The Desire to Die in Palliative Care: A sequential mixed methods study to develop a semi-structured clinical guideline

Objective: Although desire to die (DD) of varying intensity and permanence is frequent in patients with advanced cancer and those receiving palliative care, uncertainty exists concerning appropriate therapeutic responses to it. To support health professionals in dealing with patients´ potential DD, a training program and a semi-structured clinical interview guideline was developed. This study aimed for a revision of and consensus building on the clinical guideline to support proactively addressing DD and routine exploration of death and dying distress. Methods: Within a sequential mixed methods design, we invited 16 palliative patients to participate in semi-structured interviews and 377 (inter-)national experts to attend a two-round Delphi process. Interviews were analyzed using qualitative content analysis and an agreement consensus for the Delphi was determined according to predefined criteria. Results : 11 (69%) patients from different settings participated in face-to-face interviews. As key issues for conversations on DD they pointed out the relationship between professionals and patients, the setting and support from external experts, if required. A set of 149 (40%) experts (132/89% from Germany, 17/11% from 9 other countries) evaluated ten domains of the clinical guideline. There was immediate consensus on nine domains concerning conversation design, suggestions for (self-)reflection, and further recommended action. The one domain in which consensus was not achieved until the second round was “proactively addressing DD”. Conclusions : We have provided the first semi-structured clinical guideline to identify and address DD and to respond therapeutically – based on evidence, patients’ views and consensus among (inter-)national professional experts. Trial registration: German Clinical Trials Register (No. DRKS00012988).

Current recommendations suggest proactively addressing DD, 6 referencing studies which found positive effects on patients. 8,15 These effects include opening communication about emotional conditions, even in absence of DD. 16 Additionally, DD may also be assessed by validated instruments such as the Schedules of Attitudes for Hastened Death (SAHD; primary for research purposes) 8 or the Desire for Death Rating Scale (DDRS; initially developed for clinical interviews). 17 Training programs and recommendations for talking about difficult issues 18,19 may also support HPs in dealing with DD. 20 A training program regarding DD has been developed in a previous project, based on a literature review and results of focus groups with multiprofessional German PC providers. 21 This training program has been piloted and evaluated through a 32-item-scale covering the dimensions of self-confidence, skills, knowledge and attitudes. Descriptive changes indicate a major improvement in self-confidence and at least minor improvements on all dimensions after three months, with only one item concerning knowledge reaching statistical significance, though. 21 Within the development of the training program, a clinical guideline was drafted, building on published recommendations on dealing with DD 22, 23 and based on results from an interdisciplinary advisory board discussion. The draft was refined and structured within a framework analysis approach. 21 The present study aimed at further developing and refining the clinical guideline for routinely assessing death and dying distress, reacting to and (proactively) addressing DD -taking into account international expertise from a wider range of professions, patient representatives and relatives as well as the voices of patients.

Methods 6
A sequential mixed methods design was used including qualitative patient interviews and a Delphi survey with experts (see Fig. 1). Both methods were chosen to maximize the guideline's clinical relevance. More details on the background on and justification of the methods selected are described in the study protocol. 24 We applied the consolidated criteria for reporting qualitative research (COREQ) 25 and the guidelines for conducting and reporting Delphi Studies (CREDES) in PC. 26 Research was conducted according to the Declaration of Helsinki. Ethical approval for this study was obtained from the Ethics Committee of the University of Cologne (#17-265).

Patient interviews
Patient interviews were conducted to consider the perspective of palliative patients and thereby strengthen the appropriateness of the guideline. A convenience sample of 16 adult patients receiving PC was invited to participate in face-to-face interviews on the appropriate approach to DD in PC. Participants were recruited via PC providers known to the research team. A leaflet for patients and HPs included survey aims, procedure and contact details of the research team. HPs contacted the researchers in case of patients' willingness to participate. The research team approached these patients to obtain informed consent for study participation.
Interviewees chose the location of the interview, which were conducted by either one of three female (KK, VR, KB) or one male (GF) interviewers. Two researchers hold doctoral degrees (KK, GF), one was a doctoral student (VR), the other a graduate student (KB). At the time of data acquisition, all researchers worked for the Department of Palliative Medicine at the University Hospital of Cologne and were trained and experienced in conducting semi-structured interviews. This approach to data acquisition allowed the research agenda to be pre-defined to some extent and at the same time enabled respondents to freely present a range of views and new insights. 27 The interview guide was developed by a group of social scientists (KK, GF, MG), psychologists (TD, KB), a pedagogue (VR) and a physician (KMP). It assessed aspects relevant to addressing DD, desirable traits in interviewer as well as interviewees' potential personal DD. The complete interview guideline is added as an additional file (see Additional file 1). Social demographics were assessed using a brief questionnaire.

Delphi survey
In addition to patient interviews, an online Delphi survey was performed to achieve expert consensus on content and structure of the clinical guideline. We invited participants from 13 countries aiming to compose a balanced sample of 50-70 researchers and clinical practitioners. Recruitment took place through personal contacts and an internet search.
Potential participants were asked via email to take part in the survey and to propose further experts. As nurses were initially underrepresented in recruitment, we reached out to the nursing mailing list of the German Society for Palliative Medicine, which distributed our call to a selection of its members. This pushed the number of invited participants to 377.
To develop the Delphi questionnaire, we revised the existing version of the clinical guideline 21 based on our preceding patient interviews. The revised draft included three sections (B to D in Table 2) with a total of ten domains each containing a set of suggestions for HPs to take into account when (proactively) discussing DD with palliative patients. The Delphi survey was designed accordingly and conducted in co-operation with an external partner (UZ Bonn, Society for Empirical Social Research and Evaluation).
Each panelist received a link to the survey via email. During both Delphi rounds two reminders were sent out to motivate non-respondents to participate. During the first Delphi round, participants evaluated each domain of the guideline by a) rating the importance of the entire section presented on a five-point Likert scale ranging from 0 ("very unimportant") to 5 ("very important") and b) giving free text comments on the suggestions listed in the domain they just rated. In the second Delphi round, participants were asked to reevaluate the guideline in light of the results of the first round.
Participants were also fed back information about the first rounds' sample seize such as distribution of professions and the sample's experience with DD and suicidality.

Data analysis
Patient interviews were transcribed verbatim and analyzed by three researchers (KB, KK, GF) applying qualitative content analysis. 28 Inductive and deductive categories were derived and applied using the qualitative data analysis software MAXQDA 12. Inter-coder reliability was ensured by constant comparison between coders and according adjustments throughout the entire process. Transcripts were not returned to participants and they were not asked to provide feedback on the findings.
Delphi panelists' data was eligible for analysis if (inclusion criteria): panelists were patient representatives, relatives or part of the relevant occupational group (multi-professional experts in the field, reported at least ≥ 5 years of experience in dealing with DD and/or suicidality, and reached high scores in self-assessed confidence and knowledge with DD and/or suicidality (rating at least four on a 0 to 6 Likert scale on one of the four items)).
Data from the Delphi survey was analyzed statistically for quantitative results using SPSS 23 (IBM Corp., Armonk, NY, USA). Thematic analysis was applied to analyze free text comments. An a priori criterion of 80% agreement among panelists was defined to determine consensus 26 on the importance of a section, requiring a rating of at least four on a one to five Likert scale.
Further details on sampling, data collection and analysis are reported in the study protocol. 24

Patient interviews
Between 9/2017 and 1/2018, 11 (69%) patients receiving PC participated in face-to-face interviews lasting nine to 79 minutes. Reasons for patients refusing to partake or HPs judging the patient incapable of participation were (acute) deterioration of physical (3/5) or mental (2/5) health status. Table 1 presents further details on interviewees.
All but one of the interviewees reported DD either as a wish for hastened death or as acceptance of death without requiring to hasten it. A desire for physician assisted suicide was reported explicitly once, and two interviewees preferred death to deteriorating symptoms. Isolation, the feeling of being a burden, hopelessness and fear of pain were reasons reported for DD.
Deductive categories based on topics from the interview guideline were complemented with inductive coding of the interviews, producing five main categories with various subcategories: "Actively building the relationship", "Conversation partner", "Conversation context and structure", "Conversation set up and framework" and " Own Desire to Die ".

Suitability and usefulness
87% of the respondents (n = 129) valued the structure of the clinical guideline as '(very) suitable', 95% (n = 141) rated the clinical guideline to be '(very) useful' for everyday clinical practice. Table 3 displays the results of the importance ratings for Delphi round one and two.

Importance of individual domains
For all domains except 'proactively addressing DD', the a priori consensus criterion (≥80% agreement) was met during the first Delphi round (see Table 2). As only four domains did not reach 90% of agreement, we took reaching <90% to indicate potential for optimization and therefore asked our panelists to reevaluate all such domains (proactively addressing,  Table 4. No significant statistical differences in ratings for all domains were found among subsamples (e.g. gender, expertise, self-assessments).

Controversial Aspects
Aside from the relatively moderate agreement on 'proactively addressing DD' during the Delphi process, classifying DD as thoughts of terminating life prematurely, wish for assisted suicide or active euthanasia in the clinical guideline was also met with concern.
In the final version of the guideline, these classifications were nonetheless included but go hand in hand with introductory information that points out the broadness of the DD phenomenon. The introductory information in Additional file 3 shows this in more detail.

Discussion
(Proactively) addressing desire to die within trustful health professional-

patient-relationships
Since DD may be a potential way to cope with advanced disease at some point during the disease trajectory, 30 exploring it and allowing its emotional expression in conversation may be beneficial to all patients. DD can be expressed in different ways and proactively addressing it may help to clarify reasons more openly and at an earlier stage. However, in clinical practice, DD is not yet routinely assessed by HPs and talking about it is associated with discomfort. 20 Concerns about adverse effects of discussing DD, e.g. triggering suicidal thoughts in patients, are widespread 20 and became apparent in our Delphi survey. Current studies confirm that asking about suicidality causes no harm but may reduce experienced burden and distress. 31 Based on preliminary studies 20, 32 these findings can likely be extrapolated to PC. Currently, our clinical guideline is only available for HPs attending the related multi-professional training program. Within this training program, we present the aforementioned findings and the broad definition of DD prior to handing out the guideline. This is meant to alleviate concerns about iatrogenic risks of addressing DD.
A recent descriptive study employed an ad hoc semi-structured clinical interview for proactive assessment of a wish to hasten death among advanced cancer patients. 32 Participants did not experience this as distressing, but considered it important regardless of whether they were personally affected. The patients we interviewed indicated that they Framing, timing and patient-attunement of communication Although patients do not object to enquiries about potential DD, even when they do not personally have one, 32 the impact is likely to depend on how and when such conversations take place. Various guidelines have been developed for the improvement of HP-patient-communication in PC settings, e.g. "NURSE -Naming, Understanding, Respecting,

Supporting, Exploring" for difficult conversation tasks in oncology 35 and "SPIKES -A Six-
Step Protocol for Delivering Bad News" for breaking bad news. 18 An evaluation of the SPIKES implementation found that privacy positively influence patients' rating of their knowledge gained and the experience of the amount of time devoted to it as being sufficient. 36 This is consistent with our interview results that the framing of conversations is of particular importance. As receiving bad news was experienced as traumatic and a trigger for DD by some of our interviewees suffering from cancer, it may be appropriate to delay this discussion to follow-up conversations.
As has been found regarding breaking of bad news in oncology 37 and in light of the feedback from our Delphi panelists, the clinical guideline is designed to promote a patient-centered approach, giving suggestions on how to address potential DD while avoiding checklist-type interrogations. 38 As it is also stated in the guideline's usage notes, there is no need to address DD in the same way for all patients and not all thematic aspects need to be addressed in every conversation about DD,.

Appropriate (therapeutic) responses towards desire to die
The most general intervention advised with respect to every patient with DD is the initiation of an open communication about it. The clinical guideline suggests before considering any other action, HPs should aim to understand the patient's wish, its backgrounds, meanings and functions. As severe physical symptoms are also related to DD, 39 optimal symptom control is especially important in patients expressing such desire.
Discussion of the expected disease trajectory and options for treatment can also decrease anxiety. 40 For patients who experience physical suffering as unbearable, sensitively pointing out that therapy withdrawal or palliative sedation until death are legal possibilities to alleviate suffering within the statutory framework in Germany. Its administration needs to be tailored towards the individual's needs and in sensitive coordination with relatives. 6 Diagnostic clarification of depression is indispensable as it has been shown to predict 39 and moderate 41 DD. Suicidality with completed suicide at its most extreme can be a manifestation of DD, yet is not necessarily to be equated with it. Therapeutic interventions such as Meaning-Centered Therapy 45 and a supportive expressive therapy named Managing Cancer and Living Meaningfully (CALM) 42 have been shown to alleviate depressive symptoms in patients with advanced cancer. [43][44][45] Controversy concerning assisted suicide and active euthanasia Our Delphi panelists expressed only moderate agreement for directly addressing wishes for euthanasia or assisted suicide. Since euthanasia ( §216 national criminal code (StGB)) and assistance of suicide against payment ( §217 StGB) are forbidden in Germany, participants may have preferred to avoid naming such wishes because their realization would be illegal. Another concern expressed by our Delphi panelists was that acknowledging DD could pressure patients into actively seeking hastened death. Both scenarios point towards an underlying belief that enquiring about DD will cause DD to emerge. This view can lead HPs to avoid talking about it. We emphasize the importance of a clear distinction between acknowledging and agreeing with or even endorsing and supporting DD. Therefore, the legal situation in Germany is duly addressed in the mandatory training course that introduces the clinical guideline as a practical tool.

Self-protection and self-care
In addition to recommendations aimed at improving the patients' well-being, protecting the well-being of HPs is also important. Conversations about DD can be enriching but might also cause emotional stress. 46 It is therefore important for HPs to develop a sensitive grasp of their own feelings and to protect themselves from emotional overload. 46 In order to manage discussions of DD appropriately, HPs should also be aware of own values, norms and their personal stance concerning death wishes. Supervision, case meetings and everyday peer exchange can help to deal with stress or difficult situations. 47

Study Limitations
As 64% of our interview partners were cancer patients the results appear applicable to oncological settings, but the generalizability of findings onto other palliative patients may be limited. Although we aimed to compose a balanced sample of research and clinical practice perspectives for our Delphi survey, we ultimately recruited mostly practitioners, only a small number of relatives and no patient representatives. Participating HPs were multi-professional experts with more than 60% of them being nurses. Statistical analyses displayed no significant differences between professions; however the numbers in each profession may not have been sufficient to detect such differences.

Clinical Implications
As HPs are frequently confronted with DD, a semi-standardized communication guide for dealing with DD has great potential for clinical practice. The fact that more practitioners participated in the Delphi survey than originally planned shows their great interest in the topic and its particular relevance in PC. The clinical guideline thus became a tool tailored directly towards everyday practice of people working in the field. This can include all professions directly in contact with palliative patients provided they have partaken in the mandatory training course: physicians, nurses, psychologists, social and spiritual care workers as well as volunteers. We expect the guideline in conjunction with our training program to foster multi-professional competencies across all health care structures, especially on dealing with DD in patients with serious health-related suffering due to severe illness.

Conclusions
The major achievement of this study is the creation of the first consensus-based semistandardized approach for (proactively) assessing and optimally responding towards DD     clinical guideline seen to be at danger of provoking checklist type of interrogation due to bullet point setup changed interrogative clauses to instructions: ð ' Exists or existed Explore fear of death and dying ?' ð ' Exists or existed Explore thoughts related to terminating prematurely ?' complexity and changeability of desire to die (DD) in palliative patients seen to run counter to unambiguous classification added a new suggestion: ð 'In general, keep in mind: DD is complex and prone to change "manipulate" in the respective function of DD seen to be poor choice of words changed wording: ð 'Attempting to manipulate influence family or health profession "attracting attention" in the respective function of DD seen to be poor choice of words changed wording: ð ' Attracting Drawing attention to oneself and one's trouble' "treatment contracts" seen as bad practice, especially when involving handshakes for sealing the contract as it seemed to suggest "clean hands practice" changed wording, rated old and new version during round 2: ð 'Entering into a treatment contract with handshake in case latent suicidality' (32.9% agreement, M = 3.00, SD = 1. 19) ð 'Entering into a treatment contract with handshake agreemen cases of latent suicidality in order to win time for interventi