New tools for an old trade: a socio-technical appraisal of how electronic decision support is used by primary care practitioners


David Peiris, The George Institute for Global Health, Renal Division, PO Box M201, Missenden Rd, Sydney, New South Wales 2050, Australia


This article explores Australian general practitioners’ (GPs) views on a novel electronic decision support (EDS) tool being developed for cardiovascular disease management. We use Timmermans and Berg’s technology-in-practice approach to examine how technologies influence and are influenced by the social networks in which they are placed. In all, 21 general practitioners who piloted the tool were interviewed. The tool occupied an ill-defined middle ground in a dialectical relationship between GPs’ routine care and factors promoting best practice. Drawing on Lipsky’s concept of ‘street-level bureaucrats’, the tool’s ability to process workloads expeditiously was of greatest appeal to GPs. This feature of the tool gave it the potential to alter the structure, process and content of healthcare encounters. The credibility of EDS tools appears to be mediated by fluid notions of best practice, based on an expert scrutiny of the evidence, synthesis via authoritative guidelines and dissemination through trusted and often informal networks. Balanced against this is the importance of ‘soft’ forms of knowledge such as intuition and timing in everyday decision-making. This resonates with Aristotle’s theory of phronesis (practical wisdom) and may render EDS tools inconsequential if they merely process biomedical data. While EDS tools show promise in improving health practitioner performance, the socio-technical dimensions of their implementation warrant careful consideration.


Systematic reviews have shown that electronic decision support (EDS) systems may improve practitioner performance in healthcare (Garg et al. 2005, Kawamoto et al. 2005, Shojania et al. 2009). Despite the rapid proliferation of such systems and their appeal to health-systems planners there has been limited sociological inquiry into their impact. This is needed because a narrow focus on effectiveness, as gauged by practitioner performance, will not assess their influence on other dimensions of healthcare quality. Further, the demonstration of their efficacy in trial settings is likely to understate many of the real world contingencies that will impact on uptake of these technologies.

Timmermans and Berg (2003) have outlined three contrasting theoretical approaches to understanding how medical technologies influence healthcare practices: (i) technological determinism, in which innovation acts as an instrument of power imposed as part of ‘a medico-industrial complex’ onto doctors and patients who have limited agency in the process; (ii) social essentialism, in which social and political realities preferentially select particular technologies to suit their purposes and the technologies themselves are blank slates absorbing and reflecting these prevailing paradigms; and (iii) technology-in-practice, which assumes a dialectical relationship between the technologies and their users. This last approach draws on actor-network theory (Latour 2005) and views technologies as one of many actors neither devoid of ability to influence nor assuming a ‘super agency’ with extraordinary power over other actors.

Marc Berg has written extensively about the ways in which technologies such as decision support systems (Berg 1997b), patient care information systems (Berg 2001), clinical protocols (Berg 1997a), guidelines (Berg et al. 2000) and even the medical record itself (Berg and Harterink 2004) function as non-human actors in the healthcare context. Drawing mainly from ethnomethodological studies, this work allows us to appreciate how these technologies influence the ways in which organisations deliver healthcare and how human actors (doctors, nurses, patients and others) respond. By following the interactions created by these human and non-human actors we can begin to appreciate their sociological effects. Latour (2005: 12) states:

… you have ‘to follow the actors themselves’, that is try to catch up with their often wild innovations in order to learn from them what the collective existence has become in their hands, which methods they have elaborated to make it fit together, which accounts could best define the new associations that they have been forced to establish.

Health professionals’ core business of delivering quality patient care is characterised by complex knowledge processes that are not amenable to a simple process of behaviour change. While guidelines and protocols have appeal in providing professional transparency about decision-making processes, they tend to presuppose medical action as a logical sequence of steps toward a perceived, single optimum code of practice (Berg 1997a). Greatbatch et al.’s (2005) ethnographic work shows that rule-based expert systems capture only part of what experts do. Health professionals are constantly sidelining rules based on their expertise, practical limitations and judgements about their patients’ expectations and capabilities (Summerskill and Pope 2002). Berg found that decision support systems were frequently modified once implemented, often compromising substantially the original ‘scientific’ thought processes that went into their algorithms (Berg 1997b). Appreciation of these factors is important in explaining the known limited effectiveness of protocols and guidelines (Grimshaw et al. 2004).

The role of technologies as non-human actors in the healthcare encounter can pose epistemological challenges. Alongside traditional doctor-centred and emerging patient-centred agencies such as shared decision-making, culturally competent care and self-management, are third party actors with regulatory authority (May et al. 2006). One such actor is the evidence-based medicine (EBM) movement in which predominantly clinical trial and population-based evidence has come to define appropriate care. EBM is asserted to be the guiding principle by which the profession defines and regulates appropriate conduct (Armstrong 2002). The core dilemma faced by EBM is overcoming the challenge of applying population-based findings to individual healthcare encounters in which doctor and patient agencies prevail (May et al. 2006). Hence, new technologies such as point-of-care decision support are viewed optimistically by health system planners because of their potential to narrow ‘evidence-practice gaps’, bringing epidemiological and clinical trial evidence into everyday practice.

This convergence of human and non-human actors in the healthcare encounter produces a symbolic drama (May et al. 2006) whose effects are not well elaborated. In this article we draw on the technology-in-practice approach (Timmermans and Berg 2003) to develop an understanding of the ways in which EDS tools function as actors in the social context of primary care. This forms part of a broader research programme to develop an EDS tool in Australian primary care settings, described below. A previously published pilot evaluation (Peiris et al. 2009a), focused on tool design and the modifications needed to allow for its routine use by general practitioners. The motivation for this sub-study arose from a dedicated part of the interview evaluation in which general practitioners (GPs) were encouraged to give a more expansive account of the implications of the future implementation of the tool in practice. It was in this part of the interview that important data emerged relating to the dialectical relationship between routine care provided by GPs and the promotion of best practice, in particular the role of evidence, guidelines and technological tools. In this article we more thoroughly explore this relationship by examining the actors and social settings in which EDS tools are placed.

Development of an Australian EDS tool for cardiovascular disease (CVD) risk management

EDS systems are relatively underdeveloped in Australian primary care settings. In 2005 researchers at The George Institute, in partnership with several collaborators, established a research programme to develop and implement a suite of EDS tools for use in primary healthcare. A specific objective of this programme is ensuring these tools are suitable for use in private general practices and community-governed Aboriginal medical services (AMS). The first major project in this programme focuses on CVD. CVD is a collective term to include diseases such as coronary heart disease, stroke and peripheral vascular diseases. Despite major advances over the last two decades in CVD epidemiology and a plethora of clinical trials establishing the efficacy and safety of various therapeutic interventions there has been variable translation of this knowledge into practice. Multiple, complex guidelines and inadequate resources to implement them at the point of care are contributing factors to the low uptake of best practice recommendations in Australia (Peiris et al. 2009b, Webster et al. 2009). To help address this, an EDS tool was developed that synthesises recommendations from several Australian guidelines into a single CVD management algorithm to provide point-of-care recommendations.


In the pilot implementation a stand-alone version of the tool was field tested with 21 GPs working in eight private, teaching general practices and three AMSs. Sampling was purposive. We sought GPs interested in research, medical education and the provision of services to Aboriginal people. It was considered that GPs with these interests might subject the tool to vigorous scrutiny and provide recommendations for its future development. Prior to study commencement, research staff provided GPs with an orientation to the tool and interpretation of the output. In all 200 routinely attending adult patients (33% Aboriginal, 67% non-Aboriginal) within the age ranges recommended for CVD risk assessment were invited from the waiting room to participate in the study. A research assistant accessed the patient’s electronic health record and entered data into the EDS tool, thus simulating the automated data entry that would occur if the tool was fully integrated in the GP’s software system. The resultant output was printed and given directly to GPs to review during their consultation with the patient. A sample annotated output is shown in Figure S1 (this colour figure may be found in Supporting Information in the online version of this article). The manner in which GPs used the tool in the consultation was at their discretion, including whether or not to show the tool output to the patient. Each GP had outputs generated for up to 10 patients, a big enough number with which to gain an appreciation of the tool’s application in a typical working day.

At the study completion, the GPs participated in an in-depth interview evaluation and completed a survey about their professional background and attitudes to electronic information. The interviews were conducted by the lead author, a practising GP who had a working knowledge of the tool but was not involved in its development. Two interviews were conducted by phone with the remainder in the GPs’ consulting rooms. This not only created a comfortable setting but it provided contextual information about the setting in which routine healthcare encounters occur. It also allowed for easy access to computer records for patients seen in the study. The interviews were generally conducted on the day or within a few days of use of the tool and ranged from 30 to 60 minutes’ duration. The interview was semi-structured and was conducted in three parts. The interview guide is provided in Appendix 1 at the end of this paper. The first part involved an initial discussion about attitudes to the tool and its impact on the healthcare consultation. In the second part, sample tool outputs were selected in order to stimulate a collegial discussion about actual clinical scenarios and the rationale for particular management decisions. This allowed GPs to recount the ‘story’ of the healthcare encounter (Greenhalgh and Hurwitz 1999) in much the same way as cases are discussed for management purposes with colleagues and for medical education. During these discussions GPs frequently reviewed the patients’ electronic health records to elucidate the decisions made. The third and final component of the interview involved discussing recommendations for the future implementation of the tool in routine general practice.

Interview recordings were professionally transcribed and thematic content analysis was conducted following the methods outlined by Patton (2002). Analyses were conducted contemporaneously with data collection and used to inform subsequent interviews. As we began to identify the socio-technical aspects of tools to be a substantive issue, more weight was given to this in subsequent interviews. Upon completion of all interviews repeat readings of interview transcripts were conducted and the findings were then provisionally categorised by theme. These themes were jointly discussed by the research team over a series of meetings and the major thematic groupings were refined. The team comprised the lead author; a senior GP academic who has provided strategic advice on tool development from an early stage of the project; two health service researchers not involved in the project, who offered an outsider perspective to the analyses; and the project leader, who has been responsible for the broader EDS research and development programme since its inception. This insider-outsider team composition was especially useful for challenging and justifying particular thematic interpretations, and mitigating any bias arising from a single interviewer being used.

NVivo 8 (QSR International, Melbourne, Victoria) was used to assist with organising the data. Principal study findings were fed back to the participating GPs and AMS managers as a written report and oral presentations. The study was approved by both the Sydney South West Area Health Service and Aboriginal Health and Medical Research Council ethics committees. Patients and GPs gave their written consent to participate and signed agreements were obtained from the governing bodies for the three participating AMSs.

Findings and interpretation

A total of twelve male GPs and nine female GPs participated. Nine GPs worked in AMSs and the remainder in urban, teaching and general practices. The practice size varied greatly, ranging from solo operators to a large multidisciplinary service with 20 GPs. Four GPs were under 40-years old, eleven were 40–49 years old and six were over 50 years old. All spoke English as their primary language. Eighteen were Australian university graduates. High levels of professional training were attained. Fifteen held fellowship status with the Royal Australian College of General Practitioners, eleven held postgraduate diplomas and four held Master’s degrees. All but two participants regularly conducted or participated in research projects. Overall 19 participants reported use of the Internet at least once daily. Table 1 provides the survey results. Although there are limitations to self-reported data, the participants recorded a high uptake of electronic practice software features and positive attitudes to the role of computers in general practice. Five principal themes and their resonances with particular studies are discussed here. Two themes relate to how and why the tool might influence the routine care provided by GPs. They include ‘GPs as street-level bureaucrats’ and ‘communication influences’. Three further themes relate to the role of the tool in notions of best practice care. They include ‘technogovernance’, ‘mindlines’ and ‘phronesis’.

Table 1. Survey responses on attitudes to computers and sources of medical information for the 21 participating GPs
Electronic practice software features usedAlways usedSometimes usedNot used or not available
Medication prescribing201-
Automated pathology results1911
 Online billing14-7
 Electronic patient recalls1362
 Scanning of paper documents12-9
 Electronic care plans1272
 Chronic disease patient registers 7104
Effect of computers on the followingPositive/very positiveNo effectNegative/very negative
 Patient safety183-
 Practice of evidence-based medicine 1731
 Practice cost efficiencies147-
 Patient privacy1191
 Patient-doctor communication1083
Source of medical informationVery influentialSomewhat influentialNot influential
  1. * 1 missing response

  2. **2 missing responses.

  3. - no responses in the relevant category.

 Clinical guidelines from professional organisations138-
 Continuing medical education events1182
 Pharmaceutical product information in medical software1092
 GP colleagues912-
 Electronic text books 9102
 Personal internet searches 8121
 Peer-reviewed journals894
 Medical newspapers 2154
 Pharmaceutical representatives-714

GPs as street-level bureaucrats

Checkland (2004) used Lipsky’s concept of ‘street-level bureaucrats’ (Lipsky 1980) to better understand the impact of normative codes of practice such as clinical guidelines on the work of GPs. Framing GPs as street-level bureaucrats recognises their role as powerful, semi-autonomous workers directly engaging with a demanding public and having to process voluminous data to make rapid, safe and effective decisions (Lipsky 1980). The pressures of processing information efficiently were noted in this study. Despite guideline recommendations to perform regular CVD risk assessments, nearly all 19 participants reported performing these infrequently or not at all in their routine practice. A major barrier was that the most commonly used paper chart version was not practical to implement:

Interviewer: Are you a prior user of the New Zealand [paper] risk chart calculator?

Participant: Look I’ve seen it and I haven’t used it simply because where do I put it? I mean there’s enough junk on my desk and around this room already. So unless it’s in the software it’s not useful. (Interview 7: private practice male GP, over 60 years old)

In this way GPs indicated that the most supportive function of an EDS tool was if it enabled access to resources and guidelines during the healthcare encounter. On reflecting why she didn’t use guidelines more often at the point of care, one GP said:

I think that’s really been the reason why I haven’t used them more routinely because, yeah, I know the guidelines but I have them in another place. So it’s been an issue in terms of first of all getting it then and there and then applying it . . . . and all that time and stuff that’s involved with that. So I think it can only be a bonus: the fact that it’s immediately there. (Interview 5: AMS female GP, 30–39 years old)

Similarly, rather than actively not implementing evidence, GPs are more likely to be simply distracted from certain management actions by the routine pressures of the healthcare encounter:

I think that there is the scenario that [for] the people you know well you often overlook things. And it is very useful to have a tool that actually brings you back to basics again. You know, sometimes you are just so focused on things and you forget. (Interview 3: AMS female GP, 40–49 years old)

Juxtaposed against this appreciation for being brought back to the basics, one GP expressed irritation when the tool failed to acknowledge previously performed work and made blanket or seemingly obvious recommendations:

She’s [already] had her cholesterol done and then it says ‘cholesterol evaluation is recommended’… They’re telling us to suck eggs repeatedly, and I don’t like it. (Interview 15: private practice female GP, 40–49 years old)

GPs also frequently commented on the failure of onscreen prompts to effectively integrate with the natural workflow of the clinical consultation. They cautioned against EDS tools adding to the already burdensome prompts to perform various clinical tasks:

I mean, there are already a lot of pop-up windows … and adding another one isn’t going to get anyone very excited … Imagine how much better the software would be if it tracks the way you use a consultation. (Interview 4: AMS male GP, 30–39 years old)

Communication influences

The tool exerted additional important influences beyond expeditious processing. In these instances the tool punctuated conversations about CVD risk in new ways. For one GP the tool created quarantined discussions around cardiovascular risk management:

I think it was quite a good thing because you would finish the consultation about whatever that was about and then you’d almost have a separate time set for looking at cardiovascular risk … So having that piece of paper [the tool output] there gave you that conversation: ‘well now we’ve finished everything, let’s look at this’. (Interview 13: private practice male GP, 40–49 years old)

Another GP forewarned of dangers if the tool distorted priorities in the healthcare encounter. By eliciting particular types of information related to cardiovascular risk this may come at the expense of other, potentially more patient-focused information:

One of the dangers I would see with this is the encouragement of game playing … So an electronic decision support module that is only related to cardiovascular disease … could lead you to focus on getting cholesterol and things done and perhaps forget immunisations or pap smears or the housing forms because that’s what the computer is flashing up at you. (Interview 4: AMS male GP, 30–39 years old)

While some of the above observations are informed conjecture based on how GPs may implement the tool in routine care, they demonstrate how innovative technologies have the potential to synchronise the solicitation and delivery of particular types of information (Greatbatch et al. 2001). Such technologies have the capacity to focus GPs’ attention and produce particular kinds of conversations with patients, potentially displacing others. A number of GPs described how the tool not only created space for these conversations but shaped the content of their talk, broadening the existing communication devices that are deployed to discuss CVD risk management. One GP found this changed her overall communication package:

I think the biggest impact is that it changed the way I talked about what I was doing with them, in that it made it a much more slick, neat package to describe the normal screening that you do for risk management. And so I felt it was easier to deliver some description of where they’re at now. (Interview 2: AMS female GP, 40–49 years old)

Several GPs commented on how they used the classification of the person’s risk on the colour spectrum bar to influence discussions (see Figure S1). For one GP the graphic representation of risk coupled with the unquestionable authority of the computer became a device for the delivery of routine messages about adopting healthy lifestyles:

Yeah and even a coloured diagram is really helpful in being able to say, ‘Look, the computer says it’s true. This isn’t just me making up words around diabetes. Look, this is going into orange – this says “high” in red’. And there’s almost an emotional response to the colours that come back that is actually really useful compared to me saying, ‘look people with diabetes have heart attacks and strokes’. (Interview 4: AMS male GP, 30–39 years old)

In this way the tool output, and in particular the use of colour, can be viewed as a boundary object (Star and Griesemer 1989); an entity that establishes coherency between the disparate world-views of researchers, clinicians and patients but may be used differently by each of these actors. It is not merely a means of representing categories of risk on the basis of inputs to the tool algorithm. Rather, it can become an influential entry point for how health professionals frame risk and advise on actions to mitigate risk, and how patients interpret risk.


May et al. (2006) describe a process of ‘technogovernance’ in which technological innovation is used to create a differently distributed accountability from that seen in the traditional doctor–patient dyad. Non-human actors, such as EDS tools, can change the structure and direction of decision-making processes. By actively bringing guidelines into the consultation, these tools become a third epistemological authority (alongside the doctor’s and patient’s) bringing about new practices of governance.

We encountered diverse views from GPs about these technogovernance issues. A number of GPs expressed being at ease with this third presence. Although we cannot know how this may have manifested in actual practice, some of the more experienced GPs welcomed being made aware of an ideal and potentially unachievable management scenario for their patients:

I’m comfortable in that if I don’t meet a target and I’ve made a conscious decision about that I’m comfortable to live with that, I’ve been making those decisions for 20-odd years. But I’m comfortable that the guidelines are there now, I feel much safer. (Interview 17: AMS male GP, 50–59 years old)

I don’t have a problem with seeing that that’s where this person ought to be … It’s a goal and you can explain that to the patient by saying, ‘Look, this will move, if we do this, but the choice is really yours’. There’s lifestyle, there’s enjoyment of life, there’s the ideal situation and we’ll come to some consensus. (Interview 7: private practice male GP, over 60 years old)

Two GPs, however, were challenged by the tool’s perceived external authority and its agency was palpable in their accounts. For one GP the tool took on a threatening and vexatious quality ‘embarrassing’ him into better practice:

They [patients] come for condition X and the machine is tapping me on the shoulder and saying, ‘by the way, look at this’. Well I suppose one thing that crosses my mind is embarrassment. If it was so bad that you’d missed it and it said, ‘Hey you need to prescribe this like now’, I would be thinking ‘who’s running the show, the machine or me?’… Somewhere in my past I’ve always wanted to make sure that the doctor is the one who is making the decisions. (Interview 11: private practice male GP, 50–59 years old)

Another GP perceived that the tool changed the direction of engagement when her patient reviewed the recommendations and suggested to her that a medication be prescribed:

Interviewer: What about the prompts around meeting targets? I think you did in fact make changes to his blood pressure and cholesterol medicines.

GP: You made me do that! You see … [laughing], the older my patients get the less I like to interfere … So your tool has made me do an intervention which I’m not sure if it’s okay or not actually … John [pseudonym] himself wanted to see this … so that’s partly what made me change his stuff … Once a patient sees this … like you and I, we can intellectualise about it and recognise it’s just a tool, but for a patient it’s very real and they would often in fact want treatment. (Interview 14: private practice female, GP 40–49 years old)

This GP also raised concerns about the tool’s potential for use by health system bureaucracies as a mechanism to assess performance and outcome payments:

This tool … what I’d hate to see is if we ever get to outcome payments … because this hasn’t been performed within 6 months – maybe for some very valid reasons and because of other issues with the patient. So I’d hate that. (Interview 14: private practice female GP, 40–49 years old)


Related to the technogovernance issues were broader discussions about how GPs relate to evidence in practice. Gabbay and Le May’s ethnographic study found that participating GPs rarely ‘accessed, appraised and used explicit evidence directly from research or other formal sources’ (2004: 3). Rather, they relied on mindlines – tacit guidelines informed by brief readings of guidelines but, more importantly, collectively mediated through various informal networks (colleagues, opinion leaders and even their patients). Despite the limitations of self-reported data, Table 1 indicates that the most influential sources of medical information for participating GPs were specialists, professional guidelines and continuing medical education activities. More direct sources of evidence, such as peer-reviewed journals, were less influential. Although these GPs had high levels of postgraduate training and a predisposition to research, they seemed to indicate that evidence needs to be validated through trusted networks, especially colleagues and professional organisations. This was particularly important when GPs considered incorporating new evidence into practice:

There’s an element of consensus around what peers are doing. So an example would be, say, using ACE inhibitors and A2 antagonists [two blood pressure medicines] … The recommendations were to use both [in combination] and then the evidence came out that it doesn’t seem to be any more effective. My feeling is that I don’t really want to be using both of them. But I haven’t got a feel for what my peers are doing yet … and what the specialists are doing. So I’m in that discomfort zone where I’m actually fairly comfortable with what the evidence shows, but I’m not sure what’s being recommended by everyone else yet. (Interview 4: AMS male GP, 30–39 years old)

Thus, new and emerging evidence was often greeted cautiously. Until this evidence becomes embedded in particular mindlines its utility is of questionable relevance. Guidelines from trusted organisations constitute an important component to these mindlines even if they are out of date or conflict with emerging evidence:

From a GP’s point of view, you just want the guidelines. You don’t want to know about what’s coming up because what’s coming up may or may not change. You can get one trial that’s says this is a great thing and then a few years later you might get another trial that’s effectively the opposite … As a GP you can’t know about all the little bits and pieces of everything that’s not going to lead to a change in management. (Interview 12: private practice male GP, 40–49 years old)

There was also concern that premature adoption of research-based recommendations had the potential to damage the credibility of the profession. The example of delayed findings of increased risks associated with hormone replacement therapy (HRT) was cited:

I think if it’s not proven to be a guideline we’re going to look like we did with HRT. You’ve got to be pretty sure before you put it in a guideline. I imagine it’s got to be as solid as it can be. (Interview 16: private practice female GP, 40–49 years old)

If decision support tools challenge established and trusted networks they may create inconsistencies with these mindlines. The existence of multiple, sometimes conflicting and out-of-date guidelines for CVD in Australia could exacerbate these inconsistencies. One GP commented on the confusion that could occur if the authors of EDS tools became alternative expert actors who proposed recommendations that were different from those of existing experts:

It could get confusing, because I think you’d end up having the programmers or the decision-makers around the computer system becoming experts … and then there would be all sorts of quotes from professors or medical newspapers saying, ‘Well, that’s not the recommendation and we shouldn’t be doing that’. It leads to confusing messages because there isn’t actually the consensus around something until it hits the guidelines. (Interview 4: AMS male GP, 30–39 years old)


While evidence, guidelines and protocol implementations were important factors in how decisions are made, the participating GPs made a number of references to the role of wisdom, intuition and timing as being of prime importance. Their comments resonate with Aristotle’s theory of phronesis, often translated as ‘prudence’ or ‘practical wisdom’ (Flyvbjerg 2001). One GP questioned the value of CVD-related tools and felt that assessing a patient’s risk was an intuitive process that came with experience:

In general practice [tools are] not that important … I think most of us, we really treat on empirical grounds. We have a feeling that this person is at higher risk, and we just treat them … I don’t seem to have many slip through my fingers, because I see them a couple of times a year for coughs and colds etcetera. (Interview 10: private practice male GP, 50–59 years old)

In particular, the fact that management is implemented over a series of consultations can take precedence over point-of-care decision support:

I think that’s where continuity of care and the art of general practice really comes to the fore … that you actually try to meet the patient’s and the doctor’s agenda. But you know, ‘the longitudinal consultation’, I think that’s the important thing … What I would love is a tool to show the effectiveness of continuity of care. Like, John, for instance, I’ve seen him for 25 years. There is no tool like that. Continuity of care is probably … his biggest [contributor to] improvement in health. (Interview 14: private practice female GP, 40–49 years old)

Even for the few, mainly younger, GPs who performed CVD risk assessments regularly, the skill of applying the process in a meaningful way took on more prominence than the results of the assessment itself:

It takes a long time to do it [CVD risk assessments] properly … and to make people understand them. The more I did it actually, the less that I used it after … the more I realised it was an art. I think it’s great having the numbers but it’s how you apply that number to that person sitting there, is my real feeling. (Interview 19: AMS female GP, 20–29 years old)

Phronesis is characterised by an openness to change through reflection both in action during the healthcare encounter and ‘on-action’ after the encounter (Schon 1983). This reflective process was described by one AMS GP for whom awareness of population data on risk factors and inequitable health outcomes for Aboriginal peoples was influencing his previously conservative approach to treatment decisions:

My practice is changing … I’ve come from a practice where I was saying ‘let’s see how it goes’ because patients are unwilling to take a new medication or increase the dose … Whereas out here, I think you’ve just got to do it. They [Aboriginal people] are a really at-risk group … And if you say, ‘Wait, that’s okay’ you’re sending a message that you’re not that comfortable starting them on something. (Interview 17: AMS male GP, 50–59 years old)

Thus much of the decision-making process for GPs is not merely scientifically based. It is equally bounded by the social and moral values that influence the interaction between doctors and patients. As Dew et al. (2010) state, ‘the delivery of a diagnosis and a treatment plan is an interactionally complex matter that does not lend itself to the rigid following of a protocol’. Technologies that privilege ‘hard’, coded data may therefore be ineffective in influencing ‘soft’, value-laden decision-making processes.

Discussion and conclusion

This exploratory study of Australian general practitioners' attitudes to a novel electronic decision support tool for cardiovascular disease risk management provides some insights into how EDS tools might function in the primary care workplace. Our findings suggest that EDS occupies an ill-defined middle ground in a dialectical relationship between the routine care provided by GPs and a range of interrelated actors promoting best practice.

EDS tools have the potential to offer practical support to GPs for the more efficient conduct of routine care. The concept of street-level bureaucrats is helpful in understanding why care may be perceived to be enhanced. Lipsky (1980: xii) wrote: ‘The decisions of street-level bureaucrats, the routines they establish, and the devices they invent to cope with uncertainty and work pressures, effectively become the public policies they carry out’. Factors that influence the adoption of particular policies include consistency with personal vision and professional values, harmony with local practices, adequate resources and lack of competing priorities. There were several accounts in this study where GPs saw the potential for the EDS tool to support their public policies, especially through the expeditious processing of workloads. This helps explain the findings from meta-analyses of trial data in which issues such as incorporation of tools in routine work flow and availability at the time and the location of decision-making are associated with improved practitioner performance (Kawamoto et al. 2005).

Although we did not analyse the real-time use of the tool the interview accounts also indicated that GPs fashioned the tool in particular ways to make it relevant in the clinical encounter. Heath et al. (2003) suggest that technological tools are artefacts ‘made at home’ in the workplace and necessarily undergo a process of transformation according to a tacit body of practice and reasoning. Participating GPs talked of using the tool as a device to punctuate particular discussions, particularly focusing on risks of developing CVD. In this pilot the tool output came pre-prepared for use in the healthcare encounter. The future, practice-ready version of the tool will allow for real-time changing of input data during the consultation. This may alter how GPs and patients use the tool to direct the timing and flow of the clinical encounter when compared with the pilot version. Greatbatch et al. (2001) found that the computer creates structured conversational boundaries by focusing patients’ and doctors’ gaze on the screen and directing conversations via the populating of particular screen fields. We found GPs alluded to such conversational boundaries, particularly when using the tool’s risk score and colour category to explain the patient’s current and projected state of health. Consistent with the notion of the tool as a boundary object (Star and Griesemer 1989), it may be deployed and interpreted differently by the various actors that are engaging with it. Importantly, it does not necessarily privilege GP agency over that of the patient. In this study there was a striking account where a patient initiated treatment decisions based on the tool outputs, thus altering the GP’s preferred management recommendations.

Alongside the tool’s ability to support and influence routine care, we identified competing elements that can render EDS tools inconsequential. The theory of phronesis helps to appreciate the role of soft knowledge construction when GPs make decisions in practice. The notion of time as an important factor in decision-making was emphasised in this study, particularly the way in which decisions are distributed over a series of encounters between GPs and their patients (Rapley 2008). These encounters constitute a narrative that is intuitively ‘read’ by doctors (Greenhalgh and Hurwitz 1999). This narrative is not linear. The journey made between doctor and patient is marked by ‘advances and reversals, vectored progress and cyclic repetition, bursts of change and lulls of sameness’ (Charon 2000: 64). Greenhalgh and Hurwitz (1999) highlight four distinct texts that are operating in this journey: the experiential text of the patient’s life outside the healthcare encounter, the narrative text constructed by the doctor about the patient’s illness, the physical or perceptual text derived from physical examination and the instrumental text derived from tests and machines. Decision support, when derived primarily from the latter two texts, can be of secondary interest. The development of patient-specific interfaces and access to data outside of the healthcare encounter may, therefore, be effective strategies to better incorporate these broader narrative and experiential texts.

In addition to the tool’s potential to support routine care was a range of views on how it might represent a vehicle for best practice. We found that decision support tools interact with a variety of other agents of best practice, particularly EBM, guidelines and professional organisations. The participating GPs viewed best practice as a fluid process where evidence is scrutinised by trusted experts, distilled into guidelines and disseminated through trusted networks. Gabbay and le May’s (2004) conception of a non-linear incorporation of EBM along tacit mindline networks appears germane. EDS tools that are inconsistent with these mindlines may be viewed as irrelevant and potentially even untrustworthy.

Rather than a form of technological determinism (Timmermans and Berg 2003), we found that GPs actively engaged with this tool, being neither subsumed under its force nor inert to its potential influences. Most GPs did not perceive the tool as challenging their epistemological position in any meaningful way. Of note, however: the tool’s authoritarian agency was salient for two participants, one of whom expressed fear of the potential use of such tools to rate performance. It is conceivable that as these tools become organisationally embedded and promoted through national programmes their potential to influence the agency of the GP could grow. Ethnographic studies examining the implementation of the UK’s Quality and Outcomes Framework pay-for-performance programme have found that the collection of biomedical data is being prioritised over other dimensions of healthcare quality in order to meet indicator targets (Checkland et al. 2007, Grant et al. 2009). While pay-for-performance programmes may be viewed suspiciously, peak professional bodies, by contrast, appear to be perceived as trusted technogovernors. Endorsement and dissemination of tools by these organisations might be expected to enhance their credibility and diminish any sense of authoritarian agency. However, from Gabbay’s work, we see that the energising of informal networks, particularly personal accounts from other professional colleagues, is likely to be an influential factor in uptake.

Our study sample clearly has bearing on the findings presented here and their implications. The participating GPs were highly trained and generally demonstrated favourable attitudes to the role of information technology in general practice. Resistance to uptake of decision support tools may be greater among GPs who are less research- and education-oriented and less embracing of technological innovation. Alternatively, GPs with less interest in this area may be less likely to question the veracity of recommendations and be more reliant on such tools. Post hoc interviews, using a peer-to-peer method of data collection, afforded us some advantages by engaging GPs in collegial conversations but this may have predisposed GPs to reflect their ideals rather than their practice. Further, our interpretation of the tool’s influence on communication might have been different if we were able to witness actual healthcare encounters. Moreover, had we explored patients’ views different interpretations of the key themes may well have arisen. For example, while we found little difference in opinions for GPs working in Aboriginal health settings compared with those in mainstream general practice, a patient-focused enquiry might have produced quite different findings in these two settings.

Implementation in a broader range of primary care settings, involvement of the patient perspective and the use of ethnomethodology are key areas to focus on for future work in this area. In the next phase of this project the tool will be integrated into primary care software systems and there will be a specific patient interface for use both within and outside the healthcare encounter. It will be trialled over a 12-month period in a larger number of settings to those studied here. This will allow us to enlist a combination of methods, including ethnography, to understand in more detail whether and how these tools become embedded both in the workplace and the community.

The context of general practice has shifted from the surgery, in which discrete illness episodes are treated, to a complex environment where health, illness and risks are more diffusely located between the consultation room and the community (Armstrong 2004). The application of EDS tools is becoming equally complex. Other human actors (such as specialists, non-GP health professionals, lay peer supporters and families), non-human technological actors (for example, personal e-health records, mobile phones, cloud computing, personal and shared e-health records) and technogovernors (for example, e-health regulators, professional standards bodies and pay-for-performance programmes) are all shaping healthcare in new and potentially profound ways. It will be important to follow these actors carefully as our current appreciation of their roles and influence is limited. While it is apparent that EDS tools and related technologies have the potential to improve health system performance, our gaze should not be limited to only this aspect of healthcare quality. It is only through appreciation of the socio-technical dimensions of these tools that we will be able to understand their broader implications.


We thank the general practitioners and the three AMSs and their governing bodies for generously agreeing to participate in the study. The study was funded by a Pfizer Cardiovascular Lipid grant and an Australian National Health and Medical Research Council (NHMRC) Development Grant. David Peiris was supported by a scholarship from the New South Wales Clinical Excellence Commission and is now supported by a NHMRC Translating Research into Practice Fellowship. Alan Cass and Anushka Patel are supported by Senior Research Fellowships from the NHMRC.


Part 1: General overview of the EDS tool

The aim of the EDS tool is to assist GPs through the provision of decision support for the management of cardiovascular risk. I’d like to start by talking about your personal experience of the EDS tool, and then go on to ask your views about its applicability in general practice more generally.

  • a. Overall, what do you think was the impact of the EDS tool on the quality of care you were able to provide for your patients?
  • b. How useful was the EDS tool in supporting communication with your patients?
  • c. How effective was the EDS tool in assisting you to practise according to national guidelines for cardiovascular risk management?

Part 2: The EDS output

I’d now like to show you some sample printouts from patients enrolled in the study at your practice.

  • a. What did you find useful about the EDS printout?
  • b. What information was not helpful in the EDS printout?
  • c. Was there anything confusing about the printout?
  • d. How could we improve the printout?

Part 3: Implementation of the EDS in general practice

Our future plans are to integrate the EDS into the commonly used medical software in general practice.

  • a. If we do this, what do you see as its potential benefits?
  • b. Would you anticipate any disadvantages?
  • c. What barriers do you think we would face? [Probe: practical/ technical/ other]
  • d. How do you think we could improve it?
  • e. Would you personally consider using it in your practice? [Probe: why/ why not?]

Part 4: Wrap up

  • a. Are there any other issues not covered that you would like to talk about?