Implementation of the Mental Capacity Act: a national observational study of changes in place of death for older heart failure patients with or without comorbid dementia

Background: Heart failure is increasingly prevalent in the growing elderly population and commonly associated with cognitive impairment. This study compared trends in place of death (PoD) of heart failure patients with / without comorbid dementia over the period of implementation of the Mental Capacity Act (MCA) in October 2007, this legislation supporting patient-centred decision making for those with reduced agency. Methods: Analyses of death certication data for England between January 2001 and December 2018, describing the PoD and sociodemographic characteristics of all people ≥ 65 years registered with heart failure as the underlying cause of death, with / without a mention of comorbid dementia. Multiple Poisson regression modelling was used to determine the prevalence ratio (PR) of dying at home or in care homes compared to dying in hospital. Covariates included year of death, age, gender, marital status, comorbidity burden, index of multiple deprivation and urban / rural settings. Results:120,068 heart failure-related death records were included of which 8199 mentioned dementia as a contributory cause. The overall prevalence of dementia was 6.8%, the trend signicantly increasing from 5.6% to 8.0% pre- and post-MCA (p<0.0001). Dementia was coded as unspecied (78.2%), Alzheimer’s disease (13.5%) and vascular (8.3%). Those with dementia were more commonly older, female, widowed, and had more comorbidities. Pre-MCA, PoD for heart failure patients without dementia was hospital 68.2%, care homes 20.2%, 10.7% dying at home. The corresponding gures for those with comorbid dementia were 47.6%, 48.0% and 4.2%, respectively. Following MCA enforcement, PoD for those without dementia shifted from hospital to home, PR: 1.026 [95%CI: 1.024-1.029]. This trend was not signicant for those with dementia, PR: 1.001 [0.988-1.015], hospital deaths increasing. Care home deaths reduced for all, with or without dementia, PR: 0.959 [0.949-0.969], and PR: 0.996 [0.993-0.998],

such as heart failure, the MCA may be particularly important in ful lling goals of care close to the end of life. Achieving care in appropriate settings and the preferred place of death (PoD) are generally accepted as benchmarks of good quality end-of-life care, death at home or customary place of residence usually regarded as the desired option [18]. Given the relatively frequent association of cognitive impairment with heart failure, it might be expected that the decision-making processes legally constituted within the MCA would drive changes in the nal place of care and death during the terminal phase of this condition. The code of practice setting out the standards required to comply with the MCA came fully into force on October 1st, 2007 [19]. Thus, we undertook a comparative trend analysis covering the period of implementation of this legislation to discern any resultant variation in the PoD of heart failure decedents rendered vulnerable by comorbid dementia.

Study design
This was a national population-based cross-sectional study examining anonymised death registration data collated by the O ce for National Statistics (ONS) from 2001 to 2018 relating to heart failure patients resident in England.

Data Source And Study Cohort
In the UK the death certi cate is completed by the responsible clinician, civil registration of the cause of death by a relative or another quali ed informant being legally required within 5 days of medical certi cation. Sometimes a coroner assumes this role after a post-mortem examination or inquest. Following transcription of the information on the death certi cate by the recording registrar, this is digitised and uploaded to the ONS for subsequent diagnostic coding in accordance with the 10th revision of the International Statistical Classi cation of Diseases and Related Health Problems (ICD-10). This study dataset comprised all deaths registered in England from January 2001 through December 2018 of people aged ≥ 65 years for whom heart failure was recorded as the underlying cause of death. We elected to study those aged ≥ 65 years at death a priori, perceiving this subset to include the vast majority of heart failure decedents with or without dementia as a marker of cognitive impairment. Heart failure as the primary cause of death was determined by the allocation of any ICD-10: I50 code by the ONS. Designation of diminished intellectual capacity for this cohort was determined when dementia of any aetiology was mentioned as a contributory cause of death. The presence of comorbid dementia was indicated by the application of ICD-10 codes G30 [Alzheimer's disease]; F00 [dementia in Alzheimer's disease with late onset, atypical or mixed type, and unspeci ed]; F01 [vascular dementia]; F02 [dementia in other diseases classi ed elsewhere]; or F03 [unspeci ed dementia]. ONS death data acquisition and coding processes are subject to regular quality assurance. Pertinent to this study period, it should be noted that in 2011 there was a change in ONS mortality data coding practice, the previous coding of unspeci ed cerebrovascular disease when registered as a contributory cause of death being reclassi ed as vascular dementia [20].

Variables
The outcome variable PoD and the deceased's usual place of residence are both recorded on death certi cates and also codi ed by the ONS. Characterisation of PoD for this study was based on the classi cation system de ned by the National End of Life Care Intelligence Network, part of Public Health England [21]. This speci es 5 groupings of PoD: (a) Hospital, which incorporates all acute, specialist, and community hospitals whether they be National Health Service (NHS) or private, but not psychiatric hospitals; (b) Care home, including residential and nursing homes; (c) Own residence, the decedent's usual place of abode, but excludes communal living arrangements such as convents, monasteries, hostels or prisons; (d) Hospice, commonly standalone NHS or independent establishments; (e) Other places, covering psychiatric hospitals, other people's homes, communal living institutions as described above, workplaces, public spaces or roads. This grouping also applies to those declared dead on arrival at hospital, potentially relevant to some heart failure patients who succumb to sudden death. Where the PoD was unknown or unspeci ed, these data were incorporated in descriptive statistics but not considered further in multiple adjusted analyses.
Period of death as the independent variable of interest incorporated a binary indicator for the year of death pre-and post-enforcement of the MCA [0: 2001-2007; 1: 2008-2018]. Covariates included age at death, number of mentioned contributory causes, gender, marital status, socioeconomic position as measured by the Index of Multiple Deprivation, and categorisation of the location of decedents' usual residence as urban or rural based on the relevant postcode as archived on the ONS classi cation system [22,23].

Statistical analysis
Categorical and continuous variables were described using count (percentages) and means (standard deviation [SD]) as appropriate. The number of patients who died from heart failure with comorbid dementia, and the proportion of hospital deaths among heart failure decedents with or without comorbid dementia were plotted to visually determine temporal trends. These were also assessed statistically using a two-tailed Cochran-Armitage trend test.
We used Poisson regression modelling to evaluate the independent association between enforcement of the MCA and PoD. Three model sets were constructed separately for heart failure patients who died with or without comorbid dementia: home (1) versus hospital (0); care home (1) versus hospital (0); hospice (1) versus hospital (0). All covariates were forced to stay in the models to control their effects. The prevalence ratio (PR) was derived from the respective model to quantify the magnitude of association.
All analyses were performed using SAS 9.3 (SAS Institute, Cary, NC, USA). To control for Type 1 error, we applied Bonferroni correction to the alpha level. A two-sided p value of 0.008 (0.05 / 6) was considered statistically signi cant.

Results
Study sample.
Between 2001 and 2018, 120,068 people aged ≥ 65 years whose deaths were registered as directly due to heart failure were identi ed, their data subsequently included in this analysis. Table 1 describes the characteristics of these heart failure decedents. Overall, 8199 (6.8% [con dence intervals (CI) 6.7 to 7.0]) of these registrations were documented with dementia (WD) as a contributory cause, this being classi ed as unspeci ed dementia in 78.2%, Alzheimer's disease in 13.5%, and as vascular dementia in 8.3%. No other dementia subtypes were denoted. For the periods 2001-2007 and 2008-2018, pre-and post-enforcement of the MCA, the number of heart failure decedents WD was 3427 (5.6%) and 4772 (8.0%) respectively. The proportion WD gradually increased on an annual basis, this trend being statistically signi cant (Z = 18.87, p < 0.0001) as shown on Fig. 1. There was no discernible artefactual change in the general rate of dementia mentioned as a contributory cause of death associated with the 2011 change in ONS coding practice. However, additional analysis, also incorporated in Fig. 1, shows a contemporaneous and statistically signi cant reduction (p < 0.0001) in the coding of 'unspeci ed dementia' with a comparable increase in coding for 'vascular dementia'. Most of those dying from heart failure were female, this proportionately greater at 67.1% for the WD group compared to 60.3% for those with no dementia (ND). Twenty-four per-cent of the WD cohort were married at the time of death as were 27% of the ND group, widowhood being more common in the former group at 65% against 60% for those without mention of dementia. Most lived and died in urban environments (81%) and deprivation quintiles were similar for both groups.
Also demonstrated on Table 1, it is noteworthy that the rate of all heart failure-related death registrations of people ≥ 65 years, whether or not they had comorbid dementia, declined by 52% across the observation period of this study. The mean annual reduction was 369 [SD 32], this being statistically signi cant (p < 0.0001).

Trends In Place Of Death
Over the period of implementation of the MCA, comparative outcomes in PoD for these heart failure decedents are shown in Fig. 2

Discussion
In this rst national study examining variation in the PoD for heart failure patients over the period of enactment and implementation of the MCA in England, our results suggest that the decision-making and advance care planning processes enshrined in this legislation had little material effect on the ultimate site of care provision determining PoD for those with cognitive impairment manifest as dementia. In the later years of this study period when the code of practice for this legislation was operational, trend data for heart failure decedents certi ed with comorbid dementia show a modest rise in hospital deaths with fewer care home deaths. Conversely, for this study phase, there was a small reduction in hospital deaths for those without dementia who were younger and with fewer comorbidities.
The reasons behind the apparent lack of impact of this legislation may be complex, but it has been proposed that the MCA is relatively poorly applied in clinical practice. While a variety of training models have been developed to disseminate information to health professionals on the ve principles underpinning these regulations, recent reviews suggest poor understanding of when and how the provisions of the Act should be employed, with the need for clarity on the process of designating the role of surrogate decision-makers to properly ensure patients' best interests are maintained [24,25,26]. A lack of con dence of those working in acute care settings has been particularly highlighted, speci cally citing decision-making with respect to hospital discharge [27]. This issue may be especially relevant to our study observations given that we have demonstrated that the majority of those dying with heart failure in England do so in hospital, similar to data from the United States (US) [28].
At variance with the trends evident in this study, previous work has suggested a decline in the frequency of hospital deaths for those with dementia in recent years, more people dying in care homes [29]. These ndings were also based on ONS death certi cation data but included all individuals for whom dementia was mentioned as either the underlying or as a contributory cause of death. In contrast, the current study speci es heart failure as the underlying cause of death, differentiating the comparator groups by the presence or absence of dementia mentioned only as a contributory cause. While we have no information on any care transitions prior to the terminal phase for this study cohort, it is well established that the principal diagnosis is the main determinant of the site of clinical care [30], and community based primary care practitioners also appear to be relatively incognizant of patients' preferences for place of care or PoD, particularly when dealing with non-cancer diagnoses [31]. Unless policies for comfort care are clearly outlined, should people with chronic heart failure living at home or in a nursing home suddenly deteriorate, professional staff may default to arranging emergency hospital admission.
The completeness of death certi cation in the UK is regarded as relatively robust with proportionately fewer 'garbage codes' than data from many other countries [32]. However, dementia as recorded on death certi cation likely underestimates the true prevalence, and it has been suggested that studies using death certi cation alone may fail to account for 16-18% of dementia cases [33]. A variety of factors may in uence such documentation. Rates of inclusion of dementia are generally increased in those who die in institutions such as care homes compared to those dying at home, particularly if dementia is at the severe end of the clinical spectrum and includes agitation [34]. While heart failure guidelines draw attention to cognitive impairment as a comorbidity, describing all grades of this by hospital-based clinicians is reportedly poor [14], and there are diverging views on whether dying in hospital positively or negatively affects the rate of recording of dementia at the time of death certi cation [34,35]. Recent initiatives to heighten clinicians' awareness of dementia may improve matters. In 2012, NHS England introduced a quality improvement scheme through the Commissioning for Quality and Innovation (CQUIN) payment framework [36]. Acute healthcare providers were incentivised with the assurance of increased remuneration if 90% of all patients aged ≥ 75 years and whose emergency hospital admission lasted > 72 hours were screened for dementia. Further nancial gain was available if those patients whose initial assessment indicated dementia or was inconclusive were referred on for specialist review. While this dementia assessment and referral exercise was retired as a CQUIN indicator in April 2016, these conditions have been retained within the standard contract for English hospitals providing acute clinical services. It is possible that these administrative processes may have contributed in some measure to the increased mentions of dementia as certi ed for hospital decedents evident in the latter course of this study.
The relatively frequent concurrence of cognitive impairment and heart failure likely stems from various pathophysiologic features related to the latter condition combined with shared cardiovascular risk factors such as hypertension, hyperlipidaemia or dysglycaemia [7,37,38]. The haemodynamic and risk factor pro les for HFrEF and HFpEF clearly differ, but very few investigations have compared the spectrum of cognitive impairment across the range of ejection fraction phenotypes. There is a suggestion that affected domains of cognitive function may vary, but data is limited with inconsistent results [39,40].
To date, there is no evidence that evidence-based guideline-directed medical therapy (GDMT) for heart failure drives neurocognitive dysfunction [41], and indeed it has been posited that centrally acting angiotensin-converting enzyme inhibitors (ACEIs) such as perindopril or captopril, which cross the bloodbrain barrier, may slow progression of cognitive impairment in those with dementia [42]. Following the positive results of the PARADIGM-HF study demonstrating the bene ts of sacubitril / valsartan, the rst of a new class of drugs termed ARNIs (angiotensin receptor-neprilysin inhibitors) [43], this therapeutic option for HFrEF has been widely adopted. Neprilysin is a soluble metalloprotease which catalyses the degradation of natriuretic peptides (NPs), downregulation of this enzymatic activity likely increasing endogenous NP mediated natriuresis and vasodilation. However, such neprilysin inhibition might also interfere with clearance of amyloid-β protein, the vascular deposition of which results in cerebral amyloid angiopathy, a distinctive feature of Alzheimer's disease. Dementia-related adverse events were not overrepresented through 4.3 years follow-up of the relevant PARADIGM-HF study cohort compared to similar populations [44]. Nonetheless, as required by the Food and Drug Administration in the US, this potential hazard is currently being evaluated in the PERSPECTIVE study (ClinicalTrials.gov ID NCT02884206). Due to report in 2022, this trial includes a battery of neurocognitive testing and sequential 18 F-labelled orbetaben positron emission tomography to assess any longitudinal changes in cerebral amyloid plaque burden. Importantly, recent evidence shows that the hearts of some patients with Alzheimer's disease exhibit diastolic dysfunction and thickening of the interventricular septum. These features are characteristic of cardiac amyloidosis suggesting that in some individuals, amyloid-β protein may also accumulate in tissues other than the brain [45].
As inferred above, in recent years GDMT for those affected by heart failure has become increasingly sophisticated [46]. Beyond pharmacological therapy, this may include the implantation of electronic devices and durable mechanical circulatory support systems. While this complexity may be burdensome to some extent, there is also a bene cial trade off in that heart failure treatment has become more effective, which explains in part the temporal trend in declining heart failure death registrations noted over the observational period of this investigation. This is consistent with the reported 60% reduction in heart failure mortality based on analysis of death certi cation data between 1981 and 2010 in a study initially con ned to the Oxford region but expanded from 1993 to include the entire English national population [47]. Over this 30-year study of the cause of death of all men and women aged ≥ 35 years, there were parallel reductions in mortality from acute myocardial infarction and other forms of ischaemic heart disease, such gains postulated as responses to public health measures improving modi able cardiovascular risk factors, together with advances in clinical care. While this decline in cardiovascular mortality is to be celebrated, heart failure is an ambulatory care sensitive condition and remains the commonest cause of acute hospitalisation in those > 65 years [48]. Following an index heart failure admission in England, the 1-year mortality for patients discharged alive is 39.6% with a 30-day all-cause readmission rate of 19.8% [49]. Readmissions for heart failure tend to follow a tri-phasic pattern. This was apparent in a study of 8543 heart failure patients in Toronto monitored for 10 years following their index hospital admission, by which time 98.8% had died, the median survival after heart failure diagnosis being 1.75 years [50]. About 30% of all readmissions occurred within 2-months of initial hospital discharge, 50% during the 2month period leading up to death, with 15-20% taking place in the intervening 'plateau phase' of the heart failure disease trajectories. A sentinel clustering of admissions in the terminal phase of heart failure has been well described [51]. It is uncertain if the presence of dementia as a comorbidity in uences the readmission rate. Rao and colleagues followed 10,317 patients for 5 years subsequent to their rst diagnosis with heart failure between April 2008 and March 2009 using the primary care based Clinical Practice Research Datalink combined with Hospital Episode Statistics and ONS death registration data [52]. Their analysis indicated that comorbid dementia was a factor signi cantly affecting emergency hospital readmissions in only 3 of 8 regions across England.
Comparable to the epidemiological trends for heart failure, the age-adjusted prevalence and incidence of dementia may also be declining in high income countries. The Medical Research Council Cognitive Function and Ageing Studies (CFAS 1 and II) of populations living in rural Cambridgeshire and the urban environments of Newcastle and Nottingham demonstrated a 24% reduction in the prevalence of dementia in those ≥ 65 years between 1989 and 2011 [53]. Consistent with our observations, the CFAS studies also suggested that women were more commonly affected, and while the prevalence of dementia in care home residents had increased from 56-70%, most people with dementia were still living at home. Similarly, dementia events have been continuously monitored in the US based Framingham Heart Study since 1975. Recent analyses of this community cohort living in Massachusetts, predominantly of white European ancestry, have implied a 20% stepwise decline in the incidence of dementia each decade over the last 30 years [54]. The background to these cumulative decrements remains to be determined, but both the CFAS and Framingham study groups cited potential mechanisms in higher early educational attainment and attenuated vascular morbidity.
The Framingham Heart Study showed a non-signi cant reduction in Alzheimer's disease with a more overt decrease in vascular dementia. In Westernised societies, Alzheimer's is the most commonly encountered manifestation of dementia, but as an isolated pathophysiological process, this affects < 20% of those with heart failure. Rather, vascular dementia has been proposed as the likeliest associated variant, followed by mixed forms, then Alzheimer's and other speci c dementias [39]. This is at odds with the distribution of dementia subtypes noted in this study where unspeci ed dementia was most frequently mentioned on death certi cates and coded as the dominant category. A similar nding was described in a Danish study of 324,418 patients admitted with incident heart failure and tracked for 35 years against an age-and sex-matched population without heart failure selected from the Danish Civil Registration System [55]. Adelborg and colleagues found a clear association between all-cause dementia and heart failure. This was relatively weak for Alzheimer's disease, and while the vascular variant was represented, this was predominantly determined by the reported development of unspeci ed dementia. These authors proposed that some patients ostensibly exhibiting unspeci ed dementia may have been misclassi ed. However, in combining death certi cation data from sequential CFAS studies in England, all mentions of dementia as the underlying or a contributory cause of death showed a percentage distribution of subtypes of unspeci ed dementia, Alzheimer's disease and vascular dementia as 69.3%, 21.6% and 8.6% respectively [35]. These results are very similar to those noted in the current study and suggest that this proportional distribution of dementia variants is not speci c to the heart failure population.
Advance care planning offers patients the potential to receive medical treatment consistent with their expressed preferences, values and goals of care against the possibility of subsequent loss of decision-making capacity. This may help prevent needless hospital admissions and better achieve consensus on appropriate ceilings of care, avoiding exposure of patients and families to the distressing harms which sometimes accompany futile treatment escalation and burdensome invasive interventions close to the end of life. It is notable that a recent audit of end-of-life care in hospitals in England demonstrated that10% of heart failure patients were receiving mechanical ventilatory support within 24-hours of death [56].
Given the unpredictability intrinsic to the heart failure disease trajectory which challenges individual prognostication even in the late stages of this disease, and the associated multimorbidity including cognitive impairment, it might be expected that advance care planning would be central to the care of people with this condition. However, advance care planning is not routinely incorporated within heart failure care and tends to be limited to the possible withdrawal of some of the implanted devices described above, or sometimes offered as one component of the still uncommon provision of palliative care [57,58]. A review and meta-analysis of 14 randomised controlled trials of advance care planning in heart failure, mostly US based and involving 2924 individuals across a range of care structures, showed this to moderately improve the primary outcome measure in patients' quality of life, together with similarly weighted favourable effects on secondary outcomes including communication about and satisfaction with end-of-life care [59].
Advance care planning for dementia was featured as a speci c domain in the European Association of Palliative Care white paper on this condition [60], and the challenges taking this forward have been comprehensively reviewed [61,62]. Emerging themes suggest that while there is some disparity in the readiness of older people to engage in advance care planning, and the means to take this forward may also vary across the spectrum of healthcare delivery, the use of these instruments may be effective in promoting shared decision-making between patients, informal carers and professionals [63]. However, it should be noted that, in the UK, an advance care plan is not legally binding but merely an advisory statement of preferences and wishes in relation to general care and medical therapy [64].
In contrast, the MCA and the associated code of practice offer legislative protection to promote patient empowerment and safeguard their autonomy. Prepared when mental capacity is intact, patients may formulate an advance decision such as one to refuse life-sustaining treatment or, if ≥ 18 years, appoint a close person as a personal welfare lasting power of attorney (LPA) to undertake decisions on their behalf if agency is later lost. Thereafter, any clinical treatment protocol, where possible, should be in accordance with their previously documented choices and values, or these as expressed through their nominated personal welfare LPA. Importantly, we must emphasise that many people diagnosed with dementia can still make decisions about many aspects of their care, and loss of capacity should not be regarded as an all-or-none phenomenon based on that diagnostic label. Indeed, under the terms of the MCA, retention of capacity is assumed, and capacity is both decision and time speci c. For the purposes of the Act, a two-stage capacity test is applicable. To qualify through Stage 1, the individual must exhibit a demonstrable functional impairment of the mind or brain. For Stage 2, capacity is deemed to be lost if they lack the ability to ful l any of the following: (a) understand the information pertinent to the decision, (b) retain the information, (c) deliberate on that information as part of the decision-making process, and (d) communicate their decision by any means possible.
Dialogue between the patient, family and clinician is basic to shared decision-making. However, triangulation of information between this triad does not necessarily mean equitably weighted opinions, and at times patients' voices are marginalised, a dyadic interchange conducted between clinicians and family members. Such a discourse may be justi ed if the patient cannot physically contribute to meaningful shared decision-making, if capacity is deemed to be lost, legally binding instruments are not in place, and their preferences are unknown. Then, the opinion of the closest relative should be sought, their views respected, and used to inform a care plan constructed in the best interests of the patient. However, it should also be remembered that the involvement of relatives as decisional proxies may be emotionally demanding, particularly if there is tension between family members, or they are already distressed and experiencing anticipatory grief. Further, the assumption that there is always congruence between the opinions of family surrogate decision makers and those perceived of patients is awed. Rather, these are often misaligned, with at best moderate concordance between dementia / carer dyads when assessed within a hospital setting [65]. Even if decisional consensus is achieved, the discharge of hospital inpatients home may be subject to practical limitations depending on the social context. The caregiver burdens associated with heart failure and dementia have been well characterised. Informal caregivers may be unwilling or unable to reframe or enhance their roles, and it should be noted that the cohabitees of people living and dying with dementia tend to be of a similar advanced age [66]. As evident in this study, there may also be a gendering issue in that older women with heart failure and comorbid dementia are more likely to have outlived their male partners and be devoid of spousal support [67]. These intersectional aspects, constraining the social capital of those with heart failure and dementia close to the end of life, offer little prospect of mitigation [68,69].

Strengths And Weaknesses
To our knowledge, this is the rst national study to systematically examine the PoD of people dying of heart failure with dementia documented as a contributory cause of death. A major strength is that our work is based on comprehensive data collated from the gamut of clinical practice over an 18year period, but we acknowledge that we are dependent on the clinicians responsible for these heart failure decedents having made the correct diagnoses and accurately completing later death certi cation, over which there is no means of independent adjudication. Irrespective of the PoD, we have no access to information on the underlying aetiologies of heart failure, or the proportional distribution of resultant ejection fraction variants. Similarly, it is not possible to ascertain the duration of heart failure or nature of any care provision prior to the terminal phase, and whether this fatal outcome related to incident acute heart failure, worsening of chronic heart failure, or heart failure-related sudden cardiac death. It has been suggested that decisions to include dementia on death certi cates rely on medical staff regarding this as clinically signi cant [70]. This is considered more likely if dementia is relatively severe, rea rming our judgement in using the certi ed mention of dementia as a contributory cause of death to represent a valid marker of signi cant cognitive dysfunction, and therefore relevant to the aegis of the MCA.

Conclusions
In this paper, we have examined the impact of the MCA on the PoD of heart failure decedents aged ≥ 65 years whom we have shown to be increasingly affected by comorbid dementia, dying at home or usual place of residence customarily accepted as the preferred option. Our analyses of trends over the period of enactment and implementation of the code of practice relating to this legislation show little to suggest any signi cant in uence on PoD for this relatively vulnerable clinical cohort. The background to this somewhat neutral outcome is multifaceted, administration of the Act clearly challenging for a prognostically ambiguous population subject to ux in the often-nuanced scenarios typical of real-life clinical practice, a milieu dominated by the treatment imperative. Further, even if the precepts of the MCA are correctly applied, this course of action may be ineffectual in isolation. Recent evidence suggests that achieving a good death at home requires patients and their informal carers to feel secure in that setting, effected by the provision of a 24/7 responsive palliative care service, staffed by those competent in symptom relief and with good communication skills [74]. Ful lling the preferred PoD of those with heart failure and dementia might be better achieved by embedding application of the MCA within a system of anticipatory sympathetic clinical navigation across all care sectors, contingent upon effective upskilling of the relevant professionals, with good inter-agency and multidisciplinary collaboration to support and maintain appropriately con gured community-based palliative care.

Declarations Ethics and authorisation
In accordance with General Data Protection Regulations arising from the UK Data Protection Act (2018), as all patient level data was fully anonymised, no ethical approval was required. Wei Gao is accredited to analyse this data through the ONS Approved Researcher Scheme.

Availability of data and materials
Supplementary les S-1, S-2 and S-3 are available online.

Competing interests
The authors declare they have no competing interests. and not necessarily those of the NIHR or the Department of Health and Social Care. The funding bodies have no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.
Authors' contributions JMB: conception and design; organisation of the conduct of the study; interpretation of results; primary drafting and revision of the manuscript.
IJH and TAMcD: interpretation of the results; critical review and revision of the manuscript.
WG: conception and design; organisation of the conduct of the study; data acquisition and management; statistical analysis of the data; interpretation of results; drafting and revision of the manuscript. WG is guarantor of the data.
All authors approved the nal version of the paper prior to submission.