The craft of intensive care medicine


Address for correspondence: Simon Carmel, School of Health and Human Sciences, University of Essex, Wivenhoe Park, Colchester, Essex, CO4 3SQ


The practice of medicine is often represented as a dualism: is medicine a ‘science’ or an ‘art’? This dualism has been long-lasting, with evident appeal for the medical profession. It also appears to have been rhetorically powerful, for example in enabling clinicians to resist the encroachment of ‘scientific’ evidence-based medicine into core areas of medical work such as individual clinical judgement. In this article I want to make the case for a more valid conceptualisation of medical practice: that it is a ‘craft’ activity. The case I make is founded on a theoretical synthesis of the concept of craft, combined with an analysis of ethnographic observations of routine medical practice in intensive care. For this context the craft aspects of medical work can be seen in how biomedical and other types of knowledge are used in practice, the embodied skills and practical judgement of practitioners and the technological and material environment. These aspects are brought together in two conceptual dimensions for ‘craft’: first, the application of knowledge; second, interaction with the material world. Some practical and political implications of a ‘craft’ metaphor for medical practice are noted.


In this article I address the issue of conceptualising medical practice. Drawing on ethnographic data from routine medical practice in the clinical locale of intensive care, my aim is to develop a conceptually and empirically adequate understanding of clinical practice. In so doing I revisit a recurring theme in medical sociology: the problematic character of the traditional conception of medical practice as an amalgam of ‘science’ and ‘art’. My argument is that ‘craft’ is a more appropriate term for medical work than this problematic science-art dichotomy. The term craft, as I elaborate it, encapsulates both the way that there are different kinds of knowledge used and applied in medical work, and the way that medical work entails interaction with the material world. The ‘material world’, in this context, includes bodies (both clinicians’ and patients’) and technological artefacts. The term craft is intended to bring to the foreground two particular sets of skills required in medical work: using (applying) different kinds of knowledge and practical action.

The analysis of intensive care medicine I present, derived from ethnographic data, draws on Schatzki’s (2001) articulation of ‘practice theory’ for general theoretical background and on Sennett’s (2008) book The Craftsman for a more specific grounding in the sociology of work. Both these authors aim to account for materiality in relation to human activity (cf. Dant 2005) – Sennett describing his position as ‘cultural materialism’ (that physical things are not merely ‘mirrors of social norms, economic interests, religious convictions’) and Schatzki (2001: 3) describing how, in practice theory, ‘bundled activities interweave with ordered constellations of non-human entities’. Overall, Schatzki’s (2001: 2) practice theory ‘conceives of practices as embodied, materially mediated arrays of human activity’, which furthermore presumes ‘the relative unity and integrity of human agency’ (Schatzki 2001: 11). In similar vein, specifically related to work, Sennett (2008: 9) explores dimensions such as skill, commitment and judgment to demonstrate the ‘intimate connection between hand and head’. I follow these leads in wanting to eschew unnecessary theoretical divisions in my analysis of work-based practice. So far as possible – that is, rigorously tested against the empirical reality of practice – I am seeking to develop a coherent description of medical work. Medical practice is a human activity; human beings are concerned with the coherence of their activities; therefore the practice undertaken by human actors in a particular locale can, at least to some degree, be coherent and unified.

Such an emphasis on coherence and unity might be read as contrasting with contemporary science and technology studies (STS) approaches to understanding medical work. A fundamental assumption of much STS is that ‘medicine is plural and heterogeneous’ (Johnson and Berner 2010: 2); medicine and medical work is assumed to be riven with difference and multiplicities (Berg and Mol 1998, Mol 2002, Goodwin and Mort 2010). But it is important to note that the STS emphasis on difference is derived from detailed empirical research which problematises preconceptions of medical work which lack empirical warrant. My own analysis, while contrasting with contemporary STS by placing greater emphasis on the coherence of medical work, nevertheless adopts a similar approach to problematising a preconception of medical work which lacks empirical warrant.

In addition, in STS there is often an assumption that technologies as well as practices are flexible (Johnson and Berger 2010: 3). In this study, however, I make the assumption that material artefacts are relatively stable in routine clinical practice. Thus, while I acknowledge that the introduction of a technology may alter the clinical practice (Moreira and Rapley 2010), there remain important questions about routine medical practice when such technologies are no longer being introduced and have become stabilised, for example, to comprehend ‘the ways in which a particular tool or artefact is used and understood with regard to the practicalities here and now’ (Heath et al. 2003: 84).

In the rest of this article, then, I first summarise relevant sociological discussion of the nature of medical work, in particular noting some of the criticisms of the conception of medical work as a tension between science and art. I then draw on a wider literature to establish a conceptual basis for my subsequent elaboration of medical work as a craft. I note in particular two dimensions for the concept of craft: the application of knowledge and the interaction with the material world. After a short description of the methods of the study, the empirical analysis elaborates these two dimensions of craft with respect to medical work in intensive care.

Clinical practice: art and science?

The portrayal of medical practice as some kind of tension, dichotomy or amalgam of science and art is a commonplace, reflecting practical, pedagogic and political issues. Practically and pedagogically, there is evidently some difficulty experienced by trainee doctors in applying scientific biomedical knowledge (as learned at medical school) to clinical practice. This difficulty is resolved with the notion that medicine comprises scientific knowledge which is supported by artful clinical practice. The very endurance of the ‘science-art’ dichotomy indicates some kind of intuitive appeal to medical practitioners, which raises the question as to why a dichotomised conception (science-art) has proved so popular. One answer to this question is found in the rhetorical and connotative advantages of the terms science and art. Science continues to enjoy considerable popularity as the exemplary measure of knowledge in modern industrialised societies – to say that something is scientific is to accord it a privileged cultural position (Yearly 2005). But as Berg (1997) points out, it would be erroneous to assume that the term science is advantageous only in rhetorical terms. Armstrong (1977) and Berg (1997) describe a long-standing movement to incorporate the scientific method into clinical decision-making, most recently manifested in the evidence-based medicine (EBM) movement (Timmermans and Berg 2003, Dopson and Fitzgerald 2005). EBM seeks to develop a ‘hierarchy of evidence’ for medical decision-making (Sackett et al. 1996), claiming a problem with medical decision-making which is founded on anecdotes and opinion rather than scientific evidence. Thus a scientific approach has long been advocated as a remedy to the shortcomings of medical practice (Berg 1997).

Meanwhile the term ‘art’ has political advantages. For example, EBM can be seen as a challenge to the autonomy of medical decision-making (Pope 2003); and the most notable rhetorical move against EBM is to say that it cannot support the ‘subtle expertise required in daily clinical decision-making’ (Traynor 2009: 494). Used in this context, therefore, the term ‘art’ can deflect scrutiny from central medical decision-making processes. As an example of this, Downie and Macnaughton’s (2000) normative account of the nature of medical decision-making defends an autonomous space for clinical judgement. Thus they describe the ‘art of medicine’ as the:

weighing of evidence in individual cases, the interpretation of the patient’s anecdotes, and other features of the consultation; the exercise of skills, including communication skills in a manner appropriate to the individual case, and leading to the obtaining of informed consent. (Downie and Macnaughton 2000: 71).

However, it must be responded that this use of the term ‘art’ stretches its meaning beyond a conventional understanding. Although difficult to define, ‘art’ is more closely associated with notions such as taste, pleasure, beauty and emotion (Becker 1978, Graham 1997), rather than ‘weighing of evidence’ or ‘communication skills’. Downie and Macnaughton’s (2000) perpetuation of the term ‘art’ within medical decision-making can be read as a rhetorical rather than an analytical development.

Setting limits to the scope for ‘scientific’ scrutiny of medical decision-making is therefore an important advantage to characterising medical practice as an ‘art’. Another related advantage is that medicine is seen as possessing a certain ‘mystique’ (Atkinson et al. 1977), which in turn can be a powerful rhetorical resource in the reproduction of the medical profession’s high public status. These advantages go some way to explaining the endurance of the art/science dichotomy in medical pedagogy and discourse (and the dichotomy itself can donate a framework for debates within specialties –see Schlich 2007).

Overall, we can see that the political stakes in this dichotomised representation of medical work are high, bound up as they are with the central core of medical work: decision-making. But the problem with this representation, notwithstanding its rhetorical and political advantages and the programmatic stance of the evidence-based movement, is that it is simply inaccurate. Most detailed scrutiny finds that medical practice is neither very ‘scientific [nor] very artful’ (Gordon 1988: 260). As I now discuss, a more appropriate concept to apply to medical practice is that of ‘craft’.

Healthcare practice as a craft

‘Craft’ is not an entirely new conception for healthcare work, and the term has been applied to the occupational socialisation of nurses (Melia 1987), to surgical work (Pope 2002) and to psychiatry (Watson 1998). In this section I propose a more precise specification of the term ‘craft’. Drawing on the use of the term in a broad social scientific literature, I explicitly synthesise a family of meanings to establish two particular dimensions. The most straightforward meaning for craft is ‘skilled interaction with the material world’; there is an additional connotation of ‘application of knowledge’, often in thoughtful or insightful ways.

The main sense of the term ‘craft’, as used in the social science literature, is that it encompasses technical skills and manual dexterity (Lee 1981): the material world is generally altered, repaired or improved in some way. Barley and Orr (1997) distinguished ‘craft’ from technical work, arguing that manual skills and material aspects of work are common to craft work but largely absent from technical work. Furthermore, Becker (1978: 887) compared art and craft, noting that ‘craft implies practical utility, art does not’. The skill aspects of craft have also featured in observational studies of laboratory practice. Pinch (1981) for example, found that different branches of physics required different craft skills, even when the same substantive area was being investigated. Latour and Woolgar (1986), describing the micro-processes whereby scientific facts are constructed, also identified and described craft skills.

A second sense of the term ‘craft’ is the application of knowledge, which encompasses insightful judgements and interpretation. Watson (1998), for example, discussed the benefits of conceptualising psychiatry as a craft: the skills of interpretation and the influence of scientific knowledge can both be incorporated. Mills’s (1959: 195ff) ‘personal statement’ on the ‘intellectual craftsmanship’ of sociology distinguished craft from technical work. According to Mills, precise training is sufficient for the technician, but craft work requires something more. Other studies have drawn attention – in a fairly general way – to the craft knowledge inherent in science, technology and medicine (Collins and Pinch 1993, 1998, 2005). The importance of craft knowledge has also been emphasised within non-medical healthcare professions (Higgs and Titchen 2001, D’Cruz et al. 2009). These texts have sought to subtly de-emphasise the purely scientific aspects of clinical knowledge as it is used in practice. There is a sense in some articulations of ‘craft knowledge’ that the detailed theories of a formal knowledge base need not be referenced beyond an initial period of training. My conception of craft follows a similar line of reasoning, but additionally emphasising that clinical skills are subtle physical ones (Harris 2011), therefore embodiment and materiality are closely connected to work such as this (Dant 2005).

The two dimensions of craft (skilled interaction with the material world and application of knowledge) can be drawn together, as when insightfully applied knowledge is combined with dextrous manual skills. Here in particular the details of a formal knowledge base need not be referred to explicitly, since it is local and practical reasoning, applied to the problem at hand, which is paramount. A craftsperson not only formulates plans and implements them, but can change these plans with relative ease when difficulties arise. For example, Reckman (1979) studied his own craft of carpentry, and commented on the manual dexterity required and that in reformulating plans it entails judgements about the purpose of the job at hand. O’Connor (2005) described an apprenticeship of glass-blowing, where both cognitive and embodied learning were required. More specific to healthcare, Pope (2002) suggested that surgery resembles a craft in terms of how contingencies are dealt with, as well as the more obvious way that a sense of touch and manual dexterity are required.

Additionally, craft as a category can be applied to the learning of a trade, occupation or profession, and this sense carries through the dimensions identified above. Examples here have included science PhD students (Delamont and Atkinson 2001), veterinary surgeons (Pinch et al. 1997), and nurses (Melia 1987). From the social science literature, then, we can see that craft is an appropriate term for work which encompasses manual skills (interaction in the material world) and insightful judgements (the application of knowledge). It is these two dimensions of ‘craft’ which I elaborate in the empirical portion of this article.


The data for this study were collected on three intensive care units (ICUs) in the UK, through periods of direct observation (including conversational interviews in the field), supplemented by a small number of tape-recorded interviews. The particular advantage of an ethnographic approach is that the workplace was studied in situ, directly observing and recording everyday work prospectively and in close proximity (Schwartzman 1993). There were formal and informal opportunities to speak to participants directly, to clarify my own understandings and to obtain ‘explanations of [events’] meaning by participants and spectators’ (Becker and Geer 1957: 28).

Ethical approval was obtained from university and local research ethics committees. The work and activity of three ICUs was observed at different times of the week, including nights and weekends. Notes of observations and informal conversations were generally recorded openly, contemporaneously and at a low level of inference (Patton 1990: 239). They were reviewed daily and later transcribed for data analysis, supplemented by a number of in-depth interviews which were tape-recorded, transcribed and analysed. I collected around 285 hours of observational data over 48 separate visits. The ICUs were geographically spread and the hospitals in which they were located differed in their teaching status. For the data presented in this article I give the source, including a pseudonym for the name of the intensive care unit (County, Urban or Metropolitan) and the visit number, or, for the case of interviews, the staff member’s occupation and an indication for their grade. Individuals are abbreviated to a single letter (not their own initial). Field notes, where presented, have been grammatically corrected to aid legibility.

In analysis I took a mainly inductive approach, using the empirical data to build the analysis. Early analysis, conducted in tandem with periods of data collection, identified preliminary ‘sensitising concepts’ (Hammersley and Atkinson 1995: 212) which were then used to inform a search of the existing literature. Theoretical reflection in the latter stages of analysis was thus disciplined by data (Karp 1999) while making links to the two broad dimensions of ‘craft’ specified above.

Intensive Care

Before proceeding to the analysis proper there are a number of analytically relevant features of the clinical locale which need to be noted. First, intensive care is defined as a service for patients with threatened or established organ failure (DoH 1996, ICS 1997), and intensive care admissions can be either planned (typically following major surgery) or unplanned (for example following complications in an acute illness or trauma). In practice, however, most (81%) of admissions to ICUs in the UK are unplanned (ICNARC 2003). Medical practice in intensive care, therefore, is largely unpredictable and planning treatment may require several investigations and frequent reformulation of clinical objectives.

A second feature of intensive care is that patients are, more than in many other areas of healthcare, physical bodies rather than social patients. Patients who need intensive care are usually sedated and mechanically ventilated, thereby becoming, in a sense, non-social beings. Indeed, if patients are awake (and fully participating in the social realm) they are often considered inappropriate admissions as they can be cared for elsewhere in the hospital (Smith and Nielsen 1999).

Third, the materiality of intensive care is evident – machines, bodies, tubes and drugs are all pertinent features of the clinical environment. This is clearly observable on entering an ICU: equipment in the ICU can assist with, and monitor, almost every kind of bodily function.

Fourth, the ICU can be regarded as a general service which treats severe physiological abnormalities rather than a specified disease or surgical problem. The relevant knowledge base is therefore potentially very broad. (ICU doctors are sometimes described as the ‘GPs of the hospital’.) As a generalist service, intensive care is unlike a surgical specialty (defined by expertise in a particular anatomical area) or medical specialty (defined by expertise in diseases of a particular body system).

These four features of intensive care are mentioned in order that readers may assess the potential for generalising (Williams 2000) the craft metaphor to other medical specialties or contexts. In fact, these features of intensive care enable direct comparisons with accounts of surgical and anaesthetic work (Moreira 2004, 2006, Goodwin 2010) and neonatal intensive care (Mesman 2008), comparisons which I highlight in the analysis. I now come to demonstrating the appropriateness of the term craft for explicating work which encompasses the insightful application of knowledge and embodied, practical skills deployed in the material world, beginning with the way that knowledge is applied and interpreted in the clinical context.

Application of knowledge

In this section I identify how different kinds of knowledge and information are used and interpreted in medical practice. Information to which practitioners refer and which they interpret derives from diverse types of knowledge, only some of which are ‘scientific’. While technical, exact sciences (such as pharmacy and toxicology) feature in intensive care medical practice, and biomedical knowledge (such as coagulation) and technologically mediated information are clearly observed, the personal and social background of patients and information derived through the practitioners own embodied skills are also very important. All these kinds of information need to be interpreted by the practitioner and, importantly, ‘scientific’ or technological information is not necessarily given a privileged status compared with other, ‘softer’ types of information. Interpretation of information was often required immediately, indicating the need for the relevant knowledge base to be embodied – that is, readily available and accessible knowledge, contained within the practitioner’s own mind rather than in an external data source. Furthermore, the interpretive effort is greater (or at least, more visible) when drawing in the less ‘scientific’ forms of knowledge.

I start with examples of well established, formal biomedical knowledge. Within the intensive care unit there were several sources of information which seemed to exist independent of practitioners – that is, they were disembodied sources of knowledge. These included computer databases of the effects of poisons, a specialised data book on dialysis and the British National Formulary (BNF), which provides details of appropriate drug doses. The BNF was frequently referenced on ICUs, but its use seemed to require little interpretation. One explanation for this is that ‘pharmaceutical knowledge is based on exact sciences’ (Turner 1995: 143), so that its contribution to healthcare is purely technical. Another explanation is that healthcare practitioners have, in earlier training, undertaken interpretive work to learn how to access the BNF. Thus the pharmaceutical knowledge itself may not be embodied, nor is it expected to be, but the skill of reading and interpreting the BNF is an embodied skill.

An example where knowledge was expected to be embodied could be seen when a specialist registrar needed to refer to a haematology textbook to refresh her memory about blood coagulation. As she did so, she sighed as she said ‘I’ve learnt coagulation about 50 times’, implying a degree of embarrassment: she felt she should not need to look in a book whenever she needed to know about coagulation. It is important to note that she was embarrassed not because coagulation is an easy subject to grasp (on another occasion, in a ward round, a junior doctor responded to a question on coagulation as ‘I’d ask a haematologist’). The embarrassment felt by the Registrar was that she had to ‘look it up’– which is to say, it is expected that important knowledge is embodied and therefore immediately accessible to the practitioner.

Diverse kinds of evidence

Many different kinds of data about patients were called ‘evidence’ (which is, of course, a notably different conception from the ‘evidence’ of EBM, which tends to mean valid and reliable knowledge). The most precise and apparently objective ‘evidence’ were the quantitative readings from machines or laboratory reports for individual patients. Examples of these include blood tests (blood gas analyses, concentrations of serum, salts, etc. in the blood), producing numbers which needed to be interpreted immediately, especially at ward rounds. As Mesman (2008: 107) noted in her study of neonatal intensive care, in order for numbers to be effective they have to be ‘made informative’, which can imply either further number-generating activities or the embedding of quantitative data within a frame of reference. By contrast, in the adult intensive care unit the quantitative character of readings could disappear. For example, at one ward round a junior doctor started to read out the blood gas machine values; a consultant interrupted and asked her to ‘just tell me the abnormal ones’. So, she was expected to be able to interpret immediately the different figures in front of her and determine which were in the ‘normal range’ (different for each test). Thus the meaning of ‘the numbers’ (as they were called by the practitioners) can become disconnected from a ‘scientific’ value. A precise, objective reading, complete with units of measurement, instead becomes a categorical binary value: is it in the normal range or not?

A further contrast with neonatal intensive care is that there seemed to be relatively less quantitative data. Mesman (2008) described the neonatal intensive care as a ‘quantitative practice’ (qualifying this phrase and noting the importance of interpretation). But in the adult ICUs there was often discussion about evidence which was clearly not quantitative, yet regarded as clinically significant:

P (SHO) told me about a worry with the patient in bed 2. The tip of the epidural (routinely tested for bugs when removed from patients, every 3 or 7 days) had grown a bug which was resistant to the antibiotics they had been giving. This was coupled to the fact that as she was waking up she had been saying she didn’t like the light (she had a mask over her eyes), an indication for meningitis. But the Consultant had not been too bothered as the daughter had said that she normally doesn’t like the light – she has curtains drawn at home etc ... So P is not too worried ... (Urban, visit 22)

Now, while the verbal information from relatives is not considered as absolutely reliable as the ‘objective evidence’ which will be obtained from the laboratory (and in fact P still sent off for further tests), the comment from the relatives serves to lessen the urgency of further investigations and treatment; the doctors assumed for the time being that there is no meningitis. Many different kinds of data – including that obtained verbally from patients’ relatives – can become clinically significant.

An important aspect of communicating verbally is that the context is more readily apparent. Another example of the verbal communication of clinical evidence, this time between health professionals, was when a patient was referred to intensive care from the accident and emergency department. In this referral there were three key pieces of information: ‘she has a known heart problem [this was in the medical notes]; she’s just come in complaining of chest pains … but she can usually get to bingo’. I heard this formulation several times as the junior doctor in ICU spoke on the telephone to the staff in A&E, then to the ICU Consultant, then again when he went to see the patient in A&E. ‘She can usually get to bingo’ is perhaps an unusual example of the way that any piece of information can be used and passed around as a clinically relevant fact.

Conversely, once a piece of verbal information has been written down the relevant contextual information may be lost. In this case further interpretive and explanatory work is needed to ensure correct understanding of the information for clinical purposes.

F (specialist registrar): It says in the notes ‘Uses oxygen cylinder at home’, [but] that makes it sound worse than it actually is. He’s actually not too bad, only needs to use the home oxygen occasionally (Metropolitan, visit 9).

This is to say, all the various kinds of data, qualitative as well as quantitative and scientific, recorded and written down, may require contextualising and re-interpreting to be clinically relevant and useful.

In some cases practitioners referred to clinically relevant facts which were not objectively measurable at all. Here the phrase ‘clinically, I think’ was used as a way of disregarding documented evidence (from machines or in the patient’s record) altogether. A senior house officer (SHO) explained to me what the phrase ‘clinically, I think’ actually meant:

It’s a way of ignoring the numbers. It means clinical examination, including talking to the patients. For example, a patient ‘looks clinically dry’– you can tell this by examining the skin, using stethoscope. Clinical skills are not reading the numbers. For example, [in one case] – the central venous pressure was [recorded as] high, which meant that they were ‘full’ [of fluid], but by clinical examination I thought they were ‘empty’ [of fluid] (Interview, SHO, Metropolitan).

Goodwin (2010) noted the importance of information gathered via the senses in anaesthetic work, and the relationship between technologically mediated values and information gathered ‘hands-on’ has been observed in studies of surgical work. Hirschauer (1991), for example, noted how the two types of information complement one another, and Moreira (2006) described blood pressure values derived from machines and from hands-on examination as mutually parasitic. However, in the intensive care unit, as in the example above, there are occasions when hands-on clinical examination, yielding information directly derived from the practitioners own sense of touch, is privileged over the ‘hard’, objective, quantified data. It is also interesting to note how the language used connotes the importance of clinical judgment, reinforcing the privileging of information derived in this way. I note more privileging of hands-on evidence in the following section, where I examine how machines are used in clinical practice.

Interaction with the material world

In this section I consider my second dimension of craft: the interaction with the material world. I analyse how materials, in particular material medical technologies and patients’ bodies, are related to medical work in intensive care. In this part of my analysis I note particular instances of technological failure or difficulty, since these demonstrate most clearly the craft-like character of material interaction (Dant 2005). An important theme of this section is how the seriousness of the immediate clinical context informs practitioners’ reactions to machine readings. Practitioners’ willingness to ignore machine readings has been observed in other clinical contexts (Mesman 2008); in this section I specify two factors which affect whether troublesome readings are treated with more or less seriousness. The first factor is the condition of the patient and the second is the question of the technology’s use (for example, monitoring or treatment). In the latter part of this section I present some examples of the limits which the human body’s materiality presents to medical work in intensive care.

Material medical technologies

Where patients are stable, and the monitoring is routine, readings may be either ignored or routinely corrected:

(Ward round) We go into the side room, and as the doctors were about to look at the computer screen, the nurse says ‘The trace is wrong – we have to correct it manually’ (Metropolitan, visit 10).

P (SHO) is at the recently admitted patient’s bedside. The heart rate monitor shows 0, he adjusts the scale on the monitor and says ‘Heart rate of zero, methinks not’ (Urban, visit 24).

However, in more serious clinical situations, the machine readings were initially regarded more carefully and the readings were double-checked:

R (consultant) is called over to bed 4 as the patient seemed to be having a heart attack ... From the monitor, the patient’s heart rate appears to have increased from 160 to 200, causing a high-pitched bleep from the monitor. R says ‘heart rate not 200 is it’– more of a statement than a question. (Later he explained to me that the apparent (incorrect) high heart rate reading was caused by a ‘high ST-wave’ so that the machine assumed that there were two heart beats when in fact there was just one). He checked with a stethoscope (Urban, visit 15).

(Preparing for a transfer) The driver asks ‘are we ready for transfer?’ F (specialist registrar): ‘Nowhere near’... He tries to zero the instrument for measuring the blood pressure ... the machine shows low blood pressure. F puts his hand on the patient’s wrist: ‘he’s got a bounding radial pulse’– so doesn’t believe the machine reading (Metropolitan, visit 9).

So, in more clinically serious situations, more effort is put into double-checking troublesome readings. The means of double-checking and correcting machine readings are highly indicative. The ‘bounding radial pulse’, which indicates that the blood pressure cannot be low, was determined in a hands-on manner by the specialist registrar (SpR). Stethoscope ‘readings’ are an amplified version of what the practitioner would hear if he put his ear against the patient’s body. Both are basically unmediated data. Despite the presence of advanced medical technology in the ICU, the clinical skills required by practitioners are highly embodied (Harris 2011).

ICU staff must learn how and when to check and correct readings not only when equipment interacts erroneously with the human body, but also when it interacts erroneously with other equipment. Practitioners demonstrably exercise more care when at least one of the technologies concerned is being used in treating (rather than monitoring) a patient. The example I present here concerns a particular haemodialysis machine which, when in operation, made heart activity monitor readings equivalent to those as if the patient was having a heart attack. The heart monitor produced a warning: ‘Undetermined rhythm; nonspecific ST and T wave abnormality; abnormal ECG’. A senior nurse told me how the staff dealt with this:

E (sister): One of our haemofiltration machines can sometimes interfere with the ECG trace … It looks as though someone’s in irregular rhythm – atrial flutter or atrial fibrillation and it is just interference from the haemofiltration machine … There was one instance where somebody was going to treat the rhythm … but ... it was just interference. So it’s just making sure that it’s into the training programme, and orientation and the mentorship within the nursing staff and then hopefully the same thing within the medical staff. And I have a tendency to remind people if I see somebody on [the machine] ... On the ward round I will say ‘do you remember’. So hopefully you know – everybody does have that at the forefront of their minds (Interview, sister, Metropolitan).

So, on one occasion, ICU practitioners had actually responded to this ‘erroneous trace’ and were about to treat the supposedly ‘irregular rhythm’. It is clear that more effort has been taken by senior staff in this example than in the earlier examples where the patient was more stable or the equipment was used only for monitoring. In this more serious case, senior staff need to remind people so that they have it at the ‘forefront of their minds’.

A whole range of information and practical knowledge must therefore be retained by ICU workers in the course of their everyday work, including situations when machine readings can be ignored or questioned. The seriousness of the clinical context, and what machines are being used for are factors influencing the response to apparently erroneous readings.

A final aspect of the interaction with technology is that the theory of how a technology works can be unimportant in routine clinical practice:

During the ward round, consultant G: No-one understands the physics of high-frequency ventilation. [All we know for sure is that it’s something to do with] the increased surface area (County, visit 2).

Thus, it is claimed that ‘no-one understands the physics’ of how this technology interacts with the human body. The precise details of how the technology affects the lung are not strictly of interest; in clinical practice ‘knowledge’ is important only in terms of how it relates to action. As with ‘craft knowledge’, the details of a knowledge base need not be fully understood so long as there are practical uses.

The materiality of patients’ bodies

Materiality also affects clinical work in that it is body work. There are two senses of body here: work with the (clinician’s) human body and work on the (patient’s) human body. Physical skills must be learnt in apprenticeship fashion. A specific example can be drawn from my fieldwork, when a junior doctor learnt how to use a bronchoscope. The doctor initially experienced difficulties because one of the bronchi was at an unusual angle. She passed the bronchoscope to the Consultant who had been instructing her. The Consultant could be seen to have acquired superior manual skills, since he commented that the bronchus did go off at a funny angle, yet he continued with the procedure. He subsequently passed the bronchoscope back to the junior doctor for her to continue, and gave advice in terms of various ‘rules of thumb’ (e.g. ‘come out on expiration not inspiration’; ‘don’t rush if you cannot see where you are going’). Learning how to use machinery and equipment which interacts with a patient’s body must be experienced ‘on the job’ by ICU staff.

Additionally, clinical work in intensive care is also work on the human body of the patient. The material nature of the human body itself can constrain attempts at such ‘body work’ (Twigg et al. 2011). The example below follows investigations for a patient admitted to intensive care, where medical treatments which were theoretically possible were not undertaken due to material limitations.

Day 1, 9.20am. Consultant R, M (nurse) and N (SpR) are in the Side Room ... They discussed the possibility of inserting a PA [pulmonary artery] catheter, but neck problems precluded it.

Day 2, c. 2pm. Chatting to consultant W who tells me that the patient in the side room is in an ‘extraordinarily difficult situation’. The patient needs dialysis, but he is obese and with a fused neck, can’t lie flat. He is looking ‘sick to death’... Four hours later, I learnt that this patient had died.

Day 3, c. 3pm. (grand round) consultant W: ‘I feel bad, well I feel half-bad because of technical difficulties’... ‘technical challenge in every way’. P (SHO): ‘Nightmare’. (Urban, visits 8, 9, 10)

A lot of treatment in ICU does not ‘rescue’ patients: the average in-hospital death rate for ICU patients is 33% (ICNARC 2003). This begs the question, why does consultant W feel bad (‘well, half bad’) about this particular patient? I think the answer is to be found in that it was theoretically possible, within medicine’s grasp as it were, to keep this patient alive. But his body, affected by obesity and prior medical treatment which had fused his neck, precluded the necessary treatment, illustrating some implications of the materiality of intensive care work. This was called by consultant W a ‘technical challenge in every way’. But to describe it as a technical challenge was to downgrade the nature of the problem – it was, rather, a problem inherent in the natural world, that is, the physical nature of the material world. A medical practitioner may hope to implement ‘clinically indicated’ treatment(s) in any circumstances, yet physical reality (in this case, a patient’s body which was materially resistant to medical intervention) may prevent such action. There are important practical consequences, therefore, of the materiality of intensive care work; clinical judgement in intensive care must consider the material reality of the patient’s body, for example.

A further example of the limits to medical intervention could be seen in a case where there were two treatments in contradiction: a patient needed dialysis but had to be kept lying prone (face down) to improve his breathing; the lying prone precluded dialysis. The limits placed on medical work by the materiality of the human body and by limits to medical intervention resonate with Sennett’s (2008: 262) observation that ‘the good craftsman learns when it is time to stop’.

Conclusion: practical reasoning, practical skills

In this article I have discussed craft under the headings of the application of knowledge and the interaction with the material world. In summary, the application of different kinds of knowledge consists of the interpretation of different kinds of ‘evidence’. Whether documented or non-documented (for example, biographical information about the patient), evidence is part of the overall clinical context and is incorporated into clinical judgements. On the one hand, formal scientific knowledge was only infrequently observed to feature in the routine work of ICU practitioners; on the other hand, practitioners’ interpretive work was applied to a diverse array of different kinds of knowledge.

We have also seen that ICU workers demonstrate considerable skills in the material world. This relates to their own embodied manual skills and to practical work with machines and patients’ bodies. Clinical work, at least in this context, exemplifies how skills and know-how are embodied in individual practitioners (Schatzki 2001, Dant 2005). Medical work is neither ‘artistic’ nor ‘scientific’ (Gordon 1988). Rather, it resembles a craft activity (Sennett 2008): it has a practical, interpretative orientation to different kinds of knowledge, and requires embodied skills to be mastered.

This analysis of medical work has demonstrated that the concept of craft is more valid than the somewhat tired and unconvincing dichotomy of science/art. Two main implications follow. First, craft is a conception of the practice which could be helpful for junior and trainee professionals. As highlighted by Prentice (2007) in relation to surgical learning, skills development has a meaning which goes beyond the acquisition of techniques. Skills development requires a broader understanding than technical or scientific knowledge. At least in intensive care, there are more types of knowledge than simply biomedical sciences. Furthermore, medical practice is not an ‘art’, and skills development is therefore, following Sennett (2008), ‘arduous but not mysterious’.

Second, in terms of policy, the conception of medical work as craft enables a critique of public discourses which over-emphasise either the ‘art’ or the ‘science’ of clinical judgement. The complexity and subtlety of clinical judgement is not undermined by the metaphor of craft. Rather, this metaphor can acknowledge the part that science can play in clinical judgement as a resource for medical decision-making, while recognising that scientific knowledge is only part of the picture. The craft metaphor, then, undermines a conception of clinical judgement in particular, or medical practice in general, which is too narrowly scientific.

It will be noted that my conceptualisation of medical work as a craft not only differs from the traditional medical conception of art-science but also questions Nettleton et al.’s recent (2008: 345) identification of a trend towards the ‘disembodiment’ of clinical knowledge. They suggest that being a doctor ‘relies less and less on experiential knowledge and more and more on encoded and formal knowledge’– a finding which contrasts with mine. The differences could be due to the fact that Nettleton and colleagues’ analysis is based on interview data which, being in the realm of discourse, are more likely to relate to theoretical knowledge. My analysis is based on ethnography, which is more commensurate with a focus on observable embodied practices. I suggest therefore that Nettleton and colleagues’ interviewees articulated their sense of the changes to the formal knowledge base from which they are expected to draw in their practice. Of course, this aspect of clinical knowledge is difficult to observe empirically in practice, and I do acknowledge that it is a part of the background to medical work. However, with Mort and Smith (2009), I would suggest that the experiential and embodied dimensions of medical practice are still crucial, despite increasing amounts of encoded knowledge and information.

My findings resonate with other ethnographic studies in anaesthesia (Goodwin 2010), surgery (Moreira 2004, 2006) and neonatal intensive care (Mesman 2008). Undoubtedly, the ‘features of intensive care’ I identified in the methods section of this article will need to be considered carefully before generalising to other clinical contexts. Nevertheless, I suggest that the two dimensions of craft appear, prima facie, to relate to a range of medical specialties. My argument is that medical work in intensive care is very much a craft activity, which in its foreground both applies different kinds of knowledge and is a materially embodied practice.


I am grateful to the intensive care practitioners who welcomed me into their working lives. The original research, reported in Carmel (2003), was funded by an NHS Special Training Fellowship in Health Services Research (Grant number RDC 01652). I am grateful to all those who stimulated and encouraged my thinking on this topic since then, in particular Judy Green, Mary Ann Elston, Carl May, Cathy Pope, Tim Rapley, Rob Stones and the Sociology of Health and Illness reviewers and editors.