Should I stay or should I go? How healthcare professionals close encounters with people with dementia in the acute hospital setting.

Around a quarter of hospital beds in the UK are occupied by patients living with dementia (PWD), and communication impairments are common across all types of dementia, often exacerbated by the hospital environment. Unsurprisingly, healthcare professionals (HCPs) report particular challenges in caring for this patient group, whilst trying to recognise and value their personhood as per the underpinning ethos of person-centred care. However, whilst there is a growing body of research that underlines the importance of communication in dementia care, there is far less that actually examines this communication in real time interaction. Suggestions and pointers for good communication do exist, but these do not tend to be empirically derived, and sometimes conflict with empirical findings. This paper focuses on a specific area of interaction which has previously received very little attention: the way in which healthcare encounters are ended or closed. There is potentially a conflict between a pressure to manage a patient as efficiently as possible, and endeavouring to ensure person-centred care and deal with communication difficulties arising from dementia. Using conversation analysis, we examined forty-one video recordings of HCP/PWD interactions collected from an acute inpatient ward. We identify three phenomena around which there were recurring troubles in our dataset: 'open-ended pre-closings', 'mixed messages' and 'non specifics and indeterminate terms'. We conclude that moves towards closing an encounter that appear intuitive to HCPs as competent interactants, and that may represent best practice in other healthcare settings, may in fact serve to confuse a PWD and create difficulties with closings. Our findings underline the importance of examining best practice guidance as it is actually talked into being, using approaches which can unpack the interactional detail involved. They also emphasise the importance of context in the analysis of healthcare delivery, to avoid a 'one size fits all' approach.

Introduction

Around a quarter of hospital beds in the UK are occupied by patients living with dementia (RCPsych 2013). Communication impairments are common across all types of dementia (e.g. RCSLT 2014). Symptoms vary but include word retrieval difficulties, impaired comprehension of spoken or written language, poor maintenance of a topic of conversation, impaired social language e.g. loss of inhibition. In addition, people with dementia have problems of memory, judgement, insight and abstract thought (Waite et al., 2009), which can further exacerbate communication difficulties. The hospital environment can be a disorienting place for people with dementia (PWD) particularly because of unfamiliar staff, routines, and noises (Borbasi et al., 2006). Acute illness with or without added delirium (Glover et al., 2014, Whittamore et al., 2014) may also exacerbate communication problems, as will hearing or visual loss associated with older age.

Hospital staff generally feel unprepared to meet the care needs of patients with dementia (Griffiths et al., 2014). In the Alzheimer's Society Report, Counting the Cost (2009), 89% of nursing staff respondents identified working with patients with dementia as very or quite challenging, with a particular training need being around communication. Ineffective communication risks misunderstanding, failure to address needs, disappointment or distress.

The policy context for dementia services in the UK has been shaped by a drive towards ‘person-centred’ dementia care (Kitwood, 1997), with an aim to recognise and value the ‘personhood’ of PWD, a position endorsed by the National Institute for Health and Clinical Excellence (NICE, 2006). Skilled communication is central to this approach. However, whilst there is a growing body of research that underlines the importance of communication in dementia care, there is far less that actually examines this communication in real time interaction. Studies have tended to ask professionals about communication after the event, or through the use of hypothetical or general questions (e.g. Edberg et al., 2008, Hallberg and Norberg, 1995; Holst et al., 1999). Whilst suggestions and pointers for good communication do exist (e.g. Alzheimer’s Society, 2009, NHS Choices, 2017), these do not tend to be empirically derived, and sometimes conflict with empirical findings; for example the recommendation for slowed speech has not been borne out through research (Small et al., 2003, Tomoeda et al., 1990).

Whilst there are many aspects of communicating with PWD that may present challenges, this paper focuses on a specific area of interaction which has previously received very little attention: the way in which healthcare encounters are ended or closed. There is potentially a conflict between a pressure to manage a patient as efficiently as possible, and endeavouring to ensure person centred dementia care (Kitwood, 1997) and deal with communication difficulties arising from dementia. Examining how professionals conclude interactions with PWD is therefore of real importance for healthcare practice. At the same time, a detailed examination of closings in healthcare interaction has the potential to contribute to the wider field of professional/client communication.

Background

There is limited research on closings specific to healthcare interaction, but early work in conversation analysis (CA) examined how mundane interactions are brought to an end. In their pioneering paper “Opening up Closings”, Schegloff and Sacks (1973) analysed everyday telephone call interactions and described closing an interaction as an ‘achievement’, where both conversational partners must be oriented to the fact the conversation is coming to a close. This may be achieved through pre-closing indicators such as ‘so’ and ‘well’, which can indicate the potential close of a topic, e.g. “Well I'll let you get back to your books” (see Heritage (2015) for an explication of the shift implicative nature of ‘well’). If both parties are oriented to this as the closing of the interaction, it leads to what Schegloff and Sacks call a ‘terminal phase’ in which no further topics of conversation are initiated, concluding with a common closing exchange such as ‘bye-bye’.

Button (1987) describes several tools that are naturally used in conversation as ‘pre-closings’ or ‘moves towards’ closing. These include arrangement-making, and referencing back to items earlier in the conversation, or back to the original reason for the call. These pre-closings project a close to the encounter, but still allow space for new topics to be brought up. Introducing a new topic ‘re-opens’ the closing until all conversation participants orient to a closing phase once again.

What data there are from conversation analytic studies of closings in face-to- face healthcare interactions tend to come from primary care (e.g. Heath, 1986, Robinson, 2001, West, 2006). The context of primary care interactions is that it is generally the patient who has identified a problem and usually voluntarily entered into the physician's space for an appointment. The format of a primary care interaction is relatively well-defined (see Heritage and Maynard, 2006, Robinson, 2003) and likely to be well oriented to by a cognitively able patient. Heath (1986) talks of the consultation ending as ‘bringing the business to a satisfactory closure’. It is the doctor who signals the closure of the interaction either with a medical summary and arrangement-making or through printing off and handing over a prescription. Robinson (2003) notes that patients are more free to initiate actions of their own, for example questioning, after doctors have provided treatment recommendations, however this interactional environment is also a place where it is relevant to close the interaction. Although the doctor signals the closure, it is the patient who is required to orient to this and to physically leave the doctor's space (Heath, 1986). The patient usually responds to the closing signals but may bring up unmet needs or residual symptoms during this time, sometimes referred to as the ‘door handle’ or ‘by the way’ phenomenon by doctors (e.g. White et al., 1994). Robinson (2001) identifies two interactional sequences initiated by primary care physicians at the end of a consultation. The ‘future arrangement sequence’ is suggested by the doctor and requires acceptance or agreement from the patient. Alternatively, the doctor can initiate a ‘final concerns sequence’. In this the doctor asks if the patient has ‘any other problems’ or ‘any other concerns’ before moving to close the interaction. Indeed, much research advocates doctors should solicit additional concerns to counter the tendency of both doctor and patient to orient to a ‘monotopical’ visit (Robinson, 2003).

Whilst analyses of closings in primary care are a useful starting point for the analysis presented here, their application has limitations in our setting. In acute hospital interactions in the UK, usually it is the healthcare professional who enters the patient's environment (bed space), often without invitation from the patient. The patient may be unaware that there is a specific need to be addressed (for example the need to monitor blood pressure or fluid intake, or to change a patient's position to avoid pressure sores). For PWD this is exacerbated further by the fact that they may not be aware that they have medical problems, or even that they are in hospital. It is also possible that the decline in linguistic ability associated with dementia may result in a failure to understand or pick up on typical closing cues. Additionally, in the UK acute setting, termination of the encounter involves the healthcare professional in physically leaving the space of the patient.

This paper presents an analysis of data from video recordings of healthcare staff interacting with PWD on an acute hospital ward. Using conversation analysis, we investigate how these interactions reach their ending point. We discuss barriers to achieving closing in this setting, and contrast this with closing healthcare interactions in primary care. Finally, we reflect on whether there are interactional practices which can enhance the likelihood of interactionally successful closings with PWD in an acute hospital setting.

Methods

Wider study

This work is part of a wider study funded by the UK National Institute for Health Research, Health Services and Delivery Research (ref 13/114/93). The overall aim of this research was to develop and test a communication training intervention for HCPs caring for people with dementia in acute hospitals. Ethical approval was granted by the Yorkshire and Humber - Bradford Leeds Research Ethics Committee in May 2015 (15/YH/0184).

Data collection

Healthcare Professionals (HCPs) were recruited from Healthcare of the Older Person wards at a large teaching hospital in the English East Midlands. We aimed to video record 40 encounters, which was estimated would give around 6 h of recorded interaction. We recruited HCPs willing to be video recorded in advance of recruiting patient participants. HCPs were only video recorded if we were able to recruit a patient participant in their care. Forty-one HCPs were recruited including doctors, nurses (including mental health nurses) and allied health professionals (physiotherapists, speech and language therapists, and occupational therapists), and of these, 26 were video recorded for the study-see Table 1.

Table 1

Table of healthcare professional participant and data characteristics.

	Number of HCPs recruited	Number of HCPs recruitedand then filmed	Number of films collected,by professional group
Nurses	19	11	19
AHPs	11	6	10
Doctors	11	9	12
Total	41	26	41

Twenty-seven patients were recruited to the study of whom 26 were filmed. Patients could be filmed more than once, with a different HCP, so some staff and some patients appear twice in our dataset. In total 41 encounters were recorded. The average length of a recording was 9.24 min. Patient participants had a diagnosis of dementia documented in their medical notes and HCPs considered they had some level of communication difficulty. Those with a diagnosis of Parkinson's disease or who were felt to be at the end of their lives were excluded. The recruitment process included an initial assessment of the patient's mental capacity to consent to being included in the study. This was conducted by two of the authors who are both experienced clinicians. If the patient lacked capacity in this regard, a carer was asked to act as a personal consultee under section 32 of the Mental Capacity Act (2005). All patients recruited to the study lacked capacity to give informed consent.

Video and audio recordings of participating HCPs interacting with patients with dementia during everyday healthcare tasks were made on wards from September–December 2015. We sought to record routine interactions for the staff in this setting, including activities such as the patient being assisted with eating and drinking, or taking medication; routine visits by a doctor; nurse observations such as taking blood pressure; or occupational therapy assessments such as the ability to make a cup of tea. However, we did not film more intimate interactions such as washing and dressing or toileting.

Conversation analysis

Conversation analysis (CA) is a research method that originates in sociology but draws on insights from other disciplines such as psychology and linguistics (see ten Have, 2007). Its aim is to study the structure and order of naturally occurring talk in interactions. The method has been widely used to study healthcare interactions (e.g. Heritage and Maynard, 2006, Pilnick et al., 2010b). In order to apply it to the study data, recordings were initially transcribed verbatim, with sequences of interest subsequently transcribed using standard CA procedures (Jefferson, 2004). Given the limited verbal capacity of some patients, it was important also to consider non-verbal and paralinguistic features; these were noted alongside the transcriptions. Transcriptions were used alongside the original recordings as an analytic aide.

In this setting, encounters were found to typically include the following phases: opening, purpose of visit, information gathering, business (getting the patient to do something), closing. This overall structure was ‘vaguely orderly’ in Jefferson’s (1988) sense, i.e. these phases did not necessarily always occur in this order, and some were recycled during other phases, e.g. gathering information or (re)establishing a purpose could happen during the business phase. It was also noted that some phases in this context could be entirely non-verbal e.g. the business of getting a patient to drink could be accomplished solely through gesture and action. This structure differs from the interactional organisation of primary care, such as that outlined by Robinson (2003). Although they have in common establishing the reason for the visit and gathering information, in primary care doctors deliver diagnoses and provide treatment recommendations. In our encounters, diagnoses were rarely mentioned, and any treatment such as changing a dressing or other healthcare task (medical tests, eating, drinking) was completed during the encounter, rather than recommended. It was also not necessarily completed by a doctor. In addition, an important factor in a UK acute hospital setting is that a PWD may not be aware of the need to see a HCP, and will rarely have instigated the interaction.

Initial viewing of the data established that there were recurring interactional difficulties in bringing these encounters to a close, along with more successful closures. Collections of closings were then made, organised around emerging themes. These data were analysed in monthly group data sessions attended by Allwood, Pilnick, O'Brien, Goldberg and Beeke, to guide further analysis and in order to increase robustness and reliability (Sidnell, 2010).

Analysis

In the analysis that follows, we focus on three phenomena around which there were recurring troubles in our dataset. The areas have been categorised as ‘open-ended pre-closings’, ‘mixed messages’ and ‘non specifics and indeterminate terms’.

Open-ended Pre-Closings

The following are examples of HCPs closing interactions with PWD after a healthcare task has been completed. Each begins with talk that indicates that the previous activity has finished, and is followed by what could be understood as a pre-closing indicator from the HCP in a move towards ending the encounter (Schegloff and Sacks, 1973, Button, 1987). According to Schegloff and Sacks (1973), if the recipient recognises this as a closing indicator and also wishes to move towards a closing, then the conversation moves towards a ‘terminal’ ending phase. The examples below are of interest because they reveal a common pre-closing question of an ‘open-ended’ nature, one that asks if the patient has further concerns (cf Robinson, 2003). Extract 1 shows a closing which appears to happen smoothly and without interactional difficulty. All names in our data have been changed to protect anonymity.

At the start of Extract 1, in line 153, the HCP delivers an overt pre-closing that orients to the encounter being initiated by himself, and its potential to be experienced as an imposition on the patient's time (‘I'm gonna leave you be’). The patient's response is unclear, and does not appear to orient towards closing. Then, in line 160, the HCP asks ‘can I do anything for you?’, an open-ended question in the sense it comes with no topic clues or boundaries. The patient's response to this begins with a part repetition of the HCP's question ‘can you i-i-’ but the subsequent fragments of talk, to the extent that they are intelligible, appear non-contingent in relation to the question. The HCP then follows this up in line 162 with a terminal utterance, ‘’kay bye bye for now and good luck’ whilst holding the patient's hand. After an assessment of sorts from the patient, ‘I've been doing it a while’, the summarizing quality and figurative nature of which suggests some orientation to the end of a sequence (Drew and Holt, 1998), the HCP begins to move away. At this point the patient appears to acknowledge the closing and accept it with ‘okay (1.0) thank you’ (line 164); he watches as the HCP moves away and no attempt is made to re-open the interaction or return to the question of whether the doctor can do anything for the patient. The sequence of talk here by the HCP is explicit in its orientation to closing via both verbal and non-verbal means. The open nature of the question ‘can I do anything for you’ orients to person-centredness and although the PWD responds with non-contingent talk, the sequence moves towards a close which is accepted by the patient, despite the fact that the HCP walks away while he is still talking.

Extract 2 reveals an encounter where a similar open-ended question is asked as part of a closing sequence. Here the question itself explicitly orients to imminent closure of the encounter: ‘anything you want to ask me before I go?’. However, and more typically of our dataset, the patient's response orients to some difficulty in producing an appropriate answer.

After a pre-closing ‘all right then’, the HCP delivers the question of interest. At first, the patient appears to orient to its pre-closing nature with a ‘no’, but then immediately produces a question which acknowledges that she may not be best placed to assess whether there is anything she should ask. Again, the HCP's question does not contain any topic clues or boundaries to aid the recipient in answering, and the patient returns the agenda to the HCP. The HCP does not introduce any further topics, instead producing a summary of the patient's progress over lines 162-4. However, this encounter is not easily closed from this point on, as will be considered in detail in Extract 6 below in relation to ‘mixed messages’. In this section we use it only to illustrate the immediate difficulties that can occur around the use of such open-ended questions in this setting.

Extract 3 illustrates more sustained trouble around an open-ended question as part of pre-closing. The HCP has been taking ‘observations’, which include blood pressure and temperature. When he is finished, he asks the patient if she would like a hand with ‘anything else’.

At line 13, having completed the necessary tasks, the HCP provides an upshot formulation (Heritage and Watson, 1979) which captures the implication of the previous talk and signals the close of this segment of the interaction. He then produces an open-ended question: ‘do you want a hand with anything else while I'm here?’. As in the previous examples, this question can be seen as a pre-closing; the use of ‘else’ is an explicit signal that the previous business of the encounter is concluded. It does not appear to respond to any specific need or signal from the patient, since the HCP continues to type while it is produced. However, the open-ended nature of the question also offers the potential for the PWD to open a very broad range of new topics. This is followed by a 1.6 s pause before the HCP follows up with ‘or are you alright.’, where the polarity of this question indicates a preference for a positive response (Raymond, 2003); the fact that this is not produced here with a questioning intonation also supports this preference. The patient's reply in line 16 indicates a clear misunderstanding which references the open-ended remit of the question. The HCP then rephrases his original question in lines 18–19, delivering it in two parts ‘do you want a hand?’ and ‘do you want me to help with anything else or’. Despite this rephrasing of the question, the patient‘s response indicates continued difficulty. She starts with a ‘well’ followed by a 0.6 s pause and then a hesitant ‘don't think’ and another pause before saying ‘no’; this turn displays overt uncertainty. The healthcare professional repeats ‘you don't think so’ in line 21, before producing an ‘okay’ which could be seen as another pre-closing; as Beach (1993, 1995) describes, ‘okay's are frequently used by speakers to simultaneously close a prior activity while also signaling a transition to the next. Once again, however, the patient does not orient to this, but instead, in lines 23–24 makes explicit her difficulty with the question: that she does not understand the remit of it and hence does not know what a relevant response would be.

Previous conversation analytic work in primary care has highlighted design features of questions which tilt expectations towards patients raising (or not raising) additional concerns at the end of encounters. For example, Heritage et al. (2007) demonstrate how asking whether patients have ‘some’ more concerns is more likely to elicit a positive response than asking whether they have ‘any’. Heritage et al. (2007) emphasise the negative polarity of ‘any’, in that it makes a ‘no problem’ response preferable (and it is worth noting that it is the term ‘anything’ that is used in all three extracts in our data above). Robinson et al. (2016) add a further layer to this analysis by demonstrating that patients are more likely to engage in agenda setting if they are asked whether they have ‘concerns’ rather than ‘questions’, and where they are asked about additional concerns earlier in the consultation (for example after their initial problem presentation rather than in the closing phases). Heritage (2010) also concludes that question design has a profound impact on patient response, and that this generally transcends contextual factors such as the characteristics of doctor and patient, specific medical practice, etc. However, it is notable that the patients in Extracts 1–3 above do not give straightforward ‘no problem’ responses in response to the solicitation. Whilst Extracts 2 and 3 do show initial responses which correspond with the negative polarity of ‘anything’, patients subsequently problematize the question in terms of their ability to identify an appropriate response. In other words, the failure to produce an additional concern is presented as a problem of lacking knowledge, rather than a non-problem of lacking a concern.

As a result, we would argue that there are some significant differences between the primary care setting from which the findings discussed above come, and the secondary care dementia context presented here, which can impact on patient responses. In particular, these kinds of open-ended questions which require patients or clients to set an agenda have been shown to be problematic in other institutional settings where the knowledge of what counts as relevant may be unclear (e.g. Pilnick, 2002, Pilnick et al., 2010a). In the specific context of dementia care, the fact that there are no clues or boundaries embedded within the question may be particularly problematic. And as we have noted, the patient did not initiate the interaction and some patients may be unaware that they are in hospital or being spoken to by an HCP or that they may have a need for any kind of medical or nursing help.

In summary then, while open-ended questions may be viewed as good practice in healthcare generally, in the sense that they attempt to give the patient an opportunity to drive the agenda of the encounter and to raise issues of importance to them, these positive effects are not apparent in this setting. Instead, such questions can lead to interactional difficulty, and the nature of this difficulty suggests that enacting previous recommendations from primary care studies concerning the framing and positioning of the question would not necessarily overcome them. Whilst the open-ended nature of the question orients to person-centredness, it also removes any cues for an appropriate answer to be produced, and this difficulty is compounded with this group of patients, who then struggle to formulate an appropriate response. Even in encounters where the question itself does not become an issue, as in Extract 1 above, the response which is produced cannot be used to identify a specific patient concern, as it is non-contingent. An alternative possibility in this setting, which we will consider later in this paper, is that the patient does identify a topic for further discussion, but it is one which cannot be addressed in this context. Such topics then have to be closed without resolution, creating an additional interactional difficulty (see Pilnick et al., 2010a), and potential distress for the patient.

Mixed Messages

The following are examples in which the HCP appears to work towards closing an encounter either verbally or non-verbally, but then opens up the interaction again.

In Extract 4, the interaction has been focused on asking a PWD to have a drink in order to assess the safety of his swallowing, with little success.

Having failed to get the patient to demonstrate swallowing through taking a drink of water, the HCP gives a clear pre-closing in lines 298-99 ‘okay (0.4) I think I'm gonna leave you be’ and mentions consulting the patient's doctor and nurse. She then turns away from the patient, and places the glass down on a side table. However, she then turns back to look at the patient and re-opens the interaction by introducing a new topic, asking ’you alright’ (line 300) before asking more specifically about his leg, which she has earlier tried to make comfortable. Woods et al. (2015), in their work on calls to cancer helplines, note that expressed dissatisfaction from a caller in the closing phase can result in the reopening of an interaction, e.g. by revisiting the advice that has been given. However, this query from the HCP does not seem to be occasioned by any signal from the patient, who makes an unintelligible response whilst watching her replace the top portion of the bed rail. The HCP then revisits an action associated with her original healthcare task in line 305 ‘did you want the water?’; this does not seem to be prompted by any overt verbal or non-verbal action from the patient (though it may respond to the unintelligible talk at line 302). She picks up the glass again and moves it from side to side in front of the patient, apparently to check whether the patient can actually see the glass. The patient appears to reject the glass in line 308, saying ‘I already … water’. After acknowledging this refusal, the HCP asks ‘can I still give you a sip’ and moves the cup to his lips, provoking an emphatic refusal from the patient ‘I don't want any now’ (line 313). Once again the HCP acknowledges the patient's refusal, however she subsequently downgrades it to ‘you're not sure you want it now’. The patient agrees that he doesn't want it now, which she repeats, and he again responds ‘no’. After this lengthy establishment of his wishes, the HCP produces an apology, and an account for her actions: that she was ‘just checking’ (line 318). Following another turn from the patient which references the fact that as far as he is concerned the task has ‘already’ been completed (line 319) and the HCP should ‘hang onto it’ (the water), the interaction closes with the HCP thanking the patient, and placing the bed back in its original position.

This example of a HCP starting to terminate a consultation but then re-opening it despite a lack of obvious patient-led cues may indicate a desire to persist in completing a healthcare task which is seen to be of benefit to the patient. It represents a common dilemma in this setting: when to abandon the interaction even when the healthcare task has not been completed. Refusals to co-operate with or collaborate in healthcare tasks are common in this dataset (O'Brien et al., 2017), but the task may still be pursued on the basis that it is considered to be necessary and in the patient's best interest. However, there are interactional implications of re-introducing a task after an explicit pre-closing of the kind used here; at best the patient may be confused, and at worst they may continue to orient to the task as completed, and produce strong resistance.

In the next example of a mixed message, a HCP is coming towards the end of his interaction with a PWD.

The HCP's utterance at the start of this extract is ambiguous in terms of the timing of the action described: ‘we'll examine you next’ could signal that the examination of the patient will take place here and now, or that it will be carried out at another, later but unspecified time. His use of the institutional ‘we’ also makes it unclear whether he will be involved in the examination or whether it will be conducted by other staff members. However, the patient does not orient to this ambiguity as problematic, but appears to accept it, stating ‘lovely (.) I'm ‘appy’. This is accompanied by the patient extending his hand; the HCP takes it and they shake hands in what appears to be a closing gesture.

At this point of possible ‘terminal exchange’ however, the HCP offers a further pre-closing signal ‘okay’. The intonation here suggests a pre-closure; it is not produced with questioning intonation. However, Schegloff and Sacks (1973) discuss the subtlety of ‘okay’ as a preclosing; it may lead to a terminal closure if both parties orient to this, but also allows for re-opening of a prior topic or beginning a new topic. At first the patient does appear to orient to a terminal closure with ‘right’ (line 192). There follows a 0.8 s pause after which the HCP stands up. At the same time the patient returns to a topic that has dominated the previous interaction (not shown in this transcript), which is about his alcohol and cigarette consumption and the name of a Russian film star he is trying to remember.

Whilst the patient is apparently searching for the name of the Russian film star, the HCP pulls the curtains back from the patient's bed; another indicator suggesting that the interaction is coming to a close. However, when the patient produces the name ‘Richard Burton’ in line 198, the HCP returns to face the patient, leans forward onto the table between them, and makes eye contact with the patient, thereby re-engaging in the conversation. However, he produces only minimal acknowledgments that do not expand the conversation any further (line 201). Subsequently the patient raises a further new conversational topic – smoking - and the HCP re-engages more actively in the conversation, contributing the additional topic of drinking alcohol (line 203). Following a discussion of the patient's drinking habits, the HCP repeats his original pre-closing ‘okay’, adds ‘right’ and then produces a more overt closing signal ‘I'll say goodbye for now.’ (line 213) accompanied by reaching for the patient's hand. This time there is no clear orientation to this by the patient as he continues with the topic of how much alcohol he drinks, receiving minimal, close implicating responses from the HCP. He is still talking as the HCP stands up and leaves.

This example initially shows a clear orientation from the patient to the end of the interaction. The HCP stands up at this point but a delay in any further overt move towards closing (opening the curtains around the bed at this point, for example) coincides with renewed talk from the patient. As a result, the HCP actively re-engages in the interaction, allowing the patient to recycle previous interactional topics. This leads to a scenario in which a mutually agreed end-point is not achieved, and the HCP walks away whilst the patient is still talking. Of course, time pressures mean that a HCP cannot necessarily always remain with a patient who continues to talk, and our data suggest this presents a particular problem in this setting. However, as this extract shows, failing to capitalize on a mutually agreeable closing-point, and remaining with the patient for longer, can ultimately result in a more difficult closure.

Extract 6 is a further illustration of the same dilemma. This extract is a continuation of the encounter we have seen in Extract 2. In this extract a HCP is coming towards the end of an interaction in which she is discussing hospital discharge plans with a PWD.

We have already noted in our analysis of Extract 2 (lines 158-64 above) that the HCP responds to the patient's difficulty over the open-ended question with a medical summary in lines 162-64, “We're quite happy with how things are going”. This is then followed by a plan of action in lines 167-69: “We're just trying to work out somewhere for you to go when you leave hospital that's what we are waiting for now.”. In response, the patient expresses her concerns about returning home and begins to outline a possible issue with her landlord, from line 170.

The HCP aligns as a recipient of the patient's concerns, and produces a ‘no problem’ statement; that these concerns will be sorted out for her (though by whom and how remains unspecified). This statement could indicate a close of topic; it completes the discussion of the previously raised housing issues and does not raise any new ones, and the intonation suggests completion. In this position a terminal exchange could have been produced by the HCP. However, there is then a two second pause (line 182) during which the HCP continues to lean forwards and make eye contact with the patient. As Rossano (2013) highlights, speaker gaze can be used as a resource to mobilize recipient response, whereas gaze withdrawal is a resource for moving towards closure. In contrast to the finality of her statement that everything will be sorted for the patient, then, the HCP's non-verbal signals are that she is expecting the patient to produce the next turn.

The patient takes the next turn in line 183 with a quiet ‘yeah’. She reiterates her concern but then, in lines 185-88 links this to a new topic: her faith in God, subsequently expanded to her enjoyment of going to church. Throughout this extended account the HCP produces only minimal acknowledgements, until she responds with a potentially pre-closing “all right then.” in line 199, although the patient has not indicated that she is finished with the topic. The patient orients to this attempt to close the interaction with “I'm sorry if I have bored you at all”. The HCP then takes the opportunity to leave with “I'll see you later. No that's ok”, where the latter part of the turn appears to respond to the patient's concern she has bored her, followed by “bye bye”.

This extract provides clear indication that the patient experiences the closure as problematic, orienting explicitly as she does to the possibility of having bored the HCP. Once again, failing to capitalize on a first possible closing-point, and remaining with the patient for longer, results in a more difficult closure. The extract also provides another example of the ways in which there is an essential tension between person-centred practice, which emphasises that a patient should have the opportunity to voice unmet needs or concerns, and managing the opening up of a topic which cannot be adequately dealt with in this context and must therefore be closed down. This could be for practical reasons such as time, or moral or political ones, where the HCP cannot or does not want to align with expressed views or beliefs.

Non-specifics and indeterminate terms

The final category of trouble within closing sequences in these data concern the HCP's use of non-specific language or indeterminate terms, including ‘indexical’ utterances. Though all language is indexical to a degree, indexical terms here refers to terms which depend for their specific sense on the context in which they are produced, for example referring to ‘this one’ rather than naming a particular object. In Extract 7 the use of a non-specific term causes difficulties with closing.

In this extract, having delivered the open-ended ‘anything else’ question and received a clear negative response, the HCP moves to close the encounter with ‘okay I'll see you soon.’ (line 110). The PWD initially repeats part of this terminal exchange in what appears to be a reciprocal move to close. However after a 0.6 s pause, he asks explicitly about the time frame meant by the non-specific ‘soon’; does it mean a day or a week? In response to this the HCP offers a specific timing of his subsequent visit (line 113) but hedges his response with ‘probably’. This does not go unnoticed by the patient, who after acknowledging the timing of the next visit asks about the meaning of ‘probably’. Interestingly, he appears to attempt to offer a guess with ‘bloody big’, but the choice of lexical item is non-contingent in this context. The HCP restates details of the visit once more, this time removing all non-specific language to state that he ‘will’ see the patient tomorrow morning and seeking the patient's agreement to this. After the acknowledgement sequence, the patient asks the HCP to adjust his bed before the encounter ends successfully, despite the same indeterminate phrase ‘see you soon’ being used again in closing by the HCP.

Of course, it is possible that any patient, with or without dementia, who is being attended to by a variety of HCPs who come and go from a hospital ward, may ask questions designed to pin down a non-specific arrangement. Such questions may be occasioned by a desire to establish better certainty and to impose some order on an unfamiliar environment. However, the recurrence of troubles around these kinds of expressions in our dataset suggests they pose a particular problem in this setting. Extract 8 reveals another incident where non-specific language leads to interactional trouble around closing.

In this extract it is the non-specific phrase ‘I'll be around’ that causes interactional trouble. After the HCP begins to close with an explicit mention of not detaining the patient further, there is some suggestion of dispreference from the patient (‘yeah well’), to which the HCP responds with an offer to be around to answer questions. The patient's response to this offer is not thanks or an acceptance but rather an understanding check of the literal sense of the language, ‘will you?’. The patient appears to have some sense of the HCP's offer to be present, but as his subsequent query ‘around what?’ reveals, it is not clear to the patient exactly where the HCP will be. The tone of this question strongly suggests a genuine understanding check; there is no hint of sarcasm. Both patient and HCP laugh together in response to the HCP's (smiling) qualification that she will be around the ward. After this a terminal exchange is successfully concluded and the HCP leaves. A non-specific arrangement of the type ‘see you soon/later’ appears here as it did in Extract 7, in the terminal position; this suggests these kinds of non-specific terms can be used without issue following the establishment of a more precise arrangement.

A final example of trouble created around closing by an indeterminate and indexical item is seen in Extract 9.

In this extract, a physiotherapist has concluded some physical activity with a patient who has her arm in a plaster cast. She has returned to the sitting position, in a chair beside her bed, while he is standing beside her. In lines 332-33 he proposes fetching the patient a cup of tea; this is an activity that is regularly used in this setting to mark the end of an activity, and used also in other settings where difficulties may be encountered in bringing about the closing of a sequence (Heritage and Lindström, 2012). Following a question from the patient about where the tea will be placed, and her topicalising of the difficulties she has in picking up a cup of tea, in lines 340-41 the HCP agrees with the patient, referencing the plaster cast using the term ‘that’. Despite the fact that it is the patient who has introduced the difficulty of holding a cup, she does not make sense of the indexical term ‘that’ and seeks clarification. As the three examples in this section illustrate, it appears that non-specific, indeterminate and indexical terms used in interaction can cause particular issues for PWD in this setting.

Discussion

We have presented examples of closings in healthcare interactions with patients with dementia on acute hospital wards. Through this analysis of the recurring practices of employing open-ended pre-closings, producing mixed messages and using non-specific and indeterminate terms, we see a common theme of interactional trouble around closure. It is important to note, however, that these practices are not necessarily inherently interactionally problematic. A patient without cognitive impairment, who has a good understanding of why they are in hospital, may be much more readily able to respond appropriately to an open-ended pre-closing; likewise use of non-specific language and indexicals is common in both everyday and professional interaction without giving rise to particular difficulties. And as we have noted above, in relation to ‘mixed messages’, in this specific care setting there may be good reason to persist with tasks or actions which are initially refused.

Considering the practices identified in more detail, we have shown that in this setting open-ended questions seeking to elicit any further patient concerns can extend closing of the interaction in a problematic way. Specifically, some patients indicate confusion and seek clarification of the kind of answer that might be expected, or they produce non-contingent answers. They may also raise issues that cannot reasonably be responded to in a healthcare setting. However, there is a tension for HCPs in that professional training recommends asking a patient if there is anything else they wish to discuss before terminating a consultation (Walker et al., 1990). In primary care, ‘final concerns sequences’ initiated by the doctor allow the patient to raise additional concerns (Robinson, 2001). The concept of person-centred dementia care (Kitwood, 1997) also has a part to play, in that the question potentially affords the patient an opportunity to influence the encounter and demonstrates a respect for patient autonomy. In this setting however, two factors appear to render this practice problematic. The first is that, unlike in primary care encounters, the acute care patient does not seek out the interaction with a HCP motivated by a problem they wish to discuss. These are routine ward encounters, carried out in a patient's best interests, and perhaps because of that, oriented to as an imposition on the patient's time (hence HCPs producing pre-closings such as ‘I'm gonna leave you be’). The second, interlinked, factor is that of cognitive impairment; PWDs in this setting appear to have a genuine problem with understanding the purpose and scope of such questions as ‘Is there anything else you want to ask me’ in the context of an encounter that they have not initiated and given that they may not fully comprehend being acutely unwell and in hospital. In summary, in this context of acute healthcare encounters with PWD, our data suggest that ‘open-ended’ closing questions about further patient concerns may be best avoided, and that further investigation is warranted into their use.

The extracts revealing mixed messages appear to suggest it can be a challenge for a HCP to know when to leave a PWD. Some HCPs are seen to re-open an encounter at a point following verbal and nonverbal moves to close it. Sometimes this may be motivated by a wish to complete an abandoned healthcare task that is necessary (the business of the encounter) in the face of patient refusal. On other occasions this re-opening appears linked to the fact that a PWD does not orient to the HCP's pre-closing device. Finally, at least some examples reveal that the HCP does not swiftly move to a terminal exchange despite signals that the patient has oriented to pre-closing. This results in continued talk on a patient's own topic of conversation, which is often outside the scope of the encounter. In the face of this, closing the encounter becomes more problematic and can lead to the HCP walking away as the patient continues to talk, or even explicit orientation by the patient to the continued talk as being unwanted.

Finally, there is a collection of examples in these data that reveal the problematic nature of non-specific language and indeterminate terms as pre-closing devices. When faced with common pre-closing moves such as ‘I'll see you soon’, a PWD may show confusion around the timing of any future encounter. This phrase is ubiquitous in a wide range of typical closings, and its lack of specificity is what makes it a useful pre-closing device. However, in the context of acute healthcare interactions with PWD, the suggestion from our data is that concrete arrangement making (e.g. ‘I'll see you tomorrow’) is preferable. Further investigation could disambiguate whether this is just a concern for PWD or whether it is a wider issue for the acute hospital setting where patients see multiple changing HCPs across the days and weeks.

Conclusion

In summary, our data suggest that HCPs engaged in acute care on Healthcare of the Older Person wards orient to efficiently completing a healthcare task whilst leaving a patient happy and satisfied. However, moves towards closing an encounter that appear intuitive to HCPs as competent interactants, may in fact serve to confuse a PWD and create difficulties with closings. As Heritage (2011: 338) highlights, we can create problems when we transfer normative conventions “that function well in the practice of ordinary conversation into medical contexts, where they serve to limit the effectiveness of communication”. In addition, we have shown that transferring normative conventions that may represent good practice in other healthcare settings, and with patients without cognitive impairments, can also impede the effectiveness of communication in this specific setting. This in turn highlights the recurring tension in this setting between seeking to treat PWD as full agents who can collaborate in joint communicative projects, and adapting communicative practices to take impairment into account. As is evident from our data, PWD represent a wide range of communicative abilities and these abilities do not remain constant, which introduces another level of complexity to any interaction with them. Our findings underline the importance of examining best practice guidance as it is actually talked into being, using approaches which can unpack the interactional detail involved. They also emphasise the importance of context in the analysis of healthcare delivery, to avoid a ‘one size fits all’ approach. Routine practices which may be used to underpin person-centred care in other contexts, such as the use of ‘anything else?’ questions, have at their heart a desire to address patient needs in a single interactional package and to avoid neglect. However, they may at best not produce any positive outcomes in this setting and at worst may produce confusion or feelings of inadequacy in patients who lack the ability to respond to them appropriately. Practice might be improved, for example, by helping HCPs develop an awareness of the possible implications of using different closing sequences with different patient groups, and by explicitly acknowledging the difficulties that an orientation to more generic person-centred practices can create when communicating with PWD. Further work is needed to establish the particular contingencies that can arise in health care professionals' interactions with people with dementia, and to enable HCPs to successfully navigate the recurring tensions between person-centredness and workflow in this setting.

Funding acknowledgement and disclaimer

This project was funded by the National Institute for Health Research Health Services and Delivery Research Programme (project number 13/114/93). The views and opinions expressed are those of the authors and do not necessarily reflect those of the Health Research Health Services and Delivery Research Programme, NIHR, NHS or the Department of Health.