Obstacles to patient inclusion in CPR/DNAR decisions and challenging conversations: A qualitative study with internal medicine physicians in Southern Switzerland

Despite cardiopulmonary resuscitation (CPR) and do-not-attempt-resuscitation (DNAR) decisions are increasingly considered an essential component of hospital practice and patient inclusion in these conversations an ethical imperative in most cases, there is evidence that such discussions between physicians and patients/surrogate decision-makers (the person or people providing direction in decision making if a person is unable to make decisions about personal health care, e.g., family members or friends) are often inadequate, excessively delayed, or absent. We conducted a study to qualitatively explore physician-reported CPR/DNAR decision-making approaches and CPR/DNAR conversations with patients hospitalized in the internal medicine wards of the four main hospitals in Ticino, Southern Switzerland. We conducted four focus groups with 19 resident and staff physicians employed in the internal medicine unit of the four public hospitals in Ticino. Questions aimed to elicit participants’ specific experiences in deciding on and discussing CPR/DNAR with patients and their families, the stakeholders (ideally) involved in the discussion, and their responsibilities. We found that participants experienced two main tensions. On the one side, CPR/DNAR decisions were dominated by the belief that patient involvement is often pointless, even though participants favored a shared decision-making approach. On the other, despite aiming at a non-manipulative conversation, participants were aware that most CPR/DNAR conversations are characterized by a nudging communicative approach where the physician gently pushes patients towards his/her recommendation. Participants identified structural cause to the previous two tensions that go beyond the patient-physician relationship. CPR/DNAR decisions are examples of best interests assessments at the end of life. Such assessments represent value judgments that cannot be validly ascertained without patient input. CPR/DNAR conversations should be regarded as complex interventions that need to be thoroughly and regularly taught, in a manner similar to technical interventions.

Introduction than 70% of physicians defined a DNAR status without prior discussion with the patient [19]. A survey of residents in charge of 206 patients including DNAR and CPR orders in the French-speaking region of Switzerland showed that participants cited the patients' or their family's resuscitation preferences as a major justification for placing a DNAR order in 18% of cases [20]. Participants justified patients' inclusion in or exclusion from DNAR/CPR discussions using arguments based on respect for patient autonomy and decision-making capacity, as well as arguments based on their own clinical assessment of the patient [12]. Pre-pandemic evidence from the Italian-speaking region of Switzerland on physicians' reasoning regarding CPR/DNAR decisions, and on the extent to which, why and how they include patients in these decisions, is currently missing. At a policy level, a decision support tool is available to physicians working in the four main hospitals in Ticino which clarifies the circumstances under which patients should be included in CPR/DNAR decisions (Ente Ospedaliero Cantonale's internal document "D-SAN-004/B"). According to this tool, patient inclusion in the discussion is not essential if (1) the patient does not have functional limitations or major comorbidities, or if (2) the patient is approaching death. In the first case, the tool demands that a CPR order be given unless the patient explicitly refused cardiopulmonary resuscitation, while a DNAR code status should be included and implemented in the second case. Conversely, patient inclusion in the CPR/DNAR discussion is essential in all "intermediate situations", e.g., with patients affected by comorbidities for whom death is not impending and for whom CPR attempts may bring some benefit.
In this latter setting, we conducted a study to qualitatively explore physician-reported CPR/ DNAR decision-making approaches and CPR/DNAR conversations with patients hospitalized in the internal medicine wards of the four main hospitals in Ticino. In particular, the study has the objective of identifying possible tensions between the desired and actual decision-making approaches and conversations, accounting for the specificities of the current public health emergency caused by COVID-19. A tension can be defined as a situation that requires making a decision between two or more moral imperatives, neither of which is satisfactory or preferable, and where choosing one potentially results in transgressing the other(s) [21]. We decided to focus on Ticino and the hospital setting because evidence shows that advance directive completion in the Italian-speaking regions of Switzerland is at 17.9%, suggesting that the majority of CPR/DNAR decisions are addresses in the acute setting of the hospital [22].

Study design
We conducted a qualitative study using focus group interviews to explore resident and staff physicians' decision-making approaches and practices of DNAR/CPR conversations with patients in the four main hospitals in Ticino, the Italian-speaking Canton of Switzerland. Focus groups are a qualitative method in which a trained moderator interviews a small group of research participants simultaneously. For this study, we decided to employ focus group interviews because this is a rapid and convenient method to collect data from several individuals concurrently (particularly in light of physicians' demanding schedules), but also because this method capitalizes on study participants' communication exchanges in order to generate richer data [23]. In a focus group, participants are encouraged to discuss a variety of topics and exchange opinions among each other. For this reason, focus groups are particularly useful to understand participants' perspectives on the research topic, their knowledge and experiences, and can be used to explore not only what people think but also how and why they think that way [24]. While we initially opted for face-to-face focus group interviews to be held in person, several potential participants manifested difficulties in reaching the university facilities at the scheduled time. For this reason, a decision was made to conduct the focus group interviews online on Zoom (Zoom Video Communications, Inc., San Jose, CA, USA).

Sampling and recruitment
To be included in the study, participants had to be employed as either resident or staff physician in the internal medicine department of one of the four main hospitals in Ticino (Bellinzona, Locarno, Lugano or Mendrisio). We included participants with at least six months of postgraduate clinical experience to increase the chances that participants had confronted with the issues of CPR/DNAR. We excluded medical students and physicians with less than six month clinical experience. Recruitment was conducted in two phases. In a first phase, the study was officially presented by MB at a colloquium attended by resident and staff physicians employed in the various departments of internal medicine in Ticino in May 2021. In a second phase, following the study presentation, MB sent an e-mail to potential participants (N = 135) inviting them to take part to the study. The e-mail included information on the study and an explanation of the voluntarily nature of participation. Importantly, participants were reassured that their identifying information would not be shared with anyone employed by the hospital, but only retained by the study team at the Università della Svizzera italiana (MF and IF). To take part to the study, participants had to fill out an online form on Qualtrics1 (Qualtrics, Provo, UT, USA) which asked them their availability for the interview and a number of questions such as name/surname, gender, age, employment type (resident vs. staff), hospital of employment, medical specialization, years of clinical experience after obtaining the medical degree, any postgraduate training in clinical ethics, any experience in departments other than internal medicine, e-mail address and their availability. This information allowed us to allocate participants to four focus groups that were homogeneous in terms of employment type but heterogeneous in terms of age, gender, and hospital of employment.

Data collection
We conducted four online focus groups on Zoom between the 12 th and 16 th of July, 2021. Three focus groups were held with resident physicians and one with staff physicians. Interviews lasted between 42 and 60 minutes. The focus groups were moderated by MF, a female postdoctoral researcher in biomedical ethics with extensive training in qualitative research methods, and co-moderated by IF, a female research assistant with a background in cognitive psychology and extensive training in qualitative data collection and analysis. The moderators introduced themselves at the beginning of each focus group. Neither the moderator nor the co-moderator had any working relationships with the study participants. This allowed to reduce any social desirability bias in participants' reports.
During the interview, which was recorded following participants' oral informed consent, we employed an interview grid developed ad hoc for the study purposes (S1 Table). Questions aimed to elicit participants' broad definition of advance care planning, their specific experiences in making a decision on and discussing CPR/DNAR with patients and their families (including barriers, facilitators, ethical justifications for inclusion/exclusion of patients in/ from the CPR/DNAR discussion, and perspectives on both current practices and desired ones), the stakeholders (ideally) involved in the discussion and their responsibilities. At the end of each focus group, participants were asked about the reasons for participating, their participation experience, and any additional comments. This debriefing phase served to identify and mitigate potential stressful effects following the focus group discussion as participants were asked to reflect on sensitive topics such as the end-of-life.

Data analysis
The interviews were transcribed and anonymized by IF within two weeks after data collection. Shortly after each focus group, a debriefing between MF and IF occurred in which thematic insights were shared and noted in document words file placed before each transcript interview for following analysis. Subsequently, MB and MF performed a thematic analysis on the transcripts according to the analytical approach developed by Braun and Clarke [25]. After individually reading and coding the transcripts, MB and MF convened together and compared their codes to identify similarities and resolve discrepancies, develop broader thematic categories and subcategories, and highlight meaningful quotes. Finally, a report describing the study results was shared with the rest of the team for feedback. During both the data collection and the analysis phases, the team reasoned on their role in the interview encounter and the interpretation of the data, identified possible biases linked to professional or personal experiences with the end-of-life, and reflected on solutions to mitigate them. No transcripts were returned to participants for feedback. No software was used to manage the data. This manuscript adheres to the consolidated criteria for reporting qualitative research (COREQ) [26].

Ethics approval and consent to participate
The study protocol was submitted to the Ethics Committee of the Canton of Ticino (ID Req-2021-00293) which established that the protocol did not require ethics approval, since the project does not fall under Article 3 of the Swiss Human Research Act. All participants provided oral informed consent to participate in this study.

Characteristics of the sample
The final sample included 19 participants from the four main public hospitals in Ticino, of which 10 were women (see Table 1 for participants' characteristics). Of the 153 potential

General description of the extracted themes
The analysis of the transcripts yielded three main themes. The first refers to the tension that participants reported between the ideal decision-making approach regarding CPR/DNAR and advance care planning, more generally, and the actual approach to patient involvement they experienced, which is described as being largely dominated by the belief that patient involvement is often pointless. Participants reported a number of justifications for this tension and proposed practical solutions to address it. The second theme refers to the tension between what participants regarded as ideal CPR/DNAR conversations with patients and how these actually unfold in their clinical experience; participants reported that they find themselves inevitably nudging patients rather than eliciting their preferences, an approach they justify through multiple explanations and for which they suggest a number of alternative practices. The third theme refers to the structural causes of this twofold tension: these include patientprovider-, policy-, culture-, and religion-related factors. Exemplary quotes from the focus group interviews can be found in Table 2. Let's make the discourse broader, let's start dealing with what does not have to do with "now" but will have to do with it in a more holistic health-related discourse.
In the ideal world, I would like that each patient has this sheet, but it also matters when this sheet was compiled, because one thing is to do it now at twenty years old, and one thing is to do it at eighty. I will have a different idea depending on the conditions I will be. Surely you should have this sheet, even if finding everything written down sometimes is not so useful.  If they don't get there, it is useless. And at that moment we need to talk to the family doctor and share the choice with them, and if they don't have the tools, the education, at most the choice has to be imposed and then shared. We tell them that we have chosen not to resuscitate them because the quality of life that is envisaged is not vaguely comparable to that of before, and then we refine it with the relatives. Then it does not mean that we will not treat them, on the contrary! The care is assured until resuscitation, which in that case is not indicated. (  There is a whole sphere that we don't share with the patient. . . We evaluate whether to tell the patient because maybe there is no real repercussion for them. If this patient ha sa cardiac arrest, I don't resuscitate them, and they won't notice. For example, with a demented patient. (FG2)

22
Poor prognosis It also happens in everyday hospital life not to discuss the advance directives with the patient. It has happened to all of us to say "OK, this is a patient who, if they have a complication, is not resuscitated, by medical choice, because he has a number of impairments and pathologies that would not give them a chance of a good recovery". And so normally you inform them that you established this attitude of care, but we do not always discuss it with the patient. (FG1) If they have never thought about it and they come to the emergency room, we don't have time to have a good discussion, we just create anxiety in the patient. So in that case it makes sense, and it is not bad that we have to make the decision [for them], we help the patient and the family to take the responsibility. We orient the discussion and give advice on what we think might be the best solution for the patient. (FG2)

29
Desire to preserve patient's hope and the therapeutic alliance Perhaps it is because of a fear on the part of the physician to extinguish the patient's hope in treatment and in the possibility of being able to do something about their illness. Even if the physician is well aware of the fact that a disease is chronic and, as per its history, will be worse and worse, and that this worsening will not offer any possibility of treatment. They may find it hard to extinguish [the patient's] hope in this way, hard to tell the patient the way things really are. It is not uncommon for us to observe people who remain incredulous when their disease is so advanced that it precludes short-term survival. In my opinion this holds back a little bit the physicians, this fear of losing the therapeutic alliance with the patient. (FG4)

30
My reflection is on how not to take away the patient's hope. (FG1) 31 Old age combined with the presence of co-morbidities It has happened to me so many times not to talk about it because the patient is 90 years old and is hospitalised for a trivial urinary tract infection but you still have to include the advance directives in Geco [the electronic health record] and you say "he is 90 years old, he has several pathologies". . . And so you put the label "REA NO". Here maybe the patient is lucid and oriented, but you don't discuss it not because the situation is dramatic but because of all the comorbidities and the situation. (

Solutions
Define the many hopes patients may have It becomes almost easier in that area to enter into the issue of quality, to concern oneself with the patient's values and wishes, which then become the sole objective. If one took this step earlier, I would not believe that the therapeutic alliance and hope would be altered. (FG4) 35 We should define this hope. Is it a hope of recovery, hope of having a life as before, hope to be able to eat ice cream again? It really depends on the type of hope. (FG4) 36 [We should] re-establish the therapeutic alliance by re-establishing the goal according to the current condition, informing them that we are there and that something can always be done with the goal perhaps no longer to heal and return to life as before, but with the goal of living an acceptable quality of life. Reducing the objective together with the patient and informing them of their condition, where they can inform themselves by creating a therapeutic alliance between patient and doctor. (FG4)

37
Planning with the patient helps to give a little more hope. Instead, communicating a diagnosis, proposing possibilities to them with the hope of achieving something, we do not know if the first thing that goes wrong will restore hope. Planning with them what they expect and setting goals that we know are achievable for them, and not setting too high a goal allows us to work with the patient and work with them. It's certainly not something you can do in a half-hour consultation where you often don't even know the person. (  Do not ask "do you want to be resuscitated" because of course the answer is "yes"! (FG1) 52 We should never get involved in the decision. We can explain to them, we can reassure them, but the important thing is never to influence them. (FG1) And therefore the colleagues said that it was "tube no", "resuscitation no" and that the patient had to be put in palliative care. While here we were in a situation of sepsis, and he had recovered very well after Covid; and yet, this decision was set in stone and every now and then I think that this could be a possible stumbling block to this discussion. Nobody wants to make this decision that cannot be changed and that will remain forever. (FG4)

59
Keep the conversation open Sometimes there are forms or diagrams that can be useful, but in others the relationship counts a lot. Sometimes patients arrive in the emergency department with such detailed directives that they contradict each other and are difficult to understand, whereas a discussion where a multifaceted but more coherent sense prevails would be more appropriate, avoiding the "yes" and "no" that is too detailed and confusing. (FG3)

60
Consider the larger social context of the patient I think it is also important to consider the burden not only for the patient but also for the social entourage around the patient. (FG3)

61
Use words that communicate presence There is precisely the advice to never be alone, even in communicating the news, not because you are not capable but because you "carry it together".  And so one really has to take the courage sometimes to have this talk. It is often the case that those treating chronic illnesses find it hard to say that the illness will not get better but will only get worse, and so it must be decided now that this is going to be their future. And they find themselves talking about it with complete strangers. More and more choices are left to the patient, but these choices are. . . They go in a direction that is given by the physician. This is something which we inherit from Anglo-Saxon countries where we have to protect ourselves more and more with informed consent. So it seems to me that we inform patients a lot or pretend to inform them, making them take the decision, whereas it is obviously made by the physician. (FG4)

70
We take a big right when we hide something big from the patient. We do not have the right to do so even legally. We have to tell the patient how things are.   Adherence to the principle of nonmaleficence I would divide it into two aspects. One is that here is a lack of clarity on protocols, algorithms. In the sense that there are our superiors who say "OK, now it is the patient who decides". Whereas on the other hand there is a reasoning of nonmaleficence in the omission from the doctor, to give adequate advice to patients and to advise what is adequate and to take the patient down a recommended path which is often a "REA NO". Often, one is caught between these two paths and there is a lack of clarity. [ There should always be a figure who accompanies you, like a clinic leader or even a dedicated figure who comes, helps you along the way. If not, you really struggle. 90

Exposure to conversations and debriefing
It's often a "handmade" thing, in the sense that you have to be exposed to these kinds of discussions. The first few times I participated and kept my mouth shut, I listened and I also needed to see several people do it to see what techniques one and the other used. "Okay, this sentence I like a lot, okay, this example I will use it". Then at the end one builds up one's own knowledge, of course first one has to have knowledge with a structured training but it is also important to have a boss who brings the assistant, even on the first day of work and he just stands there mute.

Reassure patients that the conversation is part of routine
You try to reassure the patient about the discussion we are going to have, because it is a discussion that usually creates a lot of fear in patients. Usually they are there for a problem that is not acute but that could precipitate. Then they immediately ask you "Why so serious?". Then you start by saying, "we would like to talk to you about something, it's an issue we deal with all the patients we hospitalise". You really try to anticipate the fact that we are not in an emergency situation but that these are issues we deal with routinely in our activity. ( the question their answer was always "I don't know where I get these pneumonias", and I asked them why they think they always get pneumonias. Then they answered that they are weaker than the others and therefore have a chronic lung disease, that's how we approach questions because in this way we investigate what the patient's level of awareness is and at the same time through questions and answers given by us or self-answers. The last thing I learnt in palliative care is that we should not always feel obliged to talk, we should not be afraid of the silences between doctor and patient.

CPR/DNAR decision-making: A tension dominated by the presumption of futility
When asked what decision-making approach participants believed should characterize decisions on CPR/DNAR and advance care planning, more generally, they reported that, ideally, they ought to engage patients in a timely process to make decisions on future care plans, particularly anticipating the possibility of severe conditions and eliciting what treatment outcomes patients would consider acceptable for themselves (#1-7). One participant added that, ideally, this approach should culminate with the drafting of the patient's own advance directives, although it was noted that their usefulness is limited as they may be too specific and apt to change over time as patient's preferences change (#8). In addition, they reported that CPR/ DNAR decision-making should place the patients and their wishes at its centre, independently from the patient's age (#9-10), and physicians should aim to explore patient's understanding, expectations, wishes, and values (#11-13). However, participants reported that patient inclusion in the decision-making regarding CPR/DNAR is often believed to be pointless. Participants (particularly resident physicians) stated that patient inclusion is deemed pointless until proven valuable (#14), i.e., patients are not included by default, but there should be a good reason to do it (e.g., imminent cardiac arrest, #15). Our participants offered a number of justifications for this tension. These included  patients' presumed lack of capacity, mainly described as the inability to understand factual information (#16-22), poor prognosis (#23), young age or good health (#24-25), need to ration limited resources (#26), desire to avoid scaring the patient (#27-29), desire to preserve patient's hope and the therapeutic alliance (#30-31), old age combined with the presence of co-morbidities, even if the patient is mentally sharp and has a good prognosis (#32-33), and lack of independence (#34).
To resolve this tension between the ideal and the actual decision-making approaches, participants proposed three main solutions. The first involves the patient-provider relationship and consists in physicians making an effort to define the many hopes patients may have beyond the simple goal of healing (#35-38). The second solution calls for institutions to increase public awareness of advance care planning and advance directives (#39-40) while the third calls for a societal change in which each of us contributes to break the taboo of talking about death (#41).

CPR/DNAR conversations: A tension dominated by physicians' inevitable role of choice architect
When asked to report on the features of the ideal conversation regarding CPR/DNAR, participants reported to have a clear idea of the architecture of such a conversation. They identified the need to find the right place, time, and interlocutors to join the conversation (#42-48) and to establish rapport with the patient before the discussion takes place (#49-50). During the conversation, they reported the need to provide neutral and factual information that does not lead but rather guides the patient (#51-55); inform patients and their families, in case of an unilateral DNAR order, that this does not preclude optimal care or treatment of the underlying condition (#56-58); stress that the decision is not irreversible and can be adapted across conditions, not only with patients and their families, but also among the care team (#59); keep the conversation open, avoiding the tendency to fit every answer to either a "yes" or "no" (#60); consider the larger social context of the patients and their emotional burden (#61); use words that communicate presence (#62); tailor the communication style to the patient (#63); show humility (#64); and embrace the conversation with courage and truthfulness (#65-66). At the end of the conversation, participants reported the need to sum up its content, particularly the process through which a CPR/DNAR decision was reached (#67).
At the same time, participants were aware that discussions with patients and/or relatives or authorized representatives will always be strongly influenced by physicians' personal attitudes and preferences. In particular, they reported that such conversations tend to be inevitably leading conversations, since the way physicians present any information regarding CPR has the power to affect patients' perceptions regarding this procedure (#68). Participants agreed that, ultimately, it is the physician who makes the decision for the patient, by shaping their "choice architecture" and later presenting it as made by the patient simply because it was eventually accepted (#69-70). This persuasive approach is carried out not only by choosing certain words over others but also purposely hiding certain information or minimizing the situation (#71-72). While some call for physicians to become aware of this role as "choice architect" and the power it intrinsically entails (#69) others see this role as a legitimate part of the shared decision-making approach, in which physicians paternalistically make a decision for patients and, by later communicating it, they make it a shared one (#68). Participants also reported that physicians end up solely presenting factual information (#73) and that the time and place of the conversation are often inappropriate (#74-77). Rather than the courageous and truthful persons they ought to be, physicians reported that they end up being perceived as the "bad wolves", i.e., as heartless individuals that throw people in panic and despair whenever they engage patients in CPR/DNAR conversations (#78-80).
Our participants offered a number of justifications for this tension. These include lack of time (#81-82), fatigue and stress due to the demanding nature of their job (#83), the fact that CPR/DNAR conversations are a time-and resource-consuming task (#84), and lack of previous conversations on CPR/DNAR between patients and their general practitioners (GPs) (#85-86). One participant felt justified in purposely nudging patients to make them accept the physician's decision, as he/she felt compelled to adhere to the principle of non-maleficence (#87). According to this participant, physicians know what is best for their patients and need to prevent them from making the wrong decisions by gently leading them.
To solve this tension, participants suggested to offer timely communication training to physicians (#88), have a dedicated communication professional in the team (#89-90), ensure physicians are exposed to as many CPR/DNAR conversations as possible and offer debriefing sessions (#91), increase involvement of GPs and specialists to promote these conversations outside of the hospital setting (#92-98), reassure patients and their families that the conversation is part of routine (#99-100) and that they are not alone (#101), assess patient's interpretations of the information received (#102), and give the patient time to elaborate the information (#103-105).

The structural causes of this twofold tension
The third theme refers to the structural causes of the tensions concerning CPR/DNAR decision-making and conversations. These include factors related to the relationship between patients and providers. On the one side, participants reported that a paternalistic approach is still present where physicians' recommendation is absolute and not questioned, and that a feeling of power over patients is still widespread among physicians (#106). On the other side, they reported that a structural cause of this tension is a clinical approach that still pays excessive attention to treating the illness rather than the person (#107). At a hospital policy level, participants complained that the current regulations urging physicians to include a code status as soon as the patient is admitted place pressure on them and demand them to rapidly execute a task that would instead need extensive time to be properly performed (#108). At a societal level, two main bottlenecks thought to have both cultural and religious roots are described as slowing down progress regarding CPR/DNAR decisions and conversations. The first one is the taboo around death, a topic that is consistently avoided and which participants said they are afraid of discussing (#109-112). The second is the widespread idea of the sanctity of life, according to which we ought to extend life as long as possible using any means we have (#113-114).

Discussion
The aim of this study was to qualitatively explore physician-reported CPR/DNAR decisionmaking approaches and CPR/DNAR conversations with patients hospitalized in the internal medicine wards of the four main public hospitals in Ticino. We found that CPR/DNAR decisions were dominated by the belief that patient inclusion is often pointless, even though participants had structured views on what would make a desirable decision-making approach. Moreover, we found that participants were aware that most CPR/DNAR conversations are characterized by a nudging communicative approach where the physician pushes patients to his/her recommendation, despite the belief that conversations should be as neutral as possible. Finally, participants identified structural causes to the previous two tensions that go beyond the patient-physician relationship. In the next paragraphs, we contextualize and interpret our findings, and consider their implications in light of the study limitations.
A first finding is the tension between a desired shared decision-making approach and the belief that patient involvement in CPR/DNAR decisions is often pointless, particularly when patients are deemed as lacking decision-making capacity. This finding, which echoes previous evidence that most physicians desire patient participation in the decision-making process and yet patients are often excluded [27], suggests two beliefs and a possible ex post rationalization behavior. The first belief is that patient involvement in the decision-making process can only be justified when CPR has an acceptable predicted success rate; if CPR offers little or no likelihood of benefit, patient involvement in the decision-making process offers little or no likelihood of benefit either and, therefore, there is no need to negotiate, and a unilateral DNAR is justified. This points to a pragmatic idea of value attributed to patient involvement: since it will not change the decision, there is no point in involving patients. That physicians' own preferences may predominate in the CPR/DNAR decision-making process for their patients, and that they may not want their participation or believe that it is not necessary are both well documented [28][29][30]. Such a decision-making approach neglects two important elements. First, a decision to withhold CPR without knowing or even contrary to the patient's or surrogate decision-makers' wishes may only be justified in cases when CPR cannot reasonably be expected to yield the intended clinical benefit or achieve agreed-on goals for care, or when acute care facilities exceed functional capacity and triage decisions become necessary [17]. Even in such cases, competing concerns (e.g., risks of litigation) should be acknowledged by physicians. However, a unilateral DNAR is likely to be an ethical lapse: the decision may instead be driven by physicians' perception that it is extremely difficult to involve patients in this choice, or by the fact that they do not "believe" that CPR is worth trying. Even when it becomes ethically justifiable to unilaterally make a decision to withhold CPR, patients still have an ethical and legal right to be informed about their code status and clinicians' medical opinion and advice, and involved in other, more or less related decisions [31]. In addition, circumstances may change that make it necessary to review the CPR/DNAR status with patient input. The second element is that decisions regarding whether it is "worthwhile" to resuscitate represent, in the majority of cases, a value judgement rather than a matter of physiological probability, therefore any decisions regarding patient involvement in their CPR/DNAR decision-making will also constitute a judgment based upon a particular set of values that may contradict those of the patient affected [32]. There is wide consensus that it is extremely problematic to define the concept of futility in a medical context [33]. When recommending against CPR, physicians should approach such patients or their surrogate decision-makers with a presumption against providing CPR but also remain attentive when discussing the patient's values and goals for unique factors and concerns that might make CPR unusually beneficial [34]. The second belief suggested by our participants' reports is that decisions on whether to involve patients in the decision-making regarding their CPR/DNAR should be exclusively based on decision-making capacity criteria [35]. First, some have expressed concerns that these criteria are excessively "cognitive" and do not sufficiently attend to noncognitive aspects of choice and more demanding autonomy conditions relating to authenticity, narrative coherence, patient values, or the impact of volition, motivation, and affect or emotions (such as the emotional capacity to care) [36,37]. Second, patients are generally presumed to have decision-making capacity, but the threshold (which should be that of ordinary freedom) tends to be shifted in opportunistic ways. A patient's decision-making capacity may come under question when distortions in thinking and understanding due to a significant impairment of mental abilities are identified or suspected [38]. Only in such cases an assessment is warranted, and its results and underlying arguments need to be appropriately documented and made available to the patient, or the patient's representative, on request [38]. However, dismissing patients as lacking capacity (e.g., because they struggle to understand factual information, as many participants of this study reported), may be the symptom of a post-rationalization behavior to legitimize a posteriori unilateral CPR/DNAR decisions taken by the care team. Physicians may ascribe incapacity because they see the possibility or even the likelihood of discovering disagreement between a patient's preference and their recommendations/perspectives, and may feel ill-suited to handle such a conflict. According to the AMA, "physicians should engage patients whose capacity is impaired in decisions involving their own care to the greatest extent possible, including when the patient has previously designated a surrogate to make decisions on his or her behalf" [39]. In Switzerland, this is not only an ethical but also a legal requirement (Art. 377 of the Swiss Civil Code) [40]. Our participants were aware of these ethical and legal requirements, and many showed to be engaged with an active process of reflection on not only the reasons why, in practice, this is hardly met, but also possible solutions to resolve this tension. This finding does not surprise. Contrary to the rest of Switzerland, the Canton of Ticino borders with Italy, from which it is heavily culturally influenced. Paternalistic attitudes based on the therapeutic privilege are widespread [41]. CPR/DNAR discussion rates in this Canton are likely to be more analogous to rates detected in Italy than in German-and French-speaking Swiss Cantons [42].
A second finding is the reported tension between a desired non-manipulative conversation and the actual, leading communication style adopted by participants. This finding echoes the results of previous studies which found that, despite thoroughly trained on the importance of and ways to respect patients' autonomy, physicians unintentionally resorted to strategic forms of communication leading patients to opt for physicians' preferences [43,44]. This result points to two facts. The first is the distinction between persuasion, nudging and manipulation. Persuasion refers to "believing in something through the merit of reasons another person advances", and it is influenced by appeals to cognition rather than emotion [45], while manipulation refers to altering the quantity, quality, manner and relevance of the information presented to increase the effectiveness of a message in order to deceive or mislead the interlocutor [46]. The first may represent a valid support to patient autonomy, while the second a source of autonomy deprivation. Our participants reported to engage in a nudging communication. Nudging refers to intentionally structuring the presentation of options or decision-making environments to increase the frequency of desired choices without restricting free choice [47]. Scholars disagree on what would ethically justify nudging patients towards the decision that the physician believes to be the most desirable one. Some argue that nudging could amount to manipulation and improper exploitation of cognitive weaknesses [48][49][50], while others defend this practice as a form of "libertarian paternalism" that is not blind to cognitive biases in our judgments, and that in fact tries to "influence choices in a way that will make choosers better off, as judged by themselves", by preserving freedom of choice and supporting moral agency [47, 51,52]. Similar to all interventions that seek to direct behaviour towards certain ends, our participants were aware that nudging is problematic as it implies that physicians may impose their own judgment on what is beneficial to the patient, and may fail to address the reasons why CPR/DNAR may be the right thing for the patients. According to the SAMS guidelines on decisions on CPR, since discussions with patients and/or surrogate decision-makers will always also be influenced by personal attitudes and preferences, it is essential for physicians to be self-critical and aware of and disclose one's own position [17]. Interestingly, our participants queried themselves against not only what ends they want to achieve (i.e., avoid nudging patients towards their personal preferences) but also what persons they want to be (i.e., the courageous, honest physician who guides patients in their decision-making without concealing or distorting the information on CPR). Such reflection shows that participants' engagement of values and virtues was a central matter in their perspectives on CPR/DNAR conversation: who they are, what they recognize as important, and what they can (sustainably) achieve in relation to their CPR/DNAR conversations with patients are fundamentally intertwined. Participants reported a desire to engage in good action but they also made explicit the commitments underlying those actions. This is also stressed by the fact that they recognized the multidimensionality of the issue of nudging in CPR/DNAR conversations and reflected on multi-level solutions (e.g., involving general practitioners, offering young physicians ad hoc training to such conversations, or mobilizing social change to break the taboo around death that may hamper such discussion). The second elements suggested by this finding is that our participants may believe that there is only one, single-purpose conversation to be held with patients for whom CPR would offer little or no benefit. While a conversation on patients' CPR preferences should be initiated when CPR represents an option, other conversations should take place when a cardiac arrest is likely and CPR is not offered. In this case, neither nudging nor other persuasive communication approaches will be needed. Rather, physicians should reassure patients that they will be respected and helped achieve the goals they have in pursuing their "loyalties". This can be achieved by asking questions on their understanding of where they are with their health, their worries for the future, their hopes and priorities, and what they are (not) willing to sacrifice [53].
Finally, our participants called for a societal transformation in which each of us contributes to break the taboo of talking about death. However, it should be a priority for physicians to be able to overcome this taboo in clinical conversations, and learn how to not shy away from having difficult but important conversations about death with patients and families. Physicians often are unsure how to initiate or proceed with end-of-life discussions, mostly because medical education regarding end-of-life protocols is suboptimal or even lacking.
Our results have a number of implications which we summarise in the following list.
• First, they stress the importance of ensuring that CPR/DNAR conversations are held in the appropriate environment, with a trustworthy professional, and before the acute event presents. Ideally, GPs should be in charge of this conversation as they are likely to have established rapport with patients and can initiate these conversations outside of an emergency situation within the context of advance care planning. A national working group co-led by the SAMS and the Federal Office of Public Health (FOPH) is currently working towards improving the framework conditions and quality standards of advance care planning in Switzerland [54].
• Our results also stress the importance of recognizing that communication interventions in clinical practice should be regarded as complex interventions that need to be learned in a manner similar to technical interventions (actions that providers take and procedures they implement to deliver treatment and care to patients).
• From a policy point of view, our findings highlighted the limitations of the code status form currently implemented in the hospital where our participants were employed. Physicians land on the form as soon as the patient's electronic health record is opened, and are asked whether CPR is indicated or not. In its current shape, this form may constitute a barrier to start a CPR/DNAR conversation and may nudge physicians to unilaterally decide on the patient's code status. A better question could be "What are patient's wishes regarding CPR?". If a DNAR is entered, a question should appear asking "Has a decision been discussed?".
This study has a number of limitations. First of all, the recruitment of participants was difficult to the point that it required adapting the design from in-person to online data collection. We thus relied on a small number of participants. However, participants engaged in an active and lively conversation during all four focus group, and brought such a richness of perspectives that it was possible to identify substantial tensions in relation to the themes explored.
Second, given the low response rate, our sample was not representative of the hospital population we targeted. Furthermore, given that participation was on a voluntary basis, it may be possible that those who participated were particularly interested in the subject or with a particular discontent or tension that other non-participating physicians would have not highlighted had they joined the study. However, the issues raised by our participants remain relevant and interesting to understand the tensions experienced by resident and staff physicians and how they resolve them. Third, because of the small sample and the specific study aim, we cannot claim we reached thematic saturation. However, we relied on the concept of information power to establish saturation rather than thematic redundancy [55]. Information power indicates that the more information the sample holds that is relevant for the study, the lower number of participants is needed. Fourth, as this was the first exploratory study on CPR/DNAR decisionmaking and conversations in Italian-speaking Switzerland, we decided to restrict our target population to resident and staff physicians and exclude nurses and other professionals who have a role in CPR/DNAR decision-making and discussions. Finally, caution must be exercised in extending the findings of this study to clinical specialties other than internal medicine, outpatient settings, and different geographical areas, where different perspective on CPR/ DNAR decision-making and conversations may emerge.

Conclusions
CPR/DNAR decisions are contextual and their symbolic meaning can be highly value laden. For these reasons, they represent judgments that cannot be validly determined without the patient's contribution and that should be considered against the backdrop of a patient's whole life story. Our results reiterate that making a CPR/DNAR decision requires initiating and holding potentially difficult conversations that even highly trained practitioners may feel illprepared for and uncomfortable with. Importantly, not only did our participants seem aware that imposing their views regarding CPR/DNAR on patients would be a case of abuse of power. They also queried themselves against the tension between acknowledging that patients can view their own good differently and choose otherwise than recommended, and the actual practice of respecting their views and wishes, as well as which virtues they want to embrace.
Supporting information S1