Article Text

Download PDFPDF

Where do patients known to a community palliative care service die?
  1. Margred Capel1,
  2. Terri Gazi1,
  3. Lauren Vout2,
  4. Nicola Wilson1 and
  5. Ilora Finlay3
  1. 1George Thomas Hospice Care, Cardiff, UK
  2. 2Cardiff University School of Medicine, Cardiff, UK
  3. 3Department of Palliative Medicine, Velindre Hospital, Cardiff, UK
  1. Correspondence to Dr Margred Capel, George Thomas Hospice Care, Ty George Thomas, Whitchurch Hospital Grounds, Park Road, Whitchurch, Cardiff CF14 7BQ, UK; margred.capel{at}


Objectives The majority of people would prefer to die at home and National Health Service policy aims to support this concept. This service evaluation compared the preferred and actual place of death of patients known to a specialist community palliative care service.

Design All deaths of patients (n=788) known to the specialist palliative care service from 1 January 2009 to 31 December 2010 were examined in a service evaluation to compare the actual place of death with the preferred place of death previously identified by the patient. Triggers for admission were established when patients did not achieve this preference.

Results 69% of patients (n=263) who expressed a preference to die at home and 82% of patients (n=93) who expressed a preference to die as inpatients in the hospice fulfilled these preferences. 71% of patients (n=298) who wanted to die in their current place of residence achieved this preference.

54% of patients (n=121) who declined to express a preference for end-of-life care subsequently died in hospital, reflecting the importance of advance care planning.

Conclusions The perceived lack of social support for patients dying at home is a significant trigger for admission to a hospice. The provision of sitters to support patients dying at home may ensure people achieve their preference. Commissioners consider preferred place of care to be a marker of quality, but clinical events that precipitate admission are often outside the influence of the palliative care team.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Although the majority of people in the UK are thought to want to die at home, over half die in hospital.1 2 Government initiatives including the End-of-Life Strategy in England and the Implementation of the Palliative Care Strategy in Wales strive to improve equitably the quality of care received by patients, raising expectations among patients and their families, and professionals.3 4 As a consequence there is an increased expectation of high-quality palliative care to patients in the community by the recipients of care and other healthcare professionals.5 Services presently operate in distinct areas with changing demographic trends and a challenging financial climate.6 Local commissioning arrangements specify quality markers in funding agreements. In Wales the restructuring of the health service has incorporated primary and secondary care under the same health board. Place of death of a patient with identified palliative care needs has become a subject of interest not just to the patient and their immediate family but also to the commissioners as a quality marker of service and its impact on secondary care admissions and costs. This quality marker is outlined in the service level agreement for this community palliative care service. A service evaluation was undertaken to examine whether patients achieved their preference for place of death. The governance around the data use was agreed through the hospice governance processes. Ethical approval was not required. Studies suggest preferences for place of death may differ from place of care. Patients are known to change their preference for place of death during their illness journey.7 8 The opportunity to discuss the preferred place of death was already part of communication with the patient. The dynamic care plan enabled the changes to be recorded and the comparison drawn against the last recorded preference. A ‘death proforma’ was designed to capture the information; all team members received training prior to its introduction.

George Thomas Hospice Care is a community-based hospice specialist multidisciplinary team without inpatient beds. It is located in the Welsh capital city of Cardiff, serving a population of 250 000. It is within the geographical area served by a regional oncology centre and commissioned to accept all appropriate referrals from a discrete geographical area. One hundred and thirteen referrals (11%) between 2009 and 2010 were for patients with non-malignant disease. Referrals were accepted from healthcare professionals or patients (with the permission of the patient's general practitioner (GP)) for patients with identified palliative care needs. Out of hours generalist support for these patients was provided by the local district nursing service and GPs. District nurses were able to arrange non-qualified nurse ‘sitters’ to support patients and families at home for a maximum of two nights each week. Patients, families and healthcare professionals had access to specialist palliative medicine telephone advice at night. During Saturday and Sunday, 9–5 pm, there was access to advice and face-to-face assessments from a clinical nurse specialist (CNS) and if required a palliative medicine specialist. Inpatient hospice admission (provided by a different charity) was available locally.

The deaths of all patients known to the service in the 2 years from 1 January 2009 to 31 December 2010 were examined to compare the place of death with the preferred place of death previously stated by the patient. Preferred place of death was identified though discussion with the patient and amended as necessary during the illness journey. Place of residence was defined as their current place of living at the time of assessment, which included the patient's home, nursing home, residential home or prison. If patients were unwilling or unable to explore the subject of end-of-life care and subsequently preferred place of death this was recorded as an unwillingness to express a preference.

Additional information including the diagnosis, availability of sitters to support the family and the use of the integrated care pathway (ICP)9 for the last days of life was considered in an attempt to identify any trends in home, hospice and hospital deaths in this patient group to steer future service activity and education.


A ‘death proforma’ was completed by the CNS responsible for each patient who died between 1 January 2009 and 31 December 2010 inclusive.

Information collected prospectively included:

  • demographic details;

  • the diagnosis;

  • the preferred place of death (considered to be the last recorded preference in the case notes) and highlighted on the proforma if different from the current place of residence;

  • the actual place of death;

  • the GP practice;

  • whether the death was anticipated;

  • aspects surrounding end-of-life care, including use of ICP and drugs for symptom control, delays in commencing syringe drivers, provision of ‘sitters’.

Additional retrospective information was collected and obtained if necessary from healthcare professionals in the hospice, secondary care and family members to establish the precipitant for admission when patients had been admitted for end-of-life care to the hospice or hospital if this was not their preferred place of death. Categories for admission were grouped into:

  • a new condition (including seizure, haemorrhage, obstruction, urinary retention, acute dyspnoea, falls or other sudden events);

  • an exacerbation of an existing symptom or uncontrolled symptoms;

  • patient choice;

  • carer choice;

  • patient or family perception of a lack of support at home;

  • professional choice (GP, district nurse or CNS).


Seven hundred and eighty-eight patients known to the service died in 2009–2010 inclusive. Patients were known to the service for a median of 83 days and a mean of 160 days. Ninety-three per cent of these deaths (n=730) occurred in patients who had a diagnosis of cancer. These cancer diagnoses are reported in table 1.

Table 1

Cancer diagnoses of patients included in the evaluation

Table 2 reports the diagnoses of non-malignant diseases. Seven per cent (n=58) of deaths were patients referred to the service with these non-malignant diseases.

Table 2

Non-malignant diagnoses of patients included in the service evaluation

Fifty-four per cent (n=422) of these patients specified that their preferred place of death was their current place of living (including home, residential or nursing home and prison). Fourteen per cent (n=114) indicated they would prefer to be admitted as an inpatient to the hospice for end-of-life care; 2% (n=16) stated they would prefer to be admitted to hospital. No preference was identified in the remainder of patients (n=224). Fifty-four per cent (n=121) of patients who did not express a preference for end-of-life care (including place of death) subsequently experienced end-of-life care in hospital.

During the course of their illness 12% (n=98) changed their preference for end-of-life care; 23 patientswho initially chose to die at home subsequently selected inpatient care.

Where patients died

Figure 1 demonstrates the actual place of death of all patients. Table 3 compares the preferred place of death with the actual place of death among patients who expressed a preference (expressed as option conversion rate). Sixty-nine per cent (n=263) of patients who expressed a preference to die at home and 82% (n=93) of patients who expressed a preference to die in the hospice had their preferences met.

Figure 1

Actual place of death of patients during service evaluation.

Table 3

Preferred place of death compared with actual place of death of patient

People admitted to hospital for end-of-life care (who wanted to die at home)

Nineteen per cent (n=73) of patients expressed a preference to die at home but died in hospital. Several precipitants were cited as the reason for admission to hospital, ranging from ‘new complaint’ to a ‘perceived lack of social support’, these are summarised in table 4.

Table 4

Triggers for admission to hospital

The presumptive diagnosis of a new complaint was the main trigger for admission in 44% (n=32) of these patients. Death was anticipated in 27% (n=20) of patients at admission. The majority of referrals (63%) were instigated by the GP with hospice staff consulted in only five cases. Unavailability of a hospice bed only accounted for one admission. Lack of social support triggered fewer admissions to hospital than hospice. The identification of a ‘new complaint’ triggered more admissions to hospital than the hospice, with ‘poor symptom control’ as the main trigger for hospice admissions.

The ‘new complaints’ which precipitated admissions to hospital prior to death were broadly divided into sudden events (including haemorrhage, seizures, obstruction), complications related to treatment received (including chemo/radiotherapy, surgery, blocked stents) and sudden collapse (including stroke). The largest group, 25% (n=8) related to falls and included patients who sustained fractures. Patients admitted with a ‘new complaint’ died during their hospital admission.

People who were admitted to the hospice for end-of-life care (who wanted to die at home)

Ten percent (n=39) of patients originally expressed the preference to die at home, but died as inpatients in the hospice. Death was anticipated in 64% (n=25) of these patients at the time of admission.

The main trigger for admission of these patients was poor symptom control 38% (n=15) or a new complaint 15% (n=6). Perceived insufficient social care and support (including lack of sitters) contributed to admission for 31% (n=12) of patients. Admission was requested by the carer in 28% (n=11) of cases.

Place of death of patients referred to the service who had non-malignant disease

Fifty-eight patients with non-malignant disease known to the palliative care service died in 2009–2010, of these 67% (n=39) achieved their preferred place of death, as demonstrated in table 5.

Table 5

Preferred place of death compared with actual place of death for patients with non-malignant disease

Management of end-of-life care at place of residence

The Welsh ICP for the last days of life was used to support the care of 28% (n=83) of the patients who died in their place of residence. In 26% (n=77) of patients who died the care pathway was not initiated because the professional declined, was not using it in their practice or lacked the paperwork. Despite this, appropriate anticipatory prescribing of end-of-life care drugs occurred for 81% (n=241) of patients. There were delays in commencing the syringe driver (if used) in only 4% (n=12) of patients.

Availability of ‘sitters’ influencing place of death

Thirty-four per cent (n=130) of patients who expressed the preference to die at home requested ‘sitters’.

Sitters were available for 104of these patients and all were able to achieve their preference for death at home. Of the remaining 26 patients, 38% (n=10) were admitted while 62% (n=16) died at home.

Eighty per cent (n=104) of all patients requesting to die at home with sitters achieved their aim whereas only 57% of patients (n=143) who had not requested sitters did so.


A proforma completed upon patients' death enables data to be gathered to support the clinical activity of the palliative care team and target areas for future activity and education.7

In 2009–2010 preferences were fulfilled for 69% (n=263) of patients who wanted to die at home and 82% (n=93) of patients who preferred to die in the hospice. These figures compare favourably with results reported elsewhere.8 10 11 12

The number of patients who die at home could potentially be increased if the events which triggered acute admissions are considered as starting points to change healthcare professional behaviour through education and to target social and community care resources.

Programmes such as hospice at home or the Hospital2Home project run at the Royal Marsden Hospital have reported an 80% home death rate.13 14 Extraction of the principle themes including actively identifying, reviewing, documenting and sharing preferences for end-of-life care to achieve seamless care among all health providers is an option for all services. Reasons for admission for end-of-life care fell into two broad groups: perceived lack of social support; and a clinical event determined by the attending healthcare professional to necessitate admission. The close relationship between the ‘new complaint’ and the patient's death suggests some events are part of the dying process and may represent a missed opportunity to diagnose patients who are dying in the community.2

In this study, patients reluctant to either discuss or express their preferences for end-of-life care were more likely to be admitted to hospital and die there, highlighting the importance of advance care planning as a focus of discussion with patients. Other studies found that patients with a documented advance care plan were more likely to achieve their preference for place of death in the community.15 Good communication is essential to ensure all professionals involved in patients' care are aware of such decisions and to avoid unwarranted admissions of these patients. Patients can be unwilling to discuss their preferences, however staff inhibitions are a significant reason why place of death is not addressed in other studies.16 Staff require ongoing support and training to continue to approach this sensitive area and to maximise the opportunity for patients to discuss their preferences. This service evaluation has enabled the team to consider how the information for preferred place of death is shared with other health providers. GP awareness of an existing preference is associated with achieving a preference for death at home.17 The present system does not allow this information to be routinely shared with the patient's own GP and the out-of-hours service. The consistent use of a palliative care register across the practices may address this limitation.18 Recording this information in the out-of-hours database when information is obtained by our service could alert GPs to any preferences at the point of triage of these patients if they contact the out-of-hours service. The provision of sitters influences place of death. People who identified the need for sitters and received this support achieved their preference for a home death. Patients in our geographical area can receive support from night sitters (usually care assistants) up to two nights per week; this service is arranged through the district nursing service. For some patients and their families/carers this is insufficient support for them to be able to remain at home to die; such perceived lack of support can have a negative outcome on the bereavement of carers.19 As in other studies we found that patients living alone and without the support of a family or friend are unlikely to die in their own home, even if this was their preference.14 This raises the question of whether it is ethical to offer this option to patients living without support if, as healthcare professionals, we are unable to arrange it. Demographic factors such as age, gender and socioeconomic class are all recognised factors that can influence place of death.16 20 Increased resources can also have an impact on home death rates. Some studies21 have quantified the financial implications and savings that can be achieved if patients die at home as opposed to the acute sector/secondary care, but a major shift of resources may be needed to increase home death rates in the future.

Some of the unforeseen events precipitating admission to the acute sector, such as a blocked catheter, could more appropriately be managed by community services. Other events, such as a pneumothorax or neutropenic sepsis, probably require acute sector admission. Given the range of factors precipitating admission, it is unlikely that 100% of preferences for place of death will ever be realised.

Suggestions that a significant proportion of expected deaths in hospital could have occurred in the community2 21 raise the possibility that the precipitant to admission was part of the dying process and could have been managed at home with appropriate input. Unfortunately our community palliative care team was rarely consulted to try to avert admissions. GPs can be less familiar with complex end-of-life care, including the varying clinical scenarios and the difficulties in diagnosing patients who are dying.2 18 In Wales in 2009 a short palliative care course specifically designed for GPs was introduced to cover the theoretical principles of medicine, communication skills and scenarios experienced in the community. As with any non-mandatory training, only GPs interested in palliative medicine applied, leaving a significant majority of professionals not accessing the course despite such patients forming part of their clinical workload.

Transitions between care settings in the last phase of life potentially jeopardise end-of-life care overall: distress related to inappropriate or untimely ambulance transfers is a factor.21,,23 Encouraging generalist colleagues to use the surprise question2 in the form of ‘would you be surprised if this patient died on this admission?’ may also create an opportunity to then discuss management options in the community.

Relatively few patients with non-malignant disease were referred to our palliative care team: the majority died in their preferred place even though recognition of the terminal phase in the illness trajectory of this diverse patient group is challenging.24

For our patients, preference of a hospice death was the second most popular option, confirming it is acceptable to the general public and can easily be discussed when considering place of death in advance care planning.14 25 Use of formal documentation for advance care planning may ensure a patient's wish to die in a specific location is more likely to be respected.18

Preference for place of death often varied during a patient's illness, which has been described in other studies.22 26 The reasons behind these variations were explored with the patient at the time but not specifically covered in this evaluation. Modification of the proforma would be required to collect and collate into meaningful groups reasons for the changes in preference, which may provide additional useful information for the commissioners to support patients in the community. The influence of families' preferences for place of death may be more appropriately explored in a formal research study, which is outside of the remit of this service evaluation.

The local prison was the preferred place of death for one patient, which was achieved. Palliative care can be provided successfully in this environment.27 Following this experience the service has worked with the prison authorities to adapt the palliative care register and care pathway for use in the prison. The adaption includes preferred place of death and considers any legal implications.

In the community where multiple healthcare professionals may be involved in the care of a patient, use of the ICP for the last days of life with supporting education can aid communication in all settings, including nursing homes.28

The comparatively high proportion of patients who preferred to die in their nursing homes and were able to realise this may be the result of this project in Wales. Disappointingly one patient who wanted to die in his nursing home and whose care was supported by the care pathway was admitted to the acute sector for fluids and antibiotics by the out-of-hours GP; he died less than 2 h later. This prompted a significant events analysis.

Healthcare professionals in the community remain reluctant to use the ICP despite the community palliative care service providing support to GPs (who are the prescribers and budget holders in the community).18 We found that some GPs are reluctant to prescribe in anticipation of a patient's deterioration despite specialist palliative care involvement. However, anticipatory prescribing is a key component of symptom control for patients dying at home. Absence of the principle drugs to control the symptoms of pain, nausea, anxiety and respiratory secretions can precipitate poor symptom control, patient and family/carer distress, contact with out-of-hours services, admission to the acute sector, and family members or healthcare professionals driving around pharmacies to collect medication for administration.

Support from commissioners and regulatory authorities is essential in encouraging primary care health professionals to use the ICP and to financially support anticipatory prescribing.

Honouring the preferred place of death for a patient is a recognised important aspect of end-of-life care1 3 15 but few of the factors which influence this are within the control of the community palliative care team.7 Place of death is perhaps a more accurate quality marker of holistic community services.


The authors would like to thank A Shaw of Medical Illustrations, University Hospital of Wales, Cardiff for preparing the figure.



  • Funding The hospice receives funding from commissioners, which include Cardiff and Vale University Health Board and the Welsh Assembly Government. They had no role in the service evaluation design, collection, analysis or interpretation of the data or writing of the report.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.