In the remainder of this paper we will draw primarily on ethnographic observations from Clover surgery - which we identify as a ‘pro-technology’ surgery - to reflect on the relationships between use of the EPR, organisational practices of coding and summarising patient records and notions of ‘quality’ in health care. The term ‘backstage’ from Goffman [
34] is used to describe settings that are not directly ‘patient-facing’, where work
with and about patients (including their representations in texts and records) does not also involve
interacting with patients directly. We adopt a socio-technical perspective, recognising that one cannot meaningfully study technologies in isolation from the social situations in which they are used (or not used, or worked around) and that technologies shape – and are to some extent shaped by – human action [
19,
20,
25].
Care and ‘aftercare’ for the EPR: summarising the electronic patient record
At Clover surgery, the summarisation and Read coding of patient records was a high priority activity. Although some coding was done in clinical consultations (especially by nurses doing chronic disease management reviews) [
24], much of it occurred in the backstage regions of the surgery and was done by administrators, with guidance from doctors. Four members of administrative staff were involved, and for two of them it was their main administrative role. There was a fourteen page practice protocol on summarising medical records which was in its second version. It opened with the words:
AIM: The aim of summarising patients’ notes is to be able to easily access the past and present medical history of the patient via the computer screen. Using the protocol, information from patients’ notes is entered on screen using agreed Read codes. This enables future accurate auditing to be undertaken [Reproduced from practice protocol].
The summarisation protocol went on to explain the procedure for sorting contents of the medical record and creating the summary; a list of the types of information which should be added; guidance on how to categorise summary data (as ‘active’ or ‘significant’); three pages of “Medical Problems and Read codes” (taken from a document supplied by the local Primary Care Trust) and a page of “Common Abbreviations”. The aim of summarising records was thus described in technological and institutional terms - the ease with which information can be accessed from the computer screen, with a particular reference to enabling institutional audit. The individual patient’s history is juxtaposed with the needs of the institution for ‘auditable’ data. The implicit assumption regarding ‘accuracy’ of this account is something we had observed in detailed micro-analysis of clinical consultations (published elsewhere) there being a tendency for the EPR to become the authoritative source of the patient’s past medical history, even when the patient was co-present, sometimes offering alternative versions of events [
36].
Summarising patients’ records was regarded by administrators as responsible work demanding concentration. One of the summarisers liked to get into the office at 7:30 a.m. so that she could start summarising while the office was quiet and relatively free from distractions. Frequent reference was made (by administrators, manager, information manager and clinicians) to the amount of time and ‘care’ that went into this task. The IM + T lead explained that other surgeries did not take as much care over coding and summarising as his own staff, and one of the GPs (who was relaying concerns around the implementation of the new electronic system for transfer of records) highlighted their close attention to detail: “They [the enthusiasts for electronic transfer of records] underestimate the work that goes into record keeping…they think it just happens in the consultation but you only need to look at how much care goes into the records, by people like [name of summariser]”. In the words of the retiring secretary “At one time there was no need for all those admin people – but now there is a whole room of them – all because of electronic records”.
Care was needed not only in creating the record but also in maintaining it - ‘aftercare’. Various terms were used for this activity: “cleaning”, “feeding”, “cleansing”, “tidying” and even “computer toilet”. In Beech surgery, one doctor often went in early for morning surgery to spend time “cleaning up” the records of patients he was about to see, removing “clutter” and, if possible, reducing the number of ‘problems’ listed on the summary screen (“I can’t stand it when there are 24 active problems showing – I’ll tend to clean it up”). A “cluttered” or “clogged up” summary screen was regarded by clinical and non-clinical staff in both surgeries as something to be avoided if possible, there appearing to be an informally shared understanding that there is a limit to the number of ‘problems’ it is reasonable for a patient to have documented, a limit which related in part to the material constraints of the technology around what could be viewed conveniently on a summary screen at one time. In one consultation we observed at Clover, an elderly man returned to his GP following an X-ray of his hands which had confirmed osteoarthritis (OA) as the cause of his aching thumbs and wrists. He had been prescribed anti-inflammatory medication and was enquiring about other treatment possibilities. After the patient left the GP looked at the patient’s summary screen and commented: “No one’s put OA here as a problem… I don’t think I’m going to put it in…I just think he’s got a lot of diagnoses already” suggesting he had reached this ill-defined limit. The diagnosis was entered as free text, but not Read coded or afforded the status of ‘Problem’ on the summary screen, making it more difficult to find in future consultations (and ‘invisible’ to any audit process).
In a specific example of ‘aftercare’ of the EPR, a doctor at Clover circulated an email to all staff explaining he was manually editing over 200 patients’ EPRs. Population Manager (software integral to the EMIS clinical system which searches daily for QOF-relevant Read codes and identifies patients with ‘missing’ data items) was identifying these patients as ‘severely mentally ill and needing reviews’ (a QOF requirement) although it had become clear to him (on closer inspection of these patients’ EPRs) that this was because of the abundance of Read codes for ‘recurrent depression’ and ‘endogenous depression’. Although these codes may have been ‘fit for purpose’ at data entry, they were now being captured in automatic daily audits of the practice population for the purposes of a QOF target intended only for those with ‘severe and enduring mental illness’ (e.g. psychoses). This additional work of editing the EPRs ensured the practice was not penalised financially for failing to offer detailed health checks to patients whose clinicians recognised did not fulfil the criterion ‘severe enduring’ mental illness.
That the process of creating and maintaining summaries was resource intensive was well recognised and was specifically highlighted in one of Clover’s newsletters for patients, at the end of a section detailing areas of expenditure in the practice. It read:
Of course our main costs in keeping up to date are the employment of staff in updating our records and summarising our notes [Reproduced from Clover Newsletter].
Not only does this draw attention to the financial costs of summarising records, but it also constructs this activity as centrally important to ‘keeping up to date’ and an inevitable part of the modern GP surgery.
In Additional file
1: Box 1 we present an extract of ethnographic field notes to consider the ‘care’ that goes into summarising a patient’s record.
To get her job done (Additional file
1: Box 1) Amy drew on her knowledge of a wide range of formal and informal resources. They included texts (e.g. notebook, dictionary, Internet, different parts of the patient’s existing records, summarising protocol, her ‘memory’ cards); specific technologies (Read codes, electronic messaging system); current health policy (QOF) and locally accepted procedures; the expertises of her coding colleague and a GP. She mapped her understanding of what Mr Oliver had ‘
wrong with him’ against local practice policy about how problems should be categorised, her understanding of QOF, and unofficial norms about whom to seek help from (and in what order). Drawing on her working experience of the Read code formulary and her understanding of the relative importance of particular diagnoses she made judgements regarding what constitutes a reasonable amount of time and effort to spend seeking out particular codes and resolving particular ‘diagnostic’ dilemmas. For example, she assigned greater importance to resolving the conundrum around this patient’s possible atrial fibrillation/postural hypotension than to fine tuning the coding of the patient’s testicular operation. She also made personal notes (‘memory’ cards) about Read coding as part of an ongoing process of reflective learning.
Amy explained that she liked to summarise records as she would like her own record summarised. Another said that she felt strongly that when she was coding and summarising records she was working for the patients, whilst recognising that ‘officially’ she was working for the GPs. When summaries were received from other surgeries, they would start the process of summarising again, perceiving that summaries generated elsewhere were poorer in quality, missing important items and generally not to be trusted (“there are summaries and there are summaries”). The recent decision to start using GP2GP software for transfer of electronic records between practices had done nothing to change this. Summarisers continued to ‘start over’ with summarising, placing higher value on their own summarising judgements than those of an unknown (and anonymous) coder from a distant practice.
Local ownership of the summarising process was important to the summarisers’ sense of professional identity. They took immense pride in identifying items of history that previous summarisers had missed, ‘improving’ summaries by extending them and annotating existing Read codes with qualifying free text. In the administrative areas of the practice, the patient’s data and information in the EPR is the substrate from which administrators can carefully and creatively ‘mould the EPR into shape’ for its new organisational context. Through this process of building, extending and ‘improving’ the EPR, administrators contributed to the construction and redefinition of this new information context, building their own identities as ‘good’ summarisers as they went along.
The sense of being informally accountable to patients and officially accountable to GPs and the wider institution is something that we have also reported in the work of GP receptionists [
37] and was sometimes a source of tension. Although in their role as summarisers they didn’t meet patients in person, working only with the patients’ records (Robinson has coined the term ‘
patient inscribed’) [
38], administrators were constantly making balancing judgements about the role of the EPR in supporting individual patient care (‘
patient as person’), alongside broader institutional demands for record keeping. They had built a strong professional sense of working
for the patient within these institutional constraints. For the summarisers, a ‘good’ summary was one which was extensive, thorough and complete, in which
nothing had been missed out. The GPs, by contrast, had more modest ambitions. Administrators suspected (rightly) that some of the GPs felt they spent too much time on their summaries, producing summaries which were too detailed, when something more ‘basic’ would do. This inter-professional contest over what constituted a high quality summary was one we witnessed on numerous occasions. Even within summarisers’ own practice there was a constant tension between framing the practice of Read coding as highly responsible work (“
I think it’s really serious. I mean if I get it wrong it could have serious consequences”) and a recognition that there are limits to how long one can spend searching for a suitable Read code that may or may not exist (“
that’ll have to do”).
Until just before our fieldwork began, the GPs’ role in summarisation at Clover Surgery had been limited to making judgements about whether and how to Read code ‘new’ diagnoses which were relevant to the QOF (with implications for targets, QOF-related workload and practice funding). As much QOF performance was based on activity within the previous 15 months, this determined the definition of ‘new’ diagnoses. Summarisers would alert GPs to these potential diagnoses as they came across them, cognisant of the fact that some diagnoses (e.g. atrial fibrillation or diabetes), were more consequential than others (such as multiple sclerosis or osteoarthritis) in institutional terms, thus keeping the ‘institutional’ version of the patient in mind. Once certain Read coded diagnoses were entered into the EPR the patient became part of the ‘denominator population’ for a range of QOF targets, with commensurate demands on the organisation to meet these (or risk forfeiting financial incentives). Institutionally, higher stakes attached to these diagnoses than to other diagnoses which might be comparable, from the patient’s perspective, in terms of symptoms or impact on daily life. Accordingly, different levels of ‘care’ (delivered by different professionals) attached to different entries in the patient’s EPR in the ‘backstage’ regions of the practice, some entries (or potential entries) attracting greater scrutiny than others. This careful attention to codes (or diagnoses) which aligned with institutional goals (e.g. high QOF achievement) is something we also observed in ‘front stage’ regions of the clinic, such as nurse-led chronic disease consultations [
24]. In the next section we consider the additional socio-cultural work that gets done as the inter-professional contest over what constituted a ‘quality’ summary played out in one particular aspect of the summarising routine in Clover surgery.
The constitution of professional hierarchies and local accountabilities
In recent months, GPs at Clover had started assisting more closely with the summarising process, the aim being that Clover would achieve a ‘higher percentage of notes summarised’ (itself a QOF target, and also a requirement of GP training practices). This highlighted some differences in the approaches taken by administrators and doctors and provided an opportunity for some interesting accountability work to be enacted.
One of the summarisers had distilled Clover’s official fourteen page summarisation protocol into a simpler one page document which was given to each GP, called “
Summarising of Patients’ notes – a short overview of what we currently do!” Additional file
1: Box 2 shows some sections of this document (the italics are our own). Although introduced as a ‘
short overview of what we currently do’ and incorporating a detailed list of what ‘
we include’ the document is also replete with requests to doctors to do things in certain ways e.g. ‘
For each diagnosis please write the exact date of the 1st onset (not just the year please!’).
An understanding of this document (Additional file
1: Box 2) requires some understanding of the organisational context in which it was developed. The specific reference to ‘chickenpox’ - both in Amy’s description of her summarising role (Additional file
1: Box 1) and (again) in this short document (Additional file
1: Box 2) - is a legacy of previous organisational history. The summarisers
used to record chickenpox as a ‘Minor’ problem, but when doctors had advised them to include
only diagnoses which they regarded as ‘Significant’ in their summaries, the summarisers had responded by suggesting to doctors that in certain
particular circumstances (e.g. pregnancy) chickenpox might indeed be significant. An agreement had therefore been reached that chickenpox be included in future summaries as a ‘Significant’ problem and it had been re-categorised as ‘Significant’ ever since. Its appearance in this document is not simply an ‘
overview of what summarisers do’ but a reference to a small triumph of administrators over doctors in the organisation’s recent past.
In practice, a number of problems which summarisers might previously have defined as ‘Minor’ problems they now defined as ‘Significant’. This meant they could satisfy the requirement of the official summarisation protocol (which included: “It is vitally important that nothing is missed”), exercise their wish to be thorough, and also satisfy (by means of a workaround) the doctors’ request that problems which are important enough to be in the Summary should be ‘Significant’ ones. One of the summarisers justified this workaround further by pointing out that when doctors do home visits and take a ‘summary printout’ with them (a paper summary of the patient’s EPR), this printout lists only ‘significant’ problems, not ‘minor’ problems. Classifying problems as ‘minor’ might, she said, risk compromising care for patients in this situation. This is further evidence of the tension experienced by administrators between their formal accountability to GPs and the institution, and a sense of informal accountability to patients.
As part of the recent drive to get more summaries completed, one of the GPs had developed a paper form for his GP colleagues to complete as they selected items from the patient’s record for the Summary. There were separate sections for smears, past medical history (Significant Active and Significant Past), allergies, and immunisations. The section for ‘Immunisations’ said, in brackets alongside “(
if time permits, without this paper records have to be dug out if patient enquires)
”. This qualifying note suggests that ‘Immunisations’ were not only regarded by this GP as low priority compared to other parts of the record, but constructs meanings around the relative value of time for different members of staff. It is unlikely that lack of time would be a legitimate reason for an administrator to omit immunisations from a patient’s EPR. Indeed, administrators often entered details of all immunisations into the EPR first, before tackling other aspects, painstakingly copying vaccine batch numbers and entering codes for ‘place of procedure’ alongside, meticulous in their efforts to ensure complete immunisation records. Arguably an incomplete immunisations record would be more troublesome to nurses in their daily work than to GPs (who rarely give immunisations outside of the annual influenza campaign and rely on nurses to run the travel clinic). The labour intensive work of ‘
digging out’ patients’ notes in the middle of a busy travel clinic is work that would likely fall to nurses and receptionists. This was one of several examples we encountered of different staff groups having different intentions and assumptions shaping what constitutes the summarised record [
39].
The paper form that was part of the new summarising routine for GPs was an intermediate document. Here, GPs identified what they wanted to include in the summary, without investing any time identifying specific Read codes to capture the concept. This work of matching items on the list to appropriate Read codes was passed back to the administrators, in a move which may suggest that the GPs considered that the most important judgements lay in the selection of items for summarisation, rather than in the Read coding itself.
Amy gave each GP five sets of notes for summarising per week, keeping a record of who had been sent which notes on an Excel spreadsheet. GPs returned their completed forms to the summarisers, listing the items they wished to be entered into the patient’s EPR summary. The summarisers entered these items into the EPR one by one, choosing appropriate Read codes. Not all GPs had kept up with this workload, and whilst administrators said they were delighted that the GPs were helping them out, they were uncomfortable about the way the process was unfolding. There was concern that their protocol (Additional file
1: Box 2) was not being followed and that the GPs’ summaries were not sufficiently detailed. Some diagnoses were not being included and records of immunisations and cervical smears were sometimes incomplete (it is of note that the work of cervical screening is done almost entirely by nurses, who do the smears, and administrators who manage registration and recall). However, the administrators had (reluctantly) agreed that they would not do any further checking against the original medical notes (which would incur the very time penalties that this division of labour was supposed to address). They would simply enter Read codes for the selected items as the GPs requested. This was a source of significant tension for summarisers.
An integral part of the summarising routine was that a Read code (“notes summary on computer”) was entered into the patient’s EPR when a summary was completed. This ‘meta-data’ or ‘data about data’ was a QOF requirement and itself subject to regular audit. The administrators had discussed their concerns over different standards of summarisation with the practice manager, and as a result they had identified a different Read code (“Lloyd George
a
and Problem Summary”) which would be understood (locally) to mean that a doctor (rather than an administrator) had done the summary (this subtlety of meaning would be lost on transfer of records beyond the surgery). Administrators felt that this would cover them in the event of any queries, and for completeness they added a free text entry of the specific GP’s initials next to the Read code. The summarisers had constructed a particular notion of summarising which they cherished and which constituted their own ‘gold standard’. In this instance the EPR was being used resourcefully by administrators to facilitate surveillance of their employers (the GPs) in what seemed like a curious reversal of the usual lines of accountability.
The related organisational routines around Read coding the incoming post provided similar opportunity for contests around competing lines of accountability to be performed. Discharge letters and reports of outpatient clinic attendance received at Clover were scanned by a receptionist to produce an electronic document for circulation first to a GP and then to a coder using a document management system which was integrated with the EMIS system. The GPs read and highlighted the document electronically and decided whether any specific actions were needed, identifying these in comment boxes. Items for Read coding were highlighted in grey; a yellow highlighting function was used to make certain parts of the letters more visible to future readers. As with the summarising protocol, coders were to categorise Read codes as ‘Significant’. One GP liked to do most of his Read coding himself; another occasionally did. Most doctors relied on the coders.
Coding involved three broad dimensions of increasing complexity. The technical demands of EMIS and the document management system were relatively straightforward. Letters were re-directed, notes added, and workflows ‘terminated’ once letters were coded. Selecting Read codes which matched the grey highlighting was more troublesome and fraught with the same challenges that the summarisers encountered. Most difficult of all was managing the social complexities of this task, in a (virtual) environment in which each GP had their own preferred ways of working. A short session of coding could generate many queries to resolve (Additional file
1: Box 3).
An extract from DS’ reflective journal written just after this observation read:
I was struck by the balancing act that I had been observing – by the coder’s clear sense that she was serving both patient and doctor and always trying to gauge their interests. I realised that coding a record is not the unproblematic technical task which it is so often assumed to be, but a highly social phenomenon and one which involves interpretation and judgement at so many levels. Deciphering poor handwriting; contradictory entries in notes; diagnoses that have no Read codes at all; Read codes which seem ‘not quite right’ for the particular problem. But most of all I realised how difficult it was to make those moral judgements about whether to act on (or quietly ignore) concerns that coding may not be perfect (but can it ever be?), whether and how to craft those notes to doctors, and how to gauge how different personalities in different and particular circumstances might react to receiving such notes. As she said, all the doctors are different and do things in different ways. Mastering the coding task was much less about coding and computers and so much more about managing relationships than I might ever have imagined. How a patient’s record is coded is not only (or even mainly) about ‘capturing’ and representing specific diagnoses as bytes of data, but is a product of complex and nuanced interactions between clinicians and administrators shaped not only by the ‘facts’ of the case, but the ongoing relationships which are co-constructed alongside the ‘problem list’. [Extract from reflective journal]
Once medical judgements are no longer the unique province of the doctor, other staff members such as coders and summarisers have responsibilities which are not always socially recognised in the hierarchy of the surgery. On the one hand the GP’s authority is undermined by the potential for work to become more distributed. The EPR’s wide ‘organisational reach’ and its ready openness to surveillance by other members of the practice – such as administrators – opens up scope for the medical judgements of doctors to be scrutinised (and criticised) by them [
40]. Conversely the GP’s authority within the social environment of the practice is carefully maintained (see Additional file
1: Box 3). This is in part constituted through the social actions of GPs, but more importantly it arises from the reciprocity of social actions and interactions between GPs and administrators which serves to maintain the social hierarchy. Amongst the many interpretive judgements that Linda had to make, the trickiest part of the coding role was around how to ‘act’ in situations where ‘no action’ was recommended by the GPs. ‘
No action’ never meant that no action was taken by Linda. If anything, it was when ‘no action’ was recommended by doctors that the coding task became most perplexing. It was in these circumstances that different perspectives on what constituted a high quality summary came to the foreground and her moral sense of informal accountability towards patients jostled (and often jarred) with formal accountabilities towards the doctors and the organisation [
37]. Adding to this complexity, the social negotiations which ensued were carried out primarily within virtual networks through written text, and remained visible to anyone who chose to study the ‘audit’ trail at a later date. The coder performed her identity as a competent, conscientious, caring worker trying to meet her informal obligation to patients. At the same time she was also creative in finding ways of respectfully engaging the GPs in a new kind of exchange where professional hierarchies and local accountabilities were being constantly renegotiated and notions of what constitutes high quality Read coding were being refined and revisited in every exchange. Linda was widely acknowledged to be particularly good at her job by GPs and management alike.