CHAT as an analytical lens

CHAT has been used extensively to study workplace learning. In the medical education literature, a growing number of studies have applied CHAT to research into medical student and junior doctors’ learning.26–30 In activity-theoretical studies, the analytical focus is on multiple organisational activity systems, in other words, professional groups with their specific objects of activity, artefacts/tools, rules, communities and the division of labour between these.31 This perspective is particularly helpful when considering prescribing processes within hospitals, which are large, multifaceted and complex organisations, and where the objects, rules, community and division of labour (see online appendix 1) are often profession specific, unclear, implicit and/or fluctuating.14 29 Activity systems (eg, of doctors, pharmacists and nurses) within organisations, such as hospitals, are interdependent and at the same time in tension-laden relationships with each other, generating disturbances: A disturbance stands for deviations from the normal scripted course of events in the work process, normal being defined by plans, explicit rules and instructions, or tacitly assumed traditions. It may occur between people and their instruments (eg, care pathway guidelines), or between two or more people. Disturbances can appear in the form of a tension, obstacle, difficulty, failure, disagreement or conflict.32

In our study, the concept of disturbance will be used to explore antibiotic prescribing processes, presented as a patient’s antibiotic pathway in a hospital setting. The disturbances in care processes may hinder holistic management of patient care.29 However, instead of being viewed as error-causing phenomena, we view disturbances as an inherent feature of work processes and as drivers for change and development.33 34

The CHAT based concept of object,33–35 with its special focus on the ‘misalignment’ of multiple, often competing objects in organisations,29 can widen our understanding of why disturbances take place in antibiotic prescribing processes. For example, for a junior doctor, the specific object may be learning to conduct a prescribing-relevant task, such as writing up a drug chart while simultaneously developing as a physician, and trying to please more senior doctors in the medical hierarchy.20 The flexible aligning of different, competing objects calls for the collective reflection, negotiation and reconceptualisation of the object to enhance collaboration in the provision of patient care.29

Objectives

Our aim was to explore doctors in training experiences of common problems in the antibiotic prescribing process using CHAT. Our research questions were as follows: What are the intended stages in the antibiotic prescribing process? What are the challenges and where in the prescribing process do these occur?

Study context

Our study involved two ‘typical’ National Health   Service (NHS) hospitals, one in England and one in Scotland. Both hospitals had an overarching antibiotic policy aligned to national guidance,36 but specific treatment guidelines vary by medical specialty as a result of differing types and sites of infection. The qualitative data analysed were narratives from 33 junior doctors in their first or second year after graduating from medical school, at a stage of training called foundation year (FY) 1 or 2 in the UK. Prior to graduation, as medical students, they would have received teaching sessions, learning opportunities and assessments relating to diagnosing and managing patients with infections, including choice of antibiotics and their underpinning modes of action. However, as undergraduates, they would not have been able to prescribe antibiotics independently.

Data sources

First, in May 2017, using PubMed, Google Scholar and Google, we conducted a brief search on antibiotic prescribing, with no date restrictions to try to locate any existing antibiotic prescribing ‘process maps’. Second, we analysed existing narrative interview data1 that explored the antibiotic prescribing experiences of doctors in training overseeing the care of patients with infections in two NHS hospitals. The 33 doctors in training were interviewed once, and they engaged in 20 interviews (14 individual and 6 group interviews), arranged according to individual preferences and availability. Participants were recruited by email via deanery circulation lists, through posters and verbal presentations at educational sessions and through a snowballing approach.

Interviews were semistructured and began with an orienting question: ‘What is your understanding of antimicrobial prescribing?’ Following this, a narrative interviewing approach was used to explore FY doctors’ prescribing experiences. Participants were free to choose their own examples to discuss as a result of the narrative interview approach. Note that we did not ask questions that related to particular parts of the prescribing process. No questions were asked specifically about disturbances to the prescribing process. At the end of the interviews, the interviewer asked about the educational needs of FY doctors during their transition to clinical practice.

Consent was obtained from all the interviewees and data were anonymised. Interviews usually took place at the medical schools and ranged from 26 to 82 min (mean 58 min). The 20 interviews (approximately 19 hours 20 min of audio data) were transcribed verbatim. Further information relating to these interviews can be found in the primary publication,1 which was focused on the individual prescribing experiences of FY doctors, the current paper using CHAT as a way of analysing system-level disturbances in the prescribing process. Research ethics approval was granted by both medical schools.

Patient and public involvement

Given the focus on junior doctors’ prescribing experiences for the original study on which this secondary analysis was based, patients and the public were not involved in the design, data collection or data analysis.

Developing the antibiotic prescribing process model

The antibiotic prescribing process maps that we discovered from brief searches using PubMed, Google Scholar and Google images did not fulfil our requirement for a process model depicting the stages of antibiotic prescribing expected to occur in a typical NHS hospital. Consequently, we started to develop our own model informed by national antibiotic prescribing guidance36 37 and first-hand practice and research experience (eg, author 3 is an NHS antibiotic pharmacist). We then showed our model to hospital healthcare workers (two doctors and one pharmacist; see Acknowledgements) and two medical educators (authors of this paper) who agreed that the model reflected the antimicrobial prescribing process as it is intended in a typical NHS hospital. We then commissioned an artist to present the model visually (presented in figure 1). This process model formed the starting point for data analysis, but our understanding of the process continued to develop throughout the analytic process.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Model of the intended antibiotic prescribing process in UK hospitals. Dark blue arrows show the patient journey. Other arrow and box border colours relate to the professionals most commonly providing the input: green, pharmacy; purple, microbiology; red, nursing; pink, microbiologists, pharmacists and nurses. Black box borders represent the patient. Pale blue arrows represent ideal practice.  

Mapping of the narratives

We first identified 173 narratives relating to antibiotic prescribing within the 20 transcripts, where events were recounted as stories. Most of the narratives related to times when doctors in training narrated specific events, which we refer to as personal incident narratives (PINs); a minority related to generalised incident narratives (GINs), when multiple events tended to blur into a more generalised story. Data sitting outside the narratives were excluded since they typically did not relate to participants’ actual experiences of prescribing. Each narrative was then mapped to one or more stages in the process model. Individual narratives varied substantially in their length, richness and degree of focus on a particular stage in the process model, with some narratives referring only to one stage and others referring to multiple stages. The presence or the absence of reference to each model stage for each narrative was coded in an Excel spreadsheet. Two of the authors of this paper who were part of the previous study1 undertook the analysis to protect participant confidentiality and to ensure a deep understanding of the research context.38 39 Although two hospitals were under study, we did not look for differences between the two since previous work concluded that there was more variation between wards within a single hospital than between hospitals.1

Identifying disturbances in the process

Then, drawing from the CHAT concept of disturbance,24 32 we identified deviations (ie, tensions, obstacles, difficulties, failures, disagreements or conflicts) from the intended antibiotic prescribing processes from the doctors in training narratives. We wanted to see when, where, how and why processes ‘as narrated’ differed from the processes portrayed by the model and what types of disturbances took place, given the messy realities of complex practice. As the analysis progressed, we were able to form categories of groups of disturbances. Through discussion among all authors of this paper, consensus was reached around five main categories of disturbance, which can be considered as ‘themes’ that were identified qualitatively, and which partially overlapped and interacted.

Locating disturbances as ’hot spots' on the process model

Having mapped the narratives against the process model and identified five categories of disturbances, we then highlighted parts of the model (figure 1) in orange to emphasise that the deviations from the intended processes are most likely to occur—so called hot spots (presented in figure 2). The hot spots correspond to the process stages that were most frequently narrated and/or those where disturbances were articulated by the junior doctors.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

This is the model of antibiotic prescribing in hospitals but with the particularly problematic components (as identified by junior doctors) shown in orange. These were parts of the process mentioned in ≥25% of the narratives analysed and/or one of the five frequently narrated categories of disturbance.

Process model

The process model developed (see figure 1) depicts, holistically, the multistage antibiotic prescribing process with 31 stages and multiple practitioners in a typical NHS hospital. It starts with the admission to the hospital of a patient with an infection, moving to diagnosis, through antibiotic treatment and review in the hospital, to discharge from the hospital with antibiotics. It highlights the involvement of multiple professional groups, in other words, ‘activity systems’ of junior and senior doctors, pharmacists and nurses. Numerous ‘checks’ are provided by multiple professionals at different stages, both prior to and after antibiotic therapy has been prescribed and administered. Antibiotic treatment should be reviewed after 24–72 hours or as more information comes to light (eg, culture results or a change in the patient’s clinical condition), which provides further complexity. Doctors are the most frequent prescribers. Nurses usually administer medications, providing checks, such as confirming the patient’s allergy status. Both are often involved with specimen retrieval/collection. Clinical pharmacists usually check the appropriateness of the treatment choice based on guidelines and patient parameters, ensure therapy is physically available, and support the discharge process. All professional groups communicate with and monitor the patient, which may prompt a review of therapy. For example, once a patient is eating and drinking, the nurse or the pharmacist may ask for a review of intravenous therapy (as the patient may be ready for oral treatment). Microbiologists may also request a review of treatment (or review it themselves) in the event of positive cultures, particularly from sterile sites (eg, blood). A detailed description of each process stage and the types of healthcare professional typically involved in them are provided as online appendix 2.

Mapping of narratives

Most doctors in training narratives referred to more than one stage in the process model (median of 3 out of 31 possible stages, minimum of 0, maximum of 12, IQR of 2–5). An example narrative is provided in box 1. This narrative was chosen because it is a ‘rich but concise’ narrative that articulates several model stages, as well as two disturbance categories (see below). The percentage of narratives referring to each model stage ranged from 0% to 36%. ‘Considerations and consultation’ (stage 3) was mentioned the most frequently by far, with ‘consultation’ representing three quarters of the mentions for this stage. ‘Prescribing and checks’ (stage 5) was mentioned next most frequently, followed by ‘monitoring response’ (stage 9). Other model stages received little attention in the narratives, for example, tasks undertaken by non-medics, the discharge process and the arrangements for patients’ monitoring after leaving the hospital. We anticipate that model stages referred to frequently are those that were most memorable to our trainee doctor participants, for example, because they were inherently challenging in some way. Conversely, low levels of attention might indicate unproblematic model stages or those of least interest to them, perhaps because they were not involved in these. This could be because participants did not regard these things as part of the prescribing process, or they overlooked their potential role. They may, for example, regard discharging a patient as something separate, even though it involves prescribing an antibiotic for the patient to take at home.

Consultation challenges

The most commonly narrated disturbance was the difficulties in consulting multiple individuals, who were usually more senior doctors (eg, ward-based consultants and consultant microbiologists), and/or information sources prior to prescribing an antibiotic. This was particularly evident in relation to stages 3c and 3d, but to some extent in stages 3a and 3b (shaded orange in figure 2) and frequently related to the medical hierarchy. In the following quote, the junior doctor explains the uncomfortable situation created when ward-based consultants propose a course of action different from their own judgement or that of another person (eg, consultant microbiologist or antibiotic pharmacist):

I think it’s horrifically awkward because as the junior, you are often … stuck in the middle, because, if your consultant is telling you one thing, it’s quite difficult to go against your consultant, especially if it’s going to be bloomin’ obvious that you’ve gone against your consultant, ‘cos they are going to look on the drug chart and see that [they] weren’t getting what they asked for (Female, FY 2 doctor, Location 1, Number 5).

This example depicts how the central role of the senior doctor creates a tension in the attempt to meet the needs of doctors in training developing their expertise during the care process of a patient. In this situation, the tendency is for the opinion of the ward-based consultant, who has ultimate responsibility for the patient and is the direct supervisor for the junior doctor, to prevail. This example demonstrates how a consultation, with its aim of promoting patient-centred prescribing and care, can be challenged by the hierarchy-specific division of labour in which the junior doctor has to follow the tasks and instructions of the senior.

Lack of continuity

A second commonly narrated disturbance was the lack of continuity of care when the patient is transferred (ie, physically moves) from one care provider to the next during the stages of the prescribing process. This was particularly evident in relation to stages 13 and 14 (see figure 2, shaded in orange) and exacerbated by the frequent transfer of patients between wards within the hospitals and shift work. The doctors in training often lacked awareness of what ultimately happened to the patient, whether the antibiotic they prescribed had helped and whether it was the most appropriate choice for the patient. They had to be proactive to find out what happened to patients who moved wards or were discharged from the hospital, and this is challenging in a busy work environment, with long patient care pathways and shift working:

… when you prescribe something and then you don’t necessarily follow the patients up and you don’t receive feedback on whether you prescribed something appropriately or not. I think that’s one of the struggles because the patients come in, you see them, you go home and they’ve gone to the next ward by the following day and you are never really sure if, well I feel I am not sure if I’ve actually prescribed something appropriately sometimes (Female, FY 2 doctor, Location 1, Number 1).

In this example, we witness that as well as lacking patient outcome information, doctors in training do not always receive feedback on their prescribing decision-making process. The quote above also highlights that the provision of care takes place in multiple locations and is fragmented across multiple professional groups, adding to the complexity.

Process variation

The third commonly narrated disturbance was that the prescribing process varied when working ‘out-of-hours’ (ie, at night and at the weekend) or on different wards. This was evident with regard to the prescribing process as a whole (represented as an orange box around the whole of figure 2). This inconsistency created unpredictability, confusion, lack of confidence and fear of making mistakes among the doctors in training. Hierarchies and power relations added to the complexity as help-seeking was often experienced as problematic due to junior doctors’ lower status. In the following quote, we hear how the threshold for prescribing antibiotics was perceived to be different when working out-of-hours. This may be because doctors feel safer giving an antibiotic than not and are responding to a perceived need to err on the side of caution when working with patients or in settings that are less familiar to them, or when support and advice is less available:

I think that overnight, and over the weekends in particular, people [patients] will get given antibiotics more readily. I think that there is generally a thing of safety, it’s better for someone to be on an antibiotic than off them and I’ve seen cases where SHOs [junior doctors] will willy-nilly prescribe antibiotics without necessarily looking directly at the patient (Male, FY 1 doctor, Location 1, Number 2).

This quote highlights the challenges of working out-of-hours and shows how doctors in training perceive prescribing decisions to be different at these times. The otherwise quite strict, profession-specific and hierarchical division of labour seems to become unclear, implicit and/or fluctuating out-of-hours. The example also points to a lack of overall management of the prescribing process across time, such as from the daytime to out-of-hours. The well-framed strategies that are used by distinct activity systems (professional groups), such as the antibiotic guidelines, may not work as flexible tools for depicting the stages.

Challenges of patient handover

The fourth commonly narrated disturbance was the tangible problem of coordination of care. As in the previous category, this disturbance concerned the prescribing process as a whole (see orange box around the whole of figure 2). The most common example of this was difficulties in the handover of care between healthcare professions (eg, the day team to the night team), both within the medical profession and between professions. Such handovers mostly happened verbally, although there were some positive stories of when tools (eg, email handovers) had helped. This category overlapped with the third category already discussed, in that this often related to differences between working in-hours and out-of-hours, but the focus here is different. Disturbances in this category related to poor information flows as expressed in the following quote:

… in the hospital the policy is essentially that if you have got the responsibility of looking up the blood result then you should be prescribing it [the gentamicin] and there were quite a few incidences in my second job when I was on general surgery, well, when I was on the acute surgical receiving ward, where the person on nights didn’t chase up the blood results, and therefore they didn’t write up for the next dose of the gentamicin, and it was only picked up accidentally, or sort of incidentally, or by the pharmacist that they had had one dose and no more doses had been given (Female, FY 1 doctor, Location 2, Number 1).

The quote above highlights the specific challenges for prescribing antibiotics and, in particular, gentamicin and vancomycin, which require therapeutic drug monitoring and dose adjustment based on the levels to avoid toxicity. This example demonstrates an urgent need for healthcare professionals to coordinate their practices to achieve the desired outcomes (relating to handovers in the prescribing activity). Despite some positive stories of functional tools mediating the communication in specific parts of the process, a need for system-level tools that would improve the communication between participants, for example, between doctors working in and out of hours), was depicted.

Partial loss of the common object

The final commonly narrated disturbance was what is called ‘partial loss of object’ in activity theoretical terms, referring also to the misalignment of the objects held by the medical professional treating a patient who is subjected to the prescribing process (see the patient icon shaded in orange in figure 2). In other words, rather than the focus of the prescribing process always being on patient care, the several professional groups involved were often preoccupied with other specific, well-framed goals designated to them within the process, and the patient did not always seem to be the foremost consideration. For example, the following quote depicts the contradiction junior doctors typically said they face in training: while they are highly dependent on supervision and advice from more senior medical professionals involved in the prescribing process, they are often expected to manage the complexity alone, without sufficient support and functional tools. This contradiction diverts doctors in training attention away from the object of the activity, that is, the patient:

Yeah, so obviously I mean F1s [first year junior doctors] by day they are, you know, they, we don’t make a whole lot of independent choices I would say and, really our job is to collect tasks and make sure they are completed and then suddenly at night, everyone disappears, and we’re expected to charge around, and you know display a whole amount of extra knowledge and autonomy. So, it’s an interesting contrast, and it’s, when you first go on nights it’s very daunting, and suddenly you make these decisions, including the decisions about antibiotics, but that’s just how it is that’s the way it goes (Male, FY 1 doctor, Location 1, Number 1).

This quote is a prime example of how, in the multiorganisational field of healthcare, the historically established division of labour and the distinct objects between healthcare professions and medical specialties define the doctors in training specific positions and tasks. It also demonstrates how this junior doctor saw the object of their prescribing activity as primarily about collecting and completing discrete tasks, and trying to avoid mistakes and negative feedback, by carefully following orders given to them by their senior colleagues.

To sum up, all the disturbance categories reported in our study call for interventions to support doctors in training. These could be organisational change efforts to clarify the division of labour and the distinct objects of the professional groups within the organisation.

Methodological strengths and challenges

Our study has several strengths, including the development of the process model, mapping of narratives and in-depth analysis using an activity-theoretical framework.24 32 CHAT allowed us to analyse the disturbances and hot spots throughout the whole prescribing process, viewing them as potential drivers for change and development, an issue rarely highlighted in previous studies. The process model was developed through a literature review and our own knowledge and experience, supplemented with consultations with medical practitioners to increase the credibility and authenticity. Our model, which includes multiple actors and objects, provides a holistic view of the complexity and multidimensionality of the antibiotic prescribing process. Moreover, our findings may inform other medication prescribing processes and their development more broadly.

Our study also has some challenges. It only considers hospitalised patients with infections, so cannot inform prophylactic antibiotic use or prescribing in primary care. Many activity-theoretical studies investigate work processes and workplace interactions by carrying out lengthy participatory observations, together with interviews within the analysed context. Unfortunately, we did not have enough resources to collect new data from multiple participant groups (eg, nurses, pharmacists and patients), and to incorporate observation, we rely on narrative interview data from 33 participants. Furthermore, narratives are accounts of experiences by interviewees that may or may not map onto what actually happened.44 We also acknowledge that CHAT provides only one possible lens to the analysis and the interpretation of the prescribing process.

Implications for policy and practice

From this analysis, communication and information flows seem to be the highest priorities for targeting interventions since they are highlighted through both approaches (ie, frequently narrated stages and themes identified as a category of disturbance). Indeed, previous research has explored ways to improve the situation, for example, through communication strategies and other educational interventions.30 42 45 46

Greater scrutiny of prescribing processes by applying process models such as ours will also be beneficial to future prescribing policy and practice. Indeed, due to the high complexity and multiple types of disturbances involved, the development of the prescribing processes will require continuous, collective efforts over time. We suggest that instead of viewing disturbances as error-causing and harmful phenomena, they need to be seen as important tools towards rethinking and developing care processes. To develop practice, real patient cases (including disturbances) need to be discussed among senior and junior clinicians and other key groups of healthcare professionals and researchers. Our process model can aid in stimulating these discussions.

Within the workplace, doctors in training will still need to navigate disturbances and conflicting views and combine them into a single ‘good’ prescribing decision, underpinned by a clear rationale. More opportunities in medical school to practice these challenging conversations, and seniors articulating clear rationales for decisions would be helpful. The development of a common object (patient-centred care) between different stakeholders should be supported by collective efforts of developing interprofessional teamwork in prescribing processes, as has been suggested by others.7 29 Finally, workplace opportunities for doctors in training to receive real-time feedback on their prescribing decisions, combined with information on patient outcomes, will also be needed for the development of expertise and confidence.

Implications for future research

Our study contributes to research on prescribing processes, to doctors in training antibiotic prescribing experiences and to activity theoretical studies in medical education. Future studies would benefit from direct observation of the antibiotic prescribing process in other healthcare units and countries. From our current research, neither the doctors in training nor the patients appear to have much power to influence the prescribing process. However, once out of hospital, patients are critical decision-makers who decide whether and how to take their medications, and the threshold criteria for seeking further help. It would be useful to understand this posthospital pathway from the patient’s perspective. Future research might also explore how process maps (like ours) used via mobile technologies during real-time prescribing processes could improve communications within and between healthcare teams. Strategies to provide doctors in training with more feedback on their prescribing decisions and patient outcomes, to enhance their subsequent practice, are urgently required.