How Psychiatrists Recommend Treatment and Its Relationship with Patient Uptake

ABSTRACT

Consultations for patients with chronic mental health conditions are conceived as meetings of experts: medical and experiential, respectively. Treatment decisions, in these terms, become a joint responsibility rather than handed down ex-cathedra. One resource for constituting decisions as ‘shared’ is the treatment recommendation – decisional authority can be invoked through its design. There is concern that people diagnosed with schizophrenia are infrequently involved in treatment decisions. However, the methods psychiatrists actually employ remain undefined. This article advances our understanding of psychiatric practice by mapping alternative methods used by psychiatrists to recommend treatment in outpatient consultations in situ. First, we unpack the types of treatments psychiatrists recommend. Then, we ask how psychiatrists recommend treatment? Applying a novel coding taxonomy, informed by the conversation analytic principle that recommendations represent different social actions, we identify the distribution of alternative formulations for psychiatrists’ recommendations (pronouncements, suggestions, proposals, and offers). We also propose one linguistic dimension, personal pronouns, on which recommending actions often depend, implicative for who is projected as ‘accountable’ for the decision. Finally, we examine the relationship between action type and patient uptake: is a particular type of recommendation more likely to attract acceptance/resistance from patients? And how does this relate to decisional accountability?

Consultations for patients with chronic mental health conditions are increasingly conceived as meetings of experts: medical and experiential, respectively. Treatment decisions, in these terms, become a joint responsibility rather than handed down ex-cathedra. This is reflected in treatment guidelines for schizophrenia:

For people with newly diagnosed schizophrenia, offer oral antipsychotic medication. Provide information and discuss the benefits and side-effect profile of each drug with the service user. The choice of drug should be made by the service user and the healthcare professional together, considering [factors such as] the relative potential of individual antipsychotic drugs to cause…side effects. (NICE, 2009, p. 17)

The recommendation of this partnership approach accommodates patient choice and responsibility. Indeed, particulars of antipsychotic medication, including evidence regarding overestimation of efficacy, underestimation of toxicity, and alternative treatment options (Morrison, Hutton, Shiers, & Turkington, 2012), suggest that regimens should be conceived via preference-sensitive decisions. Side effects of these treatments can span sleepiness and slowness to motor and metabolic effects. In addition, there exists a whole spectrum of subjective effects (Wolters, Knegtering, Wiersma, & Van Den Bosch, 2003). A shared decision-making model (SDM) (Charles, Gafni, & Whelan, 1997) recognizes that patients have a relevant contribution, garnered from their first-hand experience of these effects – beyond the psychiatrist’s domain.

Patients with severe mental illness generally express a desire for more participation in decisions about psychiatric care (Jared, Drake, & Wolford, 2007; The Schizophrenia Commission, 2012). However, there may be circumstances, exclusive to this setting, that problematize the realization of this ideal. In schizophrenia, a condition characterized by episodes of delusions, hallucinations, and behavioural disturbances, the negotiation of treatment is particularly delicate. Patients’ ability to rationally evaluate treatment options may be impaired. They may distrust people when they are experiencing paranoia, experience attention deficits (Hamann et al., 2006) or have poor awareness of their illness leading to resistance of medication. Psychiatrists may then use methods to persuade patients to adhere to their treatment (Mitchell & Selmes, 2007).

A further obstacle to partnership is the immediate methodological challenge of describing what SDM actually ‘looks’ like in naturalistic communication. Scales have been developed to capture the degree to which clinicians facilitate patient involvement (e.g. Elwyn et al., 2003). However, they focus solely on the clinicians’ behaviour – not how patients respond, e.g. acceptance/resistance, or how treatment decisions are co-constructed, something that can only be understood by observing specific phases of psychiatric consultations in situ. One such phase involves psychiatrists’ treatment recommendations: decisional authority can be invoked to various extents through their design. Historically defined as ‘unilateral directives’ (Byrne & Long, 1976), under the ideals of SDM (Charles et al., 1997), recommendations should account for patients’ decisional responsibility, afforded through shared expertise. These recommendation sequences present, therefore, a useful proxy for the degree of SDM in the consultation.

In general medicine, conversation analytic research identifies the treatment recommendation as a discreet phase of the consultation, typically implemented through an adjacency pair sequence (Schegloff, 2007): the doctor’s recommendation makes relevant patient acceptance (Stivers, 2006). Stivers et al. (in press) further developed a basic taxonomy of treatment recommendations in primary care as a first step towards a more comprehensive investigation. Treatment recommendations were conceptualized as not only representing different formulations, but different social actions. These actions typically took one of four forms: Pronouncements, Suggestions, Proposals or Offers, varying along a spectrum of deontic authority to decide and recommend on the patients behalf. Yet, despite these advances in mapping decision-making practices in primary care, there has been relatively less examination of actual interactions between psychiatrists and patients and their negotiation of treatment (cf. Angell & Bolden, 2015; Bolden & Angell, 2017; Kushida & Yamakawa, 2015; Quirk, Chaplin, Lelliott, & Seale, 2012).

With the unique interactional demands created by conditions such as schizophrenia, context-specific investigation of these actions in psychiatry would be of utility. As explained by Stivers et al. (in press), treatment recommendations embody epistemic and deontic authority: both as background to their production, and as encoded to varying degrees in the design of their delivery. We might therefore speculate ‘that the action level of the recommendation might be associated with whether a patient responds to the recommendation at all, whether to accept, acknowledge or resist’ (Stivers et al., in press, p. x). Such knowledge would be clinically salient in psychiatry. Acceptance of, and resistance to, treatment has, in turn, implications for adherence and patient outcomes, a central issue of concern for chronic mental health conditions – and schizophrenia particularly (Thompson & McCabe, 2012).

This article advances our understanding of psychiatric practice by mapping the alternative actions used by psychiatrists to recommend treatment in outpatient consultations. First, we unpack the types of treatments psychiatrists recommend. Then, applying Stivers et al.’s (in press) taxonomy, we ask, how do psychiatrists recommend treatment? We further propose one linguistic dimension, personal pronouns, on which recommending actions often depend/vary, implicative for who is projected as ‘accountable’ for the decision. Finally, we examine the relationship between action type and patient response: is a particular type of recommendation more likely to attract acceptance/resistance from patients? The findings are discussed in terms of potential explanatory mechanisms and clinical implications, including comparisons with recommendations (for new treatments) in UK primary care (Stivers et al., in press). Our findings contribute to a small but growing literature on psychiatric communication (Angell & Bolden, 2015; Bergmann, 1992; Bolden & Angell, 2017; McCabe et al., 2013; McCabe, Heath, Burns, & Priebe, 2002; Quirk et al., 2012; Thompson, Howes, & McCabe, 2016; Thompson & McCabe, 2016).

Data and methods

Data from two studies conducted in a psychiatric outpatient setting were combined. One hundred and thirty-four cases were drawn from an MRC study examining clinical interaction in psychosis (McCabe et al., 2013). Thirty-six psychiatrists from outpatient and assertive outreach clinics across three centres (one urban, one semi-urban, and one rural) were randomly selected, 31 consented (86%). One hundred and sixteen cases were also drawn from a cluster randomized controlled trial (McCabe et al., 2016), which assessed a communication skills training intervention for 21 psychiatrists and 64 patients with psychosis from outpatients clinics in East London and North East London. Patients assigned to clinicians who met Diagnostic and Statistical Manual – IV7 criteria for schizophrenia or schizoaffective disorder were asked to participate. Written informed consent was obtained from those who accepted, following which their consultations were audio-visually recorded. Dialogue was transcribed verbatim: the final set of 250 transcripts formed the dataset for this study.

To be included for subsequent coding and analysis, we screened the 250 cases for pre-defined inclusion criteria, as outlined by Stivers and Barnes (in press). Recommendation–response sequences and surrounding talk were transcribed using Jeffersonian orthography, capturing micro-level features of the interaction. The coding scheme was then applied to the collected cases. Informed by the conversation analytic principle that to talk is always to ‘do’ something (Schegloff, 1996), the protocol enabled identification of the primary social actions (Pronouncements, Suggestions, Proposals, or Offers) being relied on by physicians to present recommendations and the degree of patient uptake. These actions differ broadly in terms of who is treated as the primary decision maker; and the implicit claims instantiated in their design relating to their collaborative character – if they are optional or speculative, for example. Contextual variables, selected from Stivers and Barnes (in press), were also coded including medication type and primary class.

Given the distinctive context of psychiatry relative to primary care, adaptations to the coding scheme were made through an iterative process of application and refinement. Additional features coded for included; the type of treatment decision, e.g. a new medication or dosage change, and whether the latter pertained to an increase or decrease. One recommendation category, ‘assertions’ (see Stivers et al., in press; Toerien, in press), was also removed so as to focus exclusively on ‘directive’ actions: ‘on-record’ recommendations that attempt to influence the patient recipient to perform some action (Stivers et al., in press). We also coded for more expansive use of personal pronouns, e.g. I, we, us and you (in place of the categories ‘strength of endorsement’ and ‘partnership reference’), as a further examination of our data suggested these may be systematic, and interesting, lexical choices in this setting, with relevance for decisional accountability.

Descriptive data, including frequencies, were retrieved to provide a general overview of psychiatrist recommending practices and patient uptake. Statistical analyses were conducted using SPSS 20. A Chi-square test of independence was performed to examine the relation between psychiatrist recommendation action type and patient response and also between an identified systematic linguistic feature, personal pronouns, and action type. Selected single-case examples were also presented to illustrate, qualitatively, the general numerical trends.

Results

What types of treatment do psychiatrists recommend?

Fifty-four percent (n = 135) of consultations contained psychiatrist treatment recommendations. Thirty percent (n = 40) of these recommendations related to the introduction of new treatments. More frequently, dosage alterations of existing medications were recommended, 41% (n = 56), reflecting the long-term usage often characteristic of schizophrenia. Within this subset, dosage increases and decreases were equally occurring (40%, n = 29). Alternative treatments, including psychological therapies and support groups, were recommended in 29% (n = 39) cases, consistent with the notion that pharmacology remains the mainstay of psychiatric treatment for psychosis (The Schizophrenia Commission, 2012). In remaining consultations, treatment decisions were either absent (with consultations comprising only a review) or ‘no-change’ decisions. As expected, the majority of medication recommendations (58%, n = 78) were for antipsychotics, e.g. Olanzapine, Haloperidol, Risperidone. The remaining treatments were typically those which complement antipsychotic regimens, e.g. antidepressants, anxiety medications, side effects tablets (e.g. procyclidine), sleeping tablets, and tranquilisers.

How do psychiatrists recommend treatment?

Table 1 displays the distribution of four alternative recommending actions used by psychiatrists, with data examples. As is evident, psychiatrists routinely formulated decisions as shared. Most frequently – in 31% (n = 30) of cases – psychiatrist recommendations were designed as proposals, followed by suggestions (30%, n = 29), pronouncements (25%, n = 24), and offers (14%, n = 13). Thus, while recommendations can constitute clear directives (cf. Byrne & Long, 1976), e.g. pronouncements, 75% (n = 72) of psychiatrist formulations in fact embodied a degree of patient agency.

Table 1.

Distribution of treatment recommendations.

Action type, explanation (Stivers et al., in press)	Coded example from data	New	Dosage change	Total
Pronouncement Physician asserts recommendation as instigator, decision maker and presents as already determined.	What we’re going to do is … stop the prozac and give you another antidepressant	8% (n = 8)	17% (n = 16)	25% (n = 24)
Suggestion Physician recommends as instigator but treats patient as decision maker and medication as optional.	I think that probably the best thing would be to switch to something that may have less in the way of side effects	16% (n = 15)	15% (n = 14)	30% (n = 29)
Proposal Physician recommends as instigator but decision making is treated as shared by doctor and patient. Proposals highlight the recommendation as speculative.	I don’t suppose we could persuade you to have another go at Clozapine?	8% (n = 8)	23% (n = 22)	31% (n = 30)
Offer Physician treats patient as having instigated recommendation and as the decision maker, thus treating medication as having been occasioned.	If the voices are really bothering you, I can increase your injection a little bit if you want?	9% (n = 9)	4% (n = 4)	14% (n = 13)
Total		42% (n = 40)	58% (n = 56)	100% (n = 96)

Differentiating findings by treatment type, new treatment recommendations were most frequently designed as suggestions (38%, n = 15), often encoding a marker of psychiatrist endorsement (see Stivers & Barnes, in press) while dosage change recommendations – for treatments patients had experience of – were recurrently, in 39% (n = 22) of cases, formulated as proposals, invoking patient participation through their speculative character (Stivers et al., in press). As the most dominant action type within the corpus overall, we examine an extended sequence in which a psychiatrist delivers a treatment proposal in extract 1. This proposal-resistance fragment follows a lengthy discussion regarding the patient’s psychotic experiences. The psychiatrist inquires if these feelings are worse towards the latter part of the interval between her antipsychotic depot injections – which the patient confirms. The key here is that the patient does not acknowledge these feelings may be related to her illness, i.e. ‘delusions’. She claims to be worried and puzzled by people in her village who perceive her to be ‘special’ and wish to sacrifice her to a ‘giant moth’.

Extract 1
01	DOC:	Do you ever hea:r or see anything when you’re on your o::wn
02		in the [house.
03	PAT:	[No::: no =
04	DOC:	= You ^never hear voices talking to you about ^this?
05		(0.6) ((Patient shakes head))
06	DOC:	No?
07	PAT:	Nothing at a:ll no.
08	DOC:	And you never see giant moths o:r [anything like that.
09	PAT:	[No::::: I never see giant
10		moths no.
11	DOC:	.hhhh (0.6) I mean one thought I did ^ha:::ve (.) was (0.4)
12		^whether (0.4) it’d be worth increasing your depot slightly.
13	PAT:	^No:::::::::

Following this, the doctor deploys a string of interrogatives across lines 01–08 to solicit (dis)confirmation that the patient hears voices or ‘sees’ things, a typical method of assessing psychotic symptoms. The patient discounts these possibilities with some certainty. She provides a series (lines 03, 07 09) of ‘no-problem’ (Heritage & Sorjonen, 1994, p. 8) responses that, at lines 03 and 09, are delivered before the psychiatrist has come to completion of his question: in interjacent overlap (Jefferson, 1986). To deploy a proposal for increasing treatment at this point – an intervention specifically designed to deal with the presence of these symptoms – would be at odds with the patient’s responses, thus a delicate action.

Accordingly, across lines 11 and 12, the psychiatrist makes a proposal that, in several ways, is constructed tentatively, embodying an orientation to the patient’s role and preference – and to contingencies being associated with its acceptance, e.g. the patients belief that she is not ‘delusional’. He deploys a turn initial delaying device (Pomerantz, 1984; Schegloff, 2007): a long inbreath followed by an intra-turn pause before beginning to speak. His turn is further prefaced with ‘I mean’, hearably displaying spontaneous reasoning that ‘distances’ him from, and diffuses, this potentially face-threatening action (Fox Tree & Schrock, 2002, p. 745; Maynard, 2013). Indeed, the psychiatrist continues his turn with ‘one thought I did have’, framing the proposal as somewhat transient: ‘one thought’ relative to firmer epistemic commitment, e.g. ‘I think’. The remaining turn (‘was whether it would be worth increasing your depot slightly’) is further hedged. The use of the conjunction ‘whether’ expresses a degree of doubt or choice between alternatives, i.e. to increase/not to increase the patient’s depot, thereby orienting to the recommendation as speculative. The patient provides an instant rejection of the proposal in line 13 – a flat no response – foregrounding the contingencies to which the psychiatrist had initially oriented to in the design of this proposal.

Personal pronouns and the projection of decisional responsibility in the treatment recommendation

The distribution of four alternative treatment recommendations has been mapped. However, variability between and within action categories can exist through psychiatrists’ selection of alternative constructions. Proposals, like those in extract 1, for example may be designed more or less hedged, while some suggestions may encode stronger clinician endorsement than others (Stivers et al., in press, p. x). Each of these design elements has implications for the sharing of deontic authority and responsibility. Examination of the coded recommendations revealed that a key linguistic resource in the projection of decisional accountability was the provision of personal pronouns, i.e. first person (I and we) and second person (you): the ‘central forms of referring to speaker and recipient’ (Sacks, 1992, p. 1349).

We found that psychiatrists, through pronoun use, can foreground to lesser or greater extents three constructs, implicative for who is constituted as ‘responsible’ for the decision: (1) patient preference, (2) psychiatrist endorsement and (3) clinician–patient partnership. For example, a psychiatrist could deploy one of two formulations, both interpretable as suggestions: ‘I recommend that you increase the dose’ or ‘you could try increasing the dose’. The former construction contains the pronoun I combined with a recommend verb. This personalises and endorses the prospective dosage change, relative to the latter design that only casts the patient as the agent of the activity. Alternatively, a psychiatrist could lexically invoke the doctor–patient partnership by employing the first-person plural pronoun: ‘we’, e.g. ‘we could increase the dose’. A third pertinent design choice is the foregrounding of the wants or desires of the patient. For example, the psychiatrist could deploy one of two formulations, both hearable as offers, e.g. ‘Do you want to increase your dose?’ or ‘I can increase your dose?’. While both actions propose to satisfy recipients’ needs (Curl, 2006) characteristic of offers, the syntactic design of the former foregrounds the preferences of patient, while the second only denotes the psychiatrist as the agent of the offered treatment activity: this preference-implicative pronoun is notably absent.

Despite relevant alternatives, pronoun use was relatively consistent. Table 2 displays the distribution of personal pronouns between and within alternative treatment recommendation types associated with such foregrounding of psychiatrist endorsement (e.g. I think/recommend/advise etc.), decisional partnership (e.g. we/us) or patient preference (do you want/feel etc.). Pronouns not relating to endorsement, partnership or patient preference (e.g. ‘I can increase your dose’, ‘you could do x’, where the pronouns denote the agent of the treatment activity) were excluded. Offers were typically formulated with ‘you’ (77%, n = 10), explicitly displaying dependency on patient preferences. Proposals were most frequently ‘we’ formulations (73%, n = 22), lexically invoking the psychiatrist–patient alliance. Finally, suggestions were recurrently ‘I’ formulations (66%, n = 19): personally endorsed by psychiatrists. For pronouncements, endorsement was implicitly embodied in the design ‘I am going to give you…’ (orienting to the decision as ‘complete’), but also sometimes (33%, n = 8) in a manner that lexically invoked the clinician–patient partnership, e.g. ‘we are going to…’. A Chi-square test of independence displayed a significant relationship between action type and pronoun use [χ² (12, N = 96) = 82.23, p < .001]. Examples of these systematic design choices are presented in Tables 3–6.

Table 2.

Distribution of personal pronouns.

	I (endorsement)	We (partnership)	You (preference)	/Action type total**
Offers	0% (n = 0)	0% (n = 0)	77% (n = 10)	/13
Proposals	7% (n = 2)	73% (n = 22)	0% (n = 0)	/30
Suggestions	66% (n = 19)	7% (n = 2)	0% (n = 0)	/29
Pronouncements	100% (n = 24)*	33% (n = 8)	0% (n = 0)	/24
Pronoun total	35% (n = 34)	33% (n = 32)	10% (n = 10)	/96

*Endorsement is implicitly embodied in the design rather than explicitly formulated – see later examples.

**Denominator for each.

Table 3.

Examples of treatment offers.

‘You’ formulations – foregrounding patient preference

1) DOC: .tch .hhhhhh (.)^if the voices are really bothering you: >I can< in^crea::se your injection >a little bit< if you wa::nt.

2) DOC: e::r and ^also (.)^do you want to make any changes to you:::r (.) olanzapine at all.

3) DOC: .hh >I mean< do you feel you’d like to try: a:n antidepressant a^gain see if it makes you feel any different?

4) DOC: You can try them on a regular basis if you li:ke?

Table 6.

Examples of treatment pronouncements.

‘Informing’ or ‘directing’ – endorsement embodied within the formulation

1) DOC:.hh so what we are going to do is (.) increase the olanzapi::ne.

2) DOC:. e::rm, (4.4) I’m ^going to increase a ^little bit one of the medica:tion

3) DOC: So take more of the olan::zapine in the morning.

4) DOC: What we’re going to do is … stop the prozac and give you another antidepressant

In 77% (n = 10) coded treatment offers, the patient was invoked as the central figure: containing a preference – implicative pronoun, ‘you’. The patient’s role in the decision is thereby demarcated in the psychiatrists’ formulation. Indeed, according to Sacks (1992, 1, pp. 163–8, 568–77), one of the characteristics of ‘you’ is that the listener is always included. ‘You’ is the pronoun that encodes the role of ‘recipient’ (Goodwin, 1996). Table 3 displays four examples from the corpus. The patient’s ‘role’ in each fragment relates to the expression of treatment choices. While the psychiatrist retains the agency to act in realising the offer (i.e. prescribe), the decision itself is explicitly marked as contingent on patients’ preferences, e.g. if you want (1), do you want (2), do you feel you’d like to try (3), if you like (4). The decisional responsibility in the following formulations is lexically invoked as more shared. Table 4 depicts four examples of recommendations coded as proposals.

Table 4.

Examples of treatment proposals.

‘We’ formulations: foregrounding the clinician-patient partnership

1) DOC: .hh I suppose the question is whether we should reduce the olanzapine a bit mo:::::re?

2) DOC: We might need to think about changing the olanzapine to another medication.

3) DOC: .hhh (.) how about we try another antidepressant a different kind of antidepressant?

4) DOC: I ^mean, (0.4) <wha::t if> w- a:::h e::::r (.) hhh (.) what would you::r thoughts be >about< what if I said to you we::::ll (.) we shou::d look abou:::t reducing them or stopping them.

Seventy-three percent (n = 22) of proposals were designed to invoke the doctor–patient partnership, containing the first person plural pronoun, ‘we’. As Sacks (1992) asserts, ‘we’ entails collective involvement and group membership (Sacks, 1992, 1, pp. 333–40 and 2, pp. 391–5). In this way, the ‘self’ is minimised as it becomes part of a collective. The psychiatrist in examples (1), (2), (3) and (4) constructs the decision as a joint responsibility by projecting himself and the patient as mutually implicated in its realisation. Additionally, while each formulation contains a clear prospective course of action, i.e. ‘ reduce the olanzapine ’ (1), ‘changing the olanzapine’ (2), trying a ‘different kind of antidepressant’ (3), ‘reducing…or stopping’ medication (4), it is hedged and treated as in question; ‘the question is whether’ (1); ‘might need to think about’; ‘how about we try’ (3), or tentatively ‘hypothetical’; ‘what if I said..’ (4). This highlights the recommendation as somewhat speculative. In this way, the psychiatrist avoids asserting firmly the appropriate course of action, or signalling overt endorsement. A role for patient choice is thus accommodated, while not, unlike offers, presenting it as an exclusive condition. Conversely, in the following ‘suggestion’ formulations, the psychiatrist claims an explicit view on treatment, establishing a greater degree of decisional accountability.

Sixty-six percent (n = 19) of suggestions contained ‘I’ pronoun formulations combined with reporting verbs, e.g. ‘I would really recommend’ (1), ‘I’m going to suggest’ (2), ‘I would suggest’ (3), ‘I’d like to suggest’ (4). According to Sacks (1992), ‘I’ links the talk to other parts of the talk and indexes the speaker to the here and now (1992, 1, p. 32). Moreover, ‘I’ here allows the psychiatrist to state the his/her position and provide subjectivity (Malone, 1997). ‘I’ can also be used as a means of the psychiatrist showing a degree of personal commitment and involvement: encoding a ‘personal voice’ (Wilson, 1990). Indeed, the fragments in Table 5 display the psychiatrist using I/I’m to convey ownership of the suggestion, marking the ensuing treatment action: ‘go down with the medication’ (1), ‘tail off the aripiprazole’ (2), ‘give you a prescription for procyclidine’ (3), ‘reducing sleeping tablets’ (4) as contingent on his endorsement. By displaying the epistemic grounds – his clinical opinion – that form its basis, he treats himself as partly liable to the patient (see also Costello & Roberts, 2001). However, while the psychiatrist formulates a view on what should be done thus, relative to an offer or proposal, the suggestions here are epistemically ‘upgraded’, he does not choose to formulate what will be done. The psychiatrist places patient agency (or collective agency in case 4) in the dependent/subordinate clause, whereas they place their own agency in the main clause. An orientation to patient choice is therefore still accommodated. The following ‘pronouncement’ actions however (Table 6) show an even greater commitment to the course of action: endorsement is implicitly embodied within their design.

Table 5.

Examples of treatment suggestions.

‘I’ formulations – foregrounding psychiatrist endorsement

1) DOC: e:::r e:::r I would really recommend you to decrea::se to go down with the medication.

2) DOC: .hh ^what I’m going to suggest (.) <is tha::t> (.) you: (0.4) u::::m, (0.8) tail off (.) the aripiprazole.

3) DOC: um ^I would suggest that I give you a prescription for procyclidine

4) DOC: What I’d like to suggest is i:s (0.4) that we could try: (0.4) um reducing sleeping tablets

Psychiatrists’ pronouncements projected the treatment decision as ‘complete’: actions were constructed as informings or directives. Examples 1–4 show that the doctor does not give an opinion but announces a course of action that the patient (or himself) will take. Decisional endorsement and responsibility are implicitly embodied in the design of the recommendation: the unmitigated declarative form of 1, 2, and 4 contributes to a resulting action that is unambiguously an ‘informing’’ (Heritage, 2012, p. 8). Similarly, (3) acts as a clear directive through the use of the imperative verb (‘take 5 mg of olanzapine’). By declaring a course of action, the psychiatrist claims both to know what can be done and what the patient should do (increase the olanzapine (1), increase … one of the medication (2), take more of the olanzapine (3), take another antidepressant (4)), independent of patients’ preferences. He thus strongly evokes a right to advise the patient (cf. Heritage & Sefi, 1992), a high entitlement to direct the other patient and little orientation to the contingencies on which compliance with the pronouncement may rest (Craven & Potter, 2010 p. 426). While selective inclusion of plural pronouns, (1), (4), can imply greater or lesser extents of clinician–patient ‘partnership’, the abilities or desires of the patient, at least in the action formulation itself, are precluded.

Alternative formulations (pronouncements, suggestions, proposals, offers) in this corpus can therefore be usefully conceptualised in terms of patient’s decisional responsibility, accomplished and indexed, in part, by the deployment of personal pronouns. This is graphically represented from low to high in Figure 1.

Figure 1.

Patient responsibility across recommending actions.

Is action type associated with patient uptake?

Patients have a relevant slot in the turn following a treatment recommendation in which they can implement a variety of resources – acceptance, acknowledgement, or resistance. Or they may offer no response. These all have implications for the patient’s agreement to or resistance of the psychiatrist’s action. We found a significant association between psychiatrist recommendation and patient uptake, unpacked in more detail here. Table 7) displays the frequency of each response category within the corpus, differentiated by recommendation class. Table 8) displays the distribution of new recommendations only, for comparative value with Stivers et al. (in press).

Table 7.

Distribution of patient responses: Overview of corpus.

Action type	Patient uptake
	Accepts	Acknowledges /nods	Resists	No uptake	Total
Pronouncement	8% (n = 8)	14% (n = 13)	2% (n = 2)	1% (n = 1)	25% (n = 24)
Suggestion	6% (n = 6)	18% (n = 17)	1% (n = 1)	5% (n = 5)	30% (n = 29)
Proposal	6% (n = 6)	9% (n = 9)	14% (n = 13)	2% (n = 2)	31% (n = 30)
Offer	3% (n = 3)	0% (n = 0)	8% (n = 8)	2% (n = 2)	14% (n = 13)
Total	24% (n = 23)	41% (n = 39)	25% (n = 24)	10% (n = 10)	100% (n = 96)

Table 8.

Distribution of patient responses: New recommendations only.

Action type	Patient uptake – new recommendations
	No uptake	Acknowledges/nods	Resists	Accepts	Total
Pronouncement	0% (n = 0)	18% (n = 7)	0% (n = 0)	3% (n = 1)	20% (n = 8)
Suggestion	8% (n = 3)	23% (n = 9)	0% (n = 0)	8% (n = 3)	37% (n = 15)
Proposal	3% (n = 1)	3% (n = 1)	13% (n = 5)	3% (n = 1)	20% (n = 8)
Offer	3% (n = 1)	0% (n = 0)	13% (n = 5)	8% (n = 3)	23% (n = 9)
Total	13% (n = 5)	58% (n = 23)	25% (n = 10)	20% (n = 8)	100% (n = 40)

While psychiatrists often formulated decisions as ‘shared’, patients did not routinely accept these decisions in the turn following the recommendation. Of the 96 treatment recommendations deployed by psychiatrists, over a third were simply acknowledged (41%, n = 39). As acceptance is normatively oriented to when recommending treatment (Stivers, 2006), some alternative behaviour, where acceptance is expected, may constitute ‘passive resistance’ (Heritage & Sefi, 1992). Thus, allowing a gap of silence, a stand alone head nod or producing a minimal acknowledgement token may be resources for implicitly resisting recommendations (Heritage & Sefi, 1992; Koenig, 2011 Stivers, 2005): they withhold acceptance and do not ‘constitute an undertaking to follow the advice offered’ (1992, p. 395). Extract 2 illustrates this. Here, the psychiatrist indicates that acceptance of his formulation is necessary – actively pursuing it. Following discussion of the patient’s complaint of tremor, a side effect of some antipsychotic medications, the psychiatrist deploys a suggestion for a prescription of procyclidine, a drug designed to diminish such effects. Having offered his recommendation, the psychiatrist waits for uptake (line 03) and, when none is forthcoming, he continues to adjoin additional statements, supporting his original suggestion (lines 04, 06, 08, 12).

Extract 2
01	DOC:	Um ^I would suggest that I give you a prescription for
02		procyclidine.
03		(0.4)
04	DOC:	That you can take once or twice a da::y ^if you need them.
05		(0.4)
06	DOC:	^A::nd (.) it may benefit the shake.
07		(0.2)
08	DOC:	M- most people they find most of their shaking goes away.
09	PAT:	Ri:ght.
10	DOC:	^Would that be o^kay with you::

By withholding explicit acceptance, i.e. passing on the opportunity to respond (lines 03, 05, and 07) and providing a minimal acknowledgement token (line 09), the patient has resources to orient to the psychiatrist’s recommendation as ‘incomplete’. Here, this passive resistance leads the psychiatrist to pursue patient acceptance by elaborating on the medication’s benefit and effects (line 06 and 08). When acceptance is still not forthcoming in line 09 following the addition of this supporting information, the psychiatrist proceeds to explicitly pursue acceptance in line 10 ‘would that be okay with you…’.

When we consider active resistance and passive resistance (Heritage & Sefi, 1992) together, this is the most common form of patient uptake, occurring in 84% (n = 81) of cases. Acceptance was provided only 24% (n = 23) of the time, while no uptake constituted 11% of cases (n = 10).

Differentiating the findings by action type, acknowledgements (possible passive resistance) were more likely to occur in response to the, as observed in our pronoun analysis above, ‘endorsed’ (implicitly or explicitly) action types, pronouncements and suggestions (54%, (n = 13), 59% (n = 17), respectively), while active resistance was even more concentrated: most likely to arise in response to proposals and offers (43% (n = 13), 62% (n = 8), respectively), followed by acceptance (6/30, 3/13, respectively). Active resistance was rare in response to pronouncements and suggestions (8% (n = 2), 3% (n = 1), respectively). Considering new recommendations only, to align with Stivers et al. (in press), this pattern of concentrated resistance appeared consistent. There were no instances of active resistance to pronouncements and suggestions, only occurring in response to proposals and offers (63% (n = 5), 56% (n = 5).

Because psychiatrists’ utterances containing recommendations usually have identifiable turn-constructional unit (TCU) boundaries inviting of response, a potential factor affecting response type is whether those boundaries are clearly present (Stivers & Barnes, in press). Patient responsivity may be deterred by the lack of a presence of such an opportunity space: while it does not block patient acceptance/resistance, it fails to encourage it at that juncture. However, in the majority of cases here, these TCU boundaries were clear. In only 15% of all cases did psychiatrists proceed from the recommending TCU directly to a next TCU without this opportunity space being present.

A Chi-squared test of independence confirmed that the relationship between psychiatrist treatment recommendation action type and patient uptake was significant [χ² (16, N = 96) = 54.14, p < .001]. This relationship held when considering new recommendations only [χ² (16, N = 47) = 40.47, p < .001] for comparative purposes with UK primary care (Stivers et al., in press). Moreover, there was no significant relationship between recommendation action and medication [χ² (48, N = 96) = 52.77, p = .295] or patient uptake and medication [χ² (48, N = 96) = 44.23, p = .628], a potential confounding factor. Resistance thereby seemed more likely where patients’ decisional accountability was greater – see Figure 2. To illustrate this finding, extract 3 shows an offer–resistance sequence from the data.

Extract 3
01	DOC:.	tch .hhhhhh (.)^if the voices are really bothering you: >I
02		can< in^crea::se your injection >a little bit< if you wa::nt.
03	PAT:	No:::: if (.) if you increase it then I might get side
04		effects you know

Figure 2.

Patient responsibility and resistance across recommending actions.

By designing his turn in line 01 with a conditional + main clause syntax (Curl, 2006), the psychiatrist offers an increase of antipsychotic injection, contingent on the patient’s preferences (‘if you want’) in an attempt to suppress his voice-hearing. The patient deploys a flat ‘no’ response that is significantly elongated in line 03. He not only refuses but designs his turn in a way that is structurally aligned with an agreeing response. As Pomerantz (1984) notes ‘in general agreements are performed with minimisation of gap between the prior turns’ completion and the agreement turns initiation’ (p. 65). Meanwhile, disconfirmations and rejections that are performed with delays and are interpretable as ‘reluctantly’ preformed instances of the action (Pomerantz & Heritage, 2013). This straightforward response acts to assert the patient’s epistemic rights: he claims a high degree of entitlement to reject the offer. However, this immediate refusal is qualified. The patient produces an account to supplement his rejection, formulating a causal link between a medication increase and the physical contingency of ‘side effects’ – an area of the patient’s subjective experience.

Discussion

While there is a concern that people diagnosed with schizophrenia are infrequently involved in treatment decisions and sometimes not even told their diagnosis (Bayle et al., 1999), the analyses here indicate that outpatient consultations are surprisingly ‘democratic’ decision-making forums. Psychiatrists have resources to assume a paternalistic stance when recommending treatment, i.e. ‘pronounce’ a treatment decision. Expanding research beyond treatment ‘recommendations’ per se reveals the much wider range of actions that psychiatrists employ in situ. Using a novel protocol to classify treatment recommendations in four categories, distinguishable by their collaborative character, (1) pronouncements, (2) suggestions, (3) proposals and (4) offers, we provide an overview of some of the methods available to clinicians. We find that psychiatrists attempt to balance patient involvement and the exertion of authority with respect to treatment recommendations. Recommendations for dosage changes of existing treatment were most frequently designed as proposals – projecting the decision as a joint project. Recommendations for new treatments were most frequently formulated as suggestions, which, while psychiatrist-endorsed, also presented the decision as optional, for example by placing patient agency in their subordinate clause. Given the pressure for clinicians to establish SDM as the ‘norm’ in clinical practice for patients with schizophrenia (NICE, 2009), these findings are noteworthy.

The occurrence of these action types also has comparative worth. Differences to UK primary care are apparent (Stivers et al., in press), where pronouncements were observed as the most frequent formulation (29%) for new treatment recommendations and proposals comprise only 15% of recommendations and suggestions 24%. Considering the current corpus overall, this difference can be explained by the inclusion of dosage change recommendations. One might expect the invocation of patient participation (via the speculative character of proposals (Stivers et al., in press)) to be more likely in relation to treatments patients already have experience of – often, in the case of schizophrenia, on a long-term basis. Indeed, when new recommendations alone were considered, the use of proposals was, though still higher, more comparable to UK primary care, forming 20% of total recommendations. However, new recommendations formulated as suggestions were still markedly higher in psychiatry than primary care (37% of new recommendations) and offers were also more frequent – 23% vs. 15%. This raises interesting questions as to the relative degree of patient ‘choice’ in this psychiatric outpatient setting. A more endorsed recommendation (a recommendation per se, constituting more traditional medical ‘advice’) may be required to recommend a new treatment vs. an existing one. However, relative to primary care, perhaps there are specifics to antipsychotic regimens (or medications for chronic conditions more generally), later discussed, that warrant more choice, in line with a ‘recovery’ approach to schizophrenia: the notion that ‘it may be time to reconsider the prevailing opinion that all service users with psychosis require antipsychotic medication in order to recover’ (Morrison et al., 2012, p. 1).

Alternative action formulations could be usefully conceptualized in terms of decisional responsibility: accomplished, in part, by the deployment personal pronouns. Despite relevant alternatives, first and second person pronouns (I, we, you) were significantly associated with particular recommendation classes: foregrounding patient preference (e.g. offers: ‘you’ formulations), psychiatrist endorsement (e.g. suggestions: ‘I’ formulations) or the clinician–patient partnership (e.g. proposals: ‘we’ formulations). This adds further linguistic nuance to psychiatrists’ recommending practices, highlighting one mechanism for precisely how, lexically, clinicians can go about realising alternative action types. For example, ‘with suggestions, physicians abdicate only responsibility for making the decision but retain responsibility for their recommendation….[while] in making a proposal, physicians claim responsibility for instigating the recommendation, but share with patients the responsibility for the final decision’ (Stivers et al., in press, p. x)’ – the pronouns ‘I’ and ‘we’ are two specific devices which help psychiatrists convey such degrees of personal or shared responsibility, respectively. Paradoxically, however, actions characterized by less psychiatrist responsibility – proposals and offers – were significantly more likely to be resisted by patients, while direct resistance in response to ‘endorsed’ recommendation types, pronouncements and offers, was rare. This aligns with findings in US and UK primary care, whereby patients resist pronouncements and suggestions significantly less often than other action types (11% and 20% respectively) (Stivers et al., in press).

While we cannot provide a full account of the relationship between action type and resistance in the scope of this overview, our findings point to several possible explanations that warrant further exploration. Proposals were often design as hedged, speculative, or in question, while offers were designed as contingent on patient preference. Although one might expect a greater degree of patient choice and participation, in line with the ideal of SDM, to be associated with increased acceptance, these design choices may evidence psychiatrists’ sensitivities to possible contingencies involved in the granting of (and adherence to) these actions. That is, be anticipative of patient resistance due to some prior knowledge (see extract 2, for example). The assembly of recommendations may therefore relate to considerations of recipient design: the multitude of respects in which the talk by a party in a conversation is constructed or designed in ways which display an orientation and sensitivity to the particular other(s) who are co-participants (Sacks, Schegloff, & Jefferson, 1974). As Hudak, Clark, and Raymond (2011) argue, while appearing to be the doctor’s proposal alone – recommendations may reflect a patient’s treatment preference. Action formulations projecting patient decisional responsibility (proposals and offers) may also be frequently declined because they are ‘easier’ to resist epistemically: they do not require patients to disalign with an ‘expert’ view of treatment. Pronouncements and suggestions, by contrast, have an alternative common denominator. They were both frequently designed to convey or embody endorsement, e.g. ‘I think/recommend’, denoting ‘ownership’, or formulated as informings or directives. Indeed, in primary care there was no difference in the rate at which patients resist suggestions and pronouncements, as noted perhaps ‘because physicians retain epistemic authority over the recommendation itself’ (Stivers et al., in press, p. x) in both of these action types. Meanwhile, proposals and offers, foregrounding patient preference and partnership, were more frequent in psychiatry than primary care – and thereby so was the degree of resistance.

This may be indicative of psychiatrists tailoring their recommendations to the unique circumstances of this setting. In the case of antipsychotic medication, ‘expertise’ may be somewhat subverted: the assumption that ‘doctor knows best’ is tempered by the subjectivity of side effects over which psychiatrists cannot hold true epistemic authority. Patients may need to disalign from particular courses of action accordingly. In neurology settings, Monzoni, Duncan, Grunewald, and Reuber (2011) found patients’ resistance was maximal when topics were within their own ‘epistemic remit’ (p. 196). Indeed, in extract 3 the patient displays a high degree of entitlement to decline treatment – resisting, quite directly, and drawing on the possibility of side effects to support this resistance. The frequently enduring nature of anti-psychotic treatment, vulnerable to intolerable side effects, inefficacy (Lieberman et al., 2005), and divergent evaluations by patients (Hellewell, 2002) creates a domain in which patients may retain the ‘specialist’ knowledge. Given that psychiatrists’ informedness of such effects can only be derivative, i.e. ‘2nd hand’, recognising patients’ epistemic primacy by offering or proposing treatment to ‘allow’ resistance vs. endorsing or pronouncing where this may be more difficult (at least in affiliative terms) may be mutually beneficial in terms of achieving appropriate treatment decisions.

Agreement about treatment is often assumed and cited (e.g. Martin, Williams, Haskard, & Dimatteo, 2005) as intrinsic to adherence. This has connotations of psychiatrists and patients being mutually aligned in discussions about treatment. However, patients did not routinely accept recommendations, and resistance was frequent. Thus, our findings suggest that perfect alignment will not always be reached, and in fact may not be desirable. Some degree of misalignment or conflict between the views or actions of psychiatrists and patients may be necessary if true collaboration is to occur and a variety of treatment options, ways to adhere to them, and associated contingencies are to be jointly considered (Katz, 1984). Building on this work by examining the nature of patient’s resistance in detail – and the association between decision-making sequences and subsequent adherence – might help us tailor targeted interventions to improve both psychiatrist–patient communication and patient experience of treatment.

In the 18-month Clinical Antipsychotic Trials for Intervention Effectiveness (CATIE) study (Lieberman et al., 2005), a remarkable 74% of patients discontinued medication prematurely and the most common reasons for discontinuation were patient choice, lack of effect or intolerability of side effects. A great deal of attention has focused on methods to persuade patients to adhere to treatment decisions, without sufficient acknowledgement that avoidance of sometimes complex, costly, and unpleasant regimens may be entirely rational (Mitchell, 2007). By providing a more holistic picture – mapping a range of practices psychiatrists use to initiate decisions in naturally occurring consultations – support for this perspective is offered here. While pronouncements are the most common action types in US and UK primary care (Stivers et al., in press), our findings suggest that more ‘choice’ is evident within the treatment recommendation sequence in psychiatric outpatients consultations. Such differences are in line with recent research conducted in other secondary and chronic care settings that suggest that the treatment recommendation phases may be more complex than primary care (Koenig, Wingard, Sabee, Olsher, & Vandergriff, 2014; Toerien, Shaw, & Reuber, 2013). This provides the necessary basis for a more granular, situated, analysis of action types and uptake in order to fully explain the ways in which decision-making-in-action (Rapley, 2008) is accomplished and its consequences for patient uptake.

Funding Statement

The research was supported by funding from the UK Medical Research Council (G0401323) and National Institute of Health Research (PB-PG-0408-16279).

Acknowledgment

We gratefully acknowledge the support of all patients and clinicians who participated.

Funding

The research was supported by funding from the UK Medical Research Council (G0401323) and National Institute of Health Research (PB-PG-0408-16279).