Theme 1: decision-making processes
Decisions discussed in the papers concerned resuscitation, hospitalisation, surgery, ventilation, feeding via nasogastric tube or gastrostomy, and drug therapies such as antibiotics and sedation, as well as euthanasia in jurisdictions where this was legal. Such decision-making could happen in advance (via documentary or verbal evidence) or contemporaneously. In some cases advance and contemporaneous decisions conflicted.
The papers identified the distinctive roles of families and healthcare professionals in decision-making, but less often directly considered the patients themselves. Families often had different perspectives from healthcare professionals which, if unresolved, led to families expressing dissatisfaction with the dying process they witnessed. The universal approach to resolving such differences, as well as relieving the documented burdens on families of making decisions for dependent relatives, was shared decision-making, reflecting the current ethico-legal steer . However, family opinion carried apparently different weight from one study to another, suggesting that decision-making was not being shared in a consistent manner. The patient’s own documented or verbal antecedent end-of-life decisions (by which we mean, among others, advance decisions and advance care plans, see below) might be considered, but whether these were followed depended on the outcome of medical assessment of the patient’s ‘best interests’, as well as the extent to which a family (if not agreeing with the antecedent decision) could assert their own wishes. Although rarely explicitly studied, close reading of many papers suggests that patients themselves may influence decision-making even when lacking capacity, and explicit study of this phenomena seems to represent a gap in the literature.
The papers presented clinical decision-making taking place either triadically, with the healthcare professional, patient and family member playing a role (e.g. ), or else dyadically, with the family and the healthcare professional making key decisions, albeit informed by their understanding of what the patient might have wanted (e.g. ). While it was rare to exclude any consideration of the family , dyadic conceptions often excluded the patient’s active contemporaneous involvement for practical reasons such as total patient incapacity [24, 25] or mortality . But besides these practical reasons there was a clear, although unwritten, implication that the primary agents of decision-making were families and health care professionals, rather than the patients with reduced agency. Studies that directly involved patients with impaired capacity were rare [27, 28], while some implicitly conflated the experiences of families and patients .
Family and health care professional perspectives
Family perspectives can be discordant with those of clinical observers. Families of people with dementia were often disturbed by the symptoms experienced or exhibited by their relatives at the end of life, while clinical observers viewed them as consistent with a peaceful death [30–32]. Jox et al.  found almost a quarter of relatives of patients with traumatic brain injury disagreed with their doctor’s assessment of the patient’s level of consciousness. Families of patients in the acute stage of severe traumatic brain injury expected life-saving treatment either to be almost fully restorative or to fail completely, but few accepted that the likely result was a partial restoration of physical health alongside severe mental disability, in spite of receiving such information from clinicians . Similar expectations were visible in the families of people with dementia, a long term degenerative condition [27, 35] where death by a catastrophic event (such as a cardiac arrest) was anticipated, rather than a slow and extended decline.
In some cases, healthcare professionals shared the family’s failure to acknowledge that dementia was a terminal condition. Sampson et al.  notes that many professionals in acute care were unaware that dementia was a terminal illness, and two papers [30, 36] observed that the death certificates of deceased patients who had advanced dementia often failed to note this diagnosis as a cause of death.
While doctors were in theory guided by the best interests of their patients, and empowered (in many jurisdictions) to make decisions on that basis, clinicians were aware that the variable opinions and philosophies within the healthcare team led to inconsistencies in practice, and this was considered to make any communication problems with families worse .
Impact on families
Dening et al.  noted that family carers of people with dementia felt the toll of caring and often had profound regrets about their own situation, especially if they cared for a parent rather than a spouse. Some studies also note the burdensome and poorly supported environment in which carers make decisions [33, 34, 39], both in dementia and traumatic brain injury. Others highlight the feelings of guilt or failure family carers of people with dementia bear , especially about decisions that violate relatives’ perceived wishes and the strong sense of obligation carers may feel . Relatives of people with dementia may feel particularly unsupported: one epidemiological study of end-of-life shows that presence of dementia in a dying relative is associated with family dissatisfaction with medical communication . Attention to family wellbeing may vary according to the setting, with one study noting that families’ emotional needs were more likely to be recorded in hospital and hospice documentation than in nursing home records .
A sense of obligation may lead family members to wish for more treatment for their relatives than they would choose for themselves . However, family members of people with dementia who participated in focus groups showed increased levels of regret if they had requested aggressive interventions at the end-of-life . Such discordant thinking is also seen in health care professionals’ wishes for their patients [43, 44].
Weight of family views
Where families requested treatment in contradiction of previously agreed care plans [24, 34], many studies indicated that their wishes were likely to be adhered to, even if this was in opposition to the earlier expressed wishes of the relative [24, 33, 38, 45] (note there was no evidence that families rejection of treatment carried a similar weight). While this key role for the family was widespread in dementia, traumatic brain injury and frailty, one study indicated that family members were rarely consulted  and others felt they had little influence on decision-making . This finding presumably reflects the broad scope of our literature search – one study highlights regional and international differences in medical values  – but there are also signs in the literature indicating that there can be dynamic differences in the manifestation of shared decision-making from one case to another.
Ambiguity of “shared decision-making”
It was implicit across studies of frailty [46, 47], dementia [25, 32, 35, 48–53] and traumatic brain injury [33, 54] that decisions were shared between doctors and families and/or patients. Moreover, regardless of clinical setting, doctors changing their treatment plans in response to family intervention may indicate that decision-making is to an extent shared [21, 24, 28, 55]. Yet this understanding of equitable authority is muddied by reports that family wishes were circumvented when they strayed from the (previously agreed) medical plan in palliative care settings  or where their view of what the patient might want was in opposition to a clinician’s assessment of best interests. In acute settings, while families often readily acceded to aggressive lifesaving treatments, families who opposed these were overruled by the healthcare team . The concept of shared decision-making appears ambiguous, an ambiguity which is captured in the finding from a survey of family members of patients with traumatic brain injury that a majority of families felt their role was simultaneously to make decisions and to accede to decisions made by the doctor . This sense of a blurring of the boundaries between making and agreeing to decisions is further compounded by one case study series, again of traumatic brain injury, which emphasises that decisions to withdraw nutrition and hydration were made by doctors because of their challenging nature .
The role of incapacitated patients in decisions
An important aspect of decision-making is the role that legally incapacitated patients may play. This role is easily overlooked, first because the patient is not considered to have capacity to make decisions, and secondly because many patients with high levels of dependency display overt unwillingness to engage with decision-making. One survey indicated that many people with dementia wished others, usually trusted family members, to make decisions on their behalf . Further, many people with dementia and frailty in old age did not wish to discuss end-of-life issues [21, 45] and people in the middle stages of dementia could resist change and deny problems to their families and family doctors (General Practitioners, hereafter GPs), especially if they were used to holding an authoritative role in the family ; in these cases families and GPs often put on a united front to overcome this resistance.
While a lack of formal capacity results in routine marginalisation  or even exclusion from decision-making [24, 49, 53, 58] of frail older people and people with dementia, the documented resistance of people with dementia and/or frailty in old age indicates an ability to influence decisions. The difficulties of engaging in discussion with a previously authoritative parent or relative  indicate that family dynamics may exert pressure on how decisions happen. Indeed, this speaks to findings that the identity of carers can have a measurable impact on decision-making: an epidemiological study of decisions to forego hospitalisation in advanced dementia  showed a strong association between patients having a proxy who was not a child and possessing a ‘do not hospitalise’ order, which the researchers suggest indicates that patients’ children may be less willing to limit treatment than other relations (including spouses) or court appointed proxies. Similarly, a survey by Rurup et al.  notes that, while many family members agreed with limitations to treatment for relatives at the end-of-life, children and spouses agreed less often than those with other familial relationships.
A variety of influences can be glimpsed in the literature; the fact that individuals with traumatic brain injury survive at all, ‘against the odds’, in intensive care units can be seen as grounds for hope by families [22, 33, 34]. This is significant given that Jox et al.  note that the family’s perception of the patient’s wellbeing is more frequently influential than the patient’s advance directives or past attitudes. Clearly there is a symbolic role that signs and symptoms of illness play in family reasoning. Yet overt influences on decision-making that rely on the patient’s physical and verbal behaviours are also glimpsed in the literature. One survey of attitudes to artificial nutrition and hydration noted that when nursing home residents with advanced dementia refused to eat and drink, a large minority of relatives and nurses (although fewer doctors) agreed that this refusal should be ‘respected’ ; another study noted that GPs’ decisions to hospitalise frail older persons who lacked capacity to make decisions were influenced by the patient’s verbal wishes as expressed at the time . Such observations offer a glimpse at the various ways that people with impaired decision-making ability influence the decisions of others, and such influence is recognised in law in some jurisdictions,Footnote 3 the paucity of sources in this review may suggest that this is an under-researched area in this literature.
Theme 2: limiting the intensity of medical treatment
While only one study explicitly examined the issue of a ‘good death’, many other studies took an implicit position on a good death, viewing this as a death that was not overly medicalised. The aetiology of illness dictates the type of medicalisation to be avoided, and in the conditions under investigation, medical omissions were often of life-saving drugs (commonly antibiotics) and of artificial nutrition and hydration (usually pertaining to liquid feeds that were administered via nasogastric or gastrostomy tube). While avoiding hospital or aggressive treatment was associated with the idea of a good death, many patients and carers expressed an active wish for medical intervention, especially if confronted with acute deterioration or imminent death. Thus there was a substantial paradox where not only a natural death, but also aggressive life-saving treatments, were desired. Indeed, misconceptions of the type of death a relative could expect may be compounded by under-recognition of the trajectory of degenerative conditions among medical professionals, affecting the information they communicate to families.
Patterns of medicalisation
A number of studies documented the degree of medicalisation of deaths, with one study reporting that almost 1 in 5 patients received ventilation, resuscitation or surgery in the last 48 h of life . Di Giulio et al.  observed the frequent use of restraint, antibiotics and feeding by gastric tube at hospital institutions caring for those dying with advanced dementia, and others reported a comparative reduction in these measures in nursing homes and residential care . While studies reported wide variations between hospitalisation rates from different nursing homes [60, 61], Lamberg et al.  reviewed the records of 244 nursing home residents dying with advanced dementia, and noted that about 20 % had a period of hospitalisation in the last six months of life. Similarly, relatively high proportions of frail older nursing home residents had contact with out-of-hours care and ambulance services in the month prior to their deaths . When compared to other types of long-term patients, length of hospital stay of people with dementia and cognitive impairment were lower [40, 61, 62]. Basic and Shanley  reported that people with dementia were less likely to die during hospital admission, speculating this may be because those who are hypermobile have frequent review by their doctors and low nurse-to-patient ratios. Basic and Shanley’s findings are tempered somewhat by the findings of a cohort study of dementia patients by Sampson et al.  which reported that the likelihood of mortality for dementia patient’s in acute hospitals was nevertheless high, and increased as cognitive impairment or dementia worsened. I am grateful to a peer reviewer for this information.
It is noteworthy that some studies disputed the development of a dogmatic approach to de-medicalised death. For instance, while some studies indicate that vexatious factors such as overwork, poor communication and potential for litigation played a part in clinical decisions to hospitalise nursing home patients , other authors argue that some clinical emergencies, such as electrolyte imbalances and fractures, would be inappropriate to keep out of hospital , and similar themes were highlighted by relatives of older home hospice patients who requested hospital in the final days of life . Patients may not equate a de-medicalised death with a good death; as Soskis  notes, patients from marginalised communities “are far more likely to worry that treatment providers will give up on them too soon or will not consider them worth saving”.
Given the common steer toward a death that was not unnecessarily medicalised, a major impediment to avoiding prolongation of life for its own sake is distinguishing terminal trajectories from other life threatening emergencies [21, 38, 66]. Lamberg et al.  note that almost half of decisions not to hospitalise nursing home residents with dementia took place during the last 30 days of life, while more than half of decisions to palliate rather than actively treat took place in the last 7 days of life. The relative lateness of such planning may of course be related to a lack of recognition that dementia is a cause of death in its own right, highlighted in the previous theme. Such a conclusion is in broad agreement with a finding that GPs more frequently recognise when deaths from cancer were about to take place than deaths from all other conditions . These authors note that where death is not anticipated, sudden deterioration is more likely to result in hospitalisation rather than palliative care.
What constitutes a ‘good death’?
What does, and does not, constitute a ‘good death’ is a major area of academic debate, and this was not directly examined in the scoping review. As noted above, many studies took an implicit position on the meaning of a ‘good death’ by problematising the degree to which death is medicalised. Only one study, by Bosek et al. , engaged with the literature on a good death explicitly, attempting to reconcile family members’ perspectives on the quality of death of a recently deceased relative with late-stage Alzheimer’s disease with the Steinhauser et al.  typology of good death. Bosek et al.  report that, while many families were not at the bedside when their relative died, most felt that their relative had died with dignity. This notwithstanding, more than a quarter of family members felt their relative had not experienced a good death. Impediments to a good death included undesired symptoms and, in more than half of cases, failure to fulfil the antecedent wishes of the individual. Bosek et al.  link this to the concepts of symptom control and planned death that are central to the Steinhauser et al.  typology, concluding that the cognitive impairments late stage dementia present severe challenges to achieving a good death in Steinhauser’s terms, and calling for a more proactive approach to end of life care.
Shifting perceptions of dying well
Effective symptom control may be important, but understanding what constitutes a symptom requires a particular reference point, and some studies identify the unpreparedness of family members for the dying process. For instance, a study of nurses’ and relatives’ experiences of continuous deep sedation of dying dementia patients theorised that differences in the familiarity of relatives and nurses with the dying process led to differences in their perceptions of dying as (for instance) peaceful or agonising . Several other studies had similar findings; thus studies observed antecedent wishes expressed by the patient to avoid machines or extraordinary measures  or the frequent desire for natural death among family members , but noted the lack of specific information about what such deaths would involve. Indeed, families may offer paradoxical visions of good death; thus, while electing not to treat pneumonia and other infectious diseases is sometimes seen as allowing a natural death, both Forbes et al.  and Potkins et al.  note that failing to treat such conditions was often not acceptable to members of many families of people with dementia (although, where dementia is severe, there may be a rise in acceptability). It is notable that death does not always follow withholding and withdrawing antibiotics or other treatments [34, 56]. Regardless of this, a study of home hospice patients indicates that the reality of the symptoms and sequelae of the dying process prompts some families to request hospitalisation in the final stages of life . There is also a consistent tendency among both professionals and family members toward requesting more clinical interventions for others than they would wish for themselves [42, 44].
Requests for medical interventions are also made by families of patients in prolonged or permanent vegetative states [33, 34]. While often thinking their relative has an extremely poor quality of life, many families strongly resist withdrawal of artificial nutrition and hydration [22, 33], despite this being a legally permissible avenue to a ‘natural death’ for this condition in some jurisdictions. Clinicians were more likely than families to find withdrawal of artificial nutrition and hydration acceptable [37, 43, 56].
Theme 3: antecedent end-of-life decision-making
The final theme was concerned with informing future end-of-life treatment in advance. A bewildering range of terms—some legal, some bespoke—for such future planning were used, and the term antecedent end-of-life decision-making (AEDM) is used here to include all the terms encountered. Despite evidence of international momentum in adopting AEDM practices, there is apparent reluctance among some target populations to engage in antecedent end-of-life decision-making and still large number of obstacles confronting those who do want to do so, and the literature described a range of barriers patients, families and professionals encounter when trying to act in accordance with antecedent wishes. The literature revealed that families and health care professionals find discussing death difficult, and that documentary forms of AEDM (see below) are often felt to have little influence on decisions in clinical practice. Implicit in many studies was a wide variety of attitudes to death and dying, which may speak against a single approach to end-of-life decision-making. Despite this, some studies succeeded in promoting antecedent end-of-life decision-making, and documented its benefits.
A multitude of terms
A great many terms were used to describe AEDM within the papers. In part, this was due to the diversity of this review, drawing on papers from 13 different countries. In total the studies used 16 terms: advance directive, written advance directive, advance decision, advance care plan, documented advance care wishes, anticipatory care plan, power of attorney, durable power of attorney, durable power of attorney for healthcare, living will, values history, previously expressed wishes, previously expressed patient wishes, family physician treatment order, resident and family wishes, and familial advanced plan. Some of these terms are legally defined, some are not, (or are inaccurate attempts to reference legal terms) and the latter may be unique to the studies within which they are used. It appears evident that there are areas of overlap between these terms, although there are also important differences; however, exploring these differences, or the legal frameworks of different jurisdictions, is not the focus this paper.
The most common terms were advance directives and advance care plan, understood (in these studies and more generally) to represent respectively legally binding refusals of treatment, and non-legally binding statements of wishes and preferences. The UK uses the term advance decision, rather than advance directive, which is preferred in the USA (and refers to a broader range of advance planning instruments). To avoid ambiguity, we use only the terms advance decision and advance care plan in this discussion, or where both are implied, antecedent end-of-life decision-making.Footnote 4
The abundance of different terminology used in the literature may give rise to confusion among health care professionals, and one UK study indicated that professionals were unsure of the difference between an advance decision (the legally binding refusal of treatment) and the more generic ‘advance care plan’ .
Support for antecedent end-of-life decision-making
When the period preceding death is characterised by a loss of capacity, carers and clinicians face ethical dilemmas about the types of treatment they should instigate or withhold. The papers reviewed are unanimous in considering end-of-life care based upon the patient’s antecedent wishes to be the gold standard in end-of-life treatment. While this is in line with the mores of western medicine, this review indicates that this is increasingly an international viewpoint, which is reflected in research produced in countries with little social history of such interventions [47, 69] - albeit these studies take place in Hong-Kong which is likely to be more susceptible to the influence of British practice than mainland China.
Barriers to antecedent end-of-life decision-making
Antecedent end-of life decision-making is a potential solution to the ethical dilemmas faced by clinicians and carers, yet this method of decision-making has failed to permeate practice. In the USA, where advance decisions to refuse treatment have been legally enforceable since the early 1990s and has received extensive publicity, indications are that only a third of adults have completed some form of antecedent decision . In the UK, where advance decisions were given legal force in 2005, although with little publicity, only 5 % of people are reported to have formally engaged in AEDM – with fewer taking advantage of the legal instruments .
The papers in this review bear out these figures, indicating that some patients do not want to engage in antecedent end of life decision-making, or face obstacles to doing so. People with dementia [28, 45] and frail older people [21, 47], may be reluctant to think about their deaths and consider future end-of-life wishes. This reluctance is also felt by some members of families of people with dementia [25, 39, 68] and is echoed in the responses of some families of patients with severe brain injury who may resist making decisions to limit medical treatment even when injuries are catastrophic .
Dementia studies indicate that healthcare professionals can find it difficult to engage in discussions about end-of-life wishes with patients [35, 49]. While other studies report difficulties in engaging patients with advance care planning irrespective of primary diagnosis [21, 27, 45], the symptoms of dementia may themselves cause problems with advance care planning. Dening et al.  note that the nature of the impairments people with dementia experience means that they are more likely to focus on their immediate concerns rather than on the future, and this may present an obstacle to AEDM for those in the early stages of the disease. Reluctance to discuss end-of-life wishes can be exacerbated by a lack of close relationships between staff and families, which studies reveal can be a particular problem in nursing homes due to high staff turnover [35, 38].
Variations among individuals and families
Families are not always uncomfortable talking about death and making appropriate plans , especially if they have personally witnessed situations where someone has experienced a ‘bad’ death [25, 28]. These and other studies demonstrate clear variations in attitudes toward death and dying among different clinicians [37, 58], families [48, 72] and patients , but few show these variations in values as clearly as Chan and Pang , whose study offers a typology of nursing home residents’ approaches to advance care planning. While having different end-of-life preferences, residents also varied in their decisiveness, readiness to think about their deaths, and degree of fatalism. It is clear that strategies could be developed to address some of these barriers, yet the study indicates that any homogenous approach to AEDM is likely to encounter significant numbers of patients and/or families to whom it is personally ill-suited.
The efficacy of advance care planning
Advance care plans (excluding advance decisions) have an advisory, rather than determinative, effect on decision-making and dementia studies indicated that they are overridden in the case of strong opposition from families [35, 49]. As noted earlier, family members’ perceptions of their relative’s wellbeing had more influence on their decisions than the relative’s oral advance wishes and past attitudes [35, 38], and the contemporary behaviours of incapacitated patients also influenced decision-making. There was little faith in the utility of advance care plans among some patients, families  and clinicians . Indeed, the difficulties of fitting the precepts of advance care planning to situations that involve illness trajectories other than cancer has been questioned by dementia practitioners drawn from a variety of settings .
Overriding advance decisions
Both dementia and frailty studies report advance decisions being overridden where there had been an antecedent rejection of artificial nutrition and hydration  and hospitalisation [59, 61].Footnote 5 A survey by Rurup et al.  found that many more relatives and nurses than doctors felt that advance decisions should always be respected; despite this, relatives’ satisfaction with end-of-life care was more heavily influenced by a positive, trusting relationship with the treating clinician than compliance with the relative’s prior wishes .
Even where they are respected, both advance decisions and advance care plans vary greatly in determinative quality and content. While some studies indicate moderate agreement between surrogate reports and documented patient wishes, the instructions may be non-specific, fail to address common end-of-life dilemmas (e.g. tube feeding and hospitalisation), or contain contradictory wishes or illegal requests (e.g. for euthanasia in the U.K.) [25, 50, 57, 65, 68].
Promoting advance care plans and advance decisions
A number of the included intervention studies were carried out within care settings for older people with dementia or frailty. Interventions such as staff or family education and facilitation encouraged families and patients to document antecedent end-of-life decisions [27, 39, 47, 52, 65, 66]. Many of these studies suggest that benefits arise from antecedent end-of-life decision-making. Frail older people who were involved in advance care planning showed more life satisfaction, less existential distress and increased preference stability , while GPs reported reductions in care costs and hospital admissions . Interventions caused bereaved families of persons with dementia to report increased satisfaction with end-of-life care and increased numbers of documented advance care plans  and advance decisions  in care homes. These findings notwithstanding, a study in an acute hospital setting reported that relatives were frequently unwilling to make advance care plans for family members with dementia, even with expert guidance, and concluded that relatives may be ill-suited to the task in these circumstances . Thus, this limited range of literature suggests the success of interventions to promote AEDM may depend on a variety of factors, including the setting and the participants, and this deserves further investigation.