduced. 15 The SCR was still Department of Health policy but its use in clinical practice remained limited. 16

Both we and other research teams have described numerous other instances of non-adoption, limited adoption or adoption followed by abandonment of ICTs by clinicians, especially when they were introduced as part of a large, top-down change programme. 3,17–19 As one study concluded, ‘Most [healthcare ICT] implementations fail because, despite high investments in terms of both time and financial resources, physicians simply do not use them’ (p. 309). 20

The literature on resistance to ICTs in health care is extensive. In the next few sections, we summarise that literature and introduce our own chosen approach. Immediately below, we consider the (broadly) behaviourist approach taken by many cognitive psychologists and health services researchers to ‘resistance’ to ICTs, which is still the dominant framing taken by many academics or policy-makers. In the section Multilevel models of resistance to information and communication technologies, we review multilevel models of resistance in which group and organisational factors are added to individual ones. Next (see Interactional models of resistance: the sociotechnical perspective), we consider theoretical approaches that take an organic or interactional view of how individuals in organisations use technologies (or not), focusing in particular on Lynne Markus’s classic study from 1983 on technology-system fit, Albert Cherns’ work on sociotechnical systems theory, computer-supported co-operative work (CSCW) and Carl May’s normalisation process theory.

In the section Critical studies of organisational power relations, we outline the literature from critical management studies (and, more specifically, from critical information systems studies) that considers people’s resistance to ICTs in the workplace in terms of the play of power between dominant and ‘dominated’ groups. We then briefly explain actor–network theory (ANT: Bruno Latour) and related perspectives (see Actor–network theory), including important extensions of this approach in the work of Ole Hanseth and colleagues from Norway on standardisation and Marc Berg from the Netherlands on the agency as a product of the network. We introduce the sociological work of Anthony Giddens (see Giddens’ work on structuration, modernity and technology), including structuration theory and its links to modernity and technology. Finally, we explain how an empirically-oriented extension of structuration theory developed by one of us (RS) and known as ‘strong structuration theory’ (SST) can be further enhanced for the study of technologies using selected elements of ANT (see Strong structuration theory). When we originally planned this study, we were drawing on our work on this ‘ANT–SST hybrid’, though as we explain in the Discussion (see Chapter 5), we modified our stance somewhat as our thinking developed and data analysis progressed.

The technology acceptance model

The most widely used model for considering ICT adoption in organisations is probably Davis’ ‘technology acceptance model’ (TAM), located in the cognitive psychology/human–computer interaction literature and cited by over 2000 studies since. 21 It states that a technology is more likely to be adopted by a potential user if it scores highly on a set of attributes including ‘usefulness’ (e.g. the degree to which it helps someone do their job and improves their productivity) and ‘ease of use’ (e.g. features such as user-friendliness, speed and ‘clunkiness’). The TAM, which was originally derived as an adaptation of Rogers’ diffusion of innovation theory, was subsequently extended to produce a model of information system success, seen as determined by user uptake, which in turn was determined by users’ perceptions of the quality of the system and the quality of the information it contained. 22 All of these variables are seen as interconnected and interdependent; the strength of the relationship between any of them, and the influence of particular mediating or moderating variables, will vary with the situation, but according to the model, they can (in theory at least) be measured and quantified.

Delone and McLean later added a third component (perceived service quality, such as helpdesk support) to the first column in Figure 1. 23 They also unpacked the dependent variable (‘organisational impact’) into four dimensions: task productivity (the extent to which an application improves the user’s output per unit of time); task innovation (the extent to which an application helps users to create and try out new ideas in their work); customer satisfaction (the extent to which an application helps the user ‘create value’ for customers); and management control (the extent to which the application helps to regulate work processes and performance). Although their model rests on positivist assumptions (i.e. it assumes an objective reality ‘out there’ that can be measured and whose behaviour can potentially be predicted), Delone and McLean emphasised (drawing implicitly on contingency theory) that it should not be used in a simplistic, deterministic way, as different systems will have different independent variables that impact on the dependent variable.

FIGURE 1.

The Delone–McLean model of information system success. Adapted with permission from Delone WH, McLean ER, Information systems success: the quest for the dependent variable, Information Systems Research, volume 3, number 1, March 1992. 22 Copyright 1992, the Institute for Operations Research and the Management Sciences, 5521 Research Park Drive, Catonsville, Maryland 21228, USA.

A 2010 review by Holden and Karsh of the applicability of the TAM in health care considered 16 data sets analysed in over 20 studies of clinician use of health information technology (IT) for patient care across a range of settings. 24 Certain TAM relationships (e.g. between perceived usefulness and the behavioural intention to use) were consistently found to be significant, whereas others (e.g. between perceived ease of use and behavioural intention to use) were inconsistent. The authors concluded that TAM predicts a substantial portion of the use or acceptance of health IT, but that the theory may benefit from several additions and modifications – specifically, with more attention to norms and beliefs of potential adopters of the technology.

A citation track of Davis’ original paper and Holden and Karsh’s later review showed that hundreds of empirical studies of user resistance to technology have been based on the TAM and that several dozen of these relate to ICTs in health care. 25–31 These studies invariably conceptualise the individual user of an ICT as a rational actor who seeks control over his or her work and weighs up constructs such as ‘usefulness’, ‘ease of use’ and the ‘costs’ and ‘benefits’ of change. Indeed, most such studies implicitly or explicitly combine the TAM with the Theory of Planned Behaviour. 32 They depict user resistance as an individual deficit, usually comprising a combination of parochial self-interest, misunderstanding, lack of trust, a belief that change is worthless and low tolerance for change,33 and emphasise measures through which resistance (thus defined) can be overcome – for example, via ‘engagement’ activities, incentives or punishments, firm leadership of the change project and ensuring that staff feel ‘in control’ during implementation. 34 All place a central emphasis on the individual user of an ICT, and all see user acceptance – and its converse, user resistance – as a key focus of enquiry (especially in explaining ‘failed’ ICT projects).

One author in this tradition articulated the problem of resistance thus:

. . . the major challenges to system success are often more behavioral than technical. Successfully introducing such systems into complex healthcare organizations requires an effective blend of good technical and good organizational skills. People who have low psychological ownership in a system and who vigorously resist its implementation can bring a ‘technically best’ system to its knees. However, effective leadership can sharply reduce the behavioral resistance to change--including to new technologies--to achieve a more rapid and productive introduction of informatics technology.

p. 1135

As the following sections show, many researchers have rejected such overt behaviourist framings. However, this section has shown that un-nuanced and uncritically pro-innovation applications of the TAM remain popular in some parts of the research literature. In such studies, the adoption and use of new technologies is seen as a necessarily positive step linked to the promise of greater procedural efficiency (and, implicitly, clinical effectiveness). In such traditions, research has assumed a ‘rational actor’ and been directed mainly at ‘overcoming resistance’ using behaviourist methods. Reflecting this rationalist emphasis, policy measures for introducing ICTs in health care often focus largely or exclusively on ‘overcoming [individual] resistance’ (e.g. see the Department of Health’s consultation36).

Multilevel models of resistance to information and communication technologies

A number of authors have extended the focus of the TAM by nesting it within a multilevel conceptual framework. Sabherwal et al. ,37 for example, propose context-related constructs (top management support for the system, facilitating conditions); user-related constructs (user experience with an information system, user training, user attitudes, user participation in the development of the system); and system constructs (system quality, perceived usefulness, user satisfaction, system use).

Yusof et al. 38 suggest that the Delone–McLean model lacks attention to certain factors that are particularly important in the healthcare setting – specifically, the ‘soft’ human and organisational factors such as culture, buy-in and so on. They offer a model of ‘fit’ between information systems, organisational factors, and external influences (Figure 2).

FIGURE 2.

‘Fit’ between information technology, internal organisation factors and external factors. © 2014 IEEE. Reprinted and adapted, with permission, from Yusof MM, Paul RJ, Stergeoulas L. Towards a framework for health information systems evaluation. HICSS ’06 Proceedings of the 39th Annual Hawaii International Conference on System Sciences. IEEE Digital Library; 2006. 38

Based on a series of studies of the differing fortunes of ICT systems in different hospitals,39–41 an interdisciplinary group of researchers developed a ‘triple-level’ model of resistance comprising individual (people’s perceptions of the technology), group (the extent to which the new system reinforces or threatens a group’s power base) and organisational (the extent to which the innovation aligns with organisational strategy and goals). This model is shown in Figure 3.

FIGURE 3.

Triple-level model of information system implementation, based partly on Markus. Adapted with permission from Lapointe L, Rivard S, A triple take on information system implementation, Organization Science, volume 18, number 1, 2007. 41 Copyright 2007, the Institute for Operations Research and the Management Sciences, 5521 Research Park Drive, Catonsville, Maryland 21228, USA.

In this ‘triple model’, resistance is seen not simply as a psychological deficit (obstinacy, indolence, ignorance and irrational fear of change) that needs to be overcome (e.g. via top management mandate or the psychological tactics for increasing engagement). Rather, it is a phenomenon that develops over time as users interpret the alignment (‘fit’) between a new technology and individual, group and organisational context. According to this model, when resistance generates conflict and consumes time and attention, it tends to have negative impact, but resistance may also have a positive impact when it stops organisations from investing in systems that would lead to losses in productivity and stress among staff. This is true of any organisation, but Lapointe and Ricard suggest that it may be particularly apposite in health care, where ‘fit’ between an information system and the internal properties of the organisation may be particularly poor.

While this stream of research marks an important step forward from individual-focused models, it remains uncritical of the place of technology in health care and is relatively uninterested in the external forces (political, economic, technological, professional) that give rise to ICT programmes or in the subtle but potentially important ways in which ICTs may influence the nature of clinical work and the quality of the clinical interaction.

Interactional models of resistance: the sociotechnical perspective

Back in 1983, Lynne Markus published a classic review paper in a US management journal of barriers to implementation of information systems in organisations. In it, she divided theories and models into ‘people-determined’ (resistance is explained by people’s personalities, goals, capabilities and so on); system-determined (resistance is explained by features of the technology itself and how it is implemented); and interactional (resistance is explained by what we would now call sociotechnical issues, i.e. the interaction between people and technologies). 42 ‘People-determined’ theories, said Markus, depicted the solution to resistance in terms of education, training, incentives, policies and user participation (to achieve commitment). ‘System-determined’ theories saw the solution in terms of better technical and ergonomic design, better protocols and procedures, and user participation (to achieve better design). Interactional theories suggested that the organisational problems should be addressed before the system is introduced (or even designed), and that the relationships between users and designers were all-important.

At around the same time as Markus’s paper on ‘Power, Politics and Information Systems’ was published in the USA,42 British researchers based at the Tavistock Centre in London, UK (which specialises in the study of organisational dynamics from an interpersonal perspective), were developing sociotechnical systems theory. This proposes that technologies and work practices are best codesigned using participatory methods in the workplace setting, drawing on such common-sense guiding principles as staff being ‘able to access and control the resources they need to do their jobs’, and insisting that ‘processes should be minimally-specified (e.g. stipulating ends but not means) to support adaptive local solutions’. 43 Sociotechnical theory frames resistance to ICTs in terms of poor fit between the micro-detail of work practices and the practicalities of using technology.

Sociotechnical theory, developed in the 1980s, has more recently been taken forward in a number of ways. It inspired the interdisciplinary field of inquiry known as CSCW, in which one key concept is the social–technical gap. 44 A fundamental tenet of CSCW is that the interaction between people and technologies in the workplace is, with few exceptions, far more complex than would initially appear. A key reason for this, argued Ackerman, is that there is a gap between the nuanced, flexible and often unpredictable nature of human activity and what it is possible to deliver technically. 44 This is an inherent problem and can probably never be fully ‘designed out’. It links to a more general mismatch between formal work procedures (as in a manual) and the informal routines that actually occur. Grudin, cited by Symon et al. ,45 put it thus:

Work processes can be described in two ways: the way things are supposed to work and the way they do work. Software that is designed to support standard procedures can be too brittle

p. 2545

Technical solutions, argue the CSCW community, tend to be designed to align with an idealised, formal map of work processes, which may bear limited relation to what actually happens (especially in non-standard and emergency situations). Two partial solutions to this problem are recognised by CSCW: articulation and co-evolution. Gerson and Star (cited by Goorman and Berg46) define articulation as follows:

All tasks involved in assembling, scheduling, monitoring and co-ordinating all of the steps necessary to complete a production task (patient trajectory). This means carrying through a course of action despite local contingencies, unanticipated glitches, incommensurable opinions and beliefs or inadequate knowledge of local circumstances. Every real world system is an open system . . . No formal description of a system (or plan for its work) can thus be complete . . . every real world system thus requires articulation to deal with the unanticipated contingencies that arise. Articulation resolves these inconsistencies by packaging a compromise that ‘gets the job done’ that is, that closes the system locally and temporally so that the work can go on.

p. 2046

‘Articulation’ includes both temporal co-ordination (sequencing the inputs of different actors over time) and spatial co-ordination (ensuring that the right people and artefacts are in the right place). ‘Co-evolution’ refers to the parallel and reciprocal evolution of technologies and work roles or routines. 44,47 It might be a largely unintended consequence of introducing a technology (i.e. it is inadvertently found that work roles and practices have to change), or there might be a deliberate plan for such evolution to occur (i.e. the technology is introduced along with a plan for new roles and routines). This begs the question of when the introduction of a technology can justify prospective job redesign and when it is better for technologies to be closely designed around people’s current jobs. As Pratt et al. 48 put it:

At one extreme, developers carefully design the application to fit the specific work practices of its users. Under this model, users do not change their work practices at all, because the technology accommodates their specific needs and work styles. The alternative extreme is to reshape the processes of the organization around the new application. [. . .] most applications fall in a middle ground: a mixture of supporting some existing work practices and attempting to change others.

p. 13348

It follows from these principles that ‘resistance’ to ICTs by staff in any organisation may be explained by the subtle contingencies – and inherent messiness – of work. One of the leading researchers to apply these principles to the healthcare setting is Marc Berg, who also draws on ANT (see Actor–network theory, below). Berg talks, for example, of ‘growing’ rather than ‘building’ information systems in health care and working to achieve synergy between three fundamental (re)design tasks: the technical system, the primary work process (e.g. clinical care), and the secondary work process (e.g. audit, management). 49 It is perhaps worth commenting that critiques of the NPfIT have (naively, in our view) tended to emphasise poor planning, poor contracting and lack of vision but not the inherently organic nature of information system design. 4

In a contrasting development of sociotechnical theory, Brown and Duguid50 have shown how technologies in the workplace are embedded in networks of social relationships that make their use meaningful. The detail of how to use, adapt, repair or work round technologies is learned through membership of a community of practice; this social infrastructure, local and specific to an organisation, strongly influences whether or not and how particular technologies ‘work’ in particular conditions of use.

Another broadly interactional theory of technology adoption in organisations is Carl May’s normalisation process theory, which he developed to explain the uptake (or non-uptake) of ICTs, especially telemedicine, in healthcare organisations. 51 This theory comprises four constructs: interactional workability (to what extent does the technology fit with the micro-environment of the clinical encounter?); relational integration (to what extent does it fit with the network of relationships within which the clinical encounter sits, and, especially, how does it impact on issues such as interpersonal trust?); skill-set workability (to what extent does it fit with the formal and informal division of labour between staff?); and contextual integration (to what extent does the organisation understand the innovation and agree to allocate material and human resources to its implementation?).

An interview study of NHS clinicians and managers using normalisation process theory identified that Choose and Book referral software was unpopular and little used, and suggested that the main barrier to uptake was poor interactional workability, defined as ‘the impact that a new technology has on interactions, particularly . . . consultations’; in contrast, Picture Archiving and Communication System (PACS; a digital image storage and retrieval package), was popular and widely used because it scored highly on all of the four constructs listed in the previous paragraph. 19 Normalisation process theory is presented very explicitly as a ‘rational’ model for considering the uptake (and non-uptake) of complex technology-based interventions. 52 The theory recognises the complex nature of healthcare work and the human relationships on which that work depends, but it lacks a theoretical perspective on the technology itself and remains explicitly couched in assumptions of ‘rationality’ rather than in normative concerns about what constitutes good clinical practice. 52

Critical studies of organisational power relations

Some researchers, especially those aligned with critical management studies, have emphasised the network of power relations in which the introduction and use of technologies is embedded, and especially how ICTs affect domination and power struggles in organisations. The use of ICTs to monitor work, they suggest, inevitably opens the performance of organisational members to administrative surveillance – and it makes them aware that managers may be checking up on them. Drawing on Foucault’s powerful metaphor, Doolin53 describes the impact of this ‘panopticon’ in the healthcare setting:

. . . representational, inscriptional and computational techniques associated with information systems render individuals calculable, and thus knowable and governable. Some activities are given existence and attention, while others remain unrecognized, enabling managerial knowledge to make stronger truth claims and engendering compliance in those subject to such scrutiny.

p. 34553

In such situations, staff may have only limited possibilities to resist the introduction of pervasive ICT systems. Indeed, Doolin suggests that, as suggested by Foucault’s theory of governmentality, staff may come to internalise and hence fail to question the managerialist discourse that is built into, and perpetrated by, the system.

Others, however, have proposed that in such situations, resistance may take subtle forms. In a paper called ‘The failed panopticon’, for example, Timmons describes how hospital managers colluded with nurses to resist a new information system for monitoring nursing work. 54 The system’s abstracted work plans, while in theory producing ‘action at a distance’ (i.e. the quantification and control of nursing work), were recognised by managers as failing to reflect the real work of nursing. When the system appeared to reflect poor performance on a particular ward, the lead nurse was given no more than a sham telling-off by senior managers.

In an ethnographic study in UK general practice, Winthereik55 studied how general practitioners (GPs) managed two potentially conflicting agendas: accountability to external scrutiny [e.g. via the Quality and Outcomes Framework (QOF) and other government-imposed audits and incentive schemes]; and professional autonomy (the need to feel in control of their own clinical practice). Many GPs were largely negative about the new external controls and expectations, and felt that the standards against which they were being judged were arbitrary and sociopolitically driven. Despite this, they engaged actively with accountability tools (e.g. by complying with the use of codes through which their work was being audited), and they were keen to make their work visible and demonstrate good performance against the standards. Winthereik concluded that there was no simplistic zero-sum relationship between being rendered accountable and retaining clinical autonomy. Rather, accountability and autonomy appeared to co-evolve such that new, positive forms of autonomy emerged alongside the accountability agenda.

There is a small but important literature that uses a critical feminist lens to study resistance to new technology in nursing and midwifery. Henwood and Hart,56 for example, conducted a large, multisite case study of the introduction of electronic records in maternity units from the perspective of midwives. They found that in fewer than half of the units had midwives been consulted at all about the nature of their work or the design of the IT system. In interviews and focus groups, there was a marked disinterest among many midwives, who saw the new system as something introduced by management, ‘boring’, or something they might engage with when ‘it’s all sorted out’. This resistance was palpable and widespread; midwives appeared to be actively participating in their own exclusion by defining the electronic record as part of a techno-medical world, imposed by (male) doctors, marginal to the ‘real’ work of midwifery, and detrimental to client care. One interviewee is quoted thus:

I just think it would totally ruin the whole point of being a midwife. You’re supposed to be with the woman, looking after the woman. To have a computer in the corner, where you’re away and typing into it, it’s just not part of what having a baby’s all about, to me. Because that would take over what you’re actually there for, I’m sure . . . So I wouldn’t want there to be any more technology involved – it’s supposed to be a normal process. So all this hi-tech stuff is just not what I think it’s about.

Senior midwife; p. 25756

Having first positioned the ICT as ‘masculine’ and as interfering with the ‘normal’ process of birth, this interviewee went on to distance herself (and midwifery in general) from the low-status administrative work of ‘data entry’. In her symbolic world, technology was either about ethereal, unreal aspects of care or not about care at all. This example of resistance from feminist nursing studies has some resonance with our own findings, described below.

Actor–network theory

One increasingly popular approach to incorporating the material properties of technology in a sociological analysis of ICTs is Latour’s ANT,57 which shares some of its intellectual roots with CSCW (see Interactional models of resistance: the sociotechnical perspective). We have described the key concepts and assumptions of ANT relevant to this topic in more detail elsewhere. 58 Briefly, ANT considers networks that are made up of both people and technologies. A key focus is what people and technologies become as a result of their position in a network, and the power that emerges from dynamic configurations of human and non-human actors. A nurse who takes a job in a 24-hour call centre dealing with ‘emergency’ calls from the public takes on particular characteristics and develops particular skills; his or her role is shaped, enabled and constrained by the (human and technological) properties of the telephone network – and also by the other technologies to which the system is connected (e.g. access to the electronic records of the patients who call in). In such a situation, the electronic patient record becomes a part of this particular enactment of unscheduled care.

Critics of ANT are quick to point out that technologies do not have agency, but its protagonists depict agency as a product of the network, not as something internal to either a person or a technology. In Berg’s59 words:

The elements that constitute these networks should not be seen as discrete, well-circumscribed entities with pre-fixed characteristics. Rather, those entities acquire specific characteristics, roles and tasks only as part of a network. A ‘physician’ is only a ‘physician’ in the modern western sense because of the network of which s/he is a part and which makes his/her work and responsibilities a reality. [. . .] Because of this tight interrelation between elements in a network, the introduction of a new element or the disappearance of an element (as when a hospital stops training junior residents) often reverberates throughout the healthcare practice.

p. 8959

Actor–networks (that is, networks of people and technologies) are typically highly dynamic and inherently unstable. They can be stabilised to some extent when people, technologies, standards, procedures, training, incentives and so on are aligned. When introducing a new technology, this alignment is achieved (or, at least, attempted) through what Latour called ‘translation’, which involves the four stages of problematisation (defining a problem for which a particular technology is a solution), interessement (getting others to accept this problem-solution), enrolment (defining the key roles and practices in the network), and mobilisation (engaging others in fulfilling the roles, undertaking the practices and linking with others in the network). 60

Studies of resistance to ICTs from an ANT perspective tend to focus on two main issues: the efforts made by human actors to achieve (or thwart the achievement of) particular alignments of people and technologies; and the tendency of the inbuilt properties of software (e.g. access controls or pull-down menus) to shape and constrain the possibilities open to organisational actors. 61 While other research traditions have also addressed these issues, ANT offers a novel way of theorising them. Efforts made by groups of actors to influence the codes, architecture and algorithms inscribed into particular ICTs, for example, are interpreted as political moves rather than simply attention to clinical and technical standards. 62–64

Latour has suggested, for example, that if usage of a new technology that aligns with designers’ intentions is a ‘program’, resistant individuals seek to mobilise others into enacting and stabilising an ‘anti-program’ (in which the technology may be used differently for different purposes), and this wider program–anti-program dynamic across the network should be the focus of analysis when studying resistance to any technology. 65

Studies in the ANT tradition have also shown that the tension between standardisation (which helps to stabilise the network) and contingency (which reflects and responds to local needs and priorities) can never be resolved once and for all; rather, it must be actively and creatively managed – and this gets harder as the network gets bigger. 62,66,67 In this framing, ‘resistance’ may simply reflect the practical impossibility of resolving the tension between standardisation and contingency. Hanseth and Monteiro,68 for example, have used ANT to consider the institutionalisation of technology standards – whether in health care or in other fields such as transport, telecommunications or education. Infrastructure is built on an installed base (hardware and software); the (increasingly) large installed base attracts complementary production and makes the standard cumulatively more attractive. As the installed base expands, complementary products and components emerge and this in turn increases the credibility of the standard among the key groups of human actors on which its adoption and sustainability depends. This promotes adoption, which further increases the size of the installed base. Thus, a particular network of technologies and the people who (necessarily) use them emerges and becomes progressively more stable.

Actor–network theory’s framing of resistance is perhaps best illustrated with a detailed example. Constantinides and Barrett69 drew on ANT to address what they called ‘networks of power’ [‘the close interdependencies between institutional arrangements, people, ICTs, and work practices and how these mobilize (spread and extend) each other’s strength and durability, for example, how ICTs can become standardized in organizational practices’ (p. 77)] in the introduction of a region-wide health information network in Crete. They were interested in how particular tools and activities become ‘filtered’ in the habits and routines of different communities, either by becoming integrated into everyday work practice or by being discarded as impractical. The ‘surviving’ tools and activities would become parts of large, sociotechnical systems and influence, in turn, other tools and activities. They were also interested in ‘body politics’ – that is, how different institutions, people and technologies are brought together to form different versions of the diseased body and exercise politics through different rules and resources. 70

The authors interviewed senior managers and IT professionals from the private IT company that initiated the project, healthcare professionals, and senior management from the regional health authorities. They found that political negotiations between these parties were complex and protracted, with multiple competing efforts at problematisation. One early move by the IT company was to recruit three GP enthusiasts (who were already high users of other IT solutions and champions for electronic records) and, through them, engage a ‘community of users’ of fellow GPs (what Latour would call ‘interessement’ and ‘enrolment’). This positive move was to some extent outweighed, however, by the limited interest of the Ministry of Health in the project (hence use of the system remained voluntary for GPs) and also by a large denominator of undertrained GPs whose interest in the project was at best lukewarm (lack of enrolment). Despite having recruited some influential clinical champions, the company’s power was limited to offering ‘free’ computers; it did not extend to enforcing the use of these technologies. The demise of the system was linked to two key events: an acrimonious intellectual property rights dispute between the IT company and a GP who felt that he had contributed to the design of the system, and a key clinical champion who reverted to using paper records.

At the time of that study’s publication, the actor–network was fragile and the authors were unsure whether the electronic record system (which was widely seen as ‘state of the art’ and had won an international e-health award) would be taken up further or abandoned. They conclude (a) that networks of power are key to success – and especially that an IT company should seek to mobilise the enthusiasm and influence of its clinical user base; (b) that large-scale ICT innovation requires access to heterogeneous resources, the control of which can externally influence the process of translation; and (c) that the network is fragile and dynamic [‘large-scale ICTs have distributed and multiple effects on participant organisations and need to be managed over time as use may be transformed into nonuse, resources initially gained may be later lost, and collaboration with key stakeholders may turn into conflict’ (p. 89)]. This example illustrates how individual ‘resistance’ to a particular ICT can be conceptualised as emerging as a product of the actor–network rather than as originating within the person’s ideas, beliefs, competences and so on.

As we have argued previously, some concepts from ANT, notably the notion of a fluid and potentially unstable network involving people and technologies, are extremely useful in theorising large-scale ICT systems in health care. 58 However, ANT does not include a sophisticated theory of either human agency (e.g. embracing a detailed analysis of what actors ‘know’ and how they assess particular small-scale situations) or the macro social structures in which human action is nested. The quote from Berg earlier in this section, for example, suggests that a physician is a physician ‘only’ because of his or her position in the sociotechnical network – a framing that minimises the importance of such things as professional identity, values and virtues (what some sociologists would call ‘agency’). In the next section, we consider an alternative theorisation of people, society and technology, in which the concepts of social structure and human agency are key.

Giddens’ work on structuration, modernity and technology

The theory of structuration links the macro of the social environment (‘social structures’) with the micro of human action (‘agency’) and considers how this structure–agency relationship changes over time as society becomes ‘modernised’. 71 On the one hand, agents are influenced by the context in which they operate (and also, historically, by the contexts in which they were brought up, educated and trained). On the other hand, human action has consequences (intended and unintended) that influence (and, therefore, gradually change) the external social context. People can be thought of as ‘social actors’ who are, to a greater or lesser extent, knowledgeable and reflexive. They contemplate any action by taking account of social structures such as norms (‘structures of legitimation’ – what they see as reasonable and ethical, such as assumptions about the nature of clinical excellence that underpin professional practice), meaning-systems (‘structures of signification’ – the symbolic meanings and significance which they attach to people, experiences and artefacts) and rules and regulations (‘structures of domination’ – what they see as following protocol or obeying external authority). 71

In developing structuration theory, Giddens sought to bring together two apparently contradictory traditions within sociology: on the one hand, that of interpretative scholars such as Goffman and Garfinkel, who emphasised human interpretations and action, and on the other hand, the work of structural sociologists such as Durkheim and Marx, who were more interested in the wider social forces that shape and define society. Far from being mutually exclusive or incommensurable, proposed Giddens, these two perspectives are intimately and dynamically related: structure shapes agency, and agency, in turn, shapes structure.

Giddens’ application of structuration theory focused mainly on four large-scale systems: the world capitalist economy, the international division of labour (and the role of industrialisation in this), the world military order (including the industrialisation of war), and the nation-state system (including the modern revolution in administrative techniques, surveillance and the control of information). 72 The system of medical practice, supported by technological artefacts and shared guidelines and standards, represents another large-scale social system which, while operating semiautonomously and to its own logic, also depends on and interfaces with these four systems.

Giddens also proposed that, as a result of modernity and globalisation, connections between different social contexts become networked, such that local happenings are shaped by events on the other side of the world, and vice versa. Globalised disease surveillance networks linked to national public health systems allow for co-ordinated international responses to potential pandemics, but such responses – typically dictated by distant committees – inevitably restrict the freedom of individuals to attend work, school or leisure activities locally. Similarly, guidelines and protocols developed by distant committees (and, increasingly, standardised at national or international level) restrict the capacity of clinicians to tailor their practice to local needs and circumstances.

Giddens proposed that humans, both individually and collectively, continually reflect on their predicament and try to control their future. However, because they are unable to control the reflexive responses of others, they generate new forms of risk and uncertainty, from financial instability to the threat of nuclear accidents and environmental catastrophe. To this list, we might add the inability of doctors to control the exponential increase in ‘lifestyle’ diseases linked to obesity and inactivity, the failure to date of the expanding industry of ‘risk scores’ (designed to identify high-risk individuals at a sufficiently early stage to intervene and prevent disease) to translate into tangible and scalable gains in health or life expectancy, and the numerous examples of large-scale ICT programmes in health care that have generated recurring technical glitches and operational confusion rather than the predicted efficiency gains. All these disappointments undermine trust in the ability of globalised institutions (the economy, nation-states, industry, the military – and medical science) to create the structural conditions for what Giddens called the ‘ontological security’ of individuals. 73

Social practices (everything from parenting to conducting a job interview or a clinical consultation) depend in part on actors’ personal knowledge of the local context and the network of relationships in which they live and work. Social structures are reproduced because (and to the extent that) people know how to ‘go on’ in a particular social setting. However, in the modern world, social practices become disembedded as a result of the twin social forces of modernity and globalisation, which produce (among other things) ‘expert systems’ – defined by Giddens as ‘[a] system of technical accomplishment or professional expertise that organize[s] large areas of the material and social environments in which we live today’ (p. 27) – that manipulate abstract, impersonal knowledge using new technologies. 74

Social practices are now, in large part, removed from the immediacies of context, with the relations they involve typically being stretched over large tracts of time and space. Local experiences and events are shaped by processes taking place on the other side of the world, and vice versa. These are processes, moreover, that are primarily impersonal and abstract. [. . .] Each of these abstract systems plays a part in coordinating social relations between distant and absent others. Local contexts are cross-cut and ‘emptied out’ by the power and authority of these stretched relations.

pp. 449–5075

A closely related sociological concept is that of distanciation, originally defined by Giddens as ‘[t]he stretching of social systems across time-space, on the basis of mechanisms of social and system integration’ (p. 377). 71 By ‘system integration’, Giddens is referring to the different ways in which the parts of a social system are combined and co-ordinated – activities which increasingly happen at a distance and on the basis of technologically mediated and abstracted forms of information. 76 Global disease surveillance networks replacing (or substantially augmenting) the localised monitoring and management of disease outbreaks are a good example of distanciation: in the 19th century, public health physician John Snow plotted cases of cholera by hand on a street map and thereby traced their source to the Broad St. water pump; removal of the handle of that single pump successfully curtailed the outbreak. Today’s GP or public health physician typically takes action in response to an e-mail from an authorised source that conveys the recommendations of senior experts, derived in turn from calculations undertaken on regionally, nationally or internationally-aggregated data sets. While these data sets are built from locally collected data, the direct link between documenting local cases and taking local action is replaced with an indirect one based on abstracted and distantly processed data.

Another social change characteristic of modernity is the replacement of reason with rationality. As Sayer77 (drawing on various scholars including Aristotle) has argued, rationality is distinguishable by its formal and instrumental character, its abstraction from concrete situations, and its focus on means rather than ends. For example, it is concerned with identifying the most efficient method of getting a job done (‘doing the thing right’) but is not centrally engaged with the appropriateness and reasonableness of the job itself (‘doing the right thing’). In contrast, phronesis (practical reason) is characterised by its concern with the concrete and the particular; its practical, embodied and tacit character; its focus on ends rather than means (in particular, whether the ends are ethically justified); and its focus on people rather than things.

A reasonable person is someone who takes account of the specificities of the people they interact with, their particular capacities, needs and vulnerabilities, as well as other specificities of the situation. [. . .] When we talk of having ‘reasonable expectations’ of people, we mean expectations that take into account their particular characteristics, constraints and resources, including their vulnerability and fallibility, and ‘reasonable behaviour’ also suggests some degree of emotional sensitivity to others. Further, to be a reasonable person is to be able to imagine things from other people’s standpoints – in other words, to be willing to take the standpoint of the other. [. . .] Hence to call someone ‘a reasonable person’ in such contexts suggests an ethical judgement of them.

p. 65, emphasis in original77

Despite the increasing tendency for expert systems to ‘empty out’ the significance of local events and relationships by invoking general rules that apply everywhere and anywhere, the processes of system integration must necessarily be animated by local actors working in specific local situations and dependent on local contextualities. These will include physical and material constraints (such as the availability and cost of broadband) as well as symbolic ones (such as the extent to which doctors are trusted in a particular society). While the general rules governing the expert system might reflect the impersonal logic of rationality, local actors will, to a greater or lesser extent, seek to act reasonably – that is, ethically and practically with regard to the situation at hand and their (perceived) obligations towards their fellow human beings.

Giddens famously said very little about technology in his original formulation of structuration theory apart from recognising that technologies are key components of social systems, and particularly of expert systems. Other scholars, particularly in the information systems tradition, have attempted to theorise technology from within structuration theory. Notably, Barley proposed that the introduction of a new technology provides an ‘occasion for structuring’ – that is, it creates new possibilities for changing the taken-for-granted routines and patterns of interacting (‘scripts’) between actors, potentially allowing new institutional logics (i.e. patterns of interacting that are culturally and cognitively taken for granted) to emerge. 78 However, as Barley emphasised, the technology does not determine these logics since a key component of structuration is the interpretation that human actors place on the technology and the work they choose to do with it. Barley’s early work on the introduction of computerised tomography (CT) scanners in hospitals informed Orlikowski’s ‘technology structuration theory’, which considers how organisational actors, working collaboratively around common tasks, engage in a process of adapting the meaning, properties and applications of technologies to a particular context, and a parallel process of adapting the context to accommodate the technology. 79 Both Barley and Orlikowski are organisational scholars and so have focused their research efforts at the assimilation of technologies in organisations rather than considering issues of national policy.

In sum, structuration theory offers potential to analyse human action (including the use and non-use of technologies) as social practice, influenced by – and, in turn, influencing – the norms, meaning-systems and rules of a particular social system. It potentially allows us to study the fortunes of planned change efforts either as broadly aligned with prevailing social structures (and, in particular, with actors’ internalised norms, meaning-systems and rules) or as cutting across these structures and creating the preconditions for ontological insecurity.

Strong structuration theory

Whereas Giddens formulated structuration theory in a somewhat abstract way (e.g. focusing on ‘structure’ and ‘agency’ as theoretical concepts), Stones has sought to give the theory a more empirical slant in what he has called ‘strong structuration theory’, focusing on the interpretations and actions of particular people in particular circumstances, and especially on their assessment of particular situations and their efforts to act reasonably in those situations. 80,81

Strong structuration theory’s focus on empirical cases in which individuals are considered as situated in webs of networked relations (Figure 4) resonates with ANT’s focus on dynamic and unstable networks of people and technologies described in the previous section. However, in contrast with ANT’s ‘flat ontology’, SST holds that the recursive (i.e. mutually influencing) relationship between structure and agency remains a useful concept.

FIGURE 4.

Networks of position–practice relations along with technologies in SST. Reproduced with permission from Greenhalgh and Stones. 58

Stones’ quadripartite approach to structuration theory, depicted in Figure 5, considers not only external structures (the macro-level structures that form the contextual conditions of action in any situation) but also internal structures (what agents ‘know’, albeit sometimes imperfectly and even incorrectly, about the external world), actions (observed behaviour) and outcomes (how observed behaviour in a given scenario feeds back to reinforce and/or change internal and external structures). To get an empirical handle on the structure–agency relationship, SST considers the conjuncture (defined as a critical combination of events or circumstances), of which ‘the clinical encounter’ is one example.

FIGURE 5.

Stones’ quadripartite theory of structuration, adapted to encompass a technology dimension. Reproduced with permission from Greenhalgh and Stones. 58

Strong structuration theory proposes that external structures are mediated largely through position–practices (defined as a social position and associated identity and practice), together with the network of social relations that recognise and support it (‘position–practice relations’). Agents’ internal structures may be divided into:

1. general dispositions, which include such things as sociocultural schemas, discourses and world-views, moral and practical principles, attitudes, ambitions, technical and other embodied skills, and personal values – roughly what Bourdieu82 called ‘habitus’; and

2. particular (‘conjuncturally specific’) knowledge of the strategic terrain and how one is expected to act within it, based on one’s hermeneutic understanding of external structures.

Within a particular conjuncture, action occurs when the human agent draws actively and more or less reflexively on his or her internal structures.

To study active agency, SST incorporates theories from phenomenology (the study of agents’ shifting fields and horizons of action arising from the focused activity at hand83), ethnomethodology84 and symbolic interactionism (the study of the subjective meaning and interpretation of human behaviour85).

Strong structuration theory emphasises that while each of us brings generic capabilities, dispositions and strategic knowledge to any particular conjuncture, what we actually do in that situation will depend on a host of specificities including our horizon of action and particular features of context.

The healthcare setting is heavily institutionalised: behaviour in healthcare organisations is strongly influenced by such things as regulations and other governance measures, norms, professional codes of practice and deeply-held traditions (all of which are embodied and reproduced by human agents including clinicians, administrators and patients) rather than exclusively by business concerns like efficiency and profit. The agent’s knowledge of these institutional structures (in the language of SST, the ‘strategic terrain’) may be more or less accurate and more or less adequate. Outcomes of human action in any situation may be intended or unintended, and will feed back on both external and internal structures – either preserving them faithfully or changing them as they are enacted.

As we have argued previously, SST can be enhanced by selected concepts from ANT. 58 First, technologies and human actors might usefully be conceptualised as part of the same network, and both may be thought of as having a ‘position–practice’ in that network (see Figure 5). SST can accommodate ANT’s emphasis on studying the complex and dynamic alignments in the network, including efforts by commercial companies (marketing, lobbying) to influence uptake of technologies, coalition-building within and across stakeholder groupings, and local reinterpretation of objects-in-use. However, SST rejects ANT’s assumption of ‘ontological symmetry’ between people and technologies. Humans and technologies ‘act’ in different ways, and a technology can only ever have a limited and qualified form of agency. Secondly, SST can accommodate ANT’s notion that aspects of the social order can be inscribed in technologies and this may to some extent ‘freeze’ certain position–practice relations within the technology. Thirdly, human behaviour will be shaped and constrained by complex forces in the sociotechnical network and hence may in some senses be an ‘effect’ rather than a ‘cause’ of what we are studying, though SST contends that human agency cannot be reduced entirely to network effects. Indeed, from a SST perspective, what ANT theorists call ‘the sociology of translation’ (interpreting the moves of different actors as they seek to build and stabilise a network of human actors and technologies) is likely to be productively enhanced by considering agency as sited within the human actor rather than merely as a product of the network.

Technological orientations of structuration theory have only recently begun to theorise how the material properties of technologies are coconstituted along with the social practices in which they are used. 86 Conversely, ANT has been criticised for overemphasising the material aspects of technologies at the expense of the lived relationship between humans and material objects. 87 Dant’s notion of ‘material interaction’, informed by phenomenology, in which human agents meaningfully experience, perceive and manipulate objects to achieve a task, has some parallels with the notion of sociomateriality which Barley and Orlikowski’s group have latterly begun to embrace. 88,89 It also resonates with SST’s emphasis on how active agency occurs when the actor draws on phenomenological and embodied dispositions and capabilities (see boxes 2a and 2b in Figure 5).

Figure 4 summarises the broad conceptual model for considering ICT programmes, comprising a network of position–practices (humans and technologies), which evolves over time and is influenced by macro historical and social forces (which exist more or less independently of the agents on which a particular empirical study might focus), and they contribute to the external conditions of action in any given conjuncture. Social structures are embodied and reproduced by both human agents (when they use technologies for meaningful social action) and technologies (when external structures, built into the technology, are instantiated when humans choose to use the technologies – or find they cannot use them as intended).

Having developed the above adaptation of SST using insights from ANT, we drew up a provisional list of questions to guide the study of resistance to ICTs:

Aims

Our aim was to help policy-makers and change agents move beyond the commonly used (but, we believe, impoverished) rational actor models of the (non-)uptake of ICTs and un-nuanced conceptualisations of clinicians’ resistance to them. We sought, through literature review and secondary analysis of data (collected for previous empirical studies by our team8–10,16,90–95) and building on our previous preliminary theoretical work to link SST and ANT,58 to produce a sociologically informed model that would explain ICT adoption and use. In particular, we sought to account for its non-adoption and non-use in certain social and organisational situations. Our starting position was the assumption that clinical work is a social practice, shaped and constrained by wider social influences including norms, meaning-systems and rules, and that ‘resistance’ to ICTs would be explained at least partially in terms of professionals’ ethically-motivated efforts to do their job well and deliver excellence in clinical practice – and that administrative staff in healthcare organisations were also driven, at least in part, by these ethical and professional concerns.

We wished to produce a multilevel model of resistance that unpacked this notion of excellence and which also incorporated the influence of institutional changes in the healthcare field; the meanings people attach to technologies in the workplace; the complex, situated and unpredictable nature of clinical work; the material properties and affordances of ICTs and how these create possibilities and limit what is possible for the user; and the potential for creative human action to overcome the limitations of technology and bridge the ‘model-reality gap’. We were particularly interested in how nationally mandated ICTs were adopted (or not) by local actors.

A normative anchor: what is excellence in clinical practice?

As the literature review above illustrates, many, but by no means all, previous approaches to ICTs in health care begin with the index technology and ask ‘is this technology being used by its intended adopters, and if not, how can we increase or optimise its use?’. Interactional models consider the fit between the technology, organisational routines and individual capability. ANT considers dynamic changes in, and stability of, the sociotechnical network. In this study, framed in SST, we sought to take a different starting point: the internal structures of human agents – in other words, what people know. By ‘know’, we include what they believe and what they value – and how these beliefs and values are influenced by past experience and by the norms, meaning-systems and rules of society described in the section on SST (see Chapter 1, Strong structuration theory). Thus, we sought to shift the focus from means (the use or non-use of a particular technology) to ends (the purpose for which technology is used – or for which it may be more appropriate not to use a technology).

We begin, therefore, not with a focus on technology or even on the system or network, but by depicting the clinician as a social agent and considering clinical practice primarily from a normative (i.e. professional and ethical) perspective. Professionals are strongly influenced by socially shared notions of what good practice consists of and how they should behave towards patients and towards one another.

Accordingly, a key guiding question in this study was, ‘what is excellence in clinical practice and how is the use and non-use of ICTs influenced by (changing) notions of excellence?’. In this section, we summarise the literature on good clinical practice mainly in relation to doctors (because most of the relevant studies have been done on doctors and their work). We also refer to the nursing literature where available.

Much has been written about good doctoring. Perhaps the pithiest statement of its essence is the motto of the UK Royal College of General Practitioners – cum scientia caritas (‘loving care with expert knowledge’: see www.rcgp.org) – which depicts the professional ideal of delivering the highest-quality bioscience while also attending to the human needs of the patient. While the ‘scientia’ component of good doctoring is often equated with evidence-based medicine (EBM), the original definition of EBM actually incorporates ‘caritas’ too. This definition is reproduced in full below, though only the first sentence is generally quoted.

Evidence based medicine is the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients. The practice of evidence based medicine means integrating individual clinical expertise with the best available external clinical evidence from systematic research. By individual clinical expertise we mean the proficiency and judgment that individual clinicians acquire through clinical experience and clinical practice. Increased expertise is reflected in many ways, but especially in more effective and efficient diagnosis and in the more thoughtful identification and compassionate use of individual patients’ predicaments, rights, and preferences in making clinical decisions about their care. By best available external clinical evidence we mean clinically relevant research, often from the basic sciences of medicine, but especially from patient centred clinical research into the accuracy and precision of diagnostic tests (including the clinical examination), the power of prognostic markers, and the efficacy and safety of therapeutic, rehabilitative, and preventive regimens.

p. 7196

Evidence-based medicine was acknowledged by its early protagonists as being dependent on clinical judgement and contingent on patient choice (i.e. the ‘best’ treatment is not necessarily the one shown to be most efficacious in randomised controlled trials but the one that fits a particular set of individual circumstances and aligns with the patient’s preferences and priorities). Unsurprisingly, given its disciplinary roots in clinical epidemiology, the early research literature on EBM focused on the scientific component and sought to build an evidence base of randomised controlled trials and other ‘methodologically robust’ research designs. Later, a tradition of ‘evidence-based patient choice’ emerged in which the patient was assumed to be a (more or less) rational chooser and the clinical challenge was framed as how to convey the research evidence about different treatment options in a way that supported informed patient choice. 97 However, the third component of EBM – individual clinical judgement – has not been extensively theorised within that tradition.

The US bioethicist Kathryn Montgomery, drawing on Aristotle’s notion of praxis, considers clinical practice to be an example of case-based reasoning. 98 Medicine is governed not by hard and fast laws but by competing maxims or rules of thumb; the essence of judgement is deciding which (if any) rule should be applied in a particular circumstance. Clinical judgement incorporates science (especially the results of well-conducted research) and makes use of available tools and technologies (including guidelines and decision-support algorithms that incorporate research findings). However, rather than being determined solely by these elements, clinical judgement is guided both by the scientific evidence and by the practical and ethical question ‘what is it best to do, for this individual, given these circumstances?’.

The dual commitment to ‘scientia’ and ‘caritas’ has been analysed from a philosophical perspective by the Norwegian doctor and philosopher Edvin Schei. ‘Scientia’ requires the practitioner to consider the ‘objective patient’ (i.e., the patient as expressed in terms of measurements and standardised procedures, for which objectively-assessed diagnostic tests and treatments are then considered), whereas ‘caritas’ requires attention to the ‘existential patient’ (i.e. the patient’s subjective experiences and human needs). 99 The notion of evidence-based patient choice can be incorporated in the former (objective) component of good clinical practice but there are also subjective (and intersubjective) aspects of the interaction to consider.

Unlike disease, which can be defined in terms of a typical constellation of symptoms, signs and test results, illness is a personal, lived experience that is both emotionally laden and socially meaningful (e.g. it may come with various connotations of shame and blame100). The good clinician engages reflexively with this lived experience and acts not merely as diagnostician or technical expert but also as active listener101 and professional witness. 102 Using Schei’s definition, good doctoring is ‘a relational competence, where empathic perceptiveness and creativity render doctors capable of using their personal qualities, together with the scientific and technologic tools of medicine, to provide individualized help attuned to the particular circumstances of the patient’ (p. 394). 99 This definition has obvious parallels in nursing and allied professions.

The personal qualities referred to by Schei are strongly aligned with what Aristotle called virtues – character traits that strike a balance between undesirable extremes (such as courage, which lies between the extremes of recklessness and cowardice). 103 Toon has challenged principle-based medical ethics (in which clinical practice is guided by a set of core principles such as beneficence, non-maleficence, autonomy and justice) and argued that above all else, doctors must develop their professional virtues through reflection and peer support. 104 Similarly, good nursing practice has also been defined in terms of personal virtues. 105 The UK General Medical Council acknowledges the importance of virtues in the opening paragraph of its guidance ‘Good Medical Practice’: ‘Patients need good doctors. Good doctors make the care of their patients their first concern: they are competent, keep their knowledge and skills up to date, establish and maintain good relationships with patients and colleagues, are honest and trustworthy, and act with integrity’ (p. 1). 106

Whereas the objective dimension of clinical practice has often been tightly defined (and frequently redefined) in terms of adherence to best evidence guidelines, the subjective dimension is something of a mystery107 – depending as it does on a form of knowledge that is tacit, experiential and difficult to codify. 108 The judgements made by virtuous, wise clinicians entail ethical and practical considerations not just about what to do in relation to the particular circumstances of this patient but also about how to balance the competing demands of advocacy (addressing the needs of the individual patient) and distributive justice (balancing this patient’s needs or wants against the wider needs of the population in the context of limited resources) – for example, in relation to the questions of whether or not to prescribe, whether or not to operate, whether or not (and where and how urgently) to refer, and so on. 107

Critical scholars have voiced concerns that the essence of good clinical practice is being lost as society moves from a traditional era in which medicine and nursing were viewed as vocations, health care as a public good and the sick patient as a vulnerable citizen who had a right to care (and to whom the doctor has a duty of care) to a new era of market values where medicine is a business, health care a transaction and the sick patient a customer. In this latter era, informed choice by ‘empowered’ patients is seen as the driving force for achieving excellence, as clinicians (and health services) that do not produce ‘satisfaction’ will quickly go out of business. The doctor’s role is defined either as seller of specialist services or as an information purveyor. It follows from these assumptions that a good consultation is one in which the patient or their nominated advocate has been given sufficient balanced information to make a well-informed choice. 109 The empowerment of the patient is assumed to exist, more or less, in a zero-sum relationship with the disempowerment of the doctor – with the caveat that many patients do not wish to be completely autonomous (and some do not even seek ‘shared’ decision-making).

However, in the vocational model, patient empowerment and ethical practice are all defined differently. For one thing, it is illness itself, and not medical paternalism, that makes patients vulnerable. 99 Doctors’ specialist knowledge has symbolic significance; in many cases, power is not so much seized by doctors as conferred by society (doctors symbolise hope, trust, agency and authority, making possible a powerful therapeutic alliance of reciprocal interpretation and projection110). This ‘cognitive institution’ facilitates doctor–patient interaction and produces a ‘legitimate hierarchy of domination and subordination, recognized by all participants’ (p. 397)99 – though this hierarchy is rightly renegotiated and redefined more progressively as society evolves. In this hierarchy, patients are doubly vulnerable – because they opt (or are compelled) to rely on the doctor’s skill and judgement in potentially life-threatening situations, and because they expose themselves to the potential for shame or loss of dignity as intimate secrets and body parts are revealed (with the risk of loss of face if this is met with ridicule, disbelief or indifference).

For another thing, as Mol has argued, decision-making (shared or otherwise) is only part of the challenge. 111 As health problems increasingly involve chronic, non-communicable diseases which require ongoing effort by both patients (self-management) and health professionals (periodic surveillance, management of exacerbations and long-term support of disability and impairment), so the ‘logic of choice’ (episodic, decision-focused, objective, predictable – as in a decision tree) becomes less relevant than the ‘logic of care’ (continuous, relationship-focused, intersubjective, unpredictable). The logic of care includes the role of the doctor or nurse as witness and active listener – but it also includes the practicalities of care such as the effectiveness of medication in controlling symptoms, the accessibility of the clinician at times of need, and whether or not tools and technologies introduced with the aim of supporting the process of care are usable and useful in particular situations. In this framing, ‘care’ has a substantial physical and material component as well as a socioemotional one.

Good clinical practice is thus – arguably – less about achieving equal distribution of power (as in shared decision-making) than it is about ensuring that doctors draw on their personal virtues (integrity, honesty and so on) to build a healing relationship, wield their socially conferred power and use technologies pragmatically and judiciously in the patient’s best interests. Patients are also influenced by social norms: they will conform to a greater or lesser extent with expectations of what they understand to be the ideal of a good patient – for example, constructing an account of conscientious self-care, acknowledging the needs of other patients (hence not taking up too much of the doctor’s time), deferring to the doctor (or, perhaps, exerting what seems to be a reasonable level of autonomy) and seeking to demonstrate an appropriate level of knowledge and curiosity about their condition. 112 These small acts of deference contribute further to sustaining the established power differential.

In sum, the clinical consultation is not merely an informational transaction or a set of decisions. Rather, it is a complex social encounter in a heavily institutionalised environment. It has symbolic as well as scientific and practical significance. Good clinical practice involves judgement and attention to the particularities of the patient and their situation (the ‘existential patient’) as well as up-to-date knowledge and incorporation of best scientific evidence (the ‘objective patient’). It follows – and this is the point of this discursion into professional practice – that technologies that support the latter at the expense of the former are likely to be experienced by clinicians as interfering with excellent care.

We recognise that this focus on excellence in professional practice is a radical departure from most approaches to the use and non-use of technologies in health care, but it is very resonant with the ‘moral turn’ in contemporary sociology, which rejects the widespread focus on means (e.g. the efficiency of processes) and exhorts researchers and practitioners to make ends (what the processes are oriented to achieving) their primary focus. 113,114 The emphasis of SST on values and norms within the internal structures (or ‘habitus’) of actors means that humans are viewed not only as rational actors, nodes in a network or members of a sociotechnical system but as moral beings who have commitments, desires and values, both personal and professional. It views work – especially the work of doctors – not merely as a series of co-ordinated tasks but as having symbolic significance in society. In Sayer’s words, ‘things matter to people’ – objects, actions, experiences and relationships have personal and moral significance as well as economic or instrumental worth. 77

With this in mind, our analysis set out to explore the tensions between professional morals and values (inculcated by medical education, professional identity and professional communities of practice) on the one hand and the demands made on healthcare staff in the here and now by the nationally mandated technology on the other. The nexus of ethical values embedded within the habitus of a healthcare professional is not static or unproblematic. Indeed, it may be variously ambivalent, fragmented or conflicting, reflecting the ethical tensions and inherent conflicts of healthcare practice. 99 In this study, we considered that an understanding of these values and how they inform professional notions of excellence would be a useful point of departure for illuminating those practices that are framed as ‘resistance’.

Research questions

Our research questions were:

• How does the dynamic relationship between the macro (external social structures), meso (organisational routines and logics) and micro (individual understandings, dispositions and frontline actions) explain situations in which clinicians and other actors appear to resist the use of nationally mandated ICTs in health care?

• What can we learn from the detailed analysis of small-scale episodes of resistance that will inform the design, delivery and ongoing support of health-related ICT programmes and the development of health policy?

• To what extent is ‘resistance’ a useful framing – and how might the issue of low uptake and use of ICTs be better framed?

• Where should further research effort be directed?

This generic framing of our research questions was modified as needed to match the particular practice and technology under consideration, as described in the case studies in Chapter 4.

Study design

This study had two linked components. One was a review of the literature, focusing mainly on the social sciences, on topics relevant to adoption and non-adoption of ICTs and, more widely, the phenomenon of resistance as sociologists see it. This review focused in particular on the sociology of clinical practice, especially the contemporary ‘moral turn’ (which emphasises what matters to people and what drives them in terms of ‘doing the right thing’) and on the sociology of expert systems, and is presented in the first two chapters.

In parallel with this literature review, and feeding iteratively into it, we undertook a secondary analysis of qualitative and quantitative data gathered for three previous in-depth case studies of computer use in the English NHS, all gathered in the time period 2007–10. In each of these cases, we began with a much larger primary data set and selected for detailed analysis both background material (e.g. policy documents, business plans) and small-scale examples (e.g. video capture or ethnographic field notes on real consultations) relevant to ‘compliance with’ or ‘resistance to’ the index technology.

In order to develop our theoretical model of resistance, and as the main focus of our empirical analysis, we considered a data set on the practice of referral to hospital by GPs, for which the index technology was Choose and Book. This technology had been introduced in 2004 by the English Department of Health to help GPs and their patients to book hospital outpatient appointments remotely. It was anticipated that remote booking would become standard practice once technical challenges were overcome. However, despite political pressure and financial incentives, Choose and Book remained unpopular and was used reluctantly or not at all by many practices. Policy-makers framed this as a problem of ‘clinician resistance’.

We considered Choose and Book as an expert system. Our data set comprised background documents, field notes, interviews, clinical consultations (directly observed and videotaped) and naturally occurring talk relating to referral to hospital in four general practices. We used SST, Giddens’ conceptualisation of expert systems and sensitivity to other sociological perspectives on technology, institutions, the professions and values to examine the relationship between the external environment, the evolving technology and the decisions and actions of human agents (GPs, administrators, managers and patients).

We subsequently applied this theoretical model to two further data sets. One addressed chronic disease management in general practice, for which the index technology was electronic templates built into the local GP record and linked to the pay-for-performance QOF. 115 These were used extensively by practice nurses but rarely by GPs. The other data set was from a national case study of emergency and unscheduled care encounters, for which the index technology was the SCR, a centrally stored extract drawn from a patient’s general practice record. At the time, the SCR was rarely accessed, for complex reasons (again attributed by policymakers to ‘clinician resistance’116).

The secondary analysis of empirical data progressed alongside the literature review, each feeding into the other, as explained in more detail below.

The primary data sets

We used data from two primary studies. Healthcare Electronic Records in Organisations (HERO) was a 30-month study (2007–10), funded by the Medical Research Council (MRC), which sought to explore how work practices and organisational routines were shaped and constrained by electronic patient records in four UK general practices. 91 We studied the local detailed record in use, including the clinical interaction between GPs and patients;92 template-supported chronic disease management;117 and the organisational routine for repeat prescribing. 94 We also studied the use – and, significantly, non-use – of ‘networked’ technologies introduced into general practices as part of national IT policy, notably the Choose and Book online booking system for outpatient appointments. 58

The HERO data set included 50 videotaped consultations, some but not all of which were transcribed in full, and most of which also had contemporaneous screen capture of the dynamic entries onto the electronic record (e.g. it was possible to see, in real time, what the clinician typed onto the record, including entries that were erased before the record was saved), as well as ethnographic field notes, documents (e-mails, letters, business plans, protocols, practice leaflets) and naturalistic interviews. It should be noted that the original aim of the HERO study was to explore the introduction of a much wider range of networked ICTs including electronic transfer of prescriptions between GPs and pharmacies and the SCR (see Chapter 4), but because of severe slippage these programmes did not become active in the study sites within the lifetime of the study.

Our second primary data set was Summary Care Record Independent Evaluation (SCRIE), a 3-year study funded by the Department of Health. We used a mixed-methods case study design to map the multiple stakeholders and complex story of government’s attempt to introduce a centrally stored electronic summary record for every citizen in England. 118 Creation of SCRs was slow because many GPs refused to participate in the programme or allow their patients’ data to be uploaded to the central database. Analysis of a large quantitative data set on out-of-hours encounters showed that even when SCRs were available as a one-click option, clinicians working in unscheduled care settings accessed them in only one case in five. 8 Other findings included strong opposition by a minority of citizens and health professionals to the idea of centrally held medical records,90 ignorance and confusion among the lay public,10 and conflicting interpretations of the programme by clinicians, service users, policy-makers, industry and technical designers. 8,11 The SCRIE data set included ethnographic field notes on 235 unscheduled (emergency and/or out of hours) consultations in primary and secondary care (in which the patient’s SCR was accessed in only 35) as well as field notes, national and local documents, and interviews with staff.

Selecting data sets for secondary analysis

The purpose of this secondary analysis was to draw on both the raw empirical data from the above studies and the analyses already undertaken to develop a robust and detailed theoretical approach to studying what is typically described as ‘resistance’ to ICTs by clinicians and other healthcare staff. However, rather than making the technology the focus of our analysis (arguably, a modernist bias), we chose to focus on particular social practices – referral, chronic disease surveillance and unscheduled care – and consider how the technologies linked to those practices – respectively, Choose and Book, electronic templates on the GP-held record and the SCR – were used (or not) to support that practice.

The data set for each social practice consisted of (a) background documents (e.g. national and local policies, patient information materials, training materials, correspondence, minutes of meetings, press coverage); (b) clinical consultations (observed directly or video/audio recorded); (c) other observed activities (e.g. ‘backstage’ work done by administrative staff on referrals or audits); and (d) interviews with organisational staff about their work. We did not seek to classify the clinician on a unidimensional scale as ‘complying with’ or ‘resisting’ the technology in question, as interaction (and non-interaction) with technology from the perspective of the actor might take many forms including being unaware of its existence, considering it irrelevant, being unable to access it, using it minimally, using it creatively (and not as the designers intended), and so on. Our focus in each case was to understand what the actor ‘knew’, what perceptions and dispositions they brought to the encounter and the actor’s own explanations for how the technology was used (or not).

We had already published a number of papers on technology-supported chronic disease surveillance in general practice90,91,93,119 and technology-supported unscheduled care. 8,9,11,90,116 All had covered the theme of resistance as an integral aspect of the non-adoption, partial adoption, reinvention (or customisation) and abandonment of the index technologies (electronic templates and the SCR, respectively). One publication in particular included substantial excerpts from ethnographic field notes and quotes from interviews, and an extended analysis of why clinicians chose not to use (or sometimes, found that they were unable to use) the SCR,116 though we felt that this analysis would benefit from further theorising.

In addition, we had a large number of raw data, not previously analysed in detail, which covered the practice of referral and use and non-use of the Choose and Book technology. The final data set for our secondary analysis is listed in Table 1. Overall, we placed more emphasis on analysing the Choose and Book secondary data set because this had not been previously analysed for other publications and our resource did not allow us to analyse all three data sets to the same level of detail.

Theoretical framework: strong structuration theory

As set out in detail in the literature review above, our starting position was structuration theory, originally developed by Anthony Giddens and focusing centrally on the reciprocal (mutually-shaping) relationship between social structures and human agency. 71 The original version of this theory has been widely cited, but also criticised in some circles for its alleged rigidity and relatively limited empirical application. It was subsequently extended and refined by one of us (RS) as SST, an empirically grounded refinement of Giddens’ original theory which focuses on micro-level interactions between human agents and considers how these are influenced by (and how they, in turn, influence) wider social structures and relational networks. 80 We enhanced this position using selected elements from ANT, as previously described. 58

Modernity, rationalism and the expert system

Sociologists contend that the forces of modernity include an inexorable drive for change and technological innovation (both of which tend to be viewed in a positive light as representing progress) and a growing tendency to rationalise and formalise tasks and interactions, thereby making these transparent and open to scrutiny. In this study, we were interested to apply SST to the introduction of large-scale technology systems in the NHS, many of which were explicitly linked to policy changes.

NHS policy changes in the early 2000s were strongly influenced by the mind-set of late modernity, with its various commissions and omissions. These included an overwhelmingly rationalist frame of reference; a prevailing belief that innovation and change represent progress; a perception that expert systems (see Chapter 1 for definition) will almost inevitably increase efficiency and save human labour; a failure to think seriously about the possible negative consequences of such technical systems, including their impact on valuable forms of social interaction; and a powerful orientation to a vaguely imagined ‘proximate future’ – that is, a time that is almost but not quite upon us when the index technology is fully functional and all practical, technical, ethical and political challenges have been smoothed out. 74,120

While expert systems may be designed to encapsulate professional expertise, they are nevertheless characterised by an impoverished conception of the professions. An important defining feature of expert systems is their capacity to deploy impersonal knowledge, classificatory systems and procedures to shape, monitor, standardise and render calculable the work they support. Anthropologist Mary Douglas, developing earlier insights from the sociologist Emile Durkheim, set out the argument that the production of lists, rankings and other classification systems helps to establish and then sustain social institutions by introducing conventions which serve to ‘describe the way things are’ in the social world. 121 Classification systems are fiercely negotiated and defended for precisely this reason. 122 They have long been combined with those bureaucratic forms of instrumental rationality which the sociologist Max Weber carefully described and explored. 123

The expert system is a relatively recent phenomenon, resulting from the powerful triad of classificatory systems, bureaucracy and information technology in the age of globalisation. 73,124 Such systems are driven by impersonal and abstract rules, and procedures designed to co-ordinate social relations across large distances.

Giddens proposed that these expert systems, using technology to encode information and store formal knowledge, have an inherent tendency to ‘empty out’ the content of local interactions because the technical knowledge they contain is assumed to have validity independently of any particular interaction, and to have the authority to override situational contingencies. 73 The goals and characteristics of the sociotechnical network in which Choose and Book was introduced and implemented, for example, are typical of such systems,125 and the same is true for chronic disease management templates in general practice,93 though it applies less to our third case of the SCR. The prompts and sanctions embedded within this system marginalise and downplay the in situ characteristics and contingencies of particular cases and encounters. Their ability to exert control and order – measurable, quantifiable – over distance relies on removing (or, at least, attenuating) the ability of distinctive people, relations and contexts to upset the uniform application of the rules and classificatory system embedded in the system. There is a powerful momentum towards general and universalising rules and processes, and away from the application of practical wisdom (what Aristotle called phronesis) in specific contexts.

The ‘rational chooser’ assumption

The assumption behind cognitive models of ICT adoption and use (see Chapter 1, The technology acceptance model) is that the indented user rationally weighs up the pros and cons of the technology and decides if adoption is ‘worthwhile’. This user is assumed to focus in particular on whether or not the technology is easy to use and likely to make a task more effective, efficient or profitable. Such models are often explicitly behaviourist in their framing. As we shall argue in case 2, incentive schemes oriented around paying NHS clinicians or organisations to use a new technology linked to a policy (of which the chief example is the QOF for chronic disease surveillance), for example, are known as ‘pay-for-performance schemes’,115 suggesting that the professional ‘performs’ in order to get the (financial) reward. Likewise, schemes in which organisations or clinicians are financially penalised or subjected to public release of performance data (‘named and shamed’) for not using a particular technology, implicitly rest on what are arguably simplified, behaviourist ‘rational actor’ assumptions.

In our own analysis, we deliberately did not seek to classify the clinician on a unidimensional scale as ‘complying with’ or ‘resisting’ the technology in question (or, to express this slightly differently, as having more or less ‘resistance’ to the technology), as this would have bought into a naively cognitive and rationalist view of the phenomenon.

In the case of Choose and Book, the classificatory rules and procedures driving, and embedded in, the software and its networks were informed by another oversimplified modernist image – of the sick patient as a rational chooser, able and willing to weigh up information about the range of potential options for outpatient referral and decide between them (or as potentially capable of this if provided with high-quality information and decision support). The picture assumes that managing illness consists, more or less, of making a series of objective decisions based on a limited number of decontextualised indicators and then following through on these. It follows from these assumptions that provision of statistical information on the ‘quality’ of services in a standardised format (such as a table of comparisons using star ratings on key metrics) will prompt the ‘right’ choices and that these choices will lead to the ‘best’ services winning out in a competitive market. The policy ideology was that by adjusting the regulatory structures of the NHS through the coercive introduction of choice, government would be able to engineer a shift in professional norms (from ‘doctor knows best’ to ‘informed decisions by empowered patients’) and the cultural logics that define expected behaviour (supporting informed patient choice would become ‘business as usual’ throughout the NHS). 126

The sociotechnical system of the Choose and Book initiative was also heavily influenced by the ideology of competition, and by the salutary effects that competition was said to have on cost-efficiency, patient satisfaction and patient outcomes. It is appropriate to use the term ‘ideology’ here, as Choose and Book seems to have been driven quite forcefully by faith in a hybrid which saw the well-worn belief in competition as a social mechanism allied to the novel facilitating conditions of expert systems. This was ideological in that a policy of enforced competition between hospital providers – driven, it was assumed, by the rationally-choosing patient – was pushed forward with considerable confidence, deflecting attention away from that strategy’s limited evidence base (in particular, its lack of anchoring in the demonstrable conditions, processes, possibilities and consequences relevant to its assumed success). There was no detailed surfacing or exploration of the meso- and micro-level interactions and processes that would convert the policy idea (of hospitals competing to attract outpatient referrals) into the reality of a more efficient, effective and responsive healthcare system that improved patient satisfaction and outcomes.

It was with a view to countering this misleading degree of abstraction that we sought to develop a new, more empirically grounded theory of resistance to Choose and Book (and to nationally mandated ICTs more generally) that rested far less on naive rationalist assumptions.

Applying the normative anchor: medicine’s ‘internal goods’

We began by adopting MacIntyre’s conception of the internal goods and notions of excellence intrinsic to a set of practices. 127 By ‘internal goods’, we include the Aristotlean virtues, along with the dispositions and capacities, that are valued by clinicians and which they believe are necessary to sustain standards of excellence in their profession (see Chapter 2, A normative anchor: what is excellence in clinical practice?). We then use these internal goods as a benchmark against which to judge not merely the means by which an action is taken (e.g. a referral to hospital is made either via traditional letter or the Choose and Book technology) but also the ends that are in mind when that action is taken. Importantly, much of the knowledge and judgement of healthcare professionals, particularly in primary health care, relies on being rooted in the immediacies of context.

A GP’s judgements about the referral of a patient to hospital, for example, have traditionally been informed by (a) knowledge of the patient’s personal history (both medical and social), (b) knowledge of the workings of their own practice (including the various incentives, disincentives and practicalities of different options), and (c) knowledge of local social relations, including the character of local hospitals and the clinical interests and personal style of particular consultants. The ‘expert system’ character of Choose and Book militates against using such knowledge, thus impoverishing the scope and quality of judgements. We expand on this argument using empirical examples in Chapter 4.

Medicine’s internal goods are clustered, broadly speaking, around the themes of caring, curing and comforting, and are embedded in the formal and informal codes of practice of the medical, nursing and other related professions. It has long been argued by sociologists that because they bear a commitment to the refined knowledge, ethics and values of their specialised community, professionals act as a bulwark against the impersonal march of capitalist and bureaucratic forces. 128 More recently, French sociological theorist Luc Boltanski has called for policy-makers to go beyond ‘neomanagerialism’ and engage with the moral and normative positions taken by individuals and groups on particular issues, notably the ethically motivated concerns of professionals and lobbyists. 113

Given that medicine’s professional norms are built on shared and deeply held assumptions and values, which differ significantly from the assumptions behind ‘rational economic man’ or from the narrow values of efficiency and cost-effectiveness, nationally mandated ICTs were always likely to encounter problems if their design did not (at the very least) acknowledge what could be lost if these technologies were widely implemented. While economic and resource issues need to be taken seriously, so do professional norms and the internal goods of health care.

Choose and Book, like all expert systems, will necessarily impose abstract and generalised protocols that have limited capacity to take account of local circumstances and contingencies. Similarly, templates for chronic disease management that are built around national guidelines and a pay-for-performance system will tend to impose a highly standardised structure and set of priorities on the consultation. This leads to a tendency to greatly oversimplify the nature of, and to downplay the significance of, local situations and interactions. In an area such as health care, this can have highly negative effects, as medicine is inherently exception-filled (most cases differ in some way from the ‘textbook case’) and medicine’s internal goods are not, in large part, reducible to formulaic rules and nostrums. 98,99

The problem is that abstract systems uproot social relations from their previous grounding in the immediacies of context; they are unable to take account of all that derives, in terms of well-being and the quality of relationships, from the embeddedness and indexicality of social interactions in local contexts. The intended effect is the remote co-ordination and control of healthcare interactions through a promotion of their abstract, quantifiable and divisible characteristics. However, the (partially intended but partially unrecognised) by-product is a downgrading of practitioners’ skills of responsive interaction and situated judgement, and a corresponding emptying out of the space and time in which these skills used to be deployed. This threatens to subvert highly valued internal goods intrinsic to the activity of general practice, which rely on certain kinds of ‘socially established co-operative human activity’ (p. 175)127 between doctors, nurses and administrators, in their relations both with each other and with patients.

The professional framing of medical work sees doctors (and other clinicians) as wielding their symbolic power with integrity and commitment with the patient’s best interests in mind. While these professional norms are shifting to incorporate, to varying degrees, the ideal of the active and empowered patient, they also assume that there are aspects of the unequal relation between doctors and patients whose legitimacy is socially conferred, due largely to the fact that illness makes people (in a range of ways) vulnerable and in need of society’s help. 99 To the extent that the clinician’s role includes making decisions (perhaps on a shared basis with the patient), these are not merely ‘rational’ (i.e. based on the best available medical evidence) but also practical and ethical – that is, they take account of the situated details of individual cases, including a serious engagement with ‘what matters to people’ to particular flesh-and-blood patients. 77

The issue of patients’ needs and well-being is a complex one that needs to be subjected to much more scrutiny than is given to it in the policy literature on technologies such as Choose and Book, the QOF or the SCR. Neither technologies nor the policies and processes in which they become embedded are ethically neutral, and in order to be able to judge the appropriateness and adequacy, the strengths and the shortcomings, of particular policy initiatives, it is necessary to be able to assess their impact on patient well-being. Having a clear sense of the criteria of well-being relevant to particular cases is the final piece in the toolkit necessary to begin to open up policies and their sociotechnical networks to qualitative critical scrutiny. What do patients care about? What matters to them? These questions will be foregrounded in the analysis so as to focus on ends rather than means.

Analysis of cases and development of theory

Development of theory and analysis of data from our principal data set on referral (Choose and Book) occurred concurrently, each feeding into the other. Two of the authors (TG and DS) are medical doctors with an interest in the sociology of professional practice and clinical interaction; the third (RS) is a professor of sociology who is an acknowledged expert on the work of Anthony Giddens80,129,130 and has applied Giddens’ work on expert systems to the study of cultural, political and economic phenomena outside health care. 131,132 Both the literature review and the empirical analysis were informed mainly by a series of regular discussions among the three authors, grounded in the study’s objectives, research questions and contracted deliverables and informed by the policy background and emerging empirical findings. For these meetings, TG and DS selected illustrative examples of the empirical data to share with RS, and RS in turn selected sociological texts that he felt would inform the analysis of particular items of data. DS was particularly crucial in selecting the cases for discussion as she had undertaken a PhD nested within the HERO study that had entailed close reading and analysis of transcripts selected from 54 video-recorded consultations, and hence was very familiar with this data set. 95

This iterative process drew particularly on the principles of interpretivist information systems research, especially the notion of the hermeneutic circle (constantly comparing each new item of data or theoretical insight with an emerging picture of the whole). 133,134 Using a small subsample of three or four detailed examples, we developed a preliminary analytic framework which we revised as we applied it successively to all the referrals in the relevant secondary data set (see Selecting data sets for secondary analysis). Having produced a theoretical model of resistance to ICT that explained our empirical findings on Choose and Book, we then considered the extent to which the same model also explained our previously published findings on resistance to electronic templates and the SCR.

The methodology of SST applied to the use and non-use of ICTs has been described previously,58 and is summarised above (see Chapter 1, Strong structuration theory). Applying this methodology, we focused on the conjuncture – that is, a critical combination of events and circumstances in which the human agent draws on both habitus (i.e. their internal dispositions, beliefs, values, norms and so on) and knowledge of the here-and-now situation (i.e. their assessment of the particular strategic terrain and how they are expected to act within it), and is supported or constrained by the available technologies, to inform, execute and justify a particular course of action.

For the purposes of our study on referral, we considered two kinds of conjuncture: clinical consultations in which outpatient referrals were initiated and administrative activities in which staff sought to follow through on such referrals. In both types of conjuncture, the actor (clinician or administrator) either used or chose not to use (or, sometimes, was prevented from using) Choose and Book to support the referral. For each conjuncture, we considered how the actor’s habitus (e.g. the doctor’s professional identity and code of practice, or the administrator’s efforts to do a good job) combined with their assessment of external circumstances (including the incentives and reward systems linked to Choose and Book and their perception of what would happen if they did or did not use this technology). The purpose of this was to focus in detail on the extent to which, and the ways in which, the actors in particular situations felt enabled and constrained by the material properties and capabilities of Choose and Book (including the role assumptions, categories, values and other social structures inscribed in the software).

Having developed our preliminary theoretical model on the referral (Choose and Book) data set, including a set of questions to ask of a nationally mandated healthcare technology, we then considered our two other data sets – chronic disease management (QOF templates linked to the GP record) and unscheduled care (SCR). We applied the model to relevant parts of the large empirical data sets. Because of resource constraints on this project, we did not undertake exhaustive analysis of raw data on these two further cases. Rather, we used published versions of these case studies along with unpublished interim summaries, and also returned to selected segments of raw data to check particular aspects of our analysis. In these two additional cases, we were particularly vigilant to seek disconfirming data (i.e. examples that did not fit the explanatory model of resistance developed on the referral data set) and to extend our theorisation to explain such examples.

In the presentation of results that follows, we have offered a narrative synthesis of each case using conventional case study notation and formatting,135 illustrated by quotes and extracts that were selected to illustrate theoretical points derived from analysis of the wider data set.

Description of data set

As described above, our final data set was selected from a much larger combined data set of two large empirical studies, HERO and SCRIE, conducted between 2007 and 2010. It consisted of raw data (not previously analysed or written up) pertaining to one technology-supported practice (referral) and completed reports along with interim summaries and selected data extracts from two other practices (chronic disease surveillance and unscheduled care). The material used in the three case studies is listed in Table 1 (referral – Choose and Book), Box 1 (chronic disease surveillance – QOF templates) and Box 2 (unscheduled care – the SCR).

In the sections that follow, we first present a worked-up analysis of our main case study (referral, for which the index technology was Choose and Book). We give a brief narrative of the policy background to Choose and Book, and then describe the material properties of the technology, how it was intended to be used, how it was actually used in both clinical and administrative settings (and/or why it was not used as intended by its designers and policy-makers), and what the consequences were. We then offer a new theorisation of the non-use of Choose and Book technology using our adaptation of structuration theory, and suggest that it would be fruitful to move beyond the term ‘resistance’ to conceptualise and (where appropriate) overcome the non-use of ICTs in healthcare settings.

We then apply our theoretical model to the two supplementary case studies. We describe these cases (chronic disease surveillance and unscheduled care) briefly, and suggest how and to what extent the model may be transferable to other policy-driven technology projects in healthcare. We have deliberately placed most emphasis on case study 1 in this write-up because this was where most of our analytic effort was directed.

Case study 1: referral (‘Choose and Book’)

Case study 2: chronic disease surveillance (templates linked to Quality and Outcomes Framework)

Case study 3: unscheduled care (the Summary Care Record)

Practical questions for studying resistance in real-world settings

The academic outputs from this study include a contribution to social theory and will be published mainly in medical sociology journals. 166 However, in addition, we wish to contribute to efforts by policy-makers and change agents to reduce (what are perceived as) inefficiency and waste in healthcare organisations by overcoming the widespread phenomenon of ICTs that clinicians do not use. We do not share the assumption that health systems would be maximally efficient and effective if clinicians never resisted ICT use, but that position has not prevented us from developing a preliminary set of questions that should be asked when staff ‘resist’ using nationally mandated technologies.

First, what is the nature and justification of the policy that this technology was introduced to support? Our three case studies illustrate widely different professional positions on this question from near-universally in favour (of systematic, structured chronic disease surveillance) to near-universally opposed (to the ‘choice’ policy in hospital referral), as well as a profession divided over whether or not extracts from patients’ records should be made accessible on a national remotely accessible database.

Second, what are the material properties and limitations of the technology under conditions of expected use and how do these properties shape and constrain the possibilities for action? Again, our case examples illustrate wide variation in the extent to which this factor impacted on practice – from extreme material mismatch between the technology and the task it had been introduced to support (Choose and Book) to close alignment between these (chronic disease templates), and a somewhat middle position for the SCR. An aspect of materiality is cost – including the question of opportunity cost (from where is resource diverted to fund the technology and the expert system to which it is linked?). Choose and Book was costly but not cost-effective for participating GP practices, and financial incentives cut little ice in this context. The technology to support the QOF, in contrast, cost practices little or nothing to install (or at least, it came as part of a standard software package), and promised lucrative benefits if used as intended. The cost equation for the SCR was unbalanced as those who did the work (GPs and their staff creating records) were not the ones to benefit if SCRs were used – a fact that may well have contributed to reluctance to participate in the scheme.

Third, to what extent, and in what way, are local, contextual judgements attenuated or compromised when the technology is used? This question encompasses the more specific question: to what extent does use of the technology require the user to place undue emphasis on generic, abstracted forms of knowledge and/or knowledge that has been generated at another time or place over the local, particular and of-the-moment knowledge in the clinical encounter? In this regard, all the technologies studied produced this difficulty. At the most extreme level, use of Choose and Book inevitably required a profound ‘emptying-out’ of the local situated detail of the encounter, leading to awkwardness and even absurdity in negotiating referral decisions. Chronic disease surveillance templates ‘configured the patient’ and produced pop-up prompts that were intrusive to the flow of the consultation, and the SCR offered ‘objective’ information on drugs and allergies that sometimes conflicted with the patient’s here-and-now account (and, perhaps, the bottle of medication they held in their hand) – and its defined data fields may or may not have accorded with the data needed clinically or with what the patient themselves wished to share or keep private.

Finally, what changes in social roles and relations are required or made possible when the technology is used? In particular, how does the technology ‘configure the user’ and what are the implications for professional identity and the delivery of high-quality, ethical care? Again, Choose and Book stood out as requiring GPs to adopt roles that were perceived as professionally demeaning and even unethical. Some GPs also viewed uploading SCRs as a breach of their professional code of conduct. In contrast, the roles expected by the QOF and inscribed in the chronic disease surveillance templates produced relatively minor conflicts in this regard (although they did result in new distributions of work between doctors, nursing and administrative staff).

Upstream of these questions about resistance to the technology is a preliminary, overarching question about means versus ends: have professionals’ concerns about the ends (the achievement of high-quality, ethical practice) been fully acknowledged and addressed, or have these concerns been dismissed in the pursuit of means (‘efficiency’)? Again, the means-versus-ends tension played out most dramatically in relation to Choose and Book, though it was also illustrated by both of the other case examples.

An anonymous reviewer of a previous draft of this report made the following perceptive comment, which aptly sums up the implications of the above for the frontline clinician:

Perhaps the strongest message I am taking away from [this] study is a renewed conviction that professionals need to understand a system, to learn it, and then to use it to enhance their practice rather than substitute for it.

To this we would add the following caveat: that this ‘enhancement’ of practice may only be fully realised if the design of systems becomes sufficiently customisable by professionals in their particular context of use that it becomes possible to align the technology with the professional and ethical concerns of clinicians and the needs of their patients.

Some avenues for further research

Our findings imply that if policy-makers and change agents seek to reduce ‘resistance’ to centrally mandated IT systems, there are huge potential benefits to seeking a genuine dialogue with the world of professional values at the stage of design and implementation. Efforts to strike a balance between such virtual, remote, systems and the exigencies of the local sites in which professional values are performed are likely to be rewarded by the emergence of technologies and systems that support excellent professional practice rather than – whether by accident or design – cut across the granular details of such practice. If there is one overarching conclusion to our three case studies, it is that when (and to the extent that) the design and introduction of nationally mandated IT systems fails to take account of deeply held professional norms, values and social roles, such technologies will receive (at best) limited interest and (more likely) active opposition from professionals and the people who support them in their work.

While this study has illustrated potential synergies between SST and ANT and identified some important preliminary findings (see previous paragraph), it has raised more questions than it has answered about how such things as ‘professional virtues/standards’, ‘good clinical care’ and the symbolic role of the health professions in society might be placed more centrally in research on new technologies in health care. We propose a number of additional avenues for further research.

The first avenue is methodological. While small-scale ethnographic studies undoubtedly provide a useful window into the micro-level detail of clinical practice and the material reality of frontline technology use,167 and while such ethnographic studies can also illuminate the meso level of organisational life, reviewers of an earlier draft of this report were less convinced that the data sets used in the three case studies had allowed us to gain maximum purchase on the various external (‘macro-level’) social structures that we conceptualised as influencing the human actors we had observed. As with all research that offers a significant shift in paradigm, methodology is only roughly specified in early studies and becomes more systematised and sophisticated as the paradigm develops. 168

A specific aspect of methodology that might be developed is how to achieve in-depth study of large-scale networked systems in use. A limitation of the main case study in this report (Choose and Book) was that we did not study the hospital ‘end’ of that technology (because the research was funded from a MRC grant to look at the use of electronic records in general practice, and so did not ‘follow the technology’ beyond that setting). This and similar case studies would of course be more complete if they pursued all of the different types of local setting that a networked technology reached. In an untitled online paper, Star has written powerfully on the challenges of using a small-scale technique such as ethnography to study a large-scale networked infrastructure. 169 Elsewhere, she suggests a methodology for undertaking such an ‘ethnography of infrastructure’. 170

The second avenue is further theory development. Professional identities and practices are increasingly complex, fragmented and interdependent. The three practices we studied in this secondary analysis were all relatively discrete; they all involved one individual working with the technology in a fairly autonomous way (though see DS’s previous work on the complex ‘voices’ in even these relatively simple use cases95). Many technologies in healthcare support (or, at least, are intended to support) multidisciplinary team working in which multiple professionals, as well as the patient, carer(s) and staff from beyond the healthcare sector, interact in complex ways over time. In a different study, for example, we have begun to theorise the use and non-use of assisted living technologies using a social practice lens and drawing on the theoretical insights from the present study. 171

The third avenue is application of findings. Surfacing such things as the dispositions, values, virtues and concerns of both patients and professionals and seeing how these explain the use (or not) of existing technologies may be an important starting point. However, the definitive phase of any such programme of research will be assessing how these findings can be used to inform the design of new technologies and systems that are (a) more likely to support good clinical practice and (b) less likely to be resisted. We anticipate that a codesign or corealisation approach will prove particularly helpful here. 47

Finally, there is the question of the ‘generalisablity’ of our theory of resistance raised by reviewers of this report. The three cases (three contrasting social practices linked to three very different technologies) all ‘fitted’ the theory – but this study is preliminary and was not designed or resourced to provide a definitive theoretical framework that has been fully tested across the full range of possibilities. We tentatively hypothesise that the approach we developed is likely to align with examples where resistance (e.g. by doctors, nurses, pharmacists and those working to support their practice) is rooted in professional identity and values, but that there may be other forms of resistance whose underlying drivers are not captured by this conceptual approach.

Reflections on the ‘structuration theory/actor–network theory hybrid’

We began by rejecting a conceptualisation of resistance to technology adoption that is common in the health informatics literature and policy documents (though it is by no means universally held in the academic literature more widely), in which resistance is framed in cognitive–behaviourist terms. Instead, we began from the position adopted by social scientists that the clinical consultation is a complex social encounter in a heavily institutionalised environment. From this, it follows that ‘resistance’ to ICT use is a complex phenomenon with both normative and sociomaterial components; it is unlikely to be overcome using atheoretical behaviourist measures, nor would attempts to do this be desirable.

The two coapplicants for this study, TG and RS, had met when TG, excited by the potential of ANT to theorise large-scale IT projects but also frustrated by its limitations, contacted RS, whose reading of Giddens’ structuration theory included mapping networks and how these networks changed over time (see Chapter 1, Strong structure theory). When we applied for funding for the present study, we had recently published a paper in Social Science and Medicine entitled ‘Theorising big IT programmes in healthcare: strong structuration theory meets actor–network theory’. 58 We had planned to develop our new ‘hybrid’ of SST and ANT in more detail. Here, we consider how successful that attempt was, both in general and as applied to each of our three cases.

We spent most of our time on the first case study (referral to outpatients by GPs, using – or not using – the Choose and Book technology). Let us first consider how a ‘pure’ ANT approach to this case might have unfolded. It would probably have focused on the ‘sociology of translation’, specifically on efforts by different interest groups to achieve the stages of problemisation, interessement, enrolment and mobilisation in Chapter 1 (see Actor–network theory). We would have considered Choose and Book as a ‘program’ and the efforts of detractors [such as the GPs in the e-mail correspondence we described in Chapter 4 (see Resistance articulated: some examples of correspondence and publications)] as attempting to mobilise an ‘anti-program’. 65 The focus of the analysis would have been on the network of ‘actants’ (people and technologies) and how it changed over time, and especially on interactions and artefacts that helped (or failed) to stablise the network. For example, the efforts by PCT managers to impose a particular set of ‘quality standards’ of performance rankings in the use of Choose and Book could have been presented as an attempt to use this artefact to stabilise the actor–network.

In ANT, irreversibility of a network is viewed as being achieved through ‘black boxes’ – that is, through configurations of actors (human and non-human) that are taken for granted as the way things are, and hence no longer questioned. A key contributor to this is what Latour called immutable mobiles – inscription devices (such as maps, graphs, quality standards, rankings and so on, as well as the codes and categories inscribed in software) that move within a network and its nodal points of passage, but which retain the same essential properties in different contexts, thereby allowing relations to be performed in the same way in a variety of different settings and locations. 172

Our empirical data set was small-scale, qualitative and based largely on ethnographic observation. This was because we sought to study social practices. We could have done this from an ANT perspective, as ANT is a performative theory – that is, it seeks to study what people do, and what is brought into being through practice(s). Indeed, had we wished merely to reject ANT’s controversial notion of symmetry between human actors and technologies (all of whom are ‘actants’ in the network), we could have aligned with Schatzki’s widely-used ‘practice theory’, which similarly views humans and technologies as linked in networks and studies practices, and actions as effortful, collective accomplishments. 173 Nicolini, for example, has studied the introduction of, and (indirectly) resistance to, telemonitoring using a Schatzkian practice theory lens along with ANT’s notion of the sociology of translation. 174,175

For actor–network theorists and practice theorists, stability of the network (and hence successful implementation of a big IT programme) is seen as being achieved through both people (who argue rhetorically for particular viewpoints or truth statements) and things (such as graphs, diagrams, lists and other inscriptions as well as the hidden infrastructure of hardware and software). As Bowker and Star put it:

Marx referred to technology as ‘frozen labor’ – work and its values embedded and transcribed in transportable form. Modern information technologies similarly embed and inscribe work in ways that are important for policy-makers, but which are often difficult to see. . . . Arguments, decisions and uncertainties and the processual nature of decision-making are hidden away inside a piece of technology or in a complex representation. Thus, values, opinions, and rhetoric are frozen into codes, electronic thresholds and computer applications. Extending Marx, then, we can say that software is frozen organizational and policy discourse.

p. 135122

Importantly, an ANT analysis of Choose and Book would necessarily have embraced the theory’s ‘flat ontology’ – the notion that the distinction between social structures and human agency does not exist (or, perhaps, is an unhelpful way of conceptualising reality). Rather, ‘layers’ in society are viewed as consisting of (usually, inherently unstable) ‘black boxes’ which tend to need continual effort to be maintained. ANT holds that there are no pre-existing layers but only ‘a single plane of endlessly entangled translations’ (p. 173). 176 The nearest a hard-line actor–network theorist will get to admitting the existence of social structures is the notion of the ‘black box’ (see above) – though even the latter represents only a set of stable-for-now relations that could change at any time. Callon and Latour, for example, produced the widely quoted statement back in 1986 that ‘macro-actors are micro-actors seated on top of many (leaky) black boxes’. 177 Nicolini’s notion of understanding practice by repeatedly ‘zooming in’ and ‘zooming out’ on the various ‘black boxes’ resonates with this metaphor. 175

Harris has summed up ANT’s treatment of social institutions:

. . . institutions are loci of methods of ordering, whose essential operation resides in the recurrent patterns by which the relations between humans (subjects) and non-humans (objects) are generated and maintained. These patterns are not architectonic, they do not precede or exceed the site of their operation: organizations (noun) are sustained by organization (verb).

p. 165176

In our analysis of the social practice of referral and the use and non-use of the Choose and Book technology, we returned, somewhat unfashionably perhaps, to the layered ontology of structuration theory and the concept of social institutions in the way Giddens (and others such as Bourdieu) conceptualised them. In particular, our theorisation of the practice of referral placed central emphasis on the social significance of medicine as a profession, the question of medicine’s ‘internal goods’ (MacIntyre) and the normative expectations on both clinicians and patients of how they ‘should’ behave.

As we argued early on in this report:

it is illness itself, and not medical paternalism, that makes patients vulnerable. 99 Doctors’ specialist knowledge has symbolic significance; in many cases, power is not so much seized by doctors as conferred by society (doctors symbolise hope, trust, agency and authority, making possible a powerful therapeutic alliance of reciprocal interpretation and projection110). This ‘cognitive institution’ facilitates doctor–patient interaction and produces a ‘legitimate hierarchy of domination and subordination, recognized by all participants’ (p. 397)99 – though this hierarchy is rightly renegotiated and redefined more progressively as society evolves.

This document, Chapter 2, A normative anchor: what is excellence in clinical practice?

And furthermore:

Good clinical practice is . . . less about achieving equal distribution of power (as in shared decision-making) than it is about ensuring that doctors draw on their personal virtues (integrity, honesty and so on) to build a healing relationship, wield their socially conferred power and use technologies pragmatically and judiciously in the patient’s best interests. Patients are also influenced by social norms: they will conform to a greater or lesser extent with expectations of what they understand to be the ideal of a good patient – for example, constructing an account of conscientious self-care, acknowledging the needs of other patients (hence not taking up too much of the doctor’s time), deferring to the doctor (or, perhaps, exerting what seems to be a reasonable level of autonomy) and seeking to demonstrate an appropriate level of knowledge and curiosity about their condition. 112 These small acts of deference contribute further to sustaining the established power differential.

This document, Chapter 2, A normative anchor: what is excellence in clinical practice?

This explicitly professional and normative framing of practice allowed us to begin our analysis of the referral/Choose and Book case study by considering the habitus of the doctor, nurse or other member of the healthcare team, how that habitus (the ‘internal structures’ of the individual – knowledge, attitudes, virtues, dispositions and so on) was shaped and influenced by external social structures (in particular, professional norms about what good medical practice and good nursing practice consist of and what Schei99 called the ‘cognitive institution’ of society’s expectations about health professionals), and how action resulted from the interplay of internal structures and the unfolding of here-and-now events in any conjuncture.

As set out in the discussion section for this case (see Chapter 4), this focus on structures (internal and external) thus allowed us to study ‘resistance’ to Choose and Book through the lens of ethical, professional practice (at least as perceived by its practitioners) and link with the ‘moral turn’ in sociology as set out by (among others) Sayer and Boltanski. 113 The empirical data in the Choose and Book case revealed that doctors’ (and managers’ and patients’) actions are strongly influenced by their normative perceptions of how they are expected to behave by virtue of their social position – and that these perceptions (which some have termed ‘scripts’178) are an excellent starting point for studying ‘resistance’.

As we explored the case of referral/Choose and Book, the concept of the expert system [defined and explained in detail in Chapter 1 (see Giddens’ work on structuration, modernity and technology)] became central to our analysis. In particular, this case illustrated what was, potentially, a more general phenomenon: ‘resistance’ playing out as a clash between the abstracted, generalised knowledge and ‘frozen discourses’ (see quote above, this section) of the expert system and the situated, case-based ‘ethical practices’ that doctors, nurses and even non-clinical staff felt driven to provide for their patients. The concept of expert systems also allowed us to highlight the difference between means-oriented and ends-oriented practice – a source of much tension in this particular case.

Our second case, which considered the social practice of chronic disease surveillance and the technology of ‘QOF templates’, also highlighted to some extent the tension between an expert system (in this case, ‘evidence based’ guidelines for chronic disease management linked to a national pay-for-performance scheme) and the exigencies of the local situation (in this case, the fact that patients do not behave as textbook disease cases or come with one problem at a time). Again, much of this tension could be explained in terms of internal social structures and the scripts about ‘good practice’ that guided health professionals’ behaviour.

The last case study – the social practice of unscheduled care and the SCR technology – is slightly different in that the SCR was not an ‘expert system’ in the same sense as Choose and Book or the QOF templates. Every SCR contained different data because those data were patient-specific. To some extent, however, it did inscribe a body of (more or less contested) expert knowledge – for example, knowledge about what constituted ‘immediate and necessary’ data about a patient in an emergency, what the ‘minimum data set’ of sharable fields within any GP-held record should be, or which diseases or risk states should be classified as ‘private’ or ‘sensitive’ (and, therefore, not to be shared).

At the time of our study on the adoption and non-adoption of the SCR, there was much debate about all these issues at national level (within Connecting for Health and also within the Royal Colleges – as illustrated by the attempt by the Royal College of General Practitioners to define the fields relevant for an ‘enriched’ record). 116 Arguably, committees charged with sharpening the definitions and clarifying the classification system (defining what Douglas called ‘the way things are’121) continually failed to resolve their internal disagreements precisely because such knowledge can never be generic and context-free. Rather, a data item (‘rash on penicillin’, ‘attention-deficit disorder’ – or even the person’s home address) acquires the qualities of relevance, urgency, sensitivity and so on only in relation to particular scenarios and particular contexts. In the sense that the SCR contained inscribed knowledge about which data fields were to be made available and which were irrelevant or ‘private’ (hence not to be uploaded onto the nationally accessible SCR), it was an expert system – and the reluctance of doctors and others to help create it stemmed at least partly from a lack of confidence in that expert system.

Having placed social structures so central to our analysis and used SST as our primary analytic lens, the question arises: did we draw on ANT to any significant extent? This is a question on which actor–network theorists are unlikely to agree, as ANT is a broad church whose protagonists range from purists who would consider ANT to be a ‘post-social’ (i.e. to exclude any possibility of theorising such things as social roles and identities179) to those who use its concepts eclectically and pragmatically180 and others who describe themselves as ‘post-actor–network theorists’. 181,182

We believe that ANT influenced our analysis in three ways but that each of those influences had limits. First, it gave us the general orientation towards sociotechnical networks: networks of people and technologies that are complex and (to a greater or lesser extent) unstable, and which may be stabilised through the efforts of people and/or the substance afforded by material things. However, we treated the people and the technologies very differently (as illustrated in Figure 5). We considered human agency to be important in its own right, and not to be merely a product of the network; unlike the stance taken by ANT, we viewed that agency as profoundly influenced by, and reciprocally linked to, social structures.

The second way in which ANT allowed us to extend and enhance SST was the notion that people become particular actors as a result of their position in the sociotechnical network. This is perhaps best illustrated by our second case: chronic disease management supported by QOF templates. The nurses whose work we observed became, to a significant extent, the fillers-out of the templates. Chronic disease management was perhaps not reduced to the entry of data onto the template but it was certainly oriented around that task – not least because so little time was available to address other aspects of the nurse’s work or the spontaneously-raised symptoms and concerns of the patient. However, we found the notion of a professional role as purely the product of all the relationships in the network (as with the quote from Berg in Chapter 1) inadequate to explain all of our findings. A good example of this is the nurse who negotiated with her GP employers to allocate 30 rather than 15 minutes per patient. This change was achieved through human agency, driven by dispositions and professional values: the nurse had insisted that her ‘job’ was not merely completing the templates, important and evidence-based though these were. Good nursing practice was also about providing empathic listening to the patient’s concerns and dealing with whatever issues the patient wished to raise. The actions of this nurse are, we contend, thus better explained using a layered ontology than a ‘single plane of endlessly entangled translations’.

Third, because of its focus on the materiality of technology, ANT allowed us to explore how these material properties interacted with clinicians’ internal structures (dispositions and professional values, especially in relation to the uniquely confidential nature of sensitive information about patients’ illnesses), influenced in turn by external structures (professional norms and the ‘internal goods’ of medicine and nursing) to shape and/or constrain action in any given situation – and to produce unintended consequences that fed back to influence the wider network. For example, in relation to the SCR, the very idea of a source of complete and accurate data on a patient’s medication and allergies, available ‘at the touch of a button’ to someone with appropriate authorisation, depends heavily on the button providing reliable access to the promised data fields when it is touched. Our data revealed how, even when accessing the SCR resonated with the script of good professional practice, the SCR sometimes provided nothing more than an error message. In relation to Choose and Book, the technology sometimes offered up no available outpatient slots for the desired referral. In both cases, even when the technology supported rather than conflicted with ‘good professional practice’ (and especially when it did not), its physical unreliability and/or sheer material complexity tended to generate unintended consequences for the clinician, patient or organisation. However, we found that ANT did not allow us to theorise these unintended consequences adequately, as a key finding in our data set was how these unintended consequences grated uncomfortably with clinicians’ perceptions of their professional role, their duties to their patients and their definition of excellent, ethical practice.

Conclusion

This study has confirmed the findings of previous researchers who have studied healthcare practices through a social science lens: the clinical consultation is a complex social encounter in a heavily institutionalised environment. ‘Resistance’ to ICT use is a complex phenomenon with both normative and sociomaterial components. It is, of course, unlikely to be overcome using atheoretical behaviourist measures, nor would attempts to do this be desirable. Rather, we suggest a number of questions that should be asked when staff ‘resist’ using nationally mandated technologies: (1) what is the nature and justification of the policy that this technology was introduced to support?; (2) what are the material properties and limitations of the technology under conditions of expected use (including its cost and the opportunity cost of introducing and supporting it) and how do these properties shape and constrain the possibilities for action?; (3) to what extent, and in what way, are local, contextual judgements attenuated or compromised when the technology is used?; and (4) what changes in social roles and relations are required or made possible when the technology is used? In particular, how does the technology ‘configure the user’ via inbuilt expectations for roles and behaviours, and what are the implications for professional identity and the delivery of high-quality, ethical care?

Upstream of these questions about resistance to the technology is a preliminary, overarching question about means versus ends: have professionals’ (and other staff’s) concerns about the ends (the achievement of high-quality, ethical practice) been fully acknowledged and addressed, or have these concerns been dismissed in the pursuit of means (efficiency of tasks and processes)?

Our findings are consistent with the conclusion that those who seek to reduce ‘resistance’ to centrally mandated IT systems should seek a dialogue with the world of professional values at the stage of design and implementation and show willingness to strike a balance between such virtual, remote systems and the exigencies of the local sites in which professional values are performed. However, a potential danger of this theoretical framing pointed out by reviewers of this report is that it could be mistakenly read as suggesting a somewhat naive dualism: ‘professionals’ (local) perspective good, policy-makers’ (distant) perspective bad’. It is clear that professionals, too, have their vested interests, which should be approached critically, and national policy-makers may in other cases be more flexible and biddable than the examples in this study suggest. Another way in which this research could be taken forward is to explore, in a more nuanced way, how the tension between mandated policy (introduced through technology) and professional resistance plays out.

Thanks

We thank the study participants, the members of the HERO-2 steering group, and four anonymous peer reviewers for helpful comments on earlier drafts of this report.

Funding

This secondary analysis of data was funded by the National Institute for Health Research (study no. 10/1011/01). The primary studies that produced the empirical data sets were funded by the MRC (Methodology Research Programme 07/133) and the National Institute for Health Research (CFHEP002 and CFHEP007).

Ethical approval

Ethical approval for the empirical work was obtained from Thames Valley Multicentre Research Ethics Committee in January 2007 (06/MRE12/81 and subsequent amendments) and North West 8 Multicentre Research Ethics Committee in September 2009 (09/H1013/36 and subsequent amendments).

Contributions of authors

Trisha Greenhalgh was the principal investigator on both of the primary studies that produced the empirical data sets for this secondary analysis.

Deborah Swinglehurst was the lead fieldworker on one of those studies (HERO); her PhD included data collection for the first case study (Choose and Book) and a detailed analysis of the second (chronic disease management templates in general practice), excerpts and adaptations of which are included in this report.

Rob Stones provided theoretical input, especially but not exclusively in relation to the conceptualisation of expert systems, the analysis of medicine and nursing from a normative perspective and the application of structuration theory to the empirical data.

Trisha Greenhalgh and Rob Stones conceptualised the study. All authors contributed to the data analysis and production and refinement of the theoretical model.

Trisha Greenhalgh took overall responsibility for drafting the final report and Deborah Swinglehurst and Rob Stones drafted sections and contributed to revisions.

All authors approved the final manuscript.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HS&DR programme or the Department of Health.