Interpreted consultations as ‘business as usual’? An analysis of organisational routines in general practices


Address for correspondence: Trisha Greenhalgh, Department of Primary Care and Population Sciences, University College London, 417 Holborn Union Building, Highgate Hill, London N19 5LW


UK general practices operate in an environment of high linguistic diversity, because of recent large-scale immigration and of the NHS's commitment to provide a professional interpreter to any patient if needed. Much activity in general practice is co-ordinated and patterned into organisational routines (defined as repeated patterns of interdependent actions, involving multiple actors, bound by rules and customs) that tend to be stable and to persist. If we want to understand how general practices are responding to pressures to develop new routines, such as interpreted consultations, we need to understand how existing organisational routines change. This will then help us to address a second question, which is how the interpreted consultation itself is being enacted and changing as it becomes routinised (or not) in everyday general practice. In seeking answers to these two questions, we undertook a qualitative study of narratives of interpreted primary care consultations in three London boroughs with large minority ethnic populations. In 69 individual interviews and two focus groups, we sought accounts of interpreted consultations from service users, professional interpreters, family member interpreters, general practitioners, practice nurses, receptionists, and practice managers. We asked participants to tell us both positive and negative stories of their experiences. We analysed these data by searching for instances of concepts relating to the organisational routine, the meaning of the interpreted consultation to the practice, and the sociology of medical work. Our findings identified a number of general properties of the interpreted consultation as an organisational routine, including the wide variation in the form of adoption, the stability of the routine, the adaptability of the routine, and the strength of the routine. Our second key finding was that this variation could be partly explained by characteristics of the practice as an organisation, especially whether it was traditional (small, family-run, ‘personal’ identity, typically multilingual, loose division of labour, relatively insular) or contemporary (large, bureaucratic, ‘efficient’ identity, typically monolingual, clear division of labour, richly networked). We conclude that there is a fruitful research agenda to be explored that links the organisational dimension of interpreting services with studies of clinical care and outcomes.


Large-scale immigration to the UK, especially in recent years, has meant that language barriers between patients and health professionals are now commonplace in some areas (Baker and Eversley 2000, Salt 2005). Such barriers, even when an interpreter is present (but especially if one is not) distort communication and may undermine physician-patient trust (Angelelli 2005, Green et al. 2005, Greenhalgh et al. 2006, Robb and Greenhalgh 2006). They account for failure of vulnerable groups to access the services they need (why seek help when you will not understand or be understood?) as well as for a rising proportion of medical errors such as administering medicines to which the patient is allergic, giving inappropriate vaccinations to children, and dismissing early symptoms of cancer as ‘non-specific complaints’. (Angelelli 2005, Chief Medical Officer 2001, Green et al. 2005, Greenhalgh et al. 2006, NHS Executive London Regional Office 2004). The UK National Health Service (NHS) has expressed its commitment to providing an equitable service defined by need, including providing a professional interpreter to any patient if needed (Department of Health 2004).

This study, jointly funded by a local Primary Care Trust (PCT) and Primary Care Research Network, was set up to address a perceived organisational problem in the NHS locally. Policymakers were concerned that the NHS interpreting service was inefficient (e.g. it was not easy to incorporate the booking of hourly-paid interpreters in a range of languages into the busy and time-constrained routines of GP surgeries). There was wide variation in practice (some GP surgeries made very high use of the NHS interpreting service but an estimated 50% did not use it at all), and anecdotal stories abounded of limited English speaking patients receiving suboptimal (and possibly dangerous) care because of communication difficulties. Use of professional interpreting services was said to be particularly low in single-handed practices.

Our empirical work (described below) produced rich data and our initial analysis of this dataset produced a Habermasian analysis of communication within the interpreted clinical consultation (Greenhalgh et al. 2006) and a study of the role of the interpreter in the mediation of trust between clinicians and patients (Robb and Greenhalgh 2006) – papers which align with recent sociological work on the varying and conflicting roles of the interpreter in the clinical consultation (Angelelli 2005, Green et al. 2005, Jalbert 1998, Leanza 2005). In a further analysis reported here, we sought to address the wider organisational questions relevant to the NHS locally – that is:

  • a. Why are current NHS interpreting services [perceived as] inefficient?
  • b. Why is there such variation in the use of the interpreting service between general practices, with some not using the service at all?

These questions must be placed in the context of a rising proportion of consultations that take place across a language barrier. An estimated five per cent to 25 per cent of primary care consultations in inner London either involve an interpreter or are compromised for the lack of one (Haringey and Islington Primary Care Trusts, personal communication). This is especially true for refugees and asylum seekers, many of whom have complex, serious and stigmatising conditions such as HIV or mental health problems (Robb and Greenhalgh 2006). It is a statutory requirement for both primary and secondary care organisations to provide interpreters whenever needed (Department of Health 2004). Yet, with up to 340 languages spoken at home in London alone (Baker and Eversley 2000), and with the languages needed in any area changing rapidly as new populations migrate (Salt 2005), the mismatch between ideal and actual practice is wide and (in many places) getting worse.

The provision of a professional interpreting service for clinical consultations can be conceptualised as a complex service innovation, defined as ‘a novel set of behaviours, routines and ways of working, which are directed at improving health outcomes, administrative efficiency, cost-effectiveness, or the user experience, and which are implemented by means of planned and coordinated action’ (Greenhalgh et al. 2004). The assimilation of a complex innovation in an organisation is usually a stop-start process, progressing via a series of triggers and shocks and usually incurring set-backs and obstacles that must be overcome before the innovation can be said to be ‘business as usual’ (Van de Ven et al. 1999).

This framing of interpreted consultations as organisational innovation begs the question of what it means for an innovation to be ‘routinised’. A routine can be defined as ‘a repetitive, recognizable pattern of interdependent actions, involving multiple actors’ (Feldman and Pentland 2003). In a recent systematic review, Becker (2004) suggested that the routine may be the most fruitful unit of analysis when studying organisational change, and set out its defining characteristics (see Box 1). The term ‘routine’ can refer to both the idealised, abstracted understanding of the action to take in a particular circumstance (‘ostensive’ aspects) and also to specific performances in specific times and places (‘performative’ aspects).

Table Box 1: .  Organisational routines: key characteristics (Becker 2004)
• Routines are recurrent, collective, interactive behaviour patterns
• Routines are specific (they have a history, a local context and a particular set of relations) – hence, there is no such thing as universal best practice
• Routines co-ordinate (they work by enhancing interaction among participants)
• Routines have two main purposes – cognitive (knowledge of what to do) and governance (control)
• Routines, by allowing actors to make many decisions at a subconscious level, conserve cognitive power for non-routine activities
• Routines store and pass on knowledge (especially tacit knowledge)
• The knowledge for executing routines may be distributed (everyone has similar knowledge) or dispersed (everyone knows something different; overlaps are small)
• Routines reduce uncertainty, and hence reduce the complexity of individual decisions
• Routines confer stability while containing the seeds of change (through the individual's response to feedback from previous iterations)
• Routines change in a path-dependent manner (i.e. depending on what has gone before)
• Routines are triggered by actor related factors (e.g. aspiration levels) and by external cues

Two theoretical controversies are relevant to the study of how complex innovations are routinised in healthcare organisations. The first is between ‘routines as the preservation of past practice’ and ‘routines as embodying scope for change’. Much of the traditional literature on routines has emphasised the former, and linked routines with the maintenance of stability and even organisational inertia. But more recent work has emphasised that routines are sustained and evolve through the agency and choice of individual actors, especially in response to failure or in a turbulent or threatening external environment (Feldman and Pentland 2003, Howard-Grenville 2005). This is an example of Giddens's structuration theory, which, briefly stated, holds that structures – defined as enacted sets of rules and resources that inform ongoing action – shape human actions and identity, which in turn reaffirm or modify structures (Giddens 1986).

Routines can be thought of as an ‘organisational grammar’, offering a repertoire of choices that could be made in particular circumstances, but the final decision of what to do in any actual circumstance must be made judiciously by the actor (Pentland 1995). It is here, in the tension between ostensive and performative aspects of the routine, that the scope for incremental change lies (Feldman and Pentland 2003, Orlikowski 2000). The second controversy is the role of individual agency in the enactment of routines. This can be framed as ‘routines as unconscious choices’ (‘mindlessness’) versus ‘routines as effortful accomplishments’. In periods of stability, many organisational routines are followed at a subconscious level, but sustained change requires effort. And there is much more to human agency than whether an action is ‘mindless’ or ‘mindful’. As Feldman has put it, ‘Routines are performed by people who think and feel and care. Their reactions are situated in institutional, organisational and personal contexts. Their actions are motivated by will and intention. All of these forces influence the enactment of organisational routines and create in them a tremendous potential for change’ (Feldman 2000: 614). A person who faithfully enacts (or modifies, or resists) a routine in a particular situation is influenced by a host of factors including knowledge, confidence, experience (especially the outcome of previous enactments of the routine), values, expectations, access to resources, and orientation (e.g. to past habits, problem-in-hand, or future possibilities) (Howard-Grenville 2005).

If routines, by virtue of being structured and enacted by human agents, contain the scope for their own refinement and evolution, it follows that they are heavily susceptible to human resistance and failure. If an innovation is introduced into an organisation, but does not become routinised, this may have a number of explanations. First, the people on whom the routine depends may not know what needs to be done or may lack the competence to do it – a particular challenge with technology-dependent routines (Gavetti 2005, Orlikowski 2000, Tripsas and Gavetti 2000). Secondly, they may know what needs to be done but choose not to do it because it does not fit with their identity, values, or goals (Pratt, Rockmann and Kaufmann 2006, Rao, Monin and Durand 2003). Thirdly, they may fail to interact effectively with other actors to achieve the necessary collaboration for executing the routine – because they lack social skills or organisational power, and/or because of clashes of professional culture (Barley 1986). Fourthly, there may be a variety of organisational level problems (e.g. the routine is under-resourced or poorly co-ordinated; the technology is inadequate; the new routine conflicts with other more established or critical routines; key actors lack the necessary autonomy, and so on (Howard-Grenville 2005)). Fifthly, there may be institutional constraints or drivers such as NHS regulations and policies (Scott 2001). Finally, all the above are susceptible to wider environmental forces (demographic, economic, political, legal and so on) (Figure 1 summarises these multiple levels of influence).

Figure 1.

A multi-level model of routinisation for interpreted consultations

Taking account of the literature on routinisation, our research questions might be amended to read:

  • a. In general practices where the provision of professional interpreters has been successfully routinised, to what can this success be attributed (and what influences the choice of face-to-face, video and telephone interpreting)?
  • b. In practices where routinisation of professional interpreting services has not occurred (or been abandoned), what have been the main reasons for this?
  • c. What insights can be drawn that might inform organisational support towards the policy goal of supporting effective communication with limited English speakers?

Participants and methods

The overall aim of the research was to identify and explore ways to improve communication across a language barrier in primary healthcare. Details of NHS ethical approval and institutional consents are available from the authors.


The study was undertaken in a part of London where Black and Minority Ethnic (BME) groups comprised approximately 40 per cent of the population. Participants and recruitment methods have been described in detail elsewhere (Greenhalgh et al. 2006). In summary, 83 participants (18 service users, 17 professional and nine family member interpreters, 13 GPs, 15 primary care nurses, eight receptionists and three practice managers) were interviewed in 69 individual interviews and two focus groups. We used a theoretical sampling frame for GPs to obtain maximum variety in language spoken, age, gender, length of time in UK, deprivation score, and size of practice. GPs were identified from PCT lists and approached by telephone; other practice staff were recruited by ‘snowballing’ from an index GP. Professional interpreters were recruited via the local NHS interpreting service. We recruited five service users through the NHS interpreting service; these were interviewed by NR through an independent interpreter (interviews U1 through U5). We recruited four additional service users through the informal community contacts of a Bengali social scientist; he interviewed them in their own homes in their own language (U6 through U9). Nine service users, recruited through a Turkish language health advocacy organisation, were interviewed in a focus group by two Turkish advocates that we had briefly trained in focus group techniques, with NR present as observer. Family member interpreters were recruited through GP colleagues and via the social networks of the research team; four were interviewed individually and five (by their own preference) joined a focus group.

The interview

We asked service users, health service staff, and interpreters to ‘describe a consultation in primary care that involved an interpreter’. Thus we sought to collect our data in narrative form (i.e. as an account from the narrator's perspective of how events and actions unfolded over time (Greenhalgh et al. 2005)). We used standard narrative prompts aimed at obtaining a complete and coherent story, such as ‘tell me more about that’; ‘what happened next?’ and ‘can you think a way that this story might have ended differently?’ We sought examples of consultations which, in the perception of the interviewee, had gone well and also those that had not gone well. Interview guides used for different participants are available from the authors.

Processing and analysis of data

All interviews were audio-taped, transcribed and annotated with contemporaneous field notes. Interviews in a language other than English were translated by an independent translator. Two researchers read all transcripts and coded responses independently. We applied Muller's (1999) method of narrative analysis which took the story as a whole, rather than segments of text, as its main unit of analysis. This enabled us to highlight, in addition to the themes covered in each account, the wider organisational and environmental context in which a particular consultation took place, emplotment techniques (the narrator's use of metaphors, imagery and rhetorical devices to imply causality and agency), and the patterns or inconsistencies that emerged from multiple stories about comparable events (Greenhalgh et al. 2005). The interview schedule and analysis framework was modified substantially through progressive focussing as the fieldwork unfolded (Glaser and Strauss 1967). For this paper, we focused on the organisational dimension of interpreting services. We used a grounded theory approach to move from descriptive categories to preliminary theoretical categories and thence to higher-order theoretical constructs (Table 1).

Table 1. Overview of data structure
First-order (descriptive) codes Theoretical categoryHigher-order theories
Accounts of what it's like to work in a practice
Individuals’ descriptions of what they do and why
(a) Organisational type (culture, ethos and values of general practice, identity of staff)(1)
Multi-level model of routinisation for interpreted consultations
- Individual
- Interpersonal
- Organisational
- Institutional
- Environmental
Stories of how an interpreter is booked (or of frustrations when trying to book one)
Stories (usually indirect) of individuals and practices that did not use the professional interpreting service
(b) Nature of the routine for ensuring the presence of professional interpreters in consultations
Stories of the unpredictable nature of illness and healthcare need
Stories of ‘inappropriate’ patient expectations and behaviour
Stories of administrative tensions between the practice and interpreting service
Stories of inefficiency and inadequate capacity in the interpreting service
(c) Difficulties and constraints in routinising professional interpreting services within general practice
Stories of multilingual practice staff using their skills as needed
Stories of family member interpreters being encouraged to attend with patients and welcomed by practice staff
Stories of ‘ad hoc’ solutions to interpreting need (e.g. soliciting bilingual patients from waiting room)
Stories of formal alternatives for meeting interpreting need (e.g. video interpreting)
(d) Alternative routines for meeting interpreting need outside the standard interpreting service(2)
Hypotheses about why single-handed GPs are low users of the NHS interpreting service
Stories of efforts by practice staff to make the system for booking and using professional interpreters more efficient
Stories of efforts by practice staff to make family member interpreting more efficient
Stories of practice staff ‘dragging their feet’ over interpreter booking
(e) Role of individuals in refining, shaping and resisting routines

Main findings

Approximately one in three people approached in all subgroups consented to be interviewed, reflecting the sensitivity of the research topic. GPs who did not use the interpreting service were less likely to consent to interview; these were mainly but not exclusively from single-handed practices. Despite this response bias, our sample of 19 practices included eight single-handed or husband-and-wife teams. Full details of practice characteristics are available from the authors, and we discuss the limitations of the sample in the Discussion. It should be noted that only one GP whose practice made no use of the professional interpreting service consented to an interview and allowed his staff to be interviewed. Hence we could not fully document the GP or practice perspective on whole-scale rejection of this service (which occurred in an estimated 45 per cent of all practices in the area studied).

The data structure is shown in Table 1. Our first reading of the data produced a number of descriptive codes which we subsequently linked to five theoretical categories: (a) an organisational taxonomy of general practices; (b) the nature of routines for organising professional interpreters in consultations; (c) difficulties and constraints in establishing these routines; (d) the presence of alternative routines (such as multilingual staff or family member interpreting); and (e) the role of individuals (practice staff, interpreters, and service users) in refining, shaping and resisting routines. From these categories, and incorporating insights from the literature, we developed two higher-order interpretations: (1) a multi-level model to explain the routinisation (or non-routinisation) of the use of professional interpreting services in general practices; and (2) a hypothesis to explain why some traditional practices make very limited use of this service. Below, we present the five theoretical categories and then discuss our more abstracted interpretation of the data.

Organisation of general practices: ‘traditional’ and ‘contemporary’

An early theme that emerged from our data was that some practices (single-handed or husband-and-wife and operating from an owner-occupied converted house) were run as a traditional ‘family business’, while others (typically multi-partner and based in a purpose-built health centre) were run in a more contemporary fashion with a wide range of employed and attached staff. Whilst few practices could be classified categorically as one or other of these, all could be matched to a greater or lesser extent with one of the organisational ‘ideal types’ shown in Table 2. This taxonomy accords closely with one developed independently by others (Green 1996). To map our participating practices against these ideal types, we supplemented our qualitative analysis by publicly available data on type and ownership of premises, year established, list size, number of full-time and part-time partners, diversity of clinics offered, and languages spoken by doctors.

Table 2. ‘Traditional’ and ‘contemporary’ general practices
 The traditional general practiceThe contemporary general practice
PracticeSingle handed or family unit; list size 2,000–4,000Multi-partner; list size 4,000–9,000
Offers basic (‘GMS’) services provided by a few core staffOffers multiple extra clinics and add-on services provided by a wide range of staff and contractors
PremisesConverted housePurpose built health centre
GP usually owner occupierOwned collaboratively or leased from private funder
HistoryOften established in 1960s or early 1970s following expansion of NHS services with 1964 GP ContractUsually established in late 1970s or 80s, perhaps through merging of several single handed practices
GP's identityFamily doctor =‘respected pillar of the community’, committed to job for lifeMember of staff in an efficient and caring organisation, committed as per contract
GP's link to practiceOften lives on site or locally and takes active part in local events e.g. campaigns, festivalsOften commutes from another area and has no link to the community other than via the list
Practice structure and ethosFamily business with roles and responsibilities defined loosely and informally. Appointment and status influenced by kinship e.g. senior partner is usually the oldest. Admin staff may include doctor's own relatives. Few formal systems; may not have differentiated management roles (GP or spouse may undertake these). Everyone is expected to ‘muck in’. Often on call 24 hours for own patients. Bureaucratic organisation in which roles and responsibilities are formal, differentiated and explicit and multiple systems and procedures are in place. GP's status influenced most by qualifications and other external measures of merit. There is often equity of status and parity of pay between all GPs, many of whom work part time and have outside interests. Out of hours care usually contracted out or shared in a co-operative.
Values and virtuesLoyalty, thrift, continuity, interpersonal relationships, friendliness, integrity, self sufficiency.The ‘Maxwell six’: effectiveness, acceptability, efficiency, accessibility, equity and relevance.
External linksFew and based around GP's social networks.Many and based around collaborative work practices.
StrategyDevelop and maintain good GP- patient/family relationshipsProvide comprehensive, up-to-date, evidence-based care for all (including proactive, preventive care)
Provide a good basic service to the sickPay special attention to poor, disempowered, and socially excluded patients, whose health problems are often linked to social disadvantage
Support and care for the family who have chosen you as their GPDraw on as many additional services as needed to extend the care package offered to patients
Don't waste money 
Environmental stressBroad, undifferentiated clinical agenda dealt with by generalist GPEconomies of scale compared to single handed practice
• Low
inline image
GP may pursue a specialism out of personal interestProfessional support and stimulationWeakly developed and voluntary division of labour based on interests of GPs
• HighRestriction of remit e.g.Efficiency measures e.g.
• Limit services to ‘core business’ (GMS)• Limit agenda to biomedical
• Limit agenda to biomedical• Strongly developed and enforced division of labour, incorporating ‘hierarchy of appropriateness’ (see text)
• Refer complex cases to secondary careƒ• Categorisation and triage of patients according to tasks needed to ‘process’ them
• Discourage resource-intensive patients (complex, demanding, needy, or geographically mobile) from registering• Extension of opening hours (e.g. evening assistants) to make maximum use of space
• Creative use of technology

Staff from traditional practices generally described their organisation as friendly, caring, and well liked by the patients; they emphasised a personal atmosphere and continuity of care. In these practices, the personality and commitment of the doctor – who typically worked full time and put in long hours – was often highlighted by staff, and organisational values were personalised in terms of what Dr X believes or stands for:

We definitely have a nice atmosphere, there's a nice rapport with our patients (Practice manager, PM2, traditional practice).

Staff from contemporary practices talked more about efficiency, accessibility, and equity. They rarely attributed views to particular doctors but emphasised ‘corporate’ values and principles such as evidence-based care and a commitment to redressing inequalities and supporting the vulnerable:

OK, X---- [local estate] as you know is a community and the community encompasses what I will call an ethnic mix. We have people from Somalia, Romania, Turkey, Kosovo, Kurdish and African and Afro-Caribbean descent OK. Because of the sheets of people that we have interpreting service is very, very important to this practice and from inception of this Health Centre (Receptionist, R5, contemporary practice).

GPs from more traditionally oriented practices typically described themselves as committed to the practice and working hard for the patient population. They accepted (and usually expected) senior status in the practice. Many were disinterested in activities that went beyond what they saw as the core business of traditional general practice –i.e. treating the sick as and when they presented:

. . . the point is I don't have an appointment system, I don't understand how anybody can be sick by appointment. So that's why we keep the door open here, so what am I going to do with ring and wait for two weeks for an interpreter? Usually they can make themselves understood as they bring somebody with them (Male GP, G7, traditional practice).

Most of these single-handed GPs were first generation Asian immigrants; a high proportion of their established patient list were people of the same minority ethnic origin. They had been serving their own community since before the official interpreting service had been set up. Meeting the needs of minority ethnic patients was thus integral to their identity rather than a special interest that they had to develop and declare.

GPs in more contemporary practices, on the other hand, typically presented themselves as part of a multi-disciplinary team without a formal figurehead or senior partner. They took pride in being up-to-date and evidence-based, and most had a side interest beyond General Medical Services (such as teaching, research or a clinical assistant post at a local hospital). These GPs, though less likely to be from an ethnic minority, were more likely to declare an interest in ethnic health or reducing inequalities:

I mean I enjoy the cultural differences, that's why I'm a GP in London (Female GP, G9, contemporary practice).

Receptionists and managers in traditional practices embodied the ‘family’ culture and presented an identity that was approachable, empathetic, and informal (they talked, for example, of spotting patients who looked upset and taking them aside for a cup of tea). Those in contemporary practices generally presented themselves in a more impersonal manner, with an emphasis on professionalism, efficiency, and risk management (especially maintaining order and safety in the waiting room).

Because of the marked differences in the approach to interpreted consultations taken by staff in practices that were either strongly traditional or strongly contemporary, we tried to classify all stories told by service users and interpreters according to the type of practice the story related to, though this was not possible if the participant did not remember key details.

Routines for organising professional interpreters

All but one practice (a traditional single-handed GP, discussed below) had established some sort of routine for providing professional interpreters for clinical consultations. This routine comprised four key steps, the first of which was to ascertain which language the patient spoke and ensure that this information was recorded (preferably by inserting a pop-up prompt on the patient's electronic record). This was usually the receptionist's role:

We have a book downstairs with different languages and different phrases, so we ask them, ‘What language do you speak?’[. . .] we show them that and they pick out their language, so there's basic questions on that and we go from there (Receptionist R5, contemporary practice).

Secondly, there was the booking of an interpreter to link in with a future consultation, either as a follow-up from a previous consultation (typically mediated by the interpreter in attendance for that consultation) or through a new contact by the patient (in person, by phone, or via a relative). Responsibility for the booking step varied considerably between practices. Sometimes it was done by receptionists using a faxed booking form; sometimes the clinician could book by phone:

Where I was trained before I remember we could use the telephone to contact the people if you wanted somebody to translate and we had all the numbers down there with the languages and everything next to the phone (Community nurse C3, speaking of contemporary practice).

This booking step within the practice triggered a further routine in the interpreting service – getting hold of an interpreter (typically by mobile phone), ensuring his or her availability, and confirming back with the practice. The third step in the general practice's routine, on the day of the appointment, was linking the interpreter to the correct clinician and patient – something that required key administrative data to be exchanged:

We enter the GP surgery, then we introduce ourself to the receptionist. And we give the name of the patient. And sometimes they ask which doctor. And sometimes we don't have it. The PCT sends it to us, we don't have the name of the doctor. This sometimes makes problem because some of the receptionists they are not familiar with the system. [ . . . ] And they keep asking me ‘What is the date of birth?’‘I'm sorry I don't know the date of birth’ (Female Farsi speaking interpreter, I12).

Finally, after the consultation, a staff member (clinician or administrator) completed the paperwork so that the interpreter could get paid.

For patients presenting as ‘emergencies’, receptionists used judgement to decide whether to squeeze them in on a booked clinic list or send them away, and clinicians either muddled through with no interpreter or opted to use an expensive commercial telephone interpreting service:

I mean some people . . . you sort of judge the situation. If someone's got a bit of English you know you do maybe try and work round it as much as possible. But if it is a struggle then you know you reach for Language Line. We do normally have a card up. I don't know where this one's gone today. And it's got the number on, it's got our code, so it's actually very easy. I mean you just phone the number, they usually answer immediately and then you just say you know ‘I need an interpreter in so and so language’ and give our code (Female GP, G10, contemporary practice).

There was considerable variation between practices in how well-developed this whole routine was. In general (but by no means universally), contemporary practices had stronger routines, and where these existed, staff described them with confidence and evident satisfaction. Where routines were weakly developed, both interpreters and service users expressed frustration and confusion:

The other thing that they can actually improve, when they say ‘interpreter booked’, on the computer, when they actually put it in front of the patient, that means that that should mean something, that I shouldn't actually go and remind them a hundred times ‘excuse me, isn't there – did you talk to the doctors – excuse me’ (Female Farsi speaking interpreter, I2).

Difficulties and constraints in routinising professional interpreting services

Staff in both traditional and contemporary practices described a number of constraints to the successful routinisation of professional interpreting services. The first of these was the high (and rising) demand for the service. Our field work was undertaken in London at a time (2003–2005) of rapid and profound demographic changes, including an influx of some 300,000 migrants from former Eastern European countries in the six months following their accession to the European Union, plus large asylum seeker populations arriving from Iraq, Afghanistan, and central Africa (Salt 2005). Many newly registered patients had complex and challenging problems (e.g. one interpreter estimated that over 90% of her female clients from a war-torn African country had been raped). Particularly high-maintenance for practices were the numerous families who had been placed in temporary housing and who were likely to be moved on shortly after the labour-intensive processing stage (registration, new patient medicals, catch-up immunisations and so on). Linked to these demographic changes was a growing diversity of language need (one GP documented 27 different languages spoken by attenders at his practice in the space of a fortnight), and a mismatch between the languages provided by the interpreting service and those needed by the most recent immigrants:

Recently, in the last couple of years, they have a lot of people [immigrated] from China mainland itself, but those are mainly Mandarin, and I don't do Mandarin (Female Cantonese speaking interpreter, I4).

The second constraint was the nature of clinical need, much of which was seen as being inherently unpredictable:

If we get a new birth, baby's day 10 and the new birth [home visit] needs to be done in the next two or three days, it can be difficult to get interpreters (Practice nurse, P11, contemporary practice).

Another perceived constraint was patient expectations and behaviour. Many limited English speakers failed to attend follow-up appointments booked weeks in advance, perhaps because they did not have a cultural tradition of going to the doctor when they had no symptoms:

Interviewer: Why didn't you ask in advance [for an appointment]?

Participant: Didn't need to. When I needed the doctor, I would go then, say it then (Male Sylheti speaking service user, U6, registered with traditional practice).

Conversely, such patients often attended as ‘emergencies’ for non-urgent problems. For this reason, some practices had given up booking appointments in advance altogether and operated a ‘turn up and wait’ system for seeing the doctor. Another frequently expressed difficulty was the expectation of many patients for a referral, perhaps based on past experience of primary healthcare as a ‘transit zone’ to definitive healthcare in their country of origin (Greenhalgh, Robb, and Scambler 2006), and (implicitly) a lack of awareness that the NHS has limited resources that must be distributed according to clinical need:

. . . and they [Somalian patients] always want a referral. Referral, referral. And the GPs, they have to try their best to be fairest (Female Somali speaking interpreter, I15).

Staff in many general practices perceived serious inefficiencies and inadequate capacity in the interpreting service. They described non-availability (‘they don't do Somalian’, ‘they close at 4’), failure to confirm provisional bookings, late cancellations, and interpreters turning up speaking a different language from the one that had been booked. Likewise, interpreters and interpreting service managers often highlighted organisational inefficiencies in the general practices, most often attributed to lack of time and the limited capability and autonomy of administrative staff:

Only problem I've found with all the GP surgeries, they don't – they don't want to take the – they just say: ‘We don't have time to book interpreters’, there is only a phone call. But the receptionists are really actually tied up, and they must be assisted some way, I mean or given some freedom where the interpreters have the leeway if they've got your own patients, if they’re struggling with bookings (Female Gujarati speaking interpreter, I6).

Participants were frustrated by the lack of alignment between the interpreter booking system and the practice appointments system. As described above, the former requires the receptionist to send a request and wait for confirmation of date and time, by which time the patient has gone home and may be difficult to communicate with by phone:

. . . and the fax comes back one day before the appointment that we cannot provide you an interpreter (Male GP, G12, traditional practice).

Interpreters sometimes embodied the confusion when the relevant information had not been exchanged between interpreting service and general practice (see quote from interpreter I12 earlier). A particular problem was that GP surgeries are typically only booked about three weeks ahead; for a later follow-up, patients are asked to phone back in a few weeks’ time – something that raises major practical difficulties for limited English speakers:

Participant: We were going to do a follow-up appointment and she [receptionist] said, ‘No, you can't, you have to phone [later]’, I said, ‘We can't phone because the patient doesn't know how to talk, to make an appointment talk in English and make an appointment’, I said, ‘No, no, no’, and I said, ‘Please, you know, For my sake, I am here, we’ll try to do it’ and . . and she said ‘No, we’re not, you have to phone’. So I had to give the. . . .

Interviewer: What, they wouldn't allow you to make an appointment . . . ?

Participant: A follow-up appointment. . . .

Interviewer: Then and there?

Participant: That's right, because I think it was something – I think, actually, the doctor was away on holiday or something he was going, and they couldn't make an appointment in a month's time (Female Albanian speaking interpreter, I3).

One of the most commonly cited problems was the inflexibility of the interpreting service, which (at the time of this study) restricted its services to office hours:

I had a case where I was working with the learning disabilities team and we used to visit a mother who had two children who – two children completely disabled, even physically sort of, and mentally – so they had feeding problem, the lot, and the OTs [occupational therapists], the physio[therapist] and the social worker, they wanted to assess because mother is very old now and she's not able to – they wanted to help mum to provide the services. And the feeding time for these two, well mum used to feed – they’re about 40-year-old gentleman now – was 6.30 in the evening, so they wanted to observe how mum feeds them – and they [interpreting service line manager] refused. And I've, I had been an interpreter for a long time with them and they said: ‘No, we cannot provide for that time because it's out of our office hours’ (Female Gujarati speaking interpreter, I6).

The role of individuals in overcoming (and in some cases exacerbating and perpetuating) these problems is discussed below.

Alternative routines: ad hoc interpreting or other service models

A striking feature of many traditional practices was the number of languages spoken by clinicians and other practice staff. It was not uncommon for an Asian GP to advertise three minority languages spoken fluently. Such practices often also employed bilingual staff and used these for interpreting on an ad hoc basis:

Participant: Right. Well we are a very multiethnic practice because our doctors are Indian, then we have a Somali doctor who works part time. Then we've got [other staff] Irish, Nigerian, British, one Indian too. So you know, then we've got Bangladeshi as well. So you know we've got all sorts of different things. So that way I suppose many people from small groups come to us, because if they don't have an interpreter then one of us translates for them.

Interviewer: Oh right.

Participant:  Okay? Because the doctors also speak some . . . I think it's um . . . oh crumbs . . . some Nigerian language or something like that, cos they used to work in Kenya.

Interviewer: Oh right.

Participant: Yeah I don't know what it's called.

Interviewer: But an African language.

Participant: Yeah they speak some African language too, and they speak about four or five Indian languages. And then we have the little girl who's a Bangladeshi, she can translate for the Bangladeshi patients. And then . . .

Interviewer: Who is she?

Participant: She's our receptionist (Practice Manager, PM3, traditional (husband and wife) practice).

It is apparently seen as the norm for staff in this practice not only to be bilingual or multilingual but to offer this skill as needed for ad hoc interpreting. Note that the official NHS interpreting service is not mentioned and that the practice appears to have responded to an influx of a new minority ethnic group by employing a bilingual doctor on a sessional basis.

Some practices, particularly (but not exclusively) traditional ones, routinely asked patients to bring a family member or friend to interpret. We also heard stories of practices in which it was normal for selected patients to be brought in as unpaid interpreters and even for bilingual patients to be solicited ad hoc from the waiting room:

If they’re Asian a lot of the time Mrs X--- [doctor's wife] is here, so she will help us. With others, we have got one lady who's a patient, Turkish lady, and she very kindly says like ‘Phone me up if you get stuck’, if someone just walks in. [ . . . ] If there's someone in the waiting room that speaks that language then yeah they will help out (Receptionist, R2, traditional practice).

Whereas staff in contemporary practices tended to perceive family member interpreters as ‘second best’, many staff in traditional practices preferred family members to professional interpreters and found they fitted in better with existing routines and ways of working. A common statement from staff was ‘Dr X finds the official interpreting service frustrating; he prefers the patient to bring a relative’. Clinicians in such practices consistently justified this approach on the grounds that it ‘works better’. They often believed that they were offering their patients a better service because multilingualism was seen as better than a triadic conversation, and ad hoc interpreting by on-site staff was seen as more reliable than hiring external interpreters and better aligned with practice opening hours.

In a few cases, the standard model for providing professional interpreters had been tried and deemed to have failed, so had been abandoned in favour of a different model. The following GP, an enthusiast for interpreting services who had put a lot of effort into securing a regular Turkish-speaking interpreter for a particular afternoon every week, describes how the system had unexpectedly generated more problems than it had solved:

But the sessions themselves were incredibly unpleasant and stressful because for me, and I daresay for the patients, and probably for the interpreter – because the interpreter would arrive at say two o’clock. Every Turkish patient who'd been turned away or asked to book for the previous week or two were sitting in the waiting area expecting to be seen during this special time. They invariably have of course accumulated absolutely every problem under the sun because time passes and people bring everything when there's a limited amount of opportunity. And at four o’clock the interpreter leaves – leaves through a waiting room full of angry Turkish-speaking patients, understandably angry, because the service is over (Male GP, G3, contemporary practice).

A receptionist from this practice gave a different version of the same narrative, describing a violent physical assault from one frustrated Turkish patient. The practice subsequently abandoned the booked interpreter surgery in favour of remote video-interpreting – an expensive option that was justified mainly on safety grounds.

In another example of a radical solution to the conventional service, a single-handed Asian (Tamil) GP had become frustrated that his Tamil patients persistently brought him social problems (such as housing or immigration issues) and expected him to sort these out within the biomedical consultation. After unsuccessful attempts to divert such patients to social workers and the Citizens Advice Bureau, he developed an idea to train one of his own Asian patients in a ‘customised’ link worker role (part healthcare assistant, part health promotion officer and part community worker). He expressed frustration that the PCT would not pick up the funding of this post.

The role of individuals in refining, shaping and resisting routines

We heard numerous examples of practice staff using their initiative in an attempt to develop and refine routines – and also examples of the potential wrecking power of staff who sought to resist a particular routine. Receptionists had developed ‘tricks of the trade’ for identifying what language a patient spoke, getting the interpreting service booking clerks to pick the phone up rather than take an answering machine message, and using interpreters in the most time-efficient way:

The girls are very good, they try and group book. So what happens is the interpreter might be running between all the rooms for about three hours. Even with the other surgery [which shares the same health centre] to try and make their time most useful (Female GP, G9, contemporary practice).

Interpreters, too, described numerous ‘workarounds’ that they had developed to overcome glitches in the system. For example, some (though not all) routinely gave their private phone number to patients and provided an unpaid booking service for GP appointments. This was a highly effective solution to the problem of limited English speakers being unable to ‘call back later’– but it explicitly contravened their professional code of practice and in some cases contributed to high levels of perceived job stress.

Patients, to some extent, also had some agency in shaping routines. In the following excerpt, a patient, despite being conventionally ‘disempowered’ and speaking no English, successfully cadges a lift to hospital in the interpreter's car:

I said I thought it would probably be best . . . that she had an x-ray. And then he [husband] just told the interpreter, ‘You, you, take me to the x-ray, you drive me there’, and was really quite aggressive and rather a little bit I thought – but I had a chat with the interpreter and said, ‘What do you feel about that?’, because I said, ‘In actual fact it's going to be very useful when this lady's assessed in the x-ray department, there is actually somebody can tell her to get in the right position, so I can see it may not necessarily be your job or role to provide a transport service, but I can see that it would be useful to have a continuation of interpreter to help when she gets to the hospital’. And the other thing is that she'd come regularly, so they'd come with this same interpreter on a number of times. I've looked after this family for probably about 18 months now and they've mainly come almost exclusively, 90 per cent of the consultations, with this particular interpreter (Male GP, G2, contemporary practice).

This patient's success should be measured not merely in that he got his lift to hospital, but in that his GP subsequently mused on how efficient this action had been in terms of achieving continuity of care for a frail patient across the primary-secondary care interface. It is not inconceivable that the use of interpreter as taxi and chaperone as well as translator could now become part of a more mainstream routine when patients with complex needs are referred for hospital investigations by this GP.

Interpreters told numerous stories of resistance by practice staff to implementing the interpreter booking routine:

Sometimes I've come across when I have seen the patient and they've had follow up and I say ‘My name is so and so’. Because sometimes patients say ‘Can we have you?’ because it's easier for them, the patient. It's easier for me and the doctor as well. ‘Can we have you for a follow up?’ And I write it down in my diary and everything and you know they [general practice] can't be bothered. You phone them up it's like, ‘Oh. I thought you were just going to come along’ It's like ‘No, you know the procedure, you have done it before’. ‘I am not allowed to come to appointments, although I've kept myself free . . .’‘You have to give me . . . and then I have to get a code and reference number. Without them I can't’ (Female Bengali speaking interpreter, I16).

The interpreter in the next quote described attending with a patient who had been ‘allocated’ to a GP surgery (i.e. compulsorily assigned by the PCT because no GP in the area would take the patient on). The receptionist appears to believe (incorrectly) that the practice would be charged for the interpreter and is therefore not prepared to book her for any future appointment:

So it started when I came, they were allocated GPs and we just came in, the receptionist almost, you know: ‘No, no, no, no’, and she's so . . . [ . . . ] But, you know, the practice nurse and the health visitor, they both asked: ‘Can we book you?’ So – and when we went back to the receptionists, ‘I told you already that: No, no, no, no, we haven't got funds’, and something else (Female Romanian speaking interpreter, I7, speaking of a traditional practice).

As well as the motivation of staff (and the evident wrecking power of even very junior staff in this particular routine), interpersonal relationships between different staff groups also appeared to be critical to the successful routinisation of interpreting services. Both receptionists and interpreters often spoke positively of one another, especially when their relationships were longstanding, and of creative efforts to improve services for limited English speakers that appeared to have arisen from this positive relationship:

X--- [interpreter] in particular is extremely helpful. She was aware of – in the toilet, we had notices up, you know, not to put nappies down the toilet, and this sort of thing. And she offered to do them in a variety of languages for us, which was a great help. And she done ’em very nicely, much better than our scrappy notices. And we found them very helpful. Always. Not a problem (Receptionist, R6, contemporary practice).

But examples of interpersonal tensions (especially lack of respect for one another's role and implicit status wars) were also common:

Participant: Some of them [interpreters] can be a bit rude. Mainly I've found a couple of gentlemen, Turkish gentlemen are quite rude.

Interviewer: Uhuh.

Participant: I've actually requested not to have somebody actually in the surgery no more.

Interviewer: To you or to the patient, or both?

Participant: To us. To us, just the way they speak to us (Receptionist, R4, traditional practice).


I also notice generally speaking that the lower the grade of the professional, the health professional, the more suspicious they are [of hours claimed]. If you are dealing with a professor he’ll say ‘You put the time whenever you want’, because he appreciates you (Male Arabic speaking interpreter, I17).


A multi-level model of routinisation for interpreted consultations

Figure 1 shows the model of routinisation derived from both the wider literature (Becker 2004, Edmondson et al. 2001, Feldman and Pentland 2003, Howard-Grenville 2005, Orlikowski 2000) and our empirical data. The five levels – individual, interpersonal, organisational, institutional, and environmental – interact dynamically and reciprocally with one another such that, for example, individual identity shapes interpersonal interaction, the outcome of which feeds back to influence identity.

At the individual level, the perceptions, values and goals of the actors involved in the routine are extremely important, and these are strongly linked to both organisational and individual identity. The most fundamental aspect of identity appears to be the extent to which the individual (whether GP, nurse, receptionist or manager) is him- or herself multilingual and using more than one language in daily living. Where this is the case, a patient's need for interpretation is likely to prompt a solution from the ‘lifeworld’ of the individual making the decision (i.e. the patient would be offered a multilingual staff member or ad hoc interpreter, or invited to bring a family member) (Scambler and Britten 2001). But where the individual is monolingual, this need is more likely to prompt a solution from the ‘system’– that is, a paid professional interpreter. The second dimension of individual identity is how that individual constructs and enacts their organisational role (the GP as ‘family doctor’ or ‘member of efficient multi-disciplinary team’, or the receptionist as providing ‘tea and sympathy’ or a ‘well-managed service’). As Table 2 suggests, these two dimensions of identity are often aligned, the one reinforcing the other.

The GP (perhaps unsurprisingly in this hierarchical organisation) has a powerful influence over the routinisation of interpreting. Especially in single-handed practices, decisions about how and to what extent a routine develops are individualised to the GP (‘Dr X does/does not use professional interpreters’; end of story). As has been observed previously, ‘different actors are more or less able to use a routine flexibly, and more or less able to influence whether a new kind of performance will be taken up as part of an ongoing routine’ (Howard-Grenville 2005: 619). But our data also suggest some degree of agency for reception staff in enacting the interpreting routine. This may reflect a differential status gradient amongst staff in different physical areas within the organisation (reception staff holding low status in the clinical areas but relatively high status and autonomy in the waiting room). As one receptionist put it, ‘we are the front line. To get to the doctor they've got to get past us’.

An interesting aspect of this particular routine is that the interpreter, whilst professionally qualified and trusted with confidential information, is remarkably powerless to influence the routines in most practices, as the earlier quote from interpreter I6 shows. Interpreters are independent contractors sent by an external organisation, whose official status in the practice waiting room is no higher than that of the patient. Some (though by no means all) receptionists seemed to assign them to the category of ‘intruder’.

At the interpersonal level, our data show that a successful routine depends on a level of collaborative interaction between staff that is sometimes but not always achieved. Friendship and reciprocity, built through repeated encounters over time, can oil the wheels of routines that cross professional and organisational boundaries. Conversely, lack of respect and disputes over one another's role and status will block both the implementation of the routine and its successful evolution over time.

At the organisational level, our data strongly support the findings of previous researchers that routines are almost never performed in a vacuum, but overlap with other routines (Becker 2004). Booking the interpreter must link with booking an appointment for the patient to see the GP. The interpreting service must process requests from practices, contact the interpreter to confirm availability, and send confirmation to the practice. With a typical PCT offering around 70 different languages and dialects through peripatetic interpreters who are contacted via a mobile phone (which must be switched off in the GP surgery), the sheer administrative complexity of this task should not be underestimated. Overlapping routines are especially hard to align if (as in this case) they occur at different speeds or frequencies (Becker 2004) or cross organisational boundaries (Gittell et al. forthcoming).

At an institutional level, the NHS rests on three pillars: (a) regulative pillars (formal rules, regulations and policies (Department of Health 2003, 2004), implemented via Executive Letters, Service Level Agreements and so on, and assured through clinical governance mechanisms); (b) normative pillars (professional standards and expectations, implemented through training and assured through certification, accreditation and appraisal/revalidation); and (c) cultural-cognitive pillars (constitutive schemata based on cultural understanding of ‘what goes on around here’, implemented through cognitive schemata and assured by common beliefs and shared logics of action) (Scott 2001). In terms of each of these pillars, routines can be thought of respectively as rules and protocols, roles and standards, and ‘knowing the ropes’. Our data suggest that the introduction of new policies in primary care – such as the requirement to provide interpreting services – may challenge longstanding normative values and belief systems about what a good health service is and how best to deliver it. This is especially true in primary care – partly because GP surgeries, like the corner shop, are by their nature isolated, organisationally idiosyncratic, and difficult to regulate; and also because the GP's role in the NHS has always been ambiguous, resting as it has (until recently at least) on a fiercely defended independent contractor status (Lewis 1997).

In contrast to studies of routines in many commercial companies (where the main impetus for changing routines is the agency of forward-looking individuals within the organisation (Edmondson et al. 2001, Feldman and Pentland 2003, Howard-Grenville 2005)), we found that the most powerful influence on interpreter booking routines was changes in the external environment. The profound demographic changes occurring in the few months prior to our empirical work served as a classic ‘organisational shock’ to general practices (Van de Ven, Polley, Garud et al. 1999). New immigrants often brought multiple physical, mental and emotional problems; they were not familiar with the values of the NHS or expected patient behaviour, and a small minority were aggressive or violent (something that would be particularly unacceptable in a practice with a ‘family’ ethos). In practices where GPs had served a relatively stable and compliant population for 20 years or more, the influx of recent immigrants was perceived as draining time and energy away from its ‘established customers’. The response to this shock is discussed in the next section.

Explaining the ‘we don't do interpreters’ response

There are several possible explanations for the weak or absent routines for booking professional interpreters which we documented in some practices. One is a simple size effect. Damanpour (1992) has shown in a meta-analysis that large organisations are significantly more innovative than smaller ones, probably because of a more sophisticated division of labour and greater slack resources, and research in UK general practice confirms that there is less formal organisation and less internal specialisation in single-handed practices (Campbell, Ramsay, and Green 2001, van den Hombergh, Engels, and Grol 2006). Without wishing to dismiss this explanation, and acknowledging that we were unable to access the most resistant practices in this study, we believe our data are also consistent with an explanation based on the ideal types of general practices shown in Table 2.

Faced with the ‘organisational shock’ caused by a turbulent external environment (see previous section), a policy of offering an open door and a welcoming interpreter to limited English speakers seeking to register was seen by some as potentially ‘opening the floodgates’ to large numbers of high-maintenance patients. The bottom row of Table 2 shows two possible responses as a practice moves from a period of low to high environmental stress. The traditional single-handed GP generally responds by restricting his or her remit – by discontinuing all services except core General Medical Services, increasing referrals to secondary care, and – crucially for this study – discouraging resource-intensive patients from registering in the first place.

Conversely, the contemporary general practice responds to the same shock by introducing efficiency measures such as extending and enforcing the inter-professional division of labour, most notably by employing nurses and healthcare assistants to take on some of the GPs’ work; triaging patients to the cheapest professional who can complete the tasks needed to ‘process’ them; increasing throughput (e.g. by extending opening hours); and making creative use of technology (such as the shift from face-to-face to video interpreting previously described by respondent G3). Table 2 suggests that in times of organisational stress, the resource-intensive patient (including the limited English speaker) is more vulnerable in the traditional practice than in the contemporary one, since the former might refuse even to register the patient, whereas the latter is likely to accept them but will ‘manage’ their complaints more efficiently.


To our knowledge, this study is the first to analyse interpreted consultations as a complex innovation using the organisational routine as the focal unit. As an interdisciplinary team, our different perspectives (general practice and health services research, organisational sociology, and medical sociology) led each of us to focus on different areas and notice different patterns in the data, leading to a richer understanding of the phenomena under investigation than any of us would have produced alone. The study explicitly redresses a recognised imbalance in the sociology of health and illness – the absence of a rigorous approach to the organisational dimension (Davies 2003). We have combined three background literatures – on the diffusion of innovation in healthcare (Greenhalgh, Robert, Macfarlane, Bate, and Kyriakidou 2004), organisational routines (Becker 2004, Feldman and Pentland 2003), and the changing nature of work and professional identity in general practice (Charles-Jones et al. 2003, Jones and Green 2006), along with our empirical findings, to produce an integrative model for the organisational aspects of interpreted consultations that takes the routine as its focal unit (Figure 1).

Previous work on organisational routines has been developed almost exclusively in the USA, based on empirical work in the commercial sector or large private hospitals, where the routine being studied has been either commercially crucial (e.g. microchip development (Howard-Grenville 2005)) or a complex medical investigation or procedure (Barley and Tolbert 1997, Edmondson et al. 2001, Kellogg 2005), and where the agents involved were senior professionals with a high degree of expertise and autonomy. Our own focus was much more mundane – the implementation by reception staff of a routine in general practice that is not directly (though it may be indirectly) life saving for patients. Our study thus makes a unique contribution to the literature by considering medical routines that depend heavily on people who are not highly trained and who lack autonomy and organisational power.

Two concepts are important to understanding the organisational dimension of routines: embeddedness and resourcing. A routine can be central to the workings of the organisation (e.g. booking appointments for patients to see the GPs) or it can be marginal (e.g. ordering the coffee supply). To become central, a routine must be embedded in three types of organisational structure – technological (e.g. whether the pop-up computer alert ‘interpreter needed’ is in use), cultural (e.g. social hierarchies, norms and values, especially whether the routine aligns with staff understanding of the practice's mission and their expectations for one another's behaviour – as in the first quote from respondent R5), and co-ordination/control (i.e. structures that achieve the interdependence of multiple actors and multiple routines – for example, the extent to which a receptionist responsible for booking a GP appointment also has the knowledge and autonomy to book the interpreter) (Howard-Grenville 2005). Even when practices are positive about the use of professional interpreters, and especially when they are lukewarm or negative, the booking routine may be only weakly embedded in technological and co-ordination/control structures. As one interpreter put it, ‘They are very unorganised – very, very unorganised, and I hate it’.

Shifting a routine from ‘marginal’ to ‘central’ requires not only the introduction of an appropriate technological and co-ordination infrastructure but also a shift in cultural structures such as organisational values and goals (what is our ‘core business’ and why?) and the identity of individual staff (what is my role and purpose here?). As Sewell, cited in Howard-Grenville (2005), has observed, the multiple structures operating within an organisation may ‘operate in harmony or may have conflicting claims and empowerments’– as in the interpreter who is pulled in one direction by her professional body (which exhorts staff not to give out their private phone numbers to clients) and in another by her identity as an advocate (which drives her to develop a ‘kinship’ bond and use informal as well as formal means to improve the client's experience within the system). Greater recognition of the inherent ambiguities in the interpreter role, and a package of training and support that addresses these tensions, is likely to reduce some of the current dissonance and perhaps improve interpersonal relationships with practice staff (Robb and Greenhalgh 2006).

It is a truism that routines must be adequately resourced. In organisational theory; the term ‘resources’ includes not only traditional allocative resources (money, knowledge, technical expertise), but also authoritative resources (command over things and people), and relational resources (such as trust, respect for skills and expertise, and complementarity). Difficulty in routinising the provision of professional interpreters arises because those most closely involved lack all three kinds of resources. ‘Allocating resources for interpreting’ is not as simple as assigning a budget at PCT level for paying sessional interpreters. Rather, relational resources must be addressed – for example by those with power and influence (the GPs and senior interpreting service manager) taking proactive initiative to help build relationships between those who lack these resources. In this way, alignment between the cultural and co-ordination structures within which current routines are (more or less) embedded will increase.

One hypothesis that should be explored more fully is that communication with limited English speakers may occur through a very different routine if both the clinicians and relevant administrative staff are multilingual and the practice population drawn from a small number of minority ethnic communities. In both mainland Europe and the old Commonwealth, it is the norm rather than the exception for people to be multilingual and use different languages in different social situations (for example, at work/school and at home). In relation to multilingual general practices, there has never been a systematic study of what languages are spoken and to what level of proficiency, nor has there been a systematic study of the extent to which GPs’ language proficiency matches the languages actually needed, so it seems premature to problematise the single-handed GP for providing a ‘substandard’ service to the limited English speaker.

There are some important limitations of our study design (based on 69 one-shot interviews and two focus groups). To fully understand the processes by which routines are introduced and evolve (or fail to evolve) in organisations requires longitudinal data and, ideally, a multi-method approach in which ethnographic observations are combined with interviews and documentary analysis (Barley and Tolbert 1997, Edmondson et al. 2001, Feldman and Pentland 2003, Howard-Grenville 2005). For this reason, our findings should be seen as a preliminary analysis of how the organisational, institutional and environmental contexts shape the actions and interactions that make the interpreted consultation possible, and how successive enactments of the routine in turn shape the organisational structures. We strongly recommend further empirical work, with a longitudinal dimension, to test and refine the structurational model presented here, and explore the mutually reinforcing relationships shown in Figure 1.

In conclusion, the current NHS policy of seeking to offer a professional interpreter to all patients who need it poses huge organisational challenges for general practices, especially in the context of rapid expansion in (and increasing heterogeneity of) language need. The development by general practices of strong and efficient routines for providing an effective interpreting service to primary care patients, whether through physically present professional interpreters, remote interpreters or multilingual staff, will go some way towards addressing this challenge. The well-described dangers of attempting to provide clinical care in the absence of an interpreter (see Background) suggest that the routinisation of this aspect of the service should be a clinical as well as an administrative priority. There is a potentially fruitful research agenda linking the organisational dimension of interpreting services with studies of clinical care and outcomes.

The theoretical and empirical approach described in this paper, based on structuration theory, is widely used in organisational science (see in particular the work of Orlikowski and Perlow et al. on reference list) but has rarely been applied in mainstream health-services research on innovation, which in our opinion has been compromised by weak theory and an over-emphasis on experimental methods (Greenhalgh et al. 2004). The effective study of the important and fascinating field of health service innovation demands the use of concepts and theories that do justice to its complexity. A central tenet of structuration theory is that structuration takes place simultaneously across multiple levels (Perlow et al. 2004). For this reason we believe that it has much to offer researchers of service innovation, change and transformation and that the approach described in this paper could, with appropriate modification, have much wider applications.