In this chapter we present the background, methods and results from the quantitative
cross-sectional survey of hospital inpatients, which was undertaken in two UK
hospitals. In doing so we address the following objectives:
Background
The majority of deaths in developed countries now occur in the acute hospital
setting; in England, around 58% of people currently die in acute hospitals.112 Although recent evidence
suggests a slow increase in the proportion of deaths at home in England and
Wales,19 other predictions
based on past trends estimate that only one in 10 people in the UK will die at
home by 2030, and that an expansion of inpatient facilities by one-fifth may be
required.113 The End
of Life Care Strategy for England5 has highlighted the delivery of high-quality
palliative and end-of-life care in the acute hospital setting as an area of
priority, acknowledging that a significant proportion of patients dying in acute
hospitals receive very poor care.
The identification of patients who may benefit from palliative care is recognised
as problematic. Health professionals have reported differing understandings of
what constitutes a ‘palliative care’ patient30 and when a palliative care
approach might be appropriate.
This finding is supported by evidence from our qualitative focus groups presented
in Chapter 4. There are
significant implications of a lack of consensus in identifying which patients
have palliative care needs, including poor continuity of care, inadequate
service provision and support and excess economic cost.89,114
Given the challenge that the increasing number of hospitalisations at the end of
life presents, coupled with the problem of identifying patients who would
benefit from palliative care input, it is now imperative to gain a better
understanding of the extent of palliative care need in the hospital setting to
more appropriately map services to patient need, define priorities for care and
assess the economic impact of palliative care in the acute setting.2
Methods
Survey methods
A comprehensive survey of hospital inpatients was undertaken in two UK
hospitals selected for their sociodemographic diversity. The SNGH has
> 1100 beds and serves a largely urban, economically
disadvantaged and ethnically diverse area. In contrast, the RLI has
approximately 400 beds and serves a predominantly white Caucasian
semi-rural/remote rural population. Although the term
‘census’ was originally proposed to describe this phase of
data collection, this has been replaced throughout by the term
‘survey’. This was because of concerns that the term
‘census’ implies a collection of a complete population data
set. However, this was not possible as the ethics committee stipulated that
data collection from patient notes was restricted to those patients who
consented to involvement in the study.
The survey of the SNGH was undertaken over an 11-day period in May 2010 and
the survey of the RLI took place over a 5-day period in November 2010. All
inpatient wards, with the exception of children’s wards and mother
and baby units, were included. Each ward was visited by two members of the
data collection team at some point during the survey period. Inclusion
criteria were age ≥ 18 years and resident on the ward at
0900 on the day that the ward was surveyed. Non-English-speaking patients
and deaf patients were excluded because of a lack of translation
facilities.
The approach to the inclusion of patients lacking capacity to consent for
themselves was developed in line with Mental Capacity Act 2005
guidance.10 Senior medical
and nursing staff, and relatives (when available) were consulted to identify
any patients lacking capacity to consent. Personal consultees (relatives or
close friends) were identified and, when available, were invited to
participate on behalf of patients lacking capacity.
For patients/consultees who consented to participate, the following data were
collected:
Data from patients’ hospital case notes comprising evidence
of palliative care need according to GSF prognostic indicator
criteria
57 (the
GSF prognostic indicator guide provides 11 diagnostic criteria
categories, which provide an indication of patients who might
benefit from palliative care input); reason for admission;
sociodemographic and diagnostic information; details of
comorbidities; evidence of adoption of a palliative care approach
using a list of predefined indicators (identified by health
professionals during a previous qualitative phase of this study; see
Chapter
4); number of hospital admissions in the last 12
months; discharge plans.
For each consenting patient a member of the medical staff and a
member of the nursing staff known to the patient were interviewed.
Staff were asked to provide diagnostic and admission information for
the patient. They were also asked whether they believed the patient
to have palliative care needs according to a standardised definition
(a broad and inclusive definition of palliative care was purposively
selected to maximise the potential for patient identification),
115 whether they would
be surprised if the patient died within 12 months; about the
appropriateness of the admission to hospital; and whether prognostic
discussions had taken place. When possible the member of the nursing
staff was the designated ‘named nurse’ for the
patient and the member of the medical staff was the junior
(Foundation Year 1 and 2) or senior (Specialist Trainee year 1 and
year 2) house officer or the registrar.
Patient-/consultee-completed questionnaires comprising
sociodemographic information; a service use questionnaire developed
for use with a palliative care population (Gott M, Barnes S, Payne
S, Seamark D, Small N. Department of Health, 2007, unpublished
report); and the Sheffield Profile for Assessment and Referral for
Care (SPARC).
116
SPARC is a validated holistic self-assessment tool to identify
patients who would benefit from palliative care input (see
Appendix
4). It provides scores across a range of
physical, psychological and social domains. In cases in which
consultees participated, they were asked to answer questions as they
believed that the person they were acting as consultee for would
have done.
All data were collected by a team of 30 researchers with previous experience
in health-care research as either an academic or a clinician. Data from
hospital case notes were collected by researchers with a clinical background
in medicine or nursing. All researchers attended a full-day training session
prior to the study commencing, which provided training in approaching
patients/staff, the correct use of data collection tools and procedures for
problem situations.
Data analysis
All data were recorded onto anonymised paper proformas and were subsequently
transferred onto a SPSS database version 20 (SPSS Inc., Chicago, IL, USA)
for data cleaning and analysis, with the guidance of the project
statistician (CP). Descriptive analyses were used to describe the data from
all sources.
Cohen’s kappa measure of chance-corrected agreement was used to
assess agreement between sources regarding identification of patients with
palliative care needs and appropriateness of admission to hospital. Logistic
regression analyses were used to explore the effects of a range of predictor
variables on various outcomes.
Economic analysis methods
Two palliative care consultants (BN and MB) undertook a post hoc assessment
of appropriateness of admission to hospital for patients who had been
identified with palliative care needs according to GSF criteria. They
reviewed the survey data collected from the case notes of these patients and
made a decision whether the admission was ‘potentially
avoidable’ or ‘unavoidable’ or whether there were
‘insufficient data to make a decision’. For those patients
whose admission was classed as potentially avoidable, an alternative place
of care was suggested by the consultants.
An estimated cost was then attached to each of the admissions identified by
the consultants as potentially avoidable. The cost of each hospital
admission was estimated using HRG codes. HRG codes are clinically meaningful
groupings of patient activity, based on both diagnoses and clinical
procedures undertaken. Standard costs are assigned to each HRG code and
these are used across the NHS in England to calculate the cost of a hospital
admission.117
Because of ethical restrictions, HRG codes for patients in the surveys could
not be obtained retrospectively from the trusts involved. The Sheffield
Teaching Hospitals coding team therefore undertook an ad hoc exercise to
allocate HRG codes to each admission identified as potentially avoidable
across both hospitals using the data from our survey. HRG codes and tariffs
for 2010/11 were used.
Admissions for patients who remain in hospital beyond the expected upper
length of stay for the HRG code (trim point) would normally be allocated an
additional long-stay or excess bed-day payment. In the absence of
length-of-stay data for our sample of patients, we were unable to cost for
excess bed-days.
We then estimated the costs of the alternative places of care suggested by
our consultants for those admissions deemed avoidable. The costs of
alternative places of care were taken from published national sources,
inflated to 2011 prices when necessary (see ). The cost of nursing home
care (£106 per day) was obtained from the Personal Social Services
Research Unit’s Unit Costs of Health and Social Care
2010.118 The
cost of hospice care (£325 per day) was obtained from research
commissioned by the National Audit Office.119 The cost of home care (£50 per day)
was derived from a King’s Fund report,82 which estimated the cost of home care in the
last 8 weeks of life. This cost included the following services: GPs,
district nursing, Marie Curie and Macmillan nurses, ambulance journeys,
hospice at home and hospice (inpatient) and equipment. An average daily cost
was calculated from the 8-week figure cited in the report.
Total cost of alternative care for avoidable admissions
These studies all assumed a broad societal perspective120 to take account of costs irrespective of
funding source. Although we would have also liked to include informal carer
costs and indirect costs – those incurred as a result of lost
productivity – in our calculations, this was not possible because of
a lack of data. As a point of comparison, we explored a more restrictive
cost perspective. This was the NHS and personal social services (PSS)
perspective, which considers only those costs that fall within the remit of
these two organisations. To do this we used the costs cited in a previous
study,121 which indicated
that only 43% of nursing home costs and 44% of hospice costs were funded by
the NHS or PSS. Costs were inflated to the 2011 price year when
necessary.
Results
Survey response rate
A total of 1359 inpatients were eligible for inclusion in the survey (1009
patients in Sheffield and 350 patients in Lancaster). Of the total eligible
patient population, 654 (48.1%) patients agreed to participate in the study
of whom 616 patients consented for themselves and 38 consented through a
consultee. Patient response rates were similar for the two hospitals (SNGH
46.9%, RLI 52.9%). Details of patient recruitment at the two participating
hospitals are provided in .
Details of patient recruitment at the two participating
hospitals.
Of the 654 consenting patients/consultees, complete data sets are available
for 514 patients (final response rate 37.8%). A complete data set is defined
as containing a case note review and a questionnaire completed by a member
of either the medical staff or the nursing staff, but not necessarily both.
The analyses presented in this section relate to the 514 patients with
complete data sets.
Results of the case note review
Of the 514 patients, just over one-third
(n = 185, 36.0%) met one or more of the GSF
prognostic indicator criteria for palliative care need. Of the patients
identified with palliative care needs according to GSF criteria, 53.8% were
female and the median age was 78 years, with an age range of 20–103
years. The majority of these patients were aged ≥ 65 years
(77.8%), with a considerable proportion aged ≥ 85 years
(23.2%) ().
shows demographic information for the sample of 185 patients with palliative
care needs according to GSF criteria.
Age range of patients with palliative care needs according to GSF
criteria (n = 181, no age given for
four patients).
Demographic information for patients with palliative care needs
according to GSF criteria
(n = 185)
The majority of patients (70.8%) met only one GSF criterion for palliative
care need; however, just under one-third (29.2%) met two or more criteria
(see ).
shows the breakdown of GSF prognostic indicators amongst the patient
sample.
Participant admission and diagnostic data
(n = 185)
Numbers of patients meeting each of the GSF prognostic indicators for
palliative care need (n = 185). a,
Other life-limiting illnesses included cystic fibrosis,
Huntington’s disease, asbestosis, etc.
The most common GSF prognostic indicator was frailty, with almost one-third
of patients (27%) meeting this criteria. Heart disease (20.5%), cancer
(19.5%), COPD (18.4%) and dementia (17.8%) were the next most common GSF
criteria, and were roughly equal in prevalence. Other indicators, including
stroke, renal disease and Parkinson’s disease, were less common.
provides
admission and diagnostic information for the patient group. Reason for
admission to hospital was ascertained in all but five patients (included in
the ‘other’ group). The most common reasons for admission
were falls/confusion/general frailty (14.6%), complications relating to
cancer (13.0%) and respiratory disease or exacerbation (13.0%). Patients had
a median of two comorbid conditions, with over one-third of patients having
three or more comorbidities. In the 12 months before the survey, patients
had a median of one previous hospital admission and spent a mean of 42 days
(lower quartile = 1.8, upper
quartile = 53.0) in hospital. For the majority of patients
(65.9%) there was no evidence of the adoption of a palliative care approach.
Around one-third (28.6%) of patients had a do not attempt resuscitation
(DNAR) order in place, but only a small number (8.1%) had been referred to
specialist palliative care services.
Results of medical and nursing staff assessment of palliative care
need
Medical and nursing staff were asked whether they believed patients to have
palliative care needs according to the Canadian Palliative Care Association
1997 definition.115 Nurse
questionnaires were completed for 473 patients; of these, nurses stated that
84 (17.8%) had palliative care needs. However, data from patients’
hospital case notes indicated that 174 (36.8%) of these 473 patients were
identified as having palliative care needs according to GSF criteria
(). Staff were also asked, ‘would you be surprised
if this patient died (1) during this admission and (2) in the next 12
months?’ Nursing staff would not have been surprised if the patient
died during the current admission in 74 (15.6%) cases and in the next 12
months in 180 (38.1%) cases. Medical staff questionnaires were completed for
297 patients; of these, doctors stated that 46 (15.5%) had palliative care
needs, whereas using the GSF criteria 108 (36.4%) were identified with
palliative care needs (see ). Medical staff would not have been surprised if
the patient died during the current admission in 50 (16.8%) cases and in the
next 12 months in 123 (41.4%) cases.
Agreement between nursing staff, medical staff and the GSF regarding
the identification of patients with palliative care needs
shows the
level of agreement between medical staff, nursing staff and the GSF
regarding the identification of patients with palliative care needs.
Cohen’s kappa indicates a poor agreement between nursing staff and
the GSF (n = 473,
kappa = 0.22) in terms of identifying patients with
palliative care needs. Agreement between medical staff and the GSF was also
poor (n = 297,
kappa = 0.25). Agreement between medical staff and nursing
staff regarding which patients had palliative care needs was moderate
(n = 256,
kappa = 0.42).122
Results of patient self-report data
Self-report questionnaires were completed for all 185 patients identified
with palliative care needs according to GSF criteria. Questionnaires were
completed by patients in 162 cases (87.6%) and by consultees in 23 cases
(12.4%). The SPARC questionnaire provides a self-assessment of palliative
care needs and scores variables from 0 to 3; a score of 3 on any variable
indicates that the patient merits ‘immediate attention by the
attending clinician’.116 The SPARC questionnaire contains variables in six domains:
physical symptoms, psychological symptoms, religious and spiritual issues,
independence and activity, family and social issues and treatment issues.
The majority of patients (n = 154, 83.2%)
scored 3 on at least one variable in one of the six domains. Physical
symptoms were most troublesome with 74.6% of patients scoring 3 on one or
more variable in this domain. Patients also reported high levels of
psychological symptoms (43.2%) but fewer problems relating to the other
domains ().
Consensus between patients and medical staff (kappa = 0.20,
n = 107) and between patients and
nursing staff (kappa = 0.20,
n = 173) was poor regarding identification
of palliative care need when a SPARC score of 3 on one or more variable was
used as a proxy for self-assessed palliative care need.
Sheffield Profile for Assessment and Referral for Care questionnaire
responses for patients with palliative care needs according to GSF
criteria (n = 185).
Predictors of symptom burden amongst patients with palliative care
needs
Binary logistic regression was performed to assess whether known diagnostic
factors (GSF indicator, number of comorbidities) and/or demographic factors
(age, sex, living arrangements) were able to predict symptom burden as
measured by the SPARC questionnaire. Only 183 of the 185 patients were
included in these analyses as two patients had insufficient case note data.
As only two SPARC variables included substantial numbers of patients
(psychological and physical symptoms; see ), other SPARC variables were
removed as outcome measures from the logistic regression. The two outcome
variables indicating symptom burden (physical and psychological symptoms)
were entered as binary outcome measures indicating ‘high symptom
burden’ (score of 3 on one or more variable) compared with
‘low symptom burden’ (no scores of 3 on any variables) in
the logistic regression.
Each predictor variable was entered singly and not adjusted for other
variables. The results of these analyses are presented in . The results
indicate that patients diagnosed with heart disease were less likely than
those with other diagnoses to have a high physical symptom burden [odds
ratio (OR) 0.42]. The results also indicate that patients with dementia were
more likely to have a high physical symptom burden (OR 3.94) and a high
psychological symptom burden (OR 2.88). Female sex was a significant
predictor of psychological burden (OR 2.00), and having three or more
comorbidities was also a significant predictor of psychological burden (OR
3.97). Neither age nor living arrangements were significant predictors of
symptom burden.
The results of binary logistic regression analysis to predict
psychological and physical burden
(n = 183)
Older patients aged ≥ 85 years
A key focus of this study was the care provided to older people at the end of
life in hospital. Therefore, additional analyses were undertaken on patients
who participated in the survey who were aged ≥ 85 years. Of
the 654 patients who agreed to participate in the survey, 127 (19.4%) were
aged ≥ 85 years. After data cleaning, complete data sets
were available for 110 patients aged ≥ 85 years.
shows demographic and admission data for these 110 patients. The majority
were female and the median age was 89 years. Most of the patients lived
alone and all were of white ethnic origin. Reason for admission to hospital
was obtained from hospital case notes. The most common reason for admission
(25.5%) was general frailty (including falls, confusion or general
deterioration).
Demographic and admission information for patients aged
≥ 85 years
(n = 110)
The hospital case notes of the patients aged ≥ 85 years were
examined for evidence of palliative care need according to GSF prognostic
indicator criteria. Forty-four (40.0%) patients met one or more criteria for
palliative care need. Frailty (16.4%) and dementia (13.6%) were the most
common criteria met (). The majority of these patients met just one GSF
criterion (n = 30, 68.2%), with smaller
numbers meeting two (n = 11, 25.0%) or
three (n = 3, 6.8%) criteria.
Gold Standards Framework indicators met by patients aged
≥ 85 years
(n = 44).
Hospital case notes were also examined for indicators of a transition to a
palliative care approach. Half of the patients (50%) who met GSF criteria
for palliative care need met one or more indicator of transition to
palliative care. Half of the patients had a DNAR order in place; however,
there was documentation of discussion with the patient and/or family
regarding the DNAR order in only 6 cases (13.6%). Even amongst patients who
did not meet GSF criteria, a reasonable number (16.7%) had a DNAR order in
place. Only two patients of the 44 who met GSF criteria had been referred to
specialist palliative care services; however, one patient was referred to
specialist palliative care despite not meeting any GSF criteria. The only
patients for whom advanced care plans (1.5%) or use of syringe drivers
(3.0%) had been documented did not meet GSF criteria for palliative care
need.
Predictors of transitions to palliative care
Analyses relating to transitions to palliative care were undertaken on 183 of
the patients with palliative care needs according to GSF criteria (two
patients were excluded from the analyses because of missing data on key
transition variables). Of the 183 patients, 61 (33.3%) showed evidence of
transition to a palliative care approach by meeting one or more indicator of
adoption of a palliative care approach (). Of these, 43 patients met
just one indicator, 14 patients met two indicators and four patients met
three indicators.
Indicators of adoption of a palliative care approach for patients
with evidence of transition
(n = 61)
A logistic regression was performed to assess whether various diagnostic and
demographic factors were able to predict a transition to palliative care.
Data were recorded for five indicators of transition to palliative care. As
only one individual indicator of transition to palliative care included
substantial numbers of patients (DNAR; see ), a single binary outcome
indicating any evidence of a transition to palliative care compared with no
evidence was used for the logistic regression. Among the GSF diagnostic
indicators, no patients had motor neurone disease and the frequencies for
Parkinson’s disease and multiple sclerosis were very low; therefore,
these last two indicators were combined with ‘other life-limiting
conditions’ to form a combined indicator.
shows
that, in the unadjusted logistic regression, the significant predictors of a
transition to palliative care were the GSF indicators for cancer, heart
disease and stroke, together with age and living in a residential or nursing
care home. The table also shows the final multivariate predictive model,
which comprised three GSF diagnostic indicators (cancer, dementia, stroke)
together with age. This model correctly predicted 74% of outcomes and could
not be improved significantly by adding the other GSF indicators, number of
comorbidities, sex or living arrangements. In this model, the odds of a
transition to palliative care were multiplied by an estimated 5.1 for
patients with a cancer diagnosis, by 8.0 for a stroke diagnosis and by 2.6
for a dementia diagnosis. The odds were also increased by an estimated 3%
for every additional year of age (with no evidence of significant
non-linearity in this relationship).
Relationships between potential predictors and transition to
palliative care
Potentially avoidable admissions to hospital for patients with palliative
care needs
Appropriateness of admission assessments were undertaken for 208 patients;
this was the total number of patients who were identified with palliative
care needs according to GSF criteria and for whom complete case data were
available (no medical/nursing data or patient data were required for this
stage of the analysis; this accounts for the slightly larger sample size
than that described in Results of the case note review).
The admissions of 14 (6.7%) patients were classified by our two palliative
medicine consultants as potentially avoidable. Double coding of a random
sample of 15% of the notes indicated high levels of agreement between the
consultants, using the kappa measure of chance-corrected agreement
(kappa = 0.792,
n = 30).
Of the 14 patients whose admission was assessed as potentially avoidable,
seven were male and seven were female. The median age of the patients was 84
years (range 75–97 years). Cancer was the primary diagnosis for 6
out of the 14 patients; three patients had a primary diagnosis of stroke;
and the other patients had a primary diagnosis of encephalopathy, end-stage
renal failure, Alzheimer’s disease, general frailty or hypertension.
Half of the patients lived in nursing or residential care. Of the remaining
patients, 50% lived alone and 50% cohabited. Most patients
(n = 12) were admitted to hospital
‘out of hours’ (outside 0900–1700,
Monday–Friday). Reasons for admission to hospital were
confusion/general deterioration (n = 5),
symptom control (n = 3), fall
(n = 2), stroke
(n = 2), urinary tract infection
(n = 1) and intra-abdominal catastrophe
(n = 1).
The mean cost of a potentially avoidable admission was estimated to be
£2595 (range £451–£5363), based on the HRG
tariffs assigned to each admission by the Sheffield Teaching Hospitals
coding team. The distribution of costs is shown in . The costs fall
into three broad groups: < £1000
(n = 4), between £1000 and
£4000 (n = 5) and
> £4000 (n = 5). We
can therefore estimate that the total potentially avoidable hospital cost
from potentially avoidable admissions for the period of the survey was
£36,334. The mean length of stay, based on the HRG codes assigned to
these admissions, was 16.7 days.
Distribution of estimated costs of potentially avoidable
admissions.
The most commonly recommended alternative place of care for patients was a
nursing home (n = 10). In addition, it was
considered that three patients could have been appropriately cared for in a
hospice, and one in their own home. We used the same estimate of average
length of stay in these alternative places of care as that estimated for
length of hospital stay (16.7 days). The overall cost of alternative places
of care based on this same length of stay was estimated to be
£34,807 ().
Taking into account the avoided hospital costs and the cost of providing
support in alternative locations, the estimated economic impact is a
potential cost saving of £1527 across both hospitals for the period
of the survey. There were 1359 inpatients in the two hospitals over the
survey period. This accounted for 0.9% of the total admissions over the
year. Assuming that the proportion of potentially avoidable admissions
identified during the survey period is indicative of the number who would be
identified over the course of a year, the potential annual cost saving for
the two hospitals can be estimated at just under £180,000. If the
cost perspective is restricted to NHS and PSS costs, only the cost of
avoided hospital admissions remains unchanged; the cost of alternative care
provision falls to £15,141. Based on a NHS and PSS cost perspective,
then, the potential economic impact is predicted to be a cost saving of
£21,193 for the census period or £2.5M per annum for the two
hospitals.
Discussion
Extent of palliative care need
Our results indicate that, according to the GSF prognostic indicator guide,
over one-third of hospital inpatients (36.0%) meet the criteria for
palliative care need. This figure is substantially higher than other
estimates of palliative care need in the acute hospital population. A French
survey in 1999 reported that only 13% of total hospital beds were occupied
by palliative care patients.123 In a census undertaken in 2001, Gott and colleagues124 reported that 23% of
hospital inpatients were identified as having palliative care needs. A more
recent study in 201122
reported that just 9.4% of hospital patients in Belgium were identified as
having palliative care needs (although this study excluded intensive care
units and palliative care units). All of these studies used the subjective
judgement of generalist medical and nursing staff to identify patients with
palliative care need, rather than an objective measure based on diagnostic
criteria, as used in this study. Our results show that, when using a
systematic and objective measure, the percentage of patients with identified
needs is much higher and represents a substantial proportion of the
inpatient population.
Lack of consensus between medical staff, nursing staff and Gold Standards
Framework prognostic indicators
One of the most significant findings from this survey is the lack of
concordance between medical staff, nursing staff and GSF prognostic
indicators regarding the identification of patients with palliative care
needs. Although it must be acknowledged that medical and nursing staff were
using a different definition of palliative care need than that in the GSF,
the Canadian definition was selected on the basis that it is one of the
broadest and most inclusive definitions, and is not restricted to particular
diagnostic groups. Despite this, medical and nursing staff identified far
fewer patients with palliative care needs than the GSF (15.5% and 17.4%,
respectively, vs. 36.0%). Significantly, for the majority of patients who
met GSF criteria for palliative care need (65.9%), there was no evidence of
adoption of a palliative care approach.
Even amongst patients who were expected to die within 12 months, recognition
of palliative care need was inconsistent. Medical and nursing staff judged
that less than half of the patients they expected to die within 12 months
had palliative care needs. This is despite the ‘12 months’
question constituting a key component of GSF prognostication and being
advocated in policy such as in the End of Life Care Strategy for
England.5 Data
from the SPARC tool for self-assessment of palliative care need indicate
that, of the 185 patients identified with palliative care needs according to
GSF criteria, the majority (83.2%) had problems that ‘warranted
immediate attention by an attending clinician’. Despite this,
agreement between medical and nursing staff and patients was very poor
regarding which patients had palliative care needs.
The identification of patients with palliative care needs presents a
recognised challenge.2,89,114 Recent policy5 recommends that health professionals should be
trained to identify patients approaching the end of life and to recognise
when patients are dying. However, there is a lack of consensus regarding how
these patients should be identified and this has significant implications
for quality of patient care.79,83,125,126
The challenge in agreeing a consensus of definition and identification of
palliative care needs has additional implications for generalist palliative
care providers. Recent policy and research have sought to engage more
effectively with the generalist provider;125,126 however, our survey results show that many generalists are
struggling to identify patients who might benefit from palliative care
input. Generalist palliative care is increasingly central to hospital-based
palliative care provision. Therefore, it is crucial that generalists are
provided with opportunities for greater partnership working with specialist
palliative care colleagues. More generally, there is also a clear need for a
consensus of definition and for standardised validated criteria for the
identification of patients with palliative care needs.
Older people with palliative care needs
Data from the survey suggest that older people with frailty conditions
constitute a substantial proportion of hospital inpatients with palliative
care needs. Although specialist palliative care uptake is low amongst the
frail elderly, it is unclear whether a specialist palliative care framework
is the most appropriate model for this group. The care and services provided
to older people at the end of life may best be provided by generalists such
as geriatricians, as part of a comprehensive generalist-led palliative care
framework. Given recent concerns about the level of care provided to older
people within the acute hospital setting,127,128 priority should be given to considering ways of improving
the care that older people receive at the end of life. Improving generalist
palliative care to support older patients in the community, improving
recognition of palliative care needs amongst older frail patients and
implementing models of palliative care that are appropriate for older
patients at the end of life are key priorities that need to be
addressed.
Extent of transitions to palliative care
Although UK policy5 advocates an
early and phased transition to a palliative care approach for any patient
thought to be within the last 12 months of life, our findings indicate that
only one-third (33.3%) of patients with identified palliative care needs
showed evidence of any such transition. Moreover, the majority of these were
identified as having made a transition by virtue of having a DNAR order in
place and, although a DNAR order prevents cardiopulmonary resuscitation in
the event of cardiac and/or pulmonary arrest, it does not presuppose that
any other treatments or interventions that the patient receives will be
palliative in intent.
The odds of a transition to palliative care were multiplied by an estimated
5.1 for patients with a cancer diagnosis, by 8.0 for a stroke diagnosis and
by 2.6 for a dementia diagnosis. It is well recognised that patients with a
cancer diagnosis are more likely to receive specialist palliative care than
patients with non-malignant disease.5 Our results are therefore interesting in indicating that
patients with a diagnosis of stroke are most likely to have made a
transition to palliative care. A number of policy and educational factors
may have contributed to this finding. These include the Royal College of
Physicians’ guidelines for stroke129 and the DoH’s
Stroke-Specific Educational Framework,130 which underlines the
importance of assessment and management of end-of-life care. In the case of
the dementia finding it is possible to speculate on the impact of recent
recommendations that NHS hospital policy should apply the principles of
palliative care to patients with dementia.26,131
However, National Institute for Health and Care Excellence (NICE)/Social Care
Institute for Excellence guidelines26 identify that cardiopulmonary resuscitation may not be
effective in the case of cardiopulmonary arrest for people with severe
dementia. These data may therefore be an artefact of NICE guidance as
opposed to a genuine shift towards appropriate palliative transitions for
people with dementia.
Guidance from the British Medical Association, the Resuscitation Council (UK)
and the Royal College of Nursing132 indicates that, regardless of diagnosis, advance care
planning should occur alongside the implementation of a DNAR order. Our data
indicate that this is not the case in acute hospitals in England. Indeed, no
documented evidence of advance care planning was found for this sample of
183 patients with palliative care needs according to GSF criteria. Various
barriers to advance care planning have been identified in the
literature103,133 and advanced communications
skills training programmes such as Connected have been initiated across some
clinical settings (mandatory in oncology settings), including acute
hospitals, to address these. Our findings lend weight to the need to address
the gap between policy promotion of advance care planning and current
practice. They also indicate a need to ensure that any such discussions are
recorded in clinical notes, both to facilitate clinical decision-making and
continuity of care and to facilitate future audit and research in this
area.
There are no recognised definitions of what constitutes a ‘transition
to palliative care’, despite this concept being widely used in UK
policy.5 Capturing the
timing and presence of a shift in care regime of this nature is inherently
problematic, particularly given that there is evidence to suggest that
end-of-life care discussions between clinical teams and patients and their
families go unrecorded.134 In
addition, notions of phased transitions, incorporating elements of both
palliative care and curative care, can further complicate the identification
of this shift. In the absence of any formal criteria for identifying a
transition to palliative care, we defined a transition on the basis of
findings from our earlier qualitative research phase with health
professionals. However, it is recognised that the list of indicators used in
this study may not be comprehensive and may require validation, and this
represents a limitation of this study.
Patient symptom burden
A number of variables were shown to be significant predictors of both physical
symptom burden and psychological symptom burden. Patients with heart failure
were less likely to have high physical symptom burden, whereas those with
dementia were more likely to report high levels of both physical symptom burden
and psychological symptom burden. These data provide further evidence of
elevated levels of physical symptoms amongst dementia patients, requiring
interventions of a supportive and palliative nature, and add to an existing
literature.135
Female sex was shown to be a significant predictor of psychological burden.
Although consistent with primary care studies and other evidence on common
mental disorders and sex,136 this
finding is at odds with previous evidence in relation to psychological
difficulties at the end life, which would suggest that the prevalence amongst
male patients is higher.137 A
further predictor of psychological burden was the existence of three or more
comorbidities. The relationship between multiple conditions and poor
psychological status has been demonstrated in community studies.138 Studies focusing on those
patients eligible for palliative forms of care are limited to the relationship
between disease-specific conditions, comorbidities and psychological distress.
The development of new comorbidities is associated with readmission of
palliative care patients.139
Dementia is again implicated as a significant predictor of psychological burden
(OR 2.88) and this adds further credence to the use of palliative regimes in
providing care and treatment to address psychological needs for this vulnerable
patient group.140 Furthermore, it
has been demonstrated on numerous occasions that this group of patients faces a
number of challenges in accessing palliative and supportive regimes of care and
that these challenges are both cultural and organisational.141,142 These data suggest that, in addressing the complex physical and
psychological needs of an ageing population, care teams will be increasingly
required to adapt integrated palliative approaches to a cognitively frail
hospital population.
Economic impact of potentially avoidable hospital admissions
The total cost of hospital admissions in the last year of life for adults
admitted with a primary diagnosis indicating palliative care need has been
estimated to be in the region of £1.3B.117 A lack of timely access to services in the
community may result in people with palliative care needs being unnecessarily
admitted to hospital.121 It has
been suggested that, by improving or expanding community services to allow more
people to be cared for and to die at home, there is the potential for a
proportion of hospital costs to be avoided.5 In 2010, the DoH reported that high-quality community-based
services cost no more, and can cost less, than hospital-based care.77 However, a review126 conducted as part of this study
found that the evidence base from the UK was presently too limited to support
the case for offsetting the additional costs of providing high-quality community
support through a reduction in hospital admissions for this patient group.
Only 7% of hospital admissions of patients identified as having palliative care
needs were classified as potentially avoidable. The potential cost saving of
avoiding these admissions and supporting patients in alternative places of care
in the two locations under study was estimated to be just under £180,000
per annum. The proportion of admissions identified as potentially avoidable was
low relative to the proportions reported in two recent UK studies.79,83 This difference is most likely attributable to
the fact that, in our study, not only were data relating to admission and
patient characteristics such as comorbidities, age and living arrangements
considered, we also took into account the availability of local services. If
local services were known to be inadequate to support the patient in the
community, then the admission was considered appropriate. In contrast, Abel and
colleagues79 based their
retrospective case note review on the assumption that the End of Life
Care Strategy for England5 had been fully implemented and that local services were always
available and had capacity. Employing this method, these researchers classified
one-third of all deaths in hospital as avoidable, suggesting that they could
have occurred at home. The second study was conducted by the Balance of Care
Group83 at a hospital in
Sheffield and used a similar ‘blue sky’ approach whereby
researchers assumed that alternative community facilities were always available
and had capacity. In this similarly retrospective study, 40% of admissions for
patients surveyed who died in hospital were considered to be inappropriate. The
fact that both of these studies adopted a ‘blue sky’ approach to
the availability and capacity of alternative places of care means that both face
the issue of feasibility. Equally, the fact that both studies undertook a
retrospective analysis using data collected after death means that their
researchers were judging whether the admission was avoidable given the
‘full story’, whereas our clinicians were exercising their
judgement from information available at the point of admission.
Although the 7% of hospital admissions seen as avoidable in our study may appear
low, it does reflect the lack of appropriate alternative services to hospital
admission in the two study sites for palliative care patients.
Our exploratory analysis of the economic data estimates a mean cost per avoided
admission of £2595. Our estimate is lower than that reported by both
Abel and colleagues79 and the
Balance of Care Group83 and may
have resulted in an underestimation of the cost savings generated by avoiding
admissions. The difference between the estimates cited in these two studies and
that calculated in our study may be attributable, at least in part, to
limitations in our costing data. HRG codes were assigned to our patients using a
more limited data set than would ordinarily be available. Also, because we did
not have length-of-stay data, we were unable to include excess bed-day payments,
which Abel and colleagues79 did
include in their calculations.
Differences in the patient mix under consideration in each of the studies may
also contribute to cost differentials. The adoption of a ‘blue
sky’ scenario in the studies by Abel and colleagues79 and the Balance of Care Group83 would have resulted in a greater
proportion of patients with more complex and, crucially, more expensive needs
having their admissions categorised as avoidable.
Enabling more people with palliative care needs to be cared for and to die at
home or in other community settings may result in a proportion of hospital costs
being avoided. However, a clearer understanding of the proportion of hospital
admissions that could be avoided and the economic impact of avoiding such
admissions is needed. Further research is required to explore the relationship
between admissions deemed avoidable given the situation at the point of
admission and those deemed potentially avoidable in a hypothetical ‘blue
sky’ scenario – and to demonstrate the feasibility of avoiding
such admissions in clinical practice. In addition, more robust estimates of the
cost of supporting patients in the community are also needed, in terms of both
the proportion of patients requiring different types of care and the costs
associated with such care.
Limitations
Although this study provides important evidence relating to palliative care in
acute hospitals, certain limitations must be acknowledged. Although a wide range
of experienced health professionals from primary and secondary care, and from
specialist and generalist palliative care backgrounds, were asked during the
focus group phase (see Chapter
4) to identify indicators of a palliative care transition,
we accept that the list of indicators used in this study may not be
comprehensive and may require validation. The GSF was developed as a tool for
use in primary care and has to date received no formal validation in the
hospital setting. Criticisms of the GSF include that it is a poor predictor of
mortality; therefore, its use as a tool for identifying patients with palliative
care needs in hospital should be further explored. However, amongst the 185
patients identified as meeting at least one GSF criteria, the subjective scores
on the SPARC questionnaire suggest that this group had a high level of
palliative care need. Validation of the GSF is now warranted by comparing GSF
criteria, self-assessment of palliative care need and subsequent survival.
A further limitation that must be acknowledged is the relatively low patient
response rate: only 37.8% of the total inpatient population agreed to
participate and provided a complete data set. There is also a probable response
bias as a result of the self-selected nature of the patient sample. However, as
the overwhelming reason given for non-participation was that patients felt too
ill, we believe that our sample constituted the ‘most well’ of
the inpatient population. As such, the findings presented here may underestimate
the true incidence of palliative care need in the acute hospital setting.
In 23 cases, consultees completed questionnaires on behalf of patients who lacked
capacity to consent, and responses given by consultees may not be accurate.
Therefore, caution is required in interpreting findings from the questionnaire
responses and further research should seek to compare self-assessment and
consultee assessment measures to explore consensus.
