Articles

Improving comfort around dying in elderly people: a cluster

randomised controlled trial
Kim Beernaert, Tinne Smets, Joachim Cohen, Rebecca Verhofstede, Massimo Costantini, Kim Eecloo, Nele Van Den Noortgate, Luc Deliens

Summary
Background Over 50% of elderly people die in acute hospital settings, where the quality of end-of-life care is often Published Online
suboptimum. We aimed to assess the effectiveness of the Care Programme for the Last Days of Life (CAREFuL) at May 16, 2017
http://dx.doi.org/10.1016/
improving comfort and quality of care in the dying phase in elderly people. S0140-6736(17)31265-5
See Online/Comment
Methods We did a cluster randomised controlled trial in acute geriatric wards in ten hospitals in Flemish Region, http://dx.doi.org/10.1016/
Belgium, between Oct 1, 2012, and March 31, 2015. Hospitals were randomly assigned to implementation of S0140-6736(17)31325-9
CAREFuL (CAREFuL group) or to standard care (control group) using a random number generator. Patients and End-of-Life Care Research
families were masked to interventaion allocation; hospital staff were unmasked. CAREFuL comprised a care guide Group, Vrije Universiteit
Brussel, Brussels, Belgium
for the last days of life, training, supportive documentation, and an implementation guide. Primary outcomes
(K Beernaert PhD, T Smets PhD,
were comfort around dying, measured with the End-of-Life in Dementia–Comfort Assessment in Dying Prof J Cohen PhD,
(CAD-EOLD), and symptom management, measured with the End-of-Life in Dementia–Symptom Management R Verhofstede PhD, K Eecloo MSc,
(SM-EOLD), by nurses and family carers. Analyses were by intention to treat. This trial is registered with Prof N Van Den Noortgate PhD,
Prof L Deliens PhD); End-of-Life
ClinicalTrials.gov, number NCT01890239.
Care Research Group
(K Beernaert, T Smets,
Findings 451 (11%) of 4241 beds in ten hospitals were included in the analyses. Five hospitals were randomly assigned Prof J Cohen, R Verhofstede,
to standard health care practice and five to the CAREFuL programme; 118 patients in the control group and 164 in the K Eecloo,
Prof N Van Den Noortgate,
CAREFuL group were eligible for assessment. Assessments were done for 132 (80%) of 164 patients in the CAREFuL Prof L Deliens) and Department
group and 109 (92%) of 118 in the control group by nurses, and 48 (29%) in the CAREFuL group and 23 (19%) in the of Medical Oncology
control group by family carers. Implementation of CAREFuL compared with control significantly improved nurse- (Prof L Deliens), Ghent
assessed comfort (CAD-EOLD baseline-adjusted mean difference 4·30, 95% CI 2·07–6·53; p<0·0001). No significant University, Ghent, Belgium;
Palliative Care Unit, Arcispedale
differences were noted for the CAD-EOLD assessed by family carers (baseline-adjusted mean difference –0·62, Santa Maria Nuova-IRCCS,
95% CI –6·07 to 4·82; p=0·82) or the SM-EOLD assessed by nurses (–0·41, –1·86 to 1·05; p=0·58) or by family carers Reggio Emilia, Italy
(–0·59, –3·75 to 2·57; p=0·71). (Prof M Costantini MD); and
Department of Geriatric
Medicine, Ghent University
Interpretation Although a continuous monitoring of the programme is warranted, these results suggest that Hospital, Ghent, Belgium
implementation of CAREFuL might improve care during the last days of life for patients in acute geriatric hospital wards. (Prof N Van Den Noortgate)
Correspondence to
Funding The Flemish Government Agency for Innovation by Science and Technology and the Belgian Cancer Society Dr Kim Beernaert, End-of-Life
“Kom Op Tegen Kanker”. Care Research Group, Vrije
Universiteit Brussel,
Brussels 1090, Belgium
Introduction Italy.14 In this trial, there was no significant difference in kim.beernaert@vub.ac.be
Each year, 58 million people die around the world, 53% overall quality of care between the CAREFuL and control
of whom are aged 60 years and older.1 In a population- groups, but the trial was statistically underpowered.
based study in 14 countries, between 25% (Netherlands) Findings from several non-randomised studies7,15 suggest
and 85% (South Korea) of people potentially benefiting positive effects (eg, improved symptom control) of
from palliative care died in hospital.2 In another report, end-of-life care programmes.
over half of all deaths in Britain occurred in the acute To improve the quality of end-of-life care in acute
hospital setting.3 Findings from previous studies have geriatric hospital wards,5 we developed the Care
suggested that end-of-life care for elderly people in acute Programme for the Last Days of Life (CAREFuL).16–18 This
hospital wards is suboptimum.3,4 Delivery of optimum programme is based on three existing LCP programmes
end-of-life care is challenging because dying people often (UK, the Netherlands, and Italy). Development took into
have complex and multifaceted needs.5,6 account critiques of the original LCP in the UK.19,20 We
Various initiatives have been introduced worldwide to avoided the suggestion of a care pathway because this
improve end-of-life care in acute hospital settings.5,7–9 The was deemed to lead to perceptions of the guide as a
Liverpool Care Pathway for the Dying Patient (LCP) is protocol and tick box exercise rather than an approach to
one of the most well-known programmes.10 However, a care; goals within the guide were changed into points of
Cochrane review,11,12 updated in 2016,13 identified only one attention for similar reasons. We developed an
published cluster randomised trial of the effectiveness of implementation strategy, which included training and a
an end-of-life care programme in the acute hospital guide on implementation of the programme, and
setting, namely the LCP, for patients with cancer in developed a quantitative process assessment instrument

www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5 1

 Articles
Research in context
Evidence before this study
Various initiatives have been introduced worldwide to
improve end-of-life care. However, there have been sustained
concerns about the safety of implementing such pathways or
programmes, particularly about one of the most well known—
The Liverpool Care Pathway for the Dying Patient (LCP) in the
UK. Therefore, clinicians and policy makers need end-of-life
care programmes that are evidence based. In the most recent
Cochrane review (updated search in July, 2015, of original
search done in 2009), which used the search terms
(“randomized controlled trial” or “controlled clinical trial” or
“intervention studies/experiments” or “random”) and
(”palliative care” or ”palliative care” or “end-of-life” or
”terminally ill” or “dying” or “hospice” or “end-stage”) and
(”critical pathways” or ”clinical pathway$ or care pathways” or
”integrated care pathways”), only one cluster randomised trial
assessing the effectiveness of the LCP was found, for patients
with cancer in Italy. This trial showed no significant differences
in overall quality of care between the intervention and control
groups, but was statistically underpowered. The authors of the
Cochrane review judged the study to be at high risk of bias
overall, mainly because patients were not masked to
treatment allocation and there were high rates of attrition.
Thus, little evidence exists regarding the clinical, physical,
psychological, or emotional effectiveness of end-of-life care
pathways or programmes. No search was done for randomised
to assess and monitor the quality of implementation.
Additionally, the programme was modelled to the acute
geriatric care setting.16 We aimed to assess whether
comfort and quality of care in the dying phase in elderly
people (called patients hereafter) can be improved by
implementation of CAREFuL.
Methods
Study design and participants
We did a multicentre cluster randomised controlled trial,
with hospitals as the units of randomisation and patients
the units of analysis. By designating hospitals as the
units of randomisation, we ensured that all patients
within a given ward were assigned to the same trial
group, thereby reducing the risk of cross-contamination.
The study protocol has been published previously.17
We followed the CONSORT guidelines21 to design and
report this study.
To be included in the study, hospitals had to have at
least one acute geriatric ward, and consent for inclusion
of each ward was needed from the medical and nursing
head of the ward. The acute geriatric wards in the
Flemish Region, Belgium, provide diagnostic assessment
and treatment of elderly patients (mean age 83 years;
mean hospital stay 17·5 days).22
To be included in the study, all patients (or their
representative if the patient was unable to give informed
2 www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5
controlled trials of end-of-life care programmes in intensive
care units, because the intensive care unit is a different setting
with different approaches to care compared with the acute
geriatric hospital wards in which the intervention in our study
was used.
Added value of this study
This study is, to our knowledge, the first sufficiently powered
randomised controlled trial to test the effectiveness of an
end-of-life care programme, the Care Programme for the Last
Days of Life (CAREFuL) programme. We found a statistically and
clinically significant improvement in comfort around dying
assessed by nurses after implementation of the programme.
We also found a significant positive effect of the programme on
symptoms and care needs, but a negative effect on satisfaction
with care.
Implications of all the available evidence
Our findings suggest that CAREFuL can improve the quality of
dying and quality of care in acute geriatric hospital wards;
however, further qualitative research is needed to gain a better
understanding of the effect of CAREFuL on satisfaction with
care. Further controlled implementation of the programme in
acute geriatric hospital wards is likely to improve care at the
end of life. Use and assessment of the programme in other
settings where people are cared for at the end of life might also
merit consideration.
consent) admitted to the ward were asked to give
consent to use of their personal data—collected from
medical or nursing records—in the study. There was no
informed consent procedure to participate in CAREFul
because it was an intervention at ward level, which
inevitably makes all carers and patients subject of the
intervention. Patients who gave consent and died on the
ward between Oct 1, 2012, and March 31, 2015, and had
been in hospital for more than 48 h were eligible for
inclusion in the analyses.
Ethics approval was obtained from the Central Ethics
Committee of the Vrije Universiteit Brussel (Brussels,
Belgium) and from the local ethics committees of the
participating hospitals in the Flemish Region (B.U.N.
1432012139850). The ethical issues are discussed in detail
in the protocol.17
Randomisation and masking
Before randomisation, baseline data were collected for
1 year in all participating wards, at the end of which
the included hospitals were randomly assigned to
implementation of CAREFuL (CAREFuL group) or to
standard care (control group). Because the number of
clusters to be randomised is substantially smaller in a
cluster randomised controlled trial than in trials in which
the unit of randomisation is the patient, there was a
chance of baseline imbalance between the randomised
 Articles

groups. Therefore, hospitals were stratified in terms of

the number of beds in the participating wards and the CAREFuL aim:
To improve the quality of dying and end-of-life care in patients dying on acute geriatric hospital wards
proportion of patients for whom the nurses could obtain
consent in the baseline period. A statistician outside the
research group allocated the hospitals to the CAREFuL or Targets of CAREFuL were the geriatric health-care staff; we aimed to:
control group using a random number generator. • Raise awareness of the importance of improving end-of-life care
• Prepare staff for a change in end-of-life care
Because of the nature of the intervention, hospital staff • Train staff in delivery of good end-of-life care with the support of a multi-professional care guide for the last days
could not be masked to allocation status. Patients and of life. This guide provides a comprehensive guideline of evidence-based, multidisciplinary care for the last days
and hours of life
their families received information leaflets about • Encourage staff by regularly discussing the delivered end-of-life care in staff meetings
entering the dying phase, grief and bereavement, and • Support and further educate staff in delivery of optimum end-of-life care
facilities available on the acute geriatric ward, but were
not informed about their intervention allocation and
thus were masked to the intervention.
Care guide for the last days Supportive documentation Implementation guide
of life • A manual for health-care staff Component 1: establishing the
Procedures A multi-professional document on how to use the care guide implementation project and
In the CAREFuL group, CAREFuL was implemented over that provides a guideline of care for the last days of life preparing the environment
for the last days and hours of life • An information leaflet for Component 2: preparing the
a 6-month period with the support of an implementation This guide consists of four parts: health-care staff about the care documentation
guide and supportive documentation (figure 1). The (1) A document stating when guide for the last days of life Component 3: baseline review
the care guide was started • Three leaflets for family carers Component 4: training health-
CAREFuL group were assessed for 1 year after the (2) General information about about entering the dying care staff caring for elderly
intervention, during which the care guide continued to be the patient* phase, grief and bereavement, patients in hospital
used. The control group received standard end-of-life-care (3) Points for attention—eg, “Is and facilities available on the Component 5: care guide use
the patient free of any pain?” acute geriatric ward and intensive support
immediately after the baseline assessment and were and “Are patient and family Component 6: semi-intensive
assessed for 1 year. aware of the patient’s support
condition?” If answered “No”, Component 7: assessment
All hospitals in the CAREFuL group implemented there is room to explain why Component 8: consolidation
CAREFuL. This programme involved: (1) an imple­ and what action is being Component 9: ongoing
taken. This part is checked education, training, and support
mentation guide, (2) the care guide for the last days of and written down in the
life (care guide), and (3) supportive documentation document every 4 h
(figure 1). The implementation guide incorporates (4) A checklist for informing the
family about death and
nine components that must be completed and includes a discussing practical
detailed 2-day training package to help health-care staff arrangements with them
after the patient has died
in educating and supporting their colleagues in using the Health-care staff could decide at
care guide in a correct and compassionate way (all in any time to stop the care guide
Dutch). The care guide itself is a document designed to for a particular patient

guide health-care staff in making choices in caring for

people who are dying. The care guide is initiated when
the multidisciplinary team decides that a patient is likely
to die within days or hours, on the basis of clinical
assessment. Supportive documentation such as
information leaflets for family carers about entering the
dying phase was developed to complement the care
guide. The development and content of the programme
as well as the feasibility of its implementation have been
described elsewhere.16,18
Usual care was based on the same points for attention as
the care guide (assessment of symptoms, communication,
non-pharmacological and pharmacological interventions,
and support by the specialist palliative care team).
However, standard care was given in unstructured and
heterogeneous ways between wards and patients.
Because of the vulnerability and poor health of the
patients, we did not plan to interview the patients
themselves. Primary and secondary outcomes and clinical
information were measured after death. The nurse and
physician most closely involved in the patient’s care filled
out a questionnaire within 1 week of death. For each
hospital, we allocated a staff member who was responsible
for follow-up. 6 weeks after the patient’s death, a data nurse
Figure 1: The Care Programme for the Last Days of Life
CAREFuL=Care Programme for the Last Days of Life. *Eg, diagnosis, contact information of family carers, ability to
communicate, symptoms.
(KE) sent a questionnaire to a family carer. Those who did
not respond were sent up to two reminders.16
Outcomes
The primary outcomes were comfort around dying during
the last 48 h of life using the validated End-of-Life in
Dementia–Comfort Assessment in Dying (CAD-EOLD;
range 14–42) and symptom management using a modified
version of the End-of-Life in Dementia–Symptom
Management (SM-EOLD; range 0–27), both assessed by
nurses and family carers (table 1). Key secondary outcomes
were symptoms and care needs in the last 3 days of life
assessed by nurses using the validated Palliative Care
Outcome Scale (range 0–40), satisfaction with care during
the last 48 h of life assessed by the family carer using the
validated End-of-Life in Dementia–Satisfaction With Care,
and symptom burden assessed by nurses using items
developed by the study investigators (table 1). These self-
developed items were symptoms that are common in
geriatric patients and were attention points in the care

www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5 3

 Articles

We did t tests and χ² tests to describe differences in

Scale Measured by Number of items and scoring
mean age at death, sex, and length of stay in the hospital
Primary outcomes between patients who were assessed by nurses and those
Comfort around End-of-Life in Nurse, family 14 items (scores 1–3): discomfort, pain, who were not. For differences in response between
dying in the last 48 h Dementia–Comfort carer restlessness, shortness of breath, choking,
of life23 Assessment in Dying gurgling, difficulty swallowing, fear, family carer assessment, “no-one at bedside when dying”
anxiety, crying, moaning, serenity, peace, was also analysed.
and calm. Range for the total score is We analysed differences between the control group and
14–42, with a higher score suggesting
more comfort around dying
the CAREFuL group for the primary and secondary
outcomes with a mixed linear regression model that
Symptom Modified version of Nurse, family Nine items (scores 0–3). Range for the
management in the the End-of-Life in carer total score is 0–27, with a higher score accounted for the baseline assessment and the clustered
last 48 h of life23 Dementia–Symptom suggesting less symptom control study design (ie, patients nested within hospitals).
Management scale Outcomes were analysed with hospitals as the random
Secondary outcomes factor, and group, timepoint, and their interaction as
Symptoms and care Palliative Care Nurse Ten items (scores 0–4). Range for the total fixed factors. We calculated differences in change
needs in the last Outcome Scale score is 0–40, with a higher score
3 days of life24 suggesting more symptoms and care
between post-intervention and baseline between
needs CAREFuL and control groups (interaction group × time).
Satisfaction with care End-of-Life in Family carer Ten items (scores 1–4). Range for the total Results were expressed as estimated means with
provided to the Dementia–Satisfaction score is 10–40, with a higher score 95% CIs. Comparisons were reported in terms of
patient during the With Care suggesting more satisfaction with care expected mean differences with 95% CIs. To interpret the
last 48 h of life23
magnitude of effects for the different outcomes, we
Symptom burden Self-developed items Nurse Five items (scores 1–3), with higher scores
during the last 48 h derived from the care suggesting lower symptom burden estimated effect sizes (Cohen’s d) using the baseline-
of life guide for the last days adjusted mean differences and the variance between
of life patients and between hospitals.26 Analyses were by
Table 1: Outcome measures intention to treat. All results in the main text were
analysed without a technique for missing data, except
where otherwise stated. All data were also analysed with
guide, but these symptoms were not yet included in the a technique for missing data. All analyses were done in
available validated outcome scales. Additional secondary SPSS version 22.0.
outcomes were the content of care in terms of medical and This trial is registered with ClinicalTrials.gov, number
nursing interventions in the last 48 h of life, medication NCT01890239.
use, communication between clinical staff and patients or
relatives and among clinical staff, and the level of grief of Role of the funding source
the relatives. The funders of the study had no role in study design,
We developed an assessment tool to measure the data collection, data analysis, data interpretation, or
degree to which CAREFuL was implemented in writing of the report. KB, TS, RV, and KE had full access
accordance with its protocol. Several fidelity measures to all the data in the study and all authors had final
were done during the intervention implementation responsibility for the decision to submit for publication.
process, such as whether a project steering group was
formed for the ward. Results
To recruit at least ten hospitals, we invited 31 eligible
Statistical analysis hospitals to participate in the study. 13 agreed to
The sample size was estimated to quantify the effective­ participate; however, three dropped out in the first
ness of the programme on the CAD-EOLD total score. A 3 months after inclusion (figure 2). Across the Flemish
sample size of ten clusters with a mean cluster size of Region, there are 4241 geriatric beds, of which 451 (11%)
30 was needed to achieve 80% power to detect a relevant in ten hospitals were included in the analyses. All
clinical difference of 3·2 between the control group and hospitals were non-academic (ie, regional) hospitals
the CAREFuL group assuming an intra-cluster SD of 5·8 varying from 170 to 1083 beds. 925 patients died in acute
and an intra-cluster correlation (ICC) of 0·04 using a geriatric hospital wards, 320 (35%) of whom were eligible
two-sided alpha significance level of 0·05. The ICC of for inclusion (ie, had given consent and were in hospital
0·04 corresponds to the 90% percentile of all ICCs of for more than 48 h). Of these, 293 (92%) were assessed by
follow-up studies reported by Adams and colleagues.25 nurses and 117 (37%) by family carers during the baseline
We used descriptive statistics to describe the period. After baseline assessment, five hospitals were
See Online for appendix characteristics of the study population, and χ² tests and randomly assigned to the CAREFuL group (appendix
t tests to describe differences between the control and pp 6–7); the other five continued to provide standard care.
CAREFuL groups in both the baseline and post- In the post-intervention assessment, 164 (32%) of
intervention assessments and for post-hoc non-response 506 patients who died were eligible in the CAREFuL
analyses. hospitals and 118 (36%) of 332 in the control hospitals.

4 www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5


No hospital was lost to follow-up. Assessments were done
for 132 (80%) of 164 patients in the CAREFuL group and
109 (92%) of 118 in the in the control group by nurses,
and 48 (29%) in the CAREFuL group and 23 (19%) in the
control group by family carers. On Oct 1, 2014, a decision
was made post hoc for all wards to receive the assistance
of a data nurse to recruit patients in the trial during the
assessment period, because the 1-year baseline
assessment and the 1-year post-intervention assessment
were intensive and time-consuming for the nurses.
Baseline characteristics for patients who were
assessed by nurses were similar in the control
and CAREFuL groups (table 2). Post-intervention
characteristics were also similar except for a significantly
higher number of patients dying from frailty in the
CAREFuL group (19% vs 8% in the control group;
p=0·03). In the CAREFuL group, a significant difference
between baseline and post-intervention was noted for
heart failure (28% vs 17%; p=0·027) and infection other
than pneumonia as cause of death (9% vs 19%; p=0·007).
Characteristics of deceased patients who were assessed
by family carers are described in the appendix (p 8). In
the non-response analyses, we found no differences in
mean age at death (appendix p 9). However, baseline
non-response in nurses was significantly larger for male
than female patients (p=0·01) in the control group and
for patients with a shorter length of stay in the control
group (p=0·04) and the CAREFuL group (p=0·03;
appendix p 9).
When assessed by nurses, patients in the CAREFuL
group had significantly better comfort around dying
(CAD-EOLD) than those in the control group
(baseline-adjusted mean difference 4·30, 95% CI
2·07–6·53; p<0·0001; Cohen’s d=0·78; table 3). When
analysing separate items from this scale, we found a
significant improvement in the CAREFuL group for
discomfort, pain, restlessness, shortness of breath,
choking, difficulty swallowing, fear, serenity, peace, and
calm in the last 48 h of life (all p<0·05), but not gurgling,
anxiety, crying, or moaning (appendix pp 11–12). We
found no statistically significant effect for the assessment
by family carers of comfort around dying (CAD-EOLD;
baseline-adjusted mean difference –0·62, 95% CI
–6·07 to 4·82; p=0·82; Cohen’s d=–0·10). Additionally,
we found no statistically significant effect for symptom
management (SM-EOLD) assessed by nurses (baseline-
adjusted mean difference –0·41, 95% CI –1·86 to 1·05;
p=0·58; Cohen’s d=–0·12) or family carers (–0·59,
–3·75 to 2·57; p=0·71; Cohen’s d=–0·17; table 3).
Symptoms and care needs (Palliative Care Outcome
Scale) assessed by nurses were also significantly better in
the CAREFuL than in the control group (mean difference
–2·62, 95% CI –4·96 to –0·71; p=0·009; Cohen’s d=–0·51;
table 3). We found a significant improvement for the
individual item, “Has s/he been feeling anxious or
worried about their illness or treatment?” (p=0·01;
appendix pp 11–12).
www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5 5
Articles
31 eligible hospitals (clusters) invited to participate
18 declined to participate*
13 agreed to participate
3 dropped out in the first 3 months after inclusion†
10 clusters provided baseline data
925 patients died
335 (36%) had provided informed consent to inclusion in the study
320 (35%) eligible for assessment (mean 32 per cluster [range 5–97])
293 (92%) assessed by nurses (mean 29 per cluster [range 2–93])
117 (37%) assessed by family carers (mean 13 per cluster [range 5–36])
10 clusters randomly assigned
5 assigned to standard health-care practice 5 assigned to CAREFuL
5 implemented CAREFuL
5 clusters provided post-intervention data 5 clusters provided post-intervention data
332 patients died 506 patients died
121 (36%) had given informed consent to 170 (34%) had given informed consent to
inclusion in the study inclusion in the study
118 (36%) eligible for assessment 164 (32%) eligible for assessment
(mean 24 per cluster [range 5–41]) (mean 33 per cluster [range 17–54])
109 (92%) assessed by nurses 132 (80%) assessed by nurses
(mean 22 per cluster [range 5–41]) (mean 26 per cluster [range 13–48])‡
23 (19%) assessed by family carers 48 (29%) assessed by family carers
(mean 5 per cluster [range 0–9]) (mean 10 per cluster [range 0–21])
Figure 2: Trial profile
CAREFuL=Care Programme for the Last Days of Life. *Reasons included insufficient time, understaffing of the ward or
wards, already involved in other projects, or no stable structure at the ward or wards because of new head nurse.
†Three hospitals dropped out during baseline assessment because of stringent inclusion criteria that the hospital
was unwilling to maintain for the study duration. ‡For 99 (75%) of the 132 patients who were assessed by nurses,
the care guide was started.
Conversely, satisfaction with care (End-of-Life in
Dementia–Satisfaction With Care; range 10–40)
assessed by family carers was significantly worse in the
CAREFuL group than in the control group (mean
difference –4·00, 95% CI –7·87 to –0·12; p=0·04;
Cohen’s d=–0·74). We found weak evidence of an
improved satisfaction of care post-intervention
compared with baseline in the control group, which did
not occur in the CAREFuL group (table 3). More family
carers in the control group than in the CAREFuL group
responded positively to the questions “I always knew
which doctor or nurse was in charge of my care
recipient’s care” (p=0·002) and “I feel that my care
recipient got all necessary nursing assistance” (p=0·014;
appendix pp 11–12).
For the measurement of symptom burden, compared
with the control intervention, CAREFuL significantly
 Articles

Baseline Post-intervention
Control (n=81) CAREFuL (n=212) Control (n=109) CAREFuL (n=132)
Mean age at death (years)*† 85·0 (6·23) 86·0 (6·86) 84·0 (7·52) 85·8 (6·84)
Sex*
Male 38/81 (47%) 98/211 (46%) 60/108 (56%) 70/132 (53%)
Female 43/81 (53%) 113/211 (54%) 48/108 (44%) 62/132 (47%)
Underlying cause of death‡§
Cancer 7/61 (11%) 25/204 (12%) 19/89 (21%) 14/108 (13%)
Pneumonia 19/61 (31%) 56/204 (27%) 18/89 (20%) 29/108 (27%)
Infection other than pneumonia 10/61 (16%) 18/204 (9%)¶ 12/89 (13%) 21/108 (19%)¶
Respiratory disease other than pneumonia 7/61 (11%) 16/204 (8%) 9/89 (10%) 8/108 (7%)
Heart failure 14/61 (23%) 57/204 (28%)|| 21/89 (24%) 18/108 (17%)||
Cardiovascular disease other than heart failure 7/61 (11%) 18/204 (9%) 7/89 (8%) 8/108 (7%)
Cerebrovascular disease 3/61 (5%) 16/204 (8%) 10/89 (11%) 10/108 (9%)
Advanced dementia 3/61 (5%) 7/204 (3%) 4/89 (5%) 6/108 (6%)
Frailty 6/61 (10%) 32/204 (16%) 7/89 (8%)** 20/108 (19%)**
Trauma 2/61 (3%) 4/204 (2%) 1/89 (1%) 1/108 (1%)
Other 7/61 (11%) 9/204 (4%) 10/89 (11%) 9/108 (8%)
Had dementia at time of death‡ 23/56 (41%)†† 122/186 (66%)††‡‡ 35/80 (44%)§§ 74/93 (80%)‡‡§§
Residency before hospital admission‡
Home 43/62 (69%) 151/203 (74%) 63/91 (69%) 76/110 (69%)
Nursing home or other institution 19/62 (31%) 52/203 (26%) 28/91 (31%) 34/110 (31%)
Length of hospital stay, mean number of days (SD)*¶¶ 22·3 (16·5) 24·5 (21·8) 23·3 (16·1) 28·0 (24·4)
Length of stay on acute geriatric ward‡
>4 weeks 12/60 (20%) 42/202 (21%) 19/90 (21%) 36/109 (33%)
≥2 weeks to <4 weeks 27/60 (45%) 61/202 (30%) 37/90 (41%) 41/109 (38%)
≥1 week to <2 weeks 10/60 (17%) 55/202 (27%) 22/90 (24%) 18/109 (17%)
≥72 h to <1 week 10/60 (17%) 37/202 (18%) 10/90 (11%) 11/109 (10%)
<72 h 1/60 (2%) 7/202 (3%) 2/90 (2%) 3/109 (3%)

Data are mean (SD) or n/N (%). CAREFuL=Care Programme for the Last Days of Life. *Collected by the researcher via the administrative files. †Data missing at baseline for three
patients in the CAREFuL group. ‡Reported by the physician. §Percentages can add up to more than 100 because several causes of death could be indicated by the physician in one
patient. ¶p=0·007 for the difference between post-intervention and baseline (χ² test). ||p=0·027 for the difference between post-intervention and baseline (χ² test). **p=0·03 for
the difference between groups (χ² test). ††p=0·001 for the difference between groups (χ² test). ‡‡p=0·016 for the difference between post-intervention and baseline (χ² test).
§§p<0·0001 for the difference between groups (χ² test). ¶¶Data missing post-intervention for one patient in the control group.

Table 2: Characteristics of deceased patients for whom an assessment by nurses was made

reduced troublesome mucus (baseline-adjusted mean
difference 0·27, 95% CI 0·00–0·54; p=0·047;
Cohen’s d=0·38) and vomiting (0·23, 0·05–0·41;
p=0·014; Cohen’s d=0·47; table 3). We noted no
significant differences for nausea (p=0·11), reduced
appetite (p=0·81), and fatigue (p=0·19).
Of the medical interventions stopped in the last 48 h of
life, only oxygen therapy differed between groups, with
more patients in the CAREFuL group than the control
group having their oxygen therapy stopped (odds
ratio 4·52, 95% CI 1·19–17·12; p=0·03; appendix p 13).
According to nurses, family doctors were informed about
the impending death of the patient for more patients in
the CAREFuL group than the control group (odds
ratio 2·51, 95% CI 1·06–5·95; p=0·04); no other
significant differences between groups were noted for
communication among clinical staff (appendix p 20). No
differences between groups were noted for nursing
interventions performed in the last 48 h of life (appendix
p 14), medication use (appendix p 15), communication
between clinical staff and patients (appendix p 16),
communication between clinical staff and family carers
(appendix pp 17–19), or grief (appendix p 21).
A steering group was formed in all 11 wards of the
five CAREFuL hospitals, consisting of at least one, but
mostly at least two (in nine wards), facilitators (appendix
pp 6–7). The attendance of the facilitator, and the
members of the steering group at the intensive 2-day
training varied between the wards. After the training
sessions, the facilitator, who was responsible for the
implementation of CAREFuL in the ward, organised at
least one training session on how to work with the care
guide for the last days of life. For the medical and nursing
staff adherence varied between the wards (from 0% to
100%). Physicians in general attended a training session
less often than nurses. After the training sessions, all
wards started using the care guide for the last days of life.
The proportion of dying patients cared for with use of the

6 www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5

www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5

Baseline score Post-intervention score Baseline-adjusted mean pinteraction* ICC Effect size
difference between (Cohen’s d)
groups post-
intervention (95% CI)
Control CAREFuL Control CAREFuL
Number Mean (95% CI) Number Mean (95% CI) Number Mean (95% CI) Number Mean (95% CI)
of of of of
patients patients patients patients
Primary outcomes
Comfort around dying with the CAD-EOLD (range 14–42)†
Assessed by nurses 61 32·9 (31·3 to 34·5) 182 31·9 (30·7 to 33·0) 91 31·3 (29·9 to 32·7) 110 34·6 (33·3 to 35·9) 4·30 (2·07 to 6·53) <0·0001 0·025 0·78
Assessed by family 11 30·3 (26·4 to 34·1) 65 28·7 (26·7 to 30·7) 19 32·0 (28·9 to 35·0) 31 29·8 (27·3 to 32·3) –0·62 (–6·07 to 4·82) 0·82 <0·0001 –0·10
carers
Symptom management with the SM-EOLD (range 0–27)‡
Assessed by nurses 62 5·87 (4·76 to 6·98) 146 4·81 (3·85 to 5·77) 77 5·90 (4·87 to 6·93) 109 4·44 (3·48 to 5·39) –0·41 (–1·86 to 1·05) 0·58 0·037 –0·12
Assessed by family 11 5·36 (3·26 to 7·47) 41 6·17 (5·08 to 7·26) 17 5·00 (3·31 to 6·69) 32 5·22 (3·99 to 6·45) –0·59 (–3·75 to 2·57) 0·71 0·078 –0·17
carers
Secondary outcomes
Symptoms and care 73 13·6 (8·8 to 18·36) 193 15·3 (10·4 to 20·3) 96 13·3 (8·35 to 18·3) 123 9·56 (4·09 to 15·0) –2·62 (–4·96 to –0·71) 0·009 <0·0001 –0·51
needs with the POS
assessed by nurses
(range 0–40)‡§
Satisfaction with care 20 30·3 (27·9 to 32·7) 79 30·7 (29·4 to 31·8) 22 33·5 (31·2 to 35·8) 40 29·6 (28·1 to 31·5) –4·00 (–7·87 to –0·12) 0·04 <0·0001 –0·74
with SWC-EOLD
assessed by family carers
(range 10–40)†
Self-developed items to measure symptom burden assessed by nurses (range 0–3)†
Bothersome mucus 75 2·47 (2·30 to 2·63) 202 2·35 (2·25 to 2·45) 102 2·38 (2·24 to 2·52) 126 2·54 (2·41 to 2·67) 0·27 (0·00 to 0·54) 0·047 <0·0001 0·38
Nausea 74 2·85 (2·72 to 2·99) 204 2·75 (2·64 to 2·86) 102 2·81 (2·69 to 2·93) 127 2·87 (2·75 to 2·98) 0·15 (–0·03 to 0·35) 0·11 0·026 0·31
Vomiting 75 2·91 (2·80 to 3·02) 201 2·76 (2·69 to 2·83) 102 2·82 (2·73 to 2·92) 128 2·91 (2·82 to 2·99) 0·23 (0·05 to 0·41) 0·014 0·016 0·47
Reduced appetite 73 1·69 (1·43 to 1·96) 201 1·76 (1·53 to 1·99) 100 1·77 (1·52 to 2·01) 131 1·87 (1·64 to 2·11) 0·04 (–0·27 to 0·35) 0·81 0·080 0·05
Fatigue 75 1·56 (1·31 to 1·81) 199 1·64 (1·42 to 1·86) 102 1·59 (1·36 to 1·83) 129 1·86 (1·64 to 2·08) 0·19 (–0·09 to 0·48) 0·19 0·003 0·25

All means and 95% CIs are cluster adjusted; differences are cluster and baseline adjusted. CAD-EOLD=End-of-Life in Dementia–Comfort Assessment in Dying. CAREFuL=Care Programme for the Last Days of Life. ICC=unconditional intraclass correlation
coefficient from baseline assessments. POS=Palliative Care Outcome Scale. SM-EOLD=End-of-Life in Dementia–Symptom Management. SWC-EOLD=End-of-Life in Dementia–Satisfaction With Care. *Interaction effect of group (control and CAREFuL)
and timepoint (baseline and post-intervention) calculated with a mixed linear regression model. We calculated differences in change (post-intervention minus baseline) between the CAREFuL and control groups (interaction group × time). †Higher
scores suggest better quality of dying and end-of-life care. ‡Higher scores suggest worse quality of dying and end-of-life care. §470 (88%) of 534 nurses did not fill out all the items of the POS. Therefore, we used a technique to account for missing
values.27 If at least five of ten items were filled out (by 485 [91%] of 534 nurses), we used the data; missing items were replaced with the mean of the other items for that patient. See appendix (p 10) for results for the other outcomes analysed with and
without the missing data technique.

Table 3: Cluster-adjusted mean scores and differences for the primary and secondary outcomes

Articles
7
 Articles
care guide during the 12 months after the implementation
period varied between the wards, from 38% to 100%.
The mean proportion across the wards was 60%
(SD 20·02; appendix p 7). Patients were cared for with
use of the care guide for a mean of 4·38 days (SD 5·84).
Discussion
This is, to our knowledge, the first adequately powered
cluster randomised controlled trial to test the effectiveness
of an end-of-life care programme in acute geriatric hospital
wards.11,12 We found a statistically and clinically significant
improvement in one of the primary outcomes of the trial,
comfort around dying assessed by nurses after
implementation of the programme; however, this effect
was not found by family carers. We also found a significant
positive effect on symptoms and care needs, and an
adverse effect on satisfaction with care of the family carer.
The different components of CAREFuL aim to improve
communication between the person who is dying, those
close to them, and the health professionals involved, and
to stop potentially inappropriate medications and
interventions as well as improve symptom management
through onsite professional training. These components
seem to have helped health-care professionals in
delivering care to people who are dying, especially in
symptom assessment and management—which is often
perceived to be difficult, and for which staff often do not
receive training.28 By providing the care guide, a template
of evidence-based, multidisciplinary care for the last days
and hours of life, supported by different family
information leaflets, the programme aimed to and might
have resulted in more structured end-of-life care processes.
We also found that implementation of the programme
had a negative effect on satisfaction with care. The
negative effect seems to have been due to an improvement
in satisfaction that occurred in the control group, but not
in the CAREFuL group, although no decrease was noted
either. Thus, CAREFuL did not have the desired effect on
the patient’s family (ie, satisfaction with care) and this is
a serious concern that needs to be investigated further.
However, because of the low response rate in family
carers, this result is difficult to interpret. The negative
effect might be explained by a complex combination of a
Hawthorne effect in the control group (improved
communication with family carers as a result of having
to fill in questionnaires assessing quality of care
and quality of communication) and standardisation
of procedures in the CAREFuL group. Standardisation of
care as a consequence of streamlining practice by use of
the care guide could be one reason for the positive effects
on comfort and quality of care and on symptoms and
care needs. However, such standardisation might be at
the expense of patients and families; for example, by
overlooking their concerns and having less personal
contact with them.29 Nurses in the CAREFuL group
might have presumed that giving leaflets to families
instead of taking time for communication and
8 www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5
explanation was sufficient to address their information
needs. These possible effects of standardisation of
procedures align with some of the concerns raised in the
National Care of the Dying Audit (2012)20 and the review
by Neuberger and colleagues (2013) of the LCP,19 which
included a failure to inform families that their relative
was entering the dying phase or to explain the pathway,20
and creating a tick-box exercise in which meeting
guidelines seemed more important than administering
appropriate care.19,20 In creating CAREFuL, we took into
account all the concerns about the LCP. Another possible
explanation for the negative effect of CAREFuL on
satisfaction with care might be that the training did not
pay enough attention to communication aspects in
CAREFuL. Based on our findings, we recommend that
more attention to communication with those close to the
person who is dying is needed in the further development
and implementation of the programme. Further
qualitative research through in-depth interviews with
study participants—both professionals and family
carers—is needed to gain a better understanding of the
effect of CAREFuL on satisfaction with care.
Despite the fact that designing and performing trials
involving people who are dying is challenging for
methodological, practical, and ethical reasons,30 we
developed a rigorous cluster randomised controlled trial.
The fidelity measures done during the study showed that
CAREFuL was implemented according to the protocol in
most of the wards, although physicians in wards assigned
to the CAREFuL group often did not take part in the
training sessions. However, there is a need for caution
about the positive effects of CAREFuL. Because nurses
were not masked to allocation, their outcome
measurement might have been affected (ie, detection bias
or ascertainment bias). The nurses who assessed the
quality of dying of the patients were aware of and had
implemented CAREFuL. Moreover, we found a lower
response rate among nurses in the CAREFuL group (80%)
than in the control group (92%). These differences could
suggest some bias in reporting outcomes. However, in
view of the need for retrospective assessments related to
the final days of life from key people involved in patients’
end-of-life care, the validity of proxy assessments by
nurses reported in the published work, and the anticipated
higher response rate from nurses (ie, statistical power),
we judged nurses’ assessments for the primary outcomes
as an appropriate option for our study design.
Our study has other limitations. Only 34–36% of
patients in the control and CAREFuL groups admitted to
the wards gave consent to use their personal data. During
the study period, nurses did not approach all patients
because of insufficient time. We do not have data on how
many patients were approached for informed consent,
how many refused, or the reasons for refusal. We did not
collect data to compare characteristics of the patients
who gave consent versus those who did not. We obtained
high response rates for nurses, but lower rates for family

carers, although these are comparable with postal surveys
of bereaved relatives in another study31 and might be
explained by the informed consent requirements. Nurses
reported that they were often unable to ask the family
members for informed consent shortly after the death of
the patient. In these cases, the hospital sent an informed
consent form by post to the family carers 2 weeks after
the death of the patients asking permission for the
researcher to contact them. Only family carers who gave
consent were sent a questionnaire 6 weeks after the
patient’s death. Although the voices of family members
provide a closer picture of the so-called caregiving
triangle between family and professional carers and
patients, the lower response rate of family carers makes
these outcomes difficult to interpret. Also, we were
unable to compare characteristics of nurses and family
members who assessed patients in the control and
CAREFuL groups because we had insufficient
information about the respondents.
Although we do have quantitative information, a
qualitative examination of the implementation of the
programme would help to better understand the
unexpected result of worsened family satisfaction with
care. Additionally, in complex, multifaceted interventions,
establishing which of the components is primarily
responsible for the observed effects is not possible.32
In conclusion, end-of-life care is an important part of
geriatric care and the many older people who die in acute
geriatric hospital wards have the right to the best care in
the dying phase. Findings from our cluster randomised
controlled trial showed that CAREFuL significantly
improved comfort around dying and symptoms and care
needs in the last days of life as assessed by nurses, but
negatively affected family members’ satisfaction with
care as assessed by family carers. Based on the positive
results for comfort around dying assessed by nurses,
widespread implementation of CAREFuL seems
warranted. However, our results on satisfaction with care
and the previous experiences in the UK with a similar
programme (LCP) also suggest the importance of further
exploration of the perceptions of family and patients.
Careful monitoring of the implementation of the
programme, with special attention given to the family
carers, will be important.
Contributors
KB was responsible for the integration of the data and for data analyses,
and led the writing of the manuscript. TS, JC, RV, MC, NVDN, and LD
conceived the idea of the study and JC and LD obtained funding. TS, JC,
RV, MC, KE, NVDN, and LD contributed to the planning and collection
of data. All authors contributed to the writing of the report and approved
the final version.
Declaration of interests
We declare no competing interests.
Acknowledgments
This study is part of the Flanders Study to Improve End-of-Life Care
and Evaluation Tools (FLIECE-project), a collaboration between the
Vrije Universiteit Brussel (Brussels, Belgium), Ghent University
(Ghent, Belgium), the Katholieke Universiteit Leuven (Leuven,
Belgium), and VU University Medical Centre (Amsterdam,
www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5 9
Articles
Netherlands). This study was supported by a grant from the Flemish
Government Agency for Innovation by Science and Technology
(Agentschap voor Innovatie door Wetenschap en Technologie;
SBO IWT nr. 100036) and by financial support from the Belgian
Cancer Society “Kom Op Tegen Kanker”. JC is a postdoctoral fellow of
the Research Foundation Flanders. We thank all participating
physicians, nurses, patients, and their families for providing data for
this study; Heleen Lyphout for help in collecting data; Roos Colman
for her advice on data analysis; and Jane Ruthven for her language
editing.
References
1 WHO. Preventing chronic diseases: a vital investment: Geneva:
World Health Organization, 2005.
2 Pivodic L, Pardon K, Morin L, et al. Place of death in the population
dying from diseases indicative of palliative care need:
a cross-national population-level study in 14 countries.
J Epidemiol Community Health 2016; 70: 17–24.
3 Al-Qurainy R, Collis E, Feuer D. Dying in an acute hospital setting:
the challenges and solutions. Int J Clin Pract 2009; 63: 508–15.
4 Twomey F, McDowell DK, Corcoran GD. End-of-life care for older
patients dying in an acute general hospital—can we do better?
Age Ageing 2007; 36: 462–64.
5 Hall S, Petkova H, Tsouros A, Constantini M, Higginson I.
Palliative care for older people: better practices. Copenhagen:
World Health Organization, 2011.
6 Inouye SK, Studenski S, Tinetti ME, Kuchel GA. Geriatric
syndromes: clinical, research, and policy implications of a core
geriatric concept. J Am Geriatr Soc 2007; 55: 780–91.
7 Bailey FA, Burgio KL, Woodby LL, et al. Improving processes of
hospital care during the last hours of life. Arch Intern Med 2005;
165: 1722.
8 Ellershaw J, Wilkinson S. Care of the dying: a pathway to excellence.
Oxford: Oxford University Press, 2011.
9 Bookbinder M, Blank AE, Arney E, et al. Improving end-of-life care:
development and pilot-test of a clinical pathway.
J Pain Symptom Manage 2005; 29: 529–43.
10 Ellershaw J, Ward C. Care of the dying patient: the last hours or
days of life. BMJ 2003; 326: 30–34.
11 Chan RJ, Webster J. End-of-life care pathways for improving
outcomes in caring for the dying. Cochrane Database Syst Rev 2013;
11: CD008006.
12 Chan R, Webster J. End-of-life care pathways for improving
outcomes in caring for the dying. Cochrane Database Syst Rev 2010;
1: CD008006.
13 Chan RJ, Webster J, Bowers A. End-of-life care pathways for
improving outcomes in caring for the dying.
Cochrane Database Syst Rev 2016; 2: CD008006.
14 Costantini M, Romoli V, Leo S D, et al. Liverpool Care Pathway for
patients with cancer in hospital: a cluster randomised trial.
Lancet 2014; 383: 226–37.
15 Brännström M, Fürst CJ, Tishelman C, Petzold M, Lindqvist O.
Effectiveness of the Liverpool Care Pathway for the dying in
residential care homes: an exploratory, controlled before-and-after
study. Palliat Med 2016; 30: 54–63.
16 Verhofstede R, Smets T, Cohen J, et al. Development of the Care
Programme for the Last Days of Life for older patients in acute
geriatric hospital wards: a phase 0–1 study according to the Medical
Research Council Framework. BMC Palliat Care 2015; 14: 24.
17 Verhofstede R, Smets T, Cohen J, Costantini M,
Van Den Noortgate N, Deliens L. Improving end-of-life care in acute
geriatric hospital wards using the Care Programme for the Last
Days of Life: study protocol for a phase 3 cluster randomized
controlled trial. BMC Geriatr 2015; 15: 13.
18 Verhofstede R, Smets T, Cohen J, Costantini M,
Van Den Noortgate N, Deliens L. Implementing the Care
Programme for the Last Days of Life in an acute geriatric hospital
ward: a phase 2 mixed method study. BMC Palliat Care 2016;
15: 27.
19 Neuberger J, Guthrie C, Aaronvitch D, et al. More care, less
pathway: a review of the Liverpool Care Pathway. 2013. https://
www.gov.uk/government/uploads/system/uploads/attachment_
data/file/212450/Liverpool_Care_Pathway.pdf (accessed
April 18, 2017).
 Articles
20 Marie Curie Palliative Care Institute. National Care of the Dying
Audit—hospitals. Generic report, round 2. 2009. Available:
http://www.mcpcil.org.uk/media/16531/generic_ncdah_2nd_
round_final_report[1].pdf (accessed April 18, 2017).
21 Moher D, Hopewell S, Schulz KF, et al. CONSORT 2010
explanation and elaboration: updated guidelines for reporting
parallel group randomised trials. Int J Surg 2012; 10: 28–55.
22 Agentschap Zorg en Gezondheid. Evolutie van de ligduur en
bezettingsgraad per ziekenhuisdienst. 2016. https://www.zorg-en-
gezondheid.be/evolutie-van-de-ligduur-en-bezettingsgraad-per-
ziekenhuisdienst (accessed April 18, 2017).
23 Volicer L, Hurley AC, Blasi ZV. Scales for evaluation of end-of-life
care in dementia. Alzheimer Dis Assoc Disord 2011; 15: 194–200.
24 Hearn J, Higginson IJ. Development and validation of a core
outcome measure for palliative care: the Palliative Care Outcome
Scale. Palliative Care Core Audit Project Advisory Group.
Qual Health Care 1999; 8: 219–27.
25 Adams G, Gulliford M, Ukoumunne O, Eldridge S, Chinn S,
Campbell M. Patterns of intra-cluster correlation from primary care
research to inform study design and analysis. J Clin Epidemiol 2004;
57: 785–94.
26 Vacha-Haase T, Thompson B. How to estimate and interpret various
effect sizes. J Couns Psychol 2004; 51: 473–81.
10 www.thelancet.com Published online May 16, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31265-5
27 Little RJ, D’Agostino R, Cohen ML, et al. The prevention and
treatment of missing data in clinical trials. N Engl J Med 2012;
367: 1355–60.
28 Murray-Brown F, Curtis M, Gibbins J. Survey of Foundation Year 1
doctors caring for the dying; what do they see, do and need to
perform this role, following removal of the Liverpool Care Pathway.
Palliat Med 2015; 29: 975–76.
29 Floriani C, Schramm FR. Routinization and medicalization of
palliative care: losses, gains and challenges. Palliat Support Care
2012; 10: 295–303.
30 Karlawish JHT. Conducting research that involves subjects at the
end of life who are unable to give consent. J Pain Symptom Manage
2003; 25: S14–24.
31 Gomes B, Calanzani N, Koffman J, Higginson IJ. Is dying in
hospital better than home in incurable cancer and what factors
influence this? A population-based study. BMC Med 2015; 13: 235.
32 McConnell T, O’Halloran P, Porter S, Donnelly M. Systematic
realist review of key factors affecting the successful implementation
and sustainability of the Liverpool Care Pathway for the Dying
Patient. Worldviews Evid Based Nurs 2013; 10: 218–37.