Asking about self-harm and suicide in primary care: Moral and practical dimensions

Objective: Self-harm and suicide are important topics to discuss with people experiencing mental health conditions. This study explores how such discussions unfold in practice, and how their moral and practical repercussions manifest for patients and doctors. Methods: Conversation analysis (CA) was used to examine 20 recorded examples of doctors’ questions about self-harm and suicide and their ensuing discussions with patients. Results: A tendency to frame questions about self-harm towards a ‘no’ response, to amalgamate questions around self-harm and suicide, and to limit dialogue around the protective factors offered by family and friends restricted discussion of patients’ experiences and concerns. Closed questions about thoughts and actions in the context of risk assessment resulted in missed opportunities to validate distressing thoughts. Patients responding affirmatively often did so in a way that distanced themselves from the negative stigma associated with suicide. Conclusion: The wording of questions, along with negative stigma, can make it difficult for patients to talk


Introduction
Suicide is a major public health concern, accounting for around 800,000 deaths worldwide annually, and being the second-leading cause of death among those aged 15-29 [1]. In the UK, there were 6507 registered suicides in 2018, 75 % of which were among men [2]. A repeated, cross-sectional survey study suggests that nonsuicidal self-harm (NSSH) has tripled since 2000 [3]. The 2019 UK Cross-Government Suicide Prevention Workplan noted, amongst other things, the importance of identifying suicide risk factors, including self-harm [4].
Around 45 % of people who die by suicide have seen their General Practitioner (GP) (primary care physician/doctor) in the month before death [5]. Thus, asking about suicidal thoughts in a broader context of risk is important for general practitioners seeing patients presenting with mental health concerns. Self-harm typically 1 refers to an intentional act of self-poisoning or selfinjury, irrespective of the motivation or apparent purpose of the act [6]. Berman and Silverman [7] similarly suggest that, in the UK, 'self-harm' is used to refer to both NSSH and suicidal behaviour. This is the definition we use henceforth.
Asking about self-harm is sensitive for practitioners, who worry that raising these topics could exacerbate existing suicidal thoughts, or even put these thoughts in the patient's head [8,9]. Research has identified practices that health professionals use to frame questions about self-harm in a range of settings in ways that deal with this sensitivity [8,[10][11][12].
Other research has highlighted how questions about self-harm are worded and how this influences patients' responses. In primary care, McCabe et al. [13] found that 75 % of doctors' questions were worded in a way that invited a 'no self-harm' response (e.g. "No thoughts of harming yourself?") rather than a 'self-harm' response (e.g. "Do you feel that life is not worth living?"). When questions invited a 'no self-harm' response, patients were more likely not to report self-harm (although patients were not more likely to offer a 'self-harm' response when the question invited one). A study coding transcripts, on the other hand, found that only 29 % of questions were designed for a negative response in primary care [14]. This study also found that 48 % of questions were worded in a direct, non-allusive way and that only 24 % of questions came with framing (i.e. a statement preceding the question that "serves to normalize" (p. 35) it).
To date, empirical research in primary care (comprising only three studies) has focused narrowly on how the wording of questions impacts on patients' immediate responses. Furthermore, only one study [13] has used a method that is attuned to key details in the interaction.
There has been no research on the wider discussion of self-harm within consultations. Particularly when patients respond in the affirmative. Guidelines for practitioners recommend that doctors carry out a risk assessment based on factors such as seriousness of intent, frequency and duration of thoughts, and plans for selfharming [15]. They suggest this be done in a stepwise manner, beginning with questions about suicidal ideation before moving on to questions about planning and, finally, intent [16]. It is also recommended that doctors identify 'protective factors' that could prevent patients from acting on their thoughts of self-harm (e.g. children) [8,12,15,16] and, if necessary, make a referral to secondary services.
The aim of this paper is to analyse how self-harm is discussed in primary care, focusing on (1) how doctors ask about these topics, (2) the conversations that ensue following these questions, and (3) the moral issues and conversational difficulties arising when discussing these topics.

Methods
This research was undertaken as part of the DeStress Project [17] on mental health in low-income communities.

Data
The dataset for this study comprised 52 recorded primary care consultations for mental health taken from the One in a Million data archive, along with additional demographic and other data about the participants collected at the time of recording [18]. The recordings were collected in the southwest of England between in 2014 and 2015 and feature patients from a range of socioeconomic backgrounds. Ethical clearance was obtained for the original collection of these consultations [19] and for their re-use in this study (from the NHS Cambridgeshire and Hertfordshire Research Ethics Committee). The 52 consultations were selected from the wider One in a Million dataset (300 consultations) based on their International Classification of Primary Care, Second edition (ICPC-2) codes (see Appendix 1), with the selection criteria being for consultations based on mental health conditions such as anxiety, depression, and stress. These consultations represented the entire subset of such consultations within the One in a Million archive. Patients in these consultations were at different points in their treatment; some were presenting with mental health concerns for the first time, while others had long-term, diagnosed conditions that, for example, had recently been exacerbated by life events.

Analysis
The 52 consultations were screened for cases in which doctors asked patients about suicide and/or self-harm. Both the questions themselves and the surrounding talk were transcribed using Jeffersonian [20] conventions, which capture not only what was said, but how and where it was said (see Appendix 2). The relevant consultations were also examined in full to determine whether the doctor's question marked the first mention of self-harm or if the topic had been raised earlier. In total, there were 20 questions about self-harm spread across 18 consultations (i.e. there were two consultations in which two distinct questions were asked; see Table 2). On average, doctors in these consultations had been qualified for 18 years (sd = 7.05) 2 , with an average age of 46 (sd = 5.86). Half the patients (n = 9) reported that the doctor in their recording was not their regular GP, with 4 reporting that it was, and 6 not responding.
The recordings were micro-analysed using conversation analysis (CA), an approach that considers how speakers use spoken language and non-verbal behaviour to perform actions (e.g. asking questions, making requests, making assessments). In this analysis, we draw upon research on questioning in medical interaction which shows how seemingly small differences in a question's wording can be consequential [21,22].

Doctors' questions
In all but one consultation (see extract 7 below), the doctor's question marked the first mention of self-harm within the consultation. Most questions in the data followed a similar pattern. Examples of this pattern can be seen in extracts 1-5 (Table 1).
There are three systematic features of these questions. The first is the use of the term 'self-harm' or a variant thereon. While this term might be associated with behaviours such as cutting rather than suicide, in these data it acted more as an umbrella term for both behaviours (see Berman and Silverman [7]). While we present more evidence for this below, the examples above indicate a lack of specificity as to which type of behaviour is being asked about. When doctors did separate out the two behaviours, they would often still bundle them together as part of the same question, as in extract 5.
In cases where the two behaviours were asked about separately, there was usually a clear reason for it, as in extract 6 ( Table 2).
Prior to the extract, the doctor has said that she "needs to go through some questions" with the patient, who is presenting with (previously diagnosed) bulimia. Her questions are not comparable to the other examples, then, because she is institutionally mandated to ask them in a certain way (i.e. separated out).
Indeed, she appears to be at least partially reading the questions from some sort of checklist.
The second common feature across most questions is the focus on 'thoughts' or 'feelings' of self-harm rather than the action of self-harm. When this was not the case, there was usually a reason. In extract 7 (Table 2), for example, the wording of the doctor's question ("Have you been self-harming?") is attributable to the patient having said that he has been self-harming earlier in the consultation.
The third common feature is that questions in the data were mostly polar (i.e. inviting a yes/no response) and framed negatively. A clear example of this can be seen in extract 5, above: "But you've not had any thoughts of harming yourself or suicide or anything like that." The phrasing of this question as a negative declarative statement with questioning intonation ("But you've not had any . . . ") is clearly slanted in favour of a negative response. A Table 1 Standard questions. Table 2 Non-standard questions. more subtle form of negative framing was to use negative-polarity items such as "any" (extract 3), "ever" (extract 6), and "at all" (extracts 2), all of which indicate an expectation for a negative response. 3 To summarise, questions typically: 1) Initially focused on thoughts and feelings. 2) Amalgamated enquiries about self-harm and suicide in one tightly coupled question. 3) Were polar and negatively framed.

Patient 'no' responses
'No' responses fell into two categories: unambiguous (n = 5) and ambiguous (n = 5). Unambiguous 'no' responses were emphatic, whereas ambiguous 'no' responses were typically more hesitant. Examples of unambiguous 'no' responses can be seen in extracts 8 and 9 (Table 3).
In both extracts, the patients have not only answered 'no' but have emphasised this response through repetition ("No . . . No . . . No", extract 8) and additional statements ("Absolutely not", "I enjoy my life", "I wanna be here", extract 9).
Ambiguous 'no' responses can be seen in extracts 10, 11 and 12 (Table 4). In extracts 10 and 11, the responses are delayed (e.g. the gaps at lines 2 and 5, respectively) and hesitant (e.g. the particle 'pt' at lines 3 and 6). Features such as these are well-established signals that a speaker might disagree with what is being proposed in a question [24]. In extract 12, meanwhile, the patient's response is purely non-verbal, which has been treated as weaker than a verbal response would be [25]. Given that the questions in all three of these extracts are framed for 'no' responses, the features we have identified suggest that the patients may be struggling to respond in the negative and may wish to respond in terms more nuanced than a simple 'no' would be.
Interestingly, though, these features were not treated as consequential. This can be seen in extracts 13 and 14 ( Table 5). The former is unambiguous, the latter is ambiguous, but what happens after the response is the same: the doctor changes the topic by asking a new and unrelated question. In neither case was self-harm raised again for the rest of the consultation. This pattern occurred in all but one 4 of the 'no' response cases (n = 9).

'Yes' responses
Unlike 'no' responses (which were usually standalone), 'yes' responses were always either followed up with additional information and context or were delivered in a narrative format. In some cases, this additional information upgraded the severity of the patient's response. An example can be seen in extract 15 (Table 6), where the patient delivers a 'yes' response before adding that he thinks about harming himself "all the time." In other examples, patients did the opposite by immediately downgrading the severity of their responses.
The patient in extract 16 (Table 7) responds affirmatively to the doctor's question at lines 6À7: "Yeah I've had some dark moments." She follows this up, however, by emphasising that these moments are fleeting (lines 7-8): "But I have-I tend to move out of those quite quickly." In extract 17 (Table 7), meanwhile, the patient does not offer a 'yes' but instead responds with a narrative that is inherently downplayed: "I have in some ways." Both extract 16 and 17 also, like most of the 'yes' responses in the data, feature hesitation (e.g. "U:::m::" in extract 17). Why were 'yes' responses routinely so hesitant and downgraded? In these cases, patients are reporting self-harm when the question invited a 'no self-harm' report. In going against the question, they are offering answers that, in their phrasing, "transform the question retrospectively" and thus "show which aspects of the question are problematic" [26] (p. 21). The downgraded and hesitant nature of many of the 'yes' responses can thus be attributed to their being offered to questions framed for a different response.
Another reason for patient hesitation and downgrading can be seen in extract 18 ( Table 8). The patient here describes how she felt "really really low" (line 6) and "did wanna end it" (line 7) a few days before. At lines 7-9, 10 and 12, though, she states that she was prevented from acting upon this impulse when she thought of the impact that her uncle's suicide had on her family. In downgrading Table 4 Ambiguous 'no' responses.

Table 5
After a 'no' response.
her response, then, the patient has touched upon a fundamentally moral 5 concern; she is presenting herself as someone who thinks about her family and the effect that her actions could have on them.
In extract 18, the patient's family history has made such concerns especially germane. Yet even in the absence of such history, patients presented moral concerns as a rationale for not acting on their thoughts. In extract 19 (Table 8), for example, the patient describes at lines 2-3 and 6 how she was "very tempted to take all [her] tablets" but did not do so. The doctor at line 7, asks the patient "what stopped [her]." The patient's response at lines 8 and 10-"My mum . . . [g]etting really upset"-again presents her decision as having been motivated by a moral concern about her family.
The doctor takes up and summarises that concern at line 12: "You wouldn't want to upset her." Yet it is worth noting that the moral element was always brought in by the patient rather than the doctor. Moreover, there is little evidence that doctors were especially concerned with the moral implications of the patient's suicidal thoughts. Indeed, the relevance of patients' families in this context is as a 'protective factor' [16] that would stop them from acting upon their thoughts. From a risk management perspective, then, the patient's moral concerns are transformed into something more practical-what matters is that there is something in place to prevent the patient from ending their life. 6 There were, however, rare moments when doctors did take up the moral implications of the patient killing themselves. The strongest example of this can be seen in extract 20 (Table 8), in which the doctor describes, using morally laden language, how ending one's life "rebounds on other people", noting that it is "a terrible thing to leave . . . other people with" (lines 9-10, 12, 14 and 16).
We should also emphasise that, while moral concerns were the most common stated preventative factor, this was not always the case. In extract 21 (Table 9), the doctor asks the patient if there are "some good reasons to not harm [himself]" (lines 1-2) and the patient responds by mentioning his son (line 6). However, the patient has not mentioned how having a child would prevent him from killing himself; he has not, in other words, stated outright that he is thinking in moral terms about the negative impact that his suicide would have on his son. And indeed this is not the only implication that the doctor draws out in his follow-up at lines 10, 15 and 17. He describes the patient's son as a "very good reason to stay alive", noting that "he wants a dad and you want to grow up with him." He thus shifts to a utilitarian argument that invokes benefits to both parties.
This extract is broadly similar to previous ones, therefore, in that a family member is harnessed as a protective factor in the patient's life. Instead of simply being harnessed as someone who will be upset by the patient's suicide, or lose the benefit of having a father (negative, moralistic, and utilitarian framing), they are instead harnessed as someone who enriches the patient's life and, therefore, makes it worth living (positive and non-moralistic framing, though still utilitarian). This focus on the patient's enriching role in his son's life is in line with the Interpersonal-Psychological Theory of Suicidal Behaviour [27], which suggests that feeling like a burden to others is a key factor in increased suicide risk. 7 To summarise, patients who responded affirmatively to doctor's questions about thoughts of self-harm typically did so in a way that downplayed the likelihood that they would act upon these thoughts. We have explored patients' reasons for downgrading, but there is an additional point to be made about this phenomenon: thoughts were rarely focused on as a topic in themselves.
Consider again the patient's response in extract 19 (Table 8). In the first part of this response (lines 2-4), the patient responds directly to the question about having had thoughts of self-harm.  This can be seen in her use of the words "tempted" and "feeling", both of which refer to subjective mental states. Strictly speaking, this would suffice as an answer on its own.
With her follow-up at line 6, however, the patient shifts to talking about not having acted upon these thoughts. This shift indicates that the patient is treating the question as having been about thoughts as a precursor to action and is working, in turn, to pre-emptively answer this implied question. This practice of answering more than was asked in the original question can be a way of pre-empting "negative inferences which might otherwise arise from unelaborated answers" [28] (p. 155). This is not unique in the data: as we have seen, most discussion following a 'yes' response tended to be about actions rather than thoughts.
Practically, this makes sense-risk assessment is, again, an important part of discussion around these topics [16], and it would be unusual for a doctor to ask about thoughts and feelings of selfharm without exploring the possibility that they might lead to action. However, the majority of suicidal ideation does not lead to an attempt [29], and modern 'ideation-to-action' theories of suicide emphasise the distinction between ideators and attempters [30,31]. Predominantly focusing on action over thought following a patient's response can thus have implications, as can be seen in extract 23 (Table 10) (which comes just prior to extract 20).
At lines 1-3 and 5, the doctor asks the patient if he ever makes "plans to harm [himself]" or if it's "just thoughts coming into [his] head." The patient responds that "it's just mainly thoughts" rather than plans (line 6). However, as already seen in extract 20, the patient then describes how he nonetheless finds such thoughts distressing, especially when he thinks about friends and family members.
From a risk assessment perspective, then, the patient's response is reassuring-he has indicated that he is unlikely to act upon his thoughts. However, he has also said that these thoughts are distressing for him in themselves, irrespective of the extent to which they are a precursor to action. This potential was rarely discussed in the consultations. Indeed, as can be seen across these examples, the typical pattern following a 'yes' response was for discussion to move on to the identification of protective factors. Once such factors had been successfully identified, the discussion would move away from self-harm until the end of the consultation, where the doctor would sometimes remind the patient of the procedures that they could follow (e.g. calling the surgery) if such thoughts re-occurred. In short, self-harm was overwhelmingly discussed as a practical problem rather than a matter for 'therapeutic' discussion. This is in line with Jerant et al. [32], who highlight an action-oriented approach in primary care discussions about self-harm. They also suggest that patients can find this approach to be lacking in empathy.

Discussion
In the cases we analysed, talk about self-harm was usually initiated by a doctor's question. These questions routinely had three features: an initial focus on thoughts and feelings (before asking about behaviour), grouping self-harm and suicide into one question (rather than asking about them separately), and a closed yes/no framing that usually asked the patient to report 'no thoughts of self-harm' [13].
When patients reported no self-harm, they either did this with emphatic 'no' responses or more ambiguous 'no' responses featuring delay, hesitancy, or a weak non-verbal response. From what we know about how people display disagreement in interaction, some patients displayed difficulties responding and could not provide a clear no. In these cases, further discussion to unpack these thoughts would be indicated. Nevertheless, 'no' responses saw the topic being closed down in nine out of the ten cases.
'Yes' responses were different. Sometimes, patients said 'yes' in a way that was as strong as the emphatic 'no' responses (e.g. "Yeah Table 9 Positively framed reason to stay alive. all the time", extract 21). In other cases, though, patients offered hesitant and qualified 'yes' responses that confirmed that they had been having negative thoughts while downplaying the likelihood that they would act upon them. One reason for this was disagreement with the proposal that they were not thinking about self-harm/suicide. When a patient answered in the affirmative, the discussion typically became focused on practical matters (i.e. risk assessment and management) to the neglect of the more subjective, therapeutic matters suggested by the original question.
However, the data makes it clear that, even in the absence of immediate suicide risk, patients could nonetheless find thinking about suicide to be distressing. We would suggest, therefore, that thought and action be more clearly separated out so that patients' subjective feelings can be explored and validated on their own terms rather than as precursors to action. Making this distinction within the consultation would be in line with wider attempts to distinguish suicide ideators from suicide attempters [29][30][31].
It was also clear that patients worked to distance themselves from the negative moral implications associated with suicide. Usually, this was related to their family. From a medical perspective, such concerns could act as a 'protective factor' [16]. However, there were two ways in which a patient's family could be framed as such: negative (i.e. 'I/You don't want to upset my/your family') and positive (i.e. 'I/You want to be with my/your family'). Given the prevalence of damaging stigma around the purported selfishness of suicide [33], the latter form of framing is more beneficial.
Prior to the analysis, we established the diversity of our data in terms of both demographic and other variables (e.g. strength of doctor-patient relationship, patient's current treatment stage). There is also data that we did not have access to that may be relevant to our findings (e.g. GPs' levels of experience around suicide risk assessment). We are aware that this may have weakened our findings, especially given demographic disparities (especially gender) in suicide risk. However, we would emphasise how consistent the patterns that we have observed have been despite this heterogeneity. This point also applies to the size of our sample; while the sample was relatively small, the degree of consistency we have observed within this sample leads us to believe that the conclusions that we have drawn are valid. However, the findings may still have been impacted by volunteer bias (i.e. participants who agreed to be recorded may not be representative).

Conclusions
Questions about self-harm in primary care are overwhelmingly framed for a 'no' response. This makes it difficult for patients to answer affirmatively. Adding to this difficulty is the negative stigma associated with self-harm, which patients seek to distance themselves from in their affirmative answers. A strong focus on preventing patients from acting upon their thoughts of self-harm can leave little room for discussion of those thoughts as a source of distress in themselves.

Practice implications
Discussions could be improved by asking about self-harm and suicide separately, encouraging further discussion when responses are ambiguous and acknowledging distressing thoughts as potentially problematic in themselves. Invoking negative reasons not to end one's lifesuch as the legacy it leaves for familyis problematic when it exacerbates shame and low self-worth. Such negative stigma could be reframed by exploring patients' positive reasons for wanting to stay alive.

Data access statement
The data on which this paper is based did not involve the collection of new data. Anonymised interview and focus group transcripts from participants involved in the wider DeStress study who consented to data sharing, plus additional supporting information, are available from the UK Data Service, subject to registration, at: 10.5255/UK-DA-SN-853788.