1. Defining the scope of review by concept mining and theory building

A scoping literature search was performed initially between 1 December 2021 to 24 February 2022 to clarify the purpose of this review and generate initial theories. The PubMed database and Google scholar were primarily used to seek articles relating to the study’s aims. These included systematic and non-systematic review articles, key primary studies and any other article type that had relevance to goals of care discussions in acute hospital illness. Knowledge acquisition and abductive reasoning informed the construction of seven preliminary hypotheses, constructed in a Context-Mechanism-Output (CMO) format (i.e. “if (C)…then (M)…which results in (O)” (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Context-mechanism-output configuration to generate hypotheses for realist synthesis.

https://doi.org/10.1371/journal.pone.0299933.t001

Adapted from Mitchell et al [31], Papoutsi et al [32] and Cottrell et al [33]

2. Stakeholder consultation and refinement of initial theory

The CMO hypotheses were refined following consultation with 16 key stakeholders that had clinical, research or patient experience relating to goals of care conversations and decision-making, an approach adopted in previous realist reviews [33, 34].

Stakeholders were identified within existing networks of both authors and selected based on their personal and/or professional background, their level of experience and if it was felt that they would add value to theory building. Efforts were made to involve a range of different stakeholders from different personal and professional backgrounds to give a rounded insight. Discussions were held individually and were face-to-face, via video or audio call depending on their preference. The 16 stakeholders consulted included 10 hospital-based doctors (specialising in intensive care medicine, general medicine, emergency medicine and palliative care), one specialist nurse (critical care outreach), two hospital managers, one physiotherapist and one former patient with experience of having been cared for in an ICU, and their spouse. Each consultation lasted between 30–60 minutes. Each interview involved a short presentation of the latest iteration of the CMO hypotheses and open-ended questions to ascertain their general thoughts and feelings towards these and whether or not they would recommend any modifications based on their experiences. Field notes were taken with reflective summaries and modifications made to the CMOs where appropriate. When there was doubt and/or ambiguity, agreement was sought between JG and JK. Table 2 presents the resulting CMO following consultation with all 16 stakeholders with relevant modifications made.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Proposed CMOs following scoping literature search and stakeholder engagement.

https://doi.org/10.1371/journal.pone.0299933.t002

3. Searching for and appraising the evidence

The refined CMO hypotheses informed a search strategy for the main literature search (S1 Appendix). Searches were run using Medline, Pubmed, Embase, CINAHL, PsychINFO and Scopus databases from 1946 to 14 July 2023. Purposive searching for additional relevant articles that would contribute to programme theory building and theory testing was performed iteratively. These were identified by citation tracking from papers already identified and Google Scholar. Relevant non-published grey literature was also sought using OpenGrey and Google search engines. The screening and selection of articles were based on (i) relevance: whether they contribute to theory building and testing and (ii) rigour: whether the method used to generate that particular piece of data is credible and trustworthy [35]. Although there were no geographic restrictions, papers were limited to those written in English.

Rigour screening.

The final screening process involved critical appraisal of the evidence using well-established quality appraisal checklists developed by the Joanna-Briggs Institute (JBI) [36]. JBI appraisal tools were used for analytical cross-sectional studies, case series, cohort studies, qualitative studies, quasi-experimental studies, randomised control trials, systematic reviews, text and opinion articles. For ethics-related studies, Jansen and Ellerton’s Ethics critical appraisal worksheet [37] was used and the Mixed Methods Appraisal Tool for mixed methods studies [38]. Each article was assessed against each of the criteria of the appropriate checklist according to the study design, to give a final absolute score and a relative score as a percentage of the total number of scoring criteria assessed against (S2 Appendix). Articles were given a final quality grade rating depending on relative scoring ranges of less than 60% (low), 60–79% (moderate) and greater than or equal to 80% (high) (Table 3). Expert opinion and ethical argument papers were all rated as low-quality grade rating irrespective of the absolute or relative scores. This is consistent with internationally recognised guidance of hierarchal evidence [39, 40], where expert opinion has the highest risk of bias and therefore thought to be the least trustworthy for this review.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Characteristics of included studies.

https://doi.org/10.1371/journal.pone.0299933.t003

4. Extracting data and synthesis of findings

Data analysis and synthesis processes were flexible, iterative, and creative. To maintain transparency, JG and JK kept notes from a series of meetings during which they discussed each article and its contribution to the CMOs. We used abductive reasoning for the non-observable data to create associations and to recontextualise the data, creating new plausible conclusions [30, 41]. Moving between theory and data, we used retroduction to explore, compare and explain observable patterns in data, whilst also looking for other relevant themes not captured by initial programme theories. Abductive reasoning was used to create associations between theories. For both processes, JG and JK discussed potential explanations, new findings, and strategies to refine and revise the CMOs. We retained notes and a schematic as an audit trail of decisions made. The final synthesis represents an interpretive, yet robust collation of the supporting evidence located for each of the CMOs.

i) Preconditions to receiving information

The patient’s or their family’s perception that the content discussed is considered to apply to them was an important prerequisite when engaging in any goals of care conversation [22, 58]. This was particularly relevant for patients at the end-of-life. Elderly patients with frailty and multiple comorbidities were less likely to relate to conversations relating to end-of-life care due to a lack of awareness that they are more likely to be in the last phase of life compared to those with a clearer end-of-life trajectory (e.g. people with cancer) [23].

ii) Inconsistencies in the content of information provided to patients

Inconsistencies in the content of information provided to patients were common and related to the variability of doctors’ explanation of a patient’s clinical condition, prognosis and the risks versus benefits of different treatment options [45, 84]. Conversations relating to do-not-attempt cardiopulmonary resuscitation (DNACPR) were more common than other life-sustaining interventions. Prognostic uncertainty was perceived by doctors to be difficult to manage and communicate [16, 86]. The words “death” and “dying” were infrequently mentioned [45, 79] and implied rather than expressed [79]. Moreover, palliative care-related options were not as readily discussed [45] and in some cases only discussed after patients had expressed wishes to forgo life-sustaining care [45]. Information provision was incomplete when doctors perceived the risk of information ‘overload’ or the risk of confusing the patient, particularly when the doctor considered that the topic of conversation was not immediately relevant to the patient [22]. Information provided was also considered to be inadequate if the doctor sensed rising tension or conflict with the patient and or their family due to disagreements or divergent views between them and the patient or family [22].

iii) Information not being provided in a personalised manner

Information provided by doctors commonly focussed more on medical or procedural interventions and less on broader life values [42, 43, 45, 47, 56, 58, 75, 77, 82]. Factors associated with this included the doctor’s perception that the patient was not at the end-of-life [67, 76], when a medical intervention or surgical procedure was proposed [81], the perception by the doctor that quality of life was a “softer topic” [16], the perception that medical intervention was easier to communicate than personalised life goals [89]. Language barriers presented additional challenges for doctors in providing personalised information to patients [51]. When quality of life was mentioned by patients, doctors did not always ask patients to elaborate further [45] and often reverted to how they viewed quality of life, which may differ from the quality of life as perceived by the patient.

iv) The role of decision aids

The purpose of decision aids is to assist patients in understanding their clinical situation and facilitate communication between doctor and patient. Their underlying purpose is to help patients participate in medical decisions that relate to them [51, 52]. Two studies demonstrated that decision aids allowed more consistent information to be delivered to patients. However, this was not specific to the needs of minority ethnic and or vulnerable patient groups [51, 52].

v) Understanding of information

Patient’s understanding of the information provided to them was influenced by their lack of mental capacity due to pre-existing cognitive impairment, for example, dementia [83], the effects of acute illness including, the presence of physical pain or other associated discomfort [56, 58, 73, 82]. For patients who were deemed to have mental capacity, information was often poorly or incompletely understood. This occurred when the information provided by the doctor exceeded the patient’s ability or capacity to understand and retain specific items of information. Doctor factors that were associated with information overload included the overuse of medical jargon [42, 45, 73], talking too quickly [58] and providing too much information at a specified point in time [42, 45, 58, 73]. Patient factors that were associated with a lower proficiency for understanding were low health and information literacy [51, 70], high emotional states [78, 82] and language barriers [51]. These all contributed to poor understanding between doctor and patient. Prior patient experience of hospital treatment and doctors checking understanding of any new information provided contributed to improved understanding [47]. Although doctors acknowledged the importance of patient understanding and recognised patients’ expectations, wishes and values [47], these were inconsistently explored and patients did not always feel empowered to speak up or own up in the event of a lack of understanding [42]. This resulted in patients being less informed [42].

vi) Consequences of being poorly or misinformed

Poor or misinformation resulted in poor recollection by the patient of the contents and nature of discussions about the goals of care. This had the potential to contribute to fluctuating patient treatment preferences, increase conflict between patients and their families and doctors and increase the possibility of discordant treatment [43, 75, 77].

A) Doctors

Intuition often predominates over analytical decision-making by doctors in acute hospital illnesses [54, 57, 82]. Intuition uses experience, feelings and accumulated judgments that culminate in heuristics (mental shortcuts). These are prone to error and bias which can translate into communication biases [86] or conversation avoidance [49]. Although some doctors acknowledged the importance of remaining neutral, factual and non-influential when providing information [47] this was contrary to what was observed in practice. Conversation-framing bias was reported in eight studies [42, 47, 54, 55, 58, 59, 69, 79] and was more common in situations where higher clinical uncertainty was present [69]. ‘Framing’ is a cognitive bias that relates to how a patient’s decision or thought process may be influenced by the way information is presented to them by their doctor. Evidence demonstrated that doctors sometimes persuade patients to agree with their thoughts about the goals of care [42, 54, 55, 58, 59, 69, 79]. The rationale described was superior clinical knowledge to make better ‘best interest’ decisions. It was also thought to minimise the influence of highly charged emotions experienced by families concerning acute illness that could potentially contradict the patient’s long-term goals [42, 58, 59].

Fears of repercussions were reported as a barrier to having goals of care discussions and were perceived to change the nature of such conversations. These were related to fears of causing undue suffering and distress, fears of a “difficult reaction” from family or patients, leading to conflict and fears of medico-legal repercussions [80]. As a consequence, this resulted in doctors being more hesitant about withholding life-sustaining treatment [85].

B) Influence by families

Families played a varying role in goals of care conversations. This may be influenced by different personal sociocultural beliefs about illness and the role that they (and their patient relative) were expected to play in such conversations [52]. There were examples of situations where patients viewed their family members as integral to goals of care discussions, sometimes even deferring to them for all discussions and decisions relating to their medical care. However, this was not always identified by their doctors who sometimes made efforts to speak with the patient when family members were not present [58]. Discordant views regarding treatment options also existed between patients and family members [47]. This occurred when patients and their families held different motives and priorities relating to what they viewed as desirable medical treatments and associated outcomes [43, 75, 85], for example, the family prioritising survival whilst underappreciating the trade-offs of physical and mental morbidity (e.g. delirium and psychosis) associated with life-sustaining treatment [85]. Strongly expressed views from a family member were shown to influence a patient’s thought processes relating to goals of care [80]. This was particularly relevant when the intra-family conflict was present and had the potential to harm trusting relations between doctors, family and patient [73] culminating in fewer goals of care conversations [49]. Doctors also recalled some of their patients’ wishes centred more on what their family wanted rather than them, which in some circumstances contradicted their values [46].

The impact of a family’s emotional state was intrinsically linked to the appropriate timing of conversation about goals of care [23, 54, 63, 73, 83, 85]. This was particularly relevant to crises where there was an imminent risk of death [65]. Highly charged emotions were identified as being further intensified by the presence of low health and information literacy among family members [54, 80]. This sometimes resulted in instability of expressed preferences and views based more on feelings and less on facts or balanced opinions [82]. Unrealistic expectations, usually associated with undue optimism relating to prognosis and treatment were also evident [49, 90].

c) Other influencers on Information provided

The internet and media reporting, for example, the reporting of “miracle cases”, may be influential towards patient and family misconceptions about treatments and prognosis before goals of care conversations [60, 83], potentially leading to mistrust [60, 83]. Prior discussions with doctors that had inconsistent themes were also shown to impact conversations. This was particularly relevant when patients and their families had the desire to “snatch” clinical news from different conversations, hoping to get a custom-made truth that better fitted their emotional desires regardless of their positive or negative value [87]. Language barriers and the use of translators resulted in additional biases, omissions or additions introduced by the interpreter and are thought to have potentially significant consequences on goals of care conversations [71].

(A) Skills of doctors (clinical expertise, communication and interpersonal skills) to develop trusting relationships

Appropriately skilled doctors were perceived by patients/families to have credibility, clinical expertise and a high level of competence [58, 60, 61]. Appropriate interpersonal skills were also perceived as important. These included measures such as introducing themselves, addressing patients or surrogates by their name, being personable [47, 72] and warm [61, 85], being approachable [85], listening to and understanding patients’ queries and concerns [23, 60, 63], providing information in a clear and jargon-free manner [19, 47, 61, 73, 82], showing honesty [85], compassion [47, 61, 73] and treating patients with respect and dignity [16]. The inclusion of family members in discussions (with the approval of the patient) was also deemed important in certain situations [19]. In situations when a patient lacked capacity, some family members took more of an advocacy role as the patient’s spokesperson, sometimes by default. This sometimes resulted in guilt, either expressed or implied by the family, particularly when asked for their input on potentially withdrawing or withholding life-sustaining care for those at the end-of-life. Doctors who acknowledged this and were understanding, supportive and compassionate resulted in more trusting relations and paved the way for families to discuss such issues more openly [23]. Continuity of care associated with its physical presence and frequent communication were also integral to building a relationship, rapport and mutual trust between doctors and patients [19, 23, 42, 60, 63, 72, 79, 85].

However, doctors frequently admitted lacking confidence and skills in discussing goals of care. Fear and anxiety of “taking away hope” and or getting into conflict with patients or families were reported, particularly when the patient and their families were perceived to have unrealistic expectations [16, 42, 79]. Patient or family trust was also perceived by doctors to be threatened when decisions to limit or withhold life-sustaining treatments were suggested, due to fear of abandonment and inferior care [85]. Doctors felt poorly equipped in managing and communicating prognostic uncertainty and making treatment recommendations when uncertainty was present [23, 45, 63, 76, 80, 85]. This culminated in doctors providing patients with what was viewed as sub-optimal and inconsistent information [76], requesting patients make treatment choices that risked contradicting their best interests and personal values [42, 57, 76, 81], or doctors avoiding goals of care discussions altogether [23, 49, 57, 63, 86].

Further communication skills training and support, for example, via mentoring were welcomed by doctors [19, 63, 67, 72]. However, the effectiveness of training interventions has not been established [76]. Uncertainties remain as to where the focus and content should lie, what are the most effective ways of delivering training, the optimal intensity and frequency of training interventions, how to measure their effectiveness in the short and longer term and how to engage less motivated learners [45]. The role of interpersonal accuracy and behavioural adaptability is also unclear. This relates to a doctor’s ability to recognise emotions and motivating factors or thoughts in patients (interpersonal accuracy), and adapt their communication style accordingly (behavioural adaptability) [47]. In one study, female doctors had positive correlations between interpersonal accuracy and verbal and non-verbal behavioural adaptability which translated to more positive patient consultation outcomes (for non-verbal adaptability) [50]. However, for male doctors, better interpersonal accuracy was linked to less non-verbal adaptability for unclear reasons [50].

B) Other external threats to trust

Pre-conceived views by patients and their families influenced trust in their treating doctor. Negative prior experiences with healthcare and information retrieved from the internet and other public sources of information contributed to a lower level of trust by the patient and family [60]. This was more prominent among those with relatively low health literacy, those who were less well educated, unemployed, with no medical insurance (in a fully privatised healthcare system), or those who were homeless [51]. Mistrust was also threatened when patients felt stigmatised because of their medical conditions such as human immunodeficiency virus and Acquired Immune deficiency syndrome (HIV/AIDS), psychiatric illness, including anorexia, bulimia, and substance abuse, sickle cell anaemia and other physical disabilities [51].

Cultural differences between doctors and patients and or families combined with a low level of cultural literacy (understanding) amongst doctors were detrimental to a trusting relationship [46]. Cultural differences might relate to “do not attempt cardiopulmonary resuscitation” (DNACPR) orders, the taboo around death and how death is perceived and communicated [46, 82]. However, families of culturally diverse backgrounds lacked awareness of how to communicate their cultural needs which sometimes resulted in conflict [49] or being stereotyped based on their religion or culture when their values were not always completely aligned [46].

Power differentials between doctors and patients and their families influenced trust [23, 63, 68, 80]. Furthermore, a disproportionate level of delegated power from doctor to patient and their family was considered a risk to the ongoing doctor-patient and family relationship, particularly when there were discordant views between both parties [23, 63, 68, 80].

C) The Impact of Trust and Mistrust

Mutual trust between doctors and patients and their families facilitated improved understanding and positively enhanced ongoing relationships and rapport [60, 73]. Conversely, a lack of trust or misconception about treatments translated to feelings of abandonment, neglect or inferior care by the patient and their family [60, 63]. This sometimes resulted in more invasive care that was not always perceived to be appropriate but thought necessary by the doctor to avoid further conflict [60]. Frequent end-of-life communication and conflict resolution [63] had the potential to cause a high emotional burden for the doctor, avoidance of further engagement between doctor and patient in goals of care planning, or a breakdown in mutual trust between the two respective parties [16, 23, 51, 86].

Review and further refinement of CMO hypotheses

Further analysis and reasoning suggest that if we are to understand with more confidence what happens in real-life practice in goals of care planning in acute illness, considering each CMO separately and in isolation is likely to be oversimplified. It is more appropriate to first consider the factors that influence to what extent goals of care conversations are initiated.

Here, the onus lies largely with the doctor and their motivation and willingness to engage. Our collated evidence suggests that this motivation requires “buy-in” from the doctor about the benefits of goals of care planning (“context” hypothesis 6) together with feeling appropriately trained, skilled and confident to hold such conversations (“context” hypothesis 4). Therefore hypotheses 4 and 6 are closely intertwined and share the same “mechanism", in this case, motivation, that translates to a new "outcome” of interest–initial engagement in goals of care conversations between doctor and patient. In addition, hypothesis 5 relates to a process or system, where patients most likely to benefit from goals of care conversations could be better identified (“context” hypothesis 5) to aid clinical prioritisation (“mechanism” hypothesis 5), in promoting goals of care conversations. However, existing evidence is lacking to support this theory at present.

The second consideration involves the nuances in how conversations are conducted once initial engagement has occurred. This involves a more complex relationship between the “context” and “mechanisms” of hypotheses 1 (information provision), 3 (mutual trust) and 4 (skills and confidence of doctors) than our initial theory suggested. Our proposed model (Fig 2) depicts how the “context” and “mechanisms” of these hypotheses may inter-relate and combine to achieve the final “outcome” of interest–a shared understanding between doctor and patient regarding treatment goals and priorities. In this model, it is of note that the skills, confidence and interpersonal relations of doctors ("context” hypothesis 4) play a dual role in conversation initiation and the conversation process after initiation via two separate “mechanisms” and is therefore a critical component of goals of care communication in this setting. Our data did not suggest any refinements needed for CMO2 (communication bias) or CMO 7 (organisational influence) which are both relevant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Model proposing how CMO hypotheses may be inter-related.

https://doi.org/10.1371/journal.pone.0299933.g002

Exploring the evidence further

Initiation of goals of care conversations requires doctors to value the importance of and acknowledge the benefits of a patient-centred approach [64]. Decreasing the likelihood of non-beneficial interventions, providing better patient care and allowing more open communication with the patient and their relatives were reasons cited by doctors that motivated and incentivised them to engage with patients and families about goals of care [64]. However, despite this, time pressures from competing work and the lack of any validated tools to identify patients most likely to benefit from goals of care conversations at the most appropriate times present additional challenges [23, 79, 86]. Conversations held too early risk mistrust among patients, particularly when they perceive the information provided is not relevant to them at all, or just yet [22, 58]. Conversations held too late may be limited by high emotional states at crisis points that have the potential to contradict a patient’s longer-term values and increase the risk of burdensome care [54, 65].

Evidence also supported doctors lacking confidence and competence in initiating and holding goals of care conversations often due to fear of conflict when discordant views were present or anxieties relating to discussing end-of-life issues [16, 23, 42, 45, 49, 57, 63, 76, 79–81, 84–86]. Communication skills training was frequently welcomed by doctors and studies show a range of training formats. However, evidence shows the effectiveness of any communication training is generally lacking [63, 72, 76, 83, 94, 95]

A key aim is for doctors to harbour a positive relationship with their patients and families and foster trust. Open and honest communication (in a clear way that the patient can understand), showing competence and confidence, being personable, and “warm” and continuity of care were key factors described by patients/families to gain a trusting relationship with their doctor [58, 60, 61, 73]. However, information provided by doctors was frequently inconsistent [45, 84], biased [42, 54, 55, 58, 59, 69, 79], not personalised [42, 43, 45, 56, 58, 77, 82] and often poorly or misunderstood [42, 45, 58, 73]. These can negatively impact trust leading to discordant views, all of which have the potential to result in higher burdensome non-beneficial treatment [60]. Low cultural literacy and lack of cultural awareness amongst doctors risk them failing to adapt conversations to the needs of patients from minority ethnic communities. This can be of further detriment to trusting relationships between doctor, patient and family [46, 80].

Proposals for further research, training and role allocation

Future work should focus on ways of ensuring that doctors are aware of and initiate and deliver goals of care conversations aligned to factors highlighted in this realist review. This could be accompanied by a realist evaluation [96–98] of training in this area with emphasis on how doctors are currently trained and how existing training could be modified to ensure doctors have the appropriate skills and confidence to effectively identify and communicate with patients and their families about goals of care in acute illness. Moreover, the development of validated training outcomes and patient experience measures would pave the way for the development of effective training methods and approaches [94, 95] Aside from doctor training, the role of nurses and other allied healthcare professionals requires further exploration. Our findings identify that nurses are currently underutilised and undervalued in goals of care conversations [16, 63]. Their contribution may be limited because there is no defined formal nursing role. However, nurses have considerable potential to facilitate communication between doctors and patients by enhancing mutual understanding, communicating patient concerns to doctors, and providing emotional support to patients and families. These have the potential to foster better relationships, increase the level of mutual trust and promote more timely conversations [16, 63, 78]. This is particularly relevant in a busy healthcare system where doctors are often time-pressured to have two-way communication between themselves and patients and or families. Although this review did not highlight any role for other allied healthcare professionals, for example, but not exclusive to physiotherapists, clinical psychologists, speech and language therapists and social workers; this may well be worth exploring further in future studies. Each allied health professional may offer a different perspective and skill mix, which may be of benefit to doctors, patients and their families in the context of goals of care discussions. Furthermore, this may also pave the way for multi-professional training which has been shown to improve confidence, knowledge and skills amongst healthcare professionals who underwent a training workshop in “difficult conversations” [95].

There is a need to further understand organisational and healthcare systems in addition to wider societal influences on the doctor-patient interaction. Finally, there is a need to explore the nuance of goals of care discussions. It is here that health professionals for example doctors may use a variety of ‘voices’: the ‘doctor voice’ to ask specific questions; the ‘educator voice’ to share information and help patients understand their illness, situation and treatment, and the ‘fellow human voice’ to convey empathy. By showing empathy through comments for example “I understand” or “That must be really tough”, health professionals share a ‘fellow human voice’ encouraging patients to discuss goals of care [33].

Strengths and limitations of this review

A strength of this realist review is both its explanatory and theoretical nature- to understand the complex mechanisms underlying goals of care discussions with patients with severe acute illness in hospital settings. This style of synthesis shifts the focus from specific interventions and services to broader underlying mechanisms or principles. Second, we actively incorporated key stakeholders’ views alongside published literature to refine the review focus to areas considered most pertinent to clinical practice. Twenty-seven of the studies included within the review were qualitative and five were mixed method, a strength being that studies of this type permitted salient contexts, mechanisms and outcomes to be understood in detail, particularly where ‘thick’ accounts were evident.

However, the findings and recommendations of this review would not be complete without reference to the limitations of this work. Whilst a high number of studies in this review were qualitatively orientated which allow for a deeper understanding of the key concepts, few studies made use of methodological approaches that permit wider generalisation from their findings. Second, most of the identified studies were from Global North and Westernised countries (Australia, Canada, USA, UK or other European locations) and may not translate to non-westernised societies. Third, the influence of cultural values was not explored in detail nor was any detailed analysis of the family influence and patient-family relations [99]. This might include any possible impact of family persuasion or dominance in conversations between patients and doctors and discordant views between patients and families. Fourth, many included studies were classified as being ‘weak’ to ‘moderate’ concerning their methodological design and therefore subject to potential bias (Table 3). Fifth, this review did not explore how shared understanding between patient and doctor relates to the type of patient-doctor relationship. This relates to how decisions are made once there is mutual understanding between patient and doctor of treatment goals and priorities and incorporates a spectrum of decision-making from a paternalistic approach through to pure shared decision-making [100]. This is regarded as a separate entity and warrants a separate study. Sixth, multiple ‘wrap-around’ preconditions that underpin successful shared decision-making may also be present. This necessitates a socioecological lens [101] in which a whole systems strategic approach acknowledges multiple, interconnected elements potentially exist and reside at different societal and organisational levels of influence (microsystem (person, needs and characteristics), chronosystem (dynamic influences of time), mesosystem (interactions with family/health professionals), exosystem (healthcare services/systems) and the macrosystem (societal influences). All may be necessary to consider before, during and after implementing shared decision-making conversations and warrant further exploration.