Assessments in Psychotherapy with Suicidal Patients: The Precedence of Alliance Work

In this article the aim was to explore how therapists handle suicide risk assessment in connection with patient’s experience of alliance in psychotherapy. In a naturalistic, longitudinal study, 19 suicidal patients were interviewed three times: just before they started in therapy (T1), during the early phase (T2), and again after one year (T3). 17 therapists were interviewed separately at T2 and T3. The interviews explored the personal experiences of both patients and therapists, including their handling of suicidality and the quality of their relationships. An interpretative-phenomenological approach was used to analyze the research interviews case by case, as seen from the perspective of both therapist and patient in each dyad. Because of their duties within the Norwegian Healthcare System, all therapists are obliged to follow the National Guidelines in suicide risk assessment. The results indicate that there are two pitfalls for therapists: to avoid the topic of suicidality and using the suicide risk assessment in a rigid way. Both pitfalls have disturbing effects on the working alliance between therapist and patient. Most of the therapists were able to integrate personalized assessment procedures in the running dialogue. Our argument here is that in order to integrate assessment with therapeutic work, it is important for the therapist to establish ethical responsibility with patient. A modern philosophical perspective on ethics of closeness is discussed.


Introduction
argues that assessment instruments might be particularly important for psychotherapists when a patient may be suicidal. Nevertheless, most therapists recognize that the alliance work is equally important. Researchers have concluded that the most important priority in assessing and managing suicide risk is the establishment of a caring relationship: a strong alliance between therapist and patient (Michel & Jobes, 2011). We aim to study obstacles and challenges to the assessment and management of suicidal behavior during psychotherapy. Trust and engagement are at stake within therapeutic relationships between suicide-prone patients and their therapists.
Many clinicians experience a tension between protecting themselves from legal liability and maintaining an empathic, patient-centered position (Espeland et al., 2021).
Further explorations into the ways in which procedures for risk assessment may serve to disturb/disrupt the bond aspect of the alliance should, however, include questions about what the patient thinks. In terms of alliance work, the use of assessment procedures during therapy is about how therapists and patients together may come to agree on what therapeutic tasks work for them, in combination with how to maintain and strengthen their emotional bond (Michel, 2016). The challenge for the therapist is how to keep track of the risk of suicide in each particular case and simultaneously work on the therapeutic alliance. For the patient, the corresponding challenge would be how to join efforts to live, when there is also a desire to end their lives.
In suicide prevention programs and national guidelines, the priority of employing standardized risk assessment scales is maintained-even though the sensitivity and specificity of such instruments are low (Espeland et al., 2021). In Norway, as in many other countries, National Guidelines for Suicide Prevention include procedures for assessing suicide risk, descriptions of risk factors that should be examined, which health professionals are qualified to assess patients, and the situations in which risk assessment should be completed (Norwegian Directorate of Health, 2008). It is worth noting that the guidelines does not take into regard how different contexts and treatment formats may influence their applicability.

Theoretical Concepts and Procedures for Handling the Risk of Suicide
Some researchers and clinicians have explicitly identified a potential conflict for therapists regarding where to focus their attention: on the obligatory risk assessment or on patient's experience (Espeland et al., 2021;Schechter et al., 2013). The assessment situation is at worst more therapist-centered than patient-centered because the standardized procedures may be prioritized at the expense of patient's perspective.
Waern et al. studied how psychiatrists assessed suicidal patients in a qualitative study (Waern et al., 2016). They found that psychiatrists emphasized the importance of understanding the patient, one's own reactions and how the relationship influence the risk assessment. Waern et al. state that concerns of therapists could lead to "repressive" management decisions (p. 1). Negative feelings in therapists have been explored theoretically by Gabbard (2003) and Maltberger and Buie (1974), and they have been found to be greater in both patient and therapist when patient is suicidal (Perry et al., 2013;Yaseen et al., 2013). Such feelings may also potentially disturb the therapeutic alliance. Quite a few researchers have underlined the theoretical importance of establishing an alliance with suicidal patients (Jobes & Michel, 2011;Plakun, 2009), but empirical research about how the assessment procedures affect the therapeutic alliance is lacking.
These objections to obligatory standardized risk assessment are even more potent in light of recently published meta-studies on such assessments. Specifically, several studies show that therapists cannot reliably predict suicides in a clinical setting, even if they use the best and most updated assessment instrument (Chan et al., 2016;Franklin et al., 2017). The researchers' recommendations are not unambiguous; some recommend more research on risk factors (Franklin et al., 2017), while others question the negative consequences of heavily relying on assessment instruments (Bolton et al., 2015;O´Conner et al., 2011).
Adherence is another concept that we find interesting when it comes to therapists' "translation" of suicide risk assessment procedures into the therapeutic conversation. In psychotherapy research, adherence refers to the extent to which therapists use prescribed techniques in therapy. Researchers have hypothesized that the more therapists adhere to the prescribed technique, the better the outcome will be. Unexpectedly, the evidence is not at all consistent; adherence does not necessarily determine symptom change (Wampold & Imel, 2015;Webb et al., 2010). One hypothesis is that both the working alliance and therapist qualities such as responsiveness also have to be considered. Adherence and responsiveness may in some ways appear to be opposing qualities, and this might explain why rigidly following a specific methodology does not necessarily improve outcomes (Stiles, 2009). The concept of adherence might also be applied to the use of risk assessments with suicidal patients. There might be a conflict between adhering to procedures on how to complete risk assessment according to national guidelines and being responsive to what patients have in mind.

Ethical Considerations
Historically, sanctions and penalties after a suicide, and after suicide attempts, have gradually been replaced with an understanding of suicidality as a symptom of a psychiatric illness in western countries (Barbagli, 2014). Treatment is offered to the patient as well as protection, if the psychiatric illness is severe. A part of this development is that the responsibility or blame, in the aftermath of a suicide, is potentially assigned to the health professionals. Despite meta-analyses showing that assessment instruments do not predict suicides (Chan et al., 2016;Franklin et al., 2017), the therapist's responsibility is often evaluated in terms of how well they followed standardized assessment procedures, as specified in national guidelines. The research literature has barely addressed the complicated dilemmas therapists face in therapy: how to handle assessments, suicidality, and the therapeutic alliance, as well as therapist's fear of being blamed after a potential suicide.
The ethical question is not about finding the right tool for prevention. It is rather about how to create some transparency in the relationship where there may be some discrepancy in patient's inclination to carry out suicide and therapist's efforts to elicit reasons for living. In order to study the ethical responsibility in the therapeutic relationship where suicide might be at stake, access to separate interviews from both therapists and patients in the same case is needed, which we have achieved in this study. Further, such interviews should have a broader scope than simply identifying whether assessment instruments are useful or not. To answer questions about the handling of ethical responsibility, the analyses should preferably be based on repeated interviews across the course of therapy.

Design
In this naturalistic study, we use a longitudinal design. Patients were selected because they were hospitalized due to suicide risk, after a suicide attempt, or overwhelming suicidal ideations. The invitation to participate in the research study was mediated through staff at an acute psychiatric ward. As part of the procedure in this clinic, all patients were transferred to an outpatient therapist after their discharge from hospital. All patients were offered therapeutic follow-up before, and thus independent of, their participation in the study. Once the patient was assigned to a therapist, he or she was invited to participate in the research study. In this way, the inclusion of therapists was arbitrary. They were all affiliated with an Outpatient Psychiatric Clinic (DPS) or to a psychotherapist in an independent practice. In both cases, the patient paid a modest fee per session, with an upper limit per year (according to the regulations of the Public Health Services in Norway). Patients were first interviewed when they were still in the psychiatric ward but had been assigned to a therapist (T1). The second interview with patients was scheduled after the third session with the out-patient therapist (T2). At this point in time, the patient's therapist was also interviewed. The follow-up interviews about the course of therapy were conducted after one year, or earlier if the therapy came to an end before that (T3).

Participants
Of the 26 patients asked to participate in the study, three declined and two did not receive therapy as planned. Two dropped out of the research project after T1 because they changed their mind about participation. The remaining 19 patients completed their participation in the study and were included in the analysis. 14 of the patients were women, five were men, and their ages ranged from 20 to 61 years. The average age was 38 years, and the median was 37 years.
Our knowledge about existing psychopathology of the patients was only provided by patients and therapists in the research interview. They conveyed a range of psychiatric diagnoses, such as personality disorder, bipolar disorder or depression (many with more than one diagnosis), and some with co-diagnoses of drug dependence. As expected by all having been hospitalized at a psychiatric clinic, none of the patients were free of having a diagnosis. Among the 19 patients included in the study, nine re-established contact with a therapist they already knew, and 10 were transferred to a therapist that was new to them. After one year (T3) two patients had dropped out from therapy, and one therapy was ended as therapist and patient had planned. Five patients had changed therapists for reasons outside of their control, and the new therapists were included in the interviews.
The therapists, consisting of 17, were all psychiatrists and clinical psychologists, most of them with more than five years of practice. Most of them used an eclectic or integrative model, with its main ingredients from their primary theoretical orientation, which varied between cognitive, behavioural and psychodynamic. The duration of therapy was not decided at the onset, and the ending would be based on some kind of agreement between therapist and patients.
These therapists were knowledgeable about the National Guidelines for Suicide Prevention, and had access to assessment procedures (Norwegian Directorate of Health, 2008). If a patient dies by suicide during or after psychotherapy, national authorities will investigate how the risk assessments were completed by their therapist in the aftermath. Therapists were aware of these requirements, and might therefore consider some kind of assessment as mandatory. Still, they retained some freedom as professionals in how to assess.

Interviews with Patients and with Therapists
In the first interviews when patients were still in the hospital (T1), they were invited to talk about their previous as well as their ongoing experiences in treatment. This included their relationship with the therapists, as well as their own understanding of their suicidal crisis. The interviews moved from the patients' interpretations of their earlier experiences into expectations for the future when it came to mental health services and living their lives. After their release, in the early phase of the outpatient psychotherapy (T2), the interviews were more directly focused on their participation as a patient in the ongoing work with their particular therapist-including experiences thus far and prospects for the future. The interviews also returned to suicidality as a concern in the patients' lives at present. In the follow-up interviews (T3)in most cases after one year-the interviews focused on what had happened in therapy, and what the patient made of it. This means that all interviews had a broader scope than just experiences from being assessed or not.
The first author conducted all the interviews with patients. In the last interview, the interviewer could refer to the earlier interviews, and keep track of what had changed-for better and for worse. Patients knew that the research interviews were separate from the treatment, and what they told would not in any way be communicated to their therapists or to the psychiatric ward.
The same kind of broad scope and sensitivity to details that could be relevant to direct or indirect forms of suicide risk assessment was applied in the interviews with therapists (at T2 and T3). Three experienced clinical psychologists were engaged in conducting these interviews. They had no knowledge about the actual patient in each case, which meant that the therapist had to go into detail for the interviewer. These interviews focused on what was happening in therapy as well as therapeutic interpretations and strategies. Therapists would meet the same interviewer on both occasions, which meant that the content of the first interview would serve as a point of reference for the state of affairs one year later.

Research Ethics
This study was approved by the Regional Committee for Medical Health Research Ethics in Norway (National Region South-East). We first obtained informed consent from the patients and thereafter from their therapists. All patients were offered therapy prior to, and thus independent of, their participation in the research study. In addition, the first author is an experienced clinical psychologist. She was affiliated with the Acute Psychiatric Department at the Psychiatric Hospital. Therefore, she was in a position to take action if she discovered elevated suicide risk during the research interviews with patients, to ensure that such risk was adequately handled in their treatment. All the participants presented in this article have consented to being quoted. All material are anonymized and all use of names are fictive.

Data Analysis
The first author took the lead in selecting and condensing the most relevant material from the set of five interviews from each therapy case (Malterud, 2001). All interviews were audio-recorded. As preparation for the subsequent interviews with each patient, the first author listened to the initial interviews that she had conducted herself. She then postponed listening to the interviews with the therapists until all data had been collected. Of the 19 cases, 17 were transcribed verbatim. The two remaining cases were listened to but did not seem to add new information: with the 17 cases, we concluded that the topic of study was theoretically saturated (Flick, 2018).
The aim of the analytical approach was to explore the informants' "inside perspective". How they make sense of their actions, thoughts, and feelings in being therapist and patient, respectively, in a psychotherapy where suicide might be at stake. We aimed at locating convergences and divergences in the accounts from the two parties, and the ways in which this may change over time (Flick, 2018). Based on the participants' stories, we identified relevant concepts to their experiences in order to develop a theoretical model for what assessment procedures might do to a therapeutic relationship directed at reducing or removing a possible threat to life. Our guidelines were consistent with the principles of the interpretative-phenomenological approach (Smith & Osborn, 2008). The first selection and condensation were basically all content in the interviews that was related to when and how suicidality was a topic in therapy-whether it be explicitly addressed between the two parties, or implicitly stated as in a topic one of the parties kept from the other party. Such utterances were explored further in order to determine if they at all seemed to impact the ways in which the two parties had worked together. The dialog from the research interviews was kept intact in these selected versions of relevant material from the cases.
Following the variations across cases brought forth the different ways in which therapists were able to balance seeing themselves as responsible for following the national guidelines about risk assessments on the one hand, and being able to employ their clinical judgement as a professional on the other. Only three among the 19 therapists used standardized assessment procedures at certain points during therapy. At the other end of the variation, four therapists ignored or intentionally avoided the topic of suicidality in any form. In the majority of the cases, therapists were able to integrate assessments into the therapeutic conversations. They did so in different ways, and it varied throughout the course of therapy whether or not they engaged in exploring suicide as a concern together with their patient.

When Therapist Uses Standardized Assessment Procedures
The three therapists who applied a standardized instrument in order to assess suicidality in their patient chose different procedures. The reasons for selecting an instrument, and when to apply it, actually emerged from how therapists viewed the context of their professional work. One therapist asked the patient the same standardized questions about suicidality every session. This procedure became a routine between them, and the topic was not otherwise addressed.
A second therapist introduced a SCID interview (Structural Interview of DSM Disorders) after many sessions. At this point in therapy, the therapist and the patient had touched upon suicidal ideations and actions several times and in different ways. Here, the introduction of a standardized diagnostic interview marked a deviance from their usual pattern of interaction. This initiative was motivated by questions posed by the therapist's clinical supervisor.
The third therapist also applied an assessment procedure for every session. The chosen instrument was the Beck Inventory for Depression, and this procedure was carried out by therapist alone-just after the therapy session took place. In all three cases, the life-situation of the patient did not pay a prominent role in the decision to identify the patient's "risk", nor did the quality of the present therapeutic relationship. The result of the assessment was mostly used as an effort to reassure the therapist.
These three patients interpreted therapists' efforts to register the risk of suicidal actions rather differently. In each case, their responses did not follow from the assessment method as such, but rather from their interpretations of what they were doing together. In the following section, we will present the one case in this category in more detail: We have selected to present one of the cases, Bente. With her the regular use of standardized questions came to exclude her wishes to address suicidality. Bente's male therapist worried about suicidality and was concerned about doing risk assessments properly. He had lost a patient to suicide two years earlier. Bente is a woman in her mid-twenties, diagnosed with depression, and was admitted to the psychiatric ward. Before the admission, she had been in therapy for the first time in her life, and she returned to this male therapist also after the discharge from the hospital. Her therapist thought that suicidality in general necessitated a special way of thinking about treatment in order to safeguard against suicide. He seemed to talk about suicidal patients in a general sense, and did not pay much attention to the particular ways suicidality is present in Bente's life.
Bente was bothered by the therapist's repetitive and strict way of handling risk assessment. This had an adverse effect on her trust in the therapist, but she kept her doubts to herself. She found the line of questioning by the therapist contrary to being open about the topic of suicide. Instead, she felt increasingly controlled. In the first interview (T1), Bente described the earlier therapeutic conversation-from the period before she was admitted to hospital and included in the research stud-like jumping around in ways she found very confusing. She says: B: And also, we haven't started the therapy. It has just been … I just wish we could get going with working, … there has been more focus on suicidal thoughts and plans.
The therapist admitted that he was stressed because of Bente's high level of suicide risk before her admission to the hospital. When Bente was discharged, the therapist felt the situation had abated. Even if Bente recognized that something had changed, and they now worked better with addressing difficult topics in her life, she felt that suicidality was still handled with stress and rigidity. She therefore gave up talking about it, even when the suicidal ideation became severe. According to Bente, her therapist was not sufficiently attuned to her initial wish for someone just to listen, as she described her boyfriend doing (T3). After a year (T3), the therapist said that suicide was not a topic anymore, although he assumed that it was still present in the mind of the patient.
Rather than preventing suicide, the therapist's procedures disturbed their alliance. As a result, suicidality paradoxically became silenced. A potential for transparency in the therapy was closed down.

When Suicide Assessment is Ignored, It May Also Lead to Avoidance of Suicide as a Topic During Therapy
Among all 19 therapists, four never referred to any kind of reflections concerning assessment procedures. It seemed as if they had reached a stance of non-use without having any explanation to offer. The sensitivity among these four in seeing what was at stake for the patient in this respect was actually rather different. Two of them avoided the topic in ways that made their vulnerable patient notice this avoidance as a professional habit. When we analyzed the interviews from these two patients-whom talked freely about the absence of suicidality as a topic in their therapy-it was salient that they were left in doubt about what to do about it. For these two patients, their trust in their therapist was damaged, and the distance between them increased during the course of therapy.
For the other two patients, the therapists' willingness to put aside turned out to be precisely according to their preferences. The therapists where deliberate in their decision to not assess suicide risk in each session, because they thought it hindered working on topics of change. The two patients felt their therapist to be very attentive. Putting the interviews with the patient and the therapist together, it was evident that the two parties shared some notions about what should and should not be addressed in therapy. They were building a therapeutic alliance and worked to facilitate processes of change along tracks other than just "reducing the risk of suicide".
We have selected one case where hopes and expectations of the therapist did not match the mindset of their patient. Specifically the therapist seemed to prematurely focus on self-efficacy. In Mona's case the therapist ignored suicide in ways that left her alone in shame. Mona's male therapist indicated that talking about suicidality was not part of their therapy. Using assessment instruments never occurred to him during therapy. In the research interviews, he described a sense of burden in his work situation (at T2 and T3), with many patients already on his list, and the expectations that he should take on even more patients every week. Further, he claimed that any focus on suicidality could have prevented Mona from moving forward.
Mona was a woman in her forties, with a traumatic childhood and a submissive personal style. She had been depressed since she was a young girl, and her troubles continued into her adult life, as she barely managed her work and had ongoing trouble in her intimate relationships.
Mona accepted that suicidality was not part of the conversations with her therapist during sessions. The therapist emphasized her resources for finding solutions to her trouble in life. In the last research interview (T3), he admitted that he had neglected suicidality as a concern, but he also justified his decision, as he believed evoking it as a concern might prevent progress in therapy.
Mona had discerned that suicidality was a topic not to be talked about, as it was too shameful and complicated, especially given the pressure she perceived in the working situation of her therapist. She would not be the one to take the initiative in talking about it, although she called it the "essence" of her trouble, an "unmentionable topic".
Mona was loyal to her therapist; she put the responsibility for avoiding the topic of suicidality on herself. She was not honest about her needs because she wanted to be a good patient; she wanted her therapist to think that she was doing fine in therapy. In all three of the research interviews, she only reluctantly mentioned that her therapist could have initiated talking about suicidality. However, her submissive personal style, combined with the therapist's somewhat narrowed focus in therapy, made her uncomfortable with having expectations or making demands in the therapy sessions. As a consequence, the topic of suicide was mostly avoided during therapy, although it was consistently presented in the interviews as an essential, possibly the most crucial, part of Mona's problems. Finding resources and looking forward became a kind of "as if" therapy in which Mona tried to be a model patient. She ended saying that she had not been honest and that she was the one to blame for the fact that therapy was not that helpful.

When the Therapist was Able to Integrate Assessment Procedures into the Therapeutic Conversation
In the majority of cases (12 of 19) the possibility of their patient being suicidal was something the therapists remained mindful of during and between sessions. These therapists worked to find different approaches to share this heightened awareness with the patient in ways that could engage him/ her and make it possible for the two of them to reach some kind of agreement about the present state of affairs. In this sense, the therapists' procedure for assessment was not standardized. Rather, they demonstrated making an effort to explore what where at stake when it came to creating an alliance for working together-even with the patient's potential suicidal ideations and actions as a possible intrusion in the background.
The therapists did not spend much time on deciding whether to use a standardized instrument. Still, they feared legal risks if their patient would die. They turned to the patient and tried to handle the issue so that it was transparent between the two of them. This transparency was established in order to build reciprocal trust and responsibility in the relationship.
The patients in these cases all placed some kind of trust in their therapist. From an early stage (T2)-and also later (T3)-they felt that that suicidality could be openly discussed when necessary. In some cases, considerations about when to actually talk about it and when to refrain varied from session to session. In other cases, the topic lost its intensity and "faded away" for longer periods. The patients felt free to take the initiative in talking about suicidality when it was a relevant issue for them. The therapists emphasized building trust and worked to facilitate a sense of responsibility in the patient for staying alive, or to speak up when this was overwhelmingly difficult. Several of the patients and therapists talked about how they made deals on telling when suicidal ideation became serious. Patients within this group conveyed a feeling of responsibility to stay alive. As one of the patients even nuanced that it was not only about staying alive, but that her integrity was at stake. It was for this patient essential to keep her promise toward her therapist.

Discussion
This is a longitudinal qualitative study of 19 psychotherapies with suicidal patients. The design of separately interviewing patients and therapists about their relationship, therapeutic work, handling of suicidality and risk assessment, provided the opportunity to explore how therapists implement risk assessment. The majority of therapists that did not use standardized procedures for assessment came to this conclusion because they experienced some tension between protecting themselves from legal liability and maintaining an empathic, patient-centered position towards their particular patient. Several of the therapists spontaneously talked about how the introduction of such an procedures did not resonate with the aim and scope of their therapeutic approach.
Overall the results made it clear that the guidelines for the assessment procedures did not aid therapists in making assessment integrated in the therapeutic alliance. Several therapist spontaneously talked about the struggle to find ways to assess according to guidelines at the same time as doing therapy. The guidelines seems to demand a way for working that therapists experience as imposing, limiting and estranged. We interpret the interview material from therapists to imply that the guidelines where experienced as an "authority" outside the therapeutic relationship. Our results suggest that finding solutions inside the relationship can expand the possibilities for building trust and enhance openness about suicidality. Risk assessment procedures within such a relationship have to be adapted to the patient, the context, and the chosen therapeutic methodology. This is in line with Stiles' claim about the importance of therapist's responsiveness in therapeutic conversations (Stiles, 2009), which may be applicable when it comes to understanding assessments of suicidal patients.
If the goal is to promote transparency about patient's mental pain, then it is probably therapist's sensitive attention towards their patient's inner state that matters. Following prescribed procedures about how to assess the patient may negatively interfere with this important task.
A robust therapeutic alliance enables therapist and patient to have a sense of commitment and transparency, sometimes implicit in their alliance, other times made explicit in their talks in therapy. The therapeutic alliance offers the participants a possibility to enter an "inside perspective" (Haugsgjerd et al., 2018), in which the implicit level of communication is made perceptible. This solution garners more information than an "outside perspective", which is often the focus in standardized procedures.
The therapist hopes and believes that their relationship will be gradually incorporated inside the patient, so that the therapist counts, even outside the sessions. In this way, the patient may feel loyal towards their therapist and their therapy; their conscience makes it more difficult to enact suicide. This may reflect an ethical demand, a term we find in the philosopher Knud Løgstrup's text about the responsibility that people have towards each other (Løgstrup, 1956). Trust is a priority for human beings. To establish trust in the therapeutic relationship, the therapist has to focus on the source of the patient's mental pain, and then remain present with this pain and open to possible solutions together with the patient. In such a relationship, the possibility for development, as well as openness and transparency about suicidality, may be tacitly enhanced. Patient and therapist each have their respective responsibilities; the therapist for facilitating trust in the relationship and the patient for staying alive. If the patient enacts a suicide attempt after the building of such a trustful relationship, this is a violation of the ethical agreement established between the therapist and the patient. In our material, some therapists made this deal explicit.
From the perspective of a therapist, when an ethical demand is established, with trust and transparency in the therapeutic relationship, fluctuations in relational quality are more likely to be acknowledged. If suicidal ideation or plans are at stake, these may be recognized and handled within the relationship. Trustful alliances also increase the likelihood that the patient will disclose suicidal issues spontaneously or be truthful about their inner state if their therapist asks. Bente's case shows how an inflexible handling of risk assessments can spoil the building of a trustful alliance and cause suicidality to be therapeutically unexplored. Specifically, Bente stopped telling her therapist about her suicidal ideation because she felt the topic was handled poorly, which resulted in a feeling of disappointment towards her therapist. On the other hand, ignoring the topic of suicide, as Mona's therapist did, may also cause suicidality to escape the therapeutic conversation.
Our recommendation to keep risk assessments within the therapeutic relationship does have limitations if the alliance is fragile. Nevertheless, we do not think that the solution is to place the responsibility for risk assessment onto authorities outside the relationship (as in standardized assessment procedures), especially given that the currently available assessment tools do not predict suicides in a clinical situation (Chan et al., 2016;Franklin et al., 2017). We also acknowledge the asymmetry that is always inherent in the therapeutic relationship, and, thus, the therapist always has a double responsibility: to enhance the therapeutic process for the benefit of the patient and to assess suicidality, whether the frames of therapy are sufficient or not. Nevertheless, since the therapist is not able to predict a suicide or to fully protect their patient, they cannot bear sole responsibility for keeping the patient alive; the patient also shares such responsibility. Working on the therapeutic alliance is maybe the most protective aspect in the effort to make therapy with suicidal patients safe. The patient's feelings of liability in his/her relationship to the therapist and transparency about the topic of suicide in their connection makes the topic of suicide appropriate to talk about when it is at stake. We have argued elsewhere that an open listening perspective on the therapist part probably strengthen therapeutic alliance by aligning the "private theories" patient and therapist have on problem and what is its therapeutic solution (Østlie et al., 2018).

Strengths and Limitations of the Study
We do not have objective data (audio or videotapes) to indicate how risk assessments were actually conducted in the therapy sessions, nor have we reviewed the documentation of assessments in the patient's journal systems. What we do have is both patients' and therapists' subjective experience as described in the research interviews, based on their recalled memories from the sessions. The two independent viewpoints provide an indication of what may have transpired between them during the therapy sessions. Audiotape and videotape material could only have given us data about how assessments were performed: audibly or visibly. The heart of this study is about how the participants perceived their alliance and the procedures of assessments: this is the most important issue when studying the alliance-work, and it is not possible to capture without asking the participants themselves.
Another limitation is that none of the included therapists used a standardized method of therapy for their patient. It might be that therapists who follow a specific therapeutic methodology find standardized assessments helpful if the method of assessment fits with their therapeutic one.

Conclusion
When choosing procedures for assessments of suicidality, therapists have to consider how these assessment procedures may influence therapists and the therapeutic relationship. The alliance offers access to the implicit level of communication between patient and therapist because it is the inside perspective in the relationship that is emphasized, in contrast to the outside perspective enhanced by standardized procedures. Standardized assessment procedures are not able to predict suicide in particular cases in a clinical situation. Thus, even though a strong therapeutic alliance is no guarantee against suicide in psychotherapy, it may be the best tool therapists have at their disposal. We propose the necessity of considering both the costs and benefits of implementing obligatory standardized risk assessment tools, as both sides have to be carefully and thoroughly considered. It is also important to enhance the power and status of the alliance work, as well as the ethical aspects to be considered in therapists' handling of suicidality. Future research should in our opinion aim at studying the potential that lies within the therapeutic alliance in handling suicidality. Prevention of suicide in a therapeutic setting can be researched as an extension of the already established psychotherapy research. Work with therapeutic alliance prevents suicide.