A casemix classification for those receiving specialist palliative care during their last year of life across England: the C-CHANGE research programme

The clinical challenge

People in the last year of life often suffer complex and multiple symptoms and distress because of their illness or impending death. 1–3 Functional status often deteriorates as the illness progresses, leading to increased dependency and greater care needs. 4 The families of those affected experience their own fears and losses. 5

This occurs across a wide range of different life-limiting illnesses. Symptoms, distress, functional trajectories and family support needs have been most widely studied in cancer, but other life-limiting conditions (e.g. advanced cardiac and respiratory diseases, progressive neurological conditions, dementia, end-stage kidney disease and advanced liver disease) bring equally challenging symptoms and support needs. 6,7 Family support is particularly important if bereavement is to be well negotiated8 and future psychological health maintained. 9

As the population age distribution changes, these needs are also extending over longer time periods as the length of time living with advanced illness increases. 10

Specialist palliative care: provision and inequities

Palliative care has developed to meet these complex needs of patients and families facing advanced progressive illness. It addresses physical and psychological symptoms and gives social, practical and spiritual support in the last months of life. 11 The UK ranked first in the 2015 Quality of Death Index, a measure of the quality of dying across 80 countries,12 with UK hospices leading the way internationally, providing a world-leading model of excellence in palliative care.

However, the work of UK hospices – which deliver both inpatient and home-based specialist palliative care – has often come from grass-roots and local initiatives, meaning that there are marked geographical inequities in provision across England. Much of hospice and specialist palliative care is funded from the charitable sector; in 2018, charities contributed £3 for every £1 from the NHS. 13 The balance between NHS and charitable funding varies markedly around the country;13 this and other factors has led to major inequities in palliative care provision.

A UK government-commissioned review into specialist palliative care14 found high levels of inequity in the provision of palliative and end-of-life care. There was diversity in the type and range of providers in any one geographical area15 and inequity in both the amount of funding (with money spent by the NHS in 2010 ranging from £186 to £6213 per person who died) and the source of funding (with the proportion funding coming from the NHS varying from 0% to 62%). 14 This could not be accounted for by variations in the populations served. 14

Even more marked variation in the commissioning of palliative services across England has been demonstrated more recently. 16 A ‘north–south’ divide has been identified,17 with substantial differences in the average time from referral to death between providers in the north and south of England. This is important: the benefits of specialist palliative care are notably greater if delivered earlier in the illness trajectory. 18

Older people and those with non-cancer diagnoses are also less likely to access specialist palliative care,19,20 often resulting in a poor match between individual needs, the resources provided to meet those needs and the health outcomes achieved.

Substantial differences in the approach to commissioning services suggests provision rather than population demographics account for a large component of the variation. 16

The research gap

One major challenge in improving our ability to better match resources to needs is the lack of research into what accounts for these apparent inequities in provision and – in particular – a lack of a standardised way to assess and monitor palliative care needs. With growing constraints on resources, it is becoming imperative for individual-level needs to be mapped accurately to reduce inequities, match resources to needs and improve the quality, consistency of provision, and efficiency of palliative care. This has been endorsed as a high priority nationally,14 but there is very limited evidence available to inform the best ways to match individual-level needs to resources. Most importantly, there is no ‘standard’ way to assess, capture and report the palliative care needs of individual people with advanced illness. This programme of research directly addresses this research gap.

What we already know

Illness in the last year of life places major resource burden on the NHS. Up to 20% of healthcare expenditure is spent on the last year of life. 21 Total NHS expenditure in 2017/18 was around £122B in England22 – roughly £2200 per person – indicating that around £25B per year is spent on health care in the last year of life. Much of this is on acute care, but it is estimated that palliative care is needed for about 75% of all those approaching death. 23

The amount of health and social care resources spent on those in the last year of their life is increasing and will increase further in coming years. Ageing populations, greater co-morbidity and longer chronic disease trajectories are all increasing this demand. 24,25 However, this is not a simple relationship; commonly used approximations of health, such as age or mortality, are not enough to capture the complex dynamics in healthcare requirements. 26 Population ageing particularly increases expenditures on acute and long-term care. 27 In this context, it is increasingly important to recognise palliative care needs and ensure they are effectively addressed with the best possible use of scarce resources.

Palliative care needs

Assessment of patients’ needs plays an important role in improving outcomes in palliative care. Patient-reported outcomes – where ‘outcome’ here is defined as a ‘change in current or future health status following intervention’28 – are a way of measuring the changes in patients’ health that they themselves perceive over time. These outcomes allow for assessment of intervention effectiveness. In terms of addressing needs, there is increasing evidence that home, hospital and inpatient specialist palliative care is effective and significantly improves patient outcomes, particularly for cancer patients with reduced pain and other symptoms, reduced anxiety, and reduced hospital admissions. 29 Similar evidence is emerging for patients with non-malignant disease. 30–32

Further systematic reviews,29,33,34 including meta-analyses and Cochrane reviews,35,36 and randomised controlled trials of specialist palliative care services30,37 provide consistent evidence of the effectiveness of palliative care services in improving symptoms, reducing family burden, improving satisfaction with care and preventing depression. There is wide acknowledgement that palliative care should be available to all, based on need not diagnosis and extending across settings,38,39 and these principles underpin the NHS End of Life Care Strategy. 40

Given the complexity of problems experienced in the last year of life, it is not surprising that a high proportion of healthcare resources is spent during this time: as much as 20% of all healthcare expenditure. 21 However, some of this may be poorly spent: hospital admissions may be prevented by better anticipatory symptom control, better family support and earlier facilitation of advance discussions about patient and family preferences for treatment and care. 41,42 Robust studies demonstrate improved patient outcomes and reduced costs when palliative care is provided early30,37,43 and a systematic review of trials of palliative care interventions shows greatest positive impact on quality of life when these interventions are provided early. 18

Casemix classifications: a potential way forward

Within health care, there is increasing use of casemix classifications to help gain a better understanding of how healthcare resources might be allocated in an equitable yet efficient way. Casemix classifications are based on patient-level criteria, which allow the grouping of patients into classes in terms of the resources needed to meet their needs. 44 Casemix is defined as ‘a means to classify patients into groups in order to provide a useful measure to make meaningful performance comparisons, to cost healthcare, or to fund it’ (information from NHS Digital, licenced under the current version of the Open Government Licence). 45 Casemix classifications are defined as ‘a system assigning each person [patient] into a hierarchical system of groups or classes, according to various individual casemix criteria. Each group or class is associated with higher or lower resource requirement to meet needs.’46

The USA developed such casemix classifications, including diagnosis-related groups (DRGs)47 and these have been used to develop prospective payment systems globally. 48 DRGs are a useful classification of healthcare needs driven by the diagnosis, but have been shown to be inappropriate for some areas of health care, such as mental health,49 primary care50 and palliative care. 46,51

Palliative care needs are not diagnosis-driven, but instead by factors such as functional status, physical symptoms and emotional burden. 51 Palliative care needs a consistent method of classifying types of patients with complexity of needs, treatment and costs, using casemix criteria. 52–54 Healthcare Resource Groups (HRGs) underpin the main casemix classification adopted in England. 44

It is necessary therefore to identify those with more complex palliative needs and requiring more resources. 55 An Australian casemix classification for palliative care was developed in 1997, empirically tested and progressively refined over time. 52–54 The Australian casemix classification consists of classes defined by five criteria most strongly predictive of resource use: Phase of Illness, problem severity, functional status and dependency, age and model of care. 55 Full class definition and categorizations are available at https://ahsri.uow.edu.au/pcoc/index.html (accessed 12 December 2021). Its implementation proved it was possible to consistently and routinely collect casemix data nationally;56 this has enabled consistent casemix adjustment in outcome measurement, with year-on-year improvement in outcomes at a national level and a funding model which matches patient’s needs. 51,57 Palliative care funding pilots in the UK have also suggested casemix data may be useful for these purposes. 58 However, it is unclear whether or not, and how, any existing palliative care classification can be easily applied to the UK to address unmet needs and reduce inequities.

What we do not know

Although casemix classifications have been widely used to manage resources across health care, they have rarely been applied to palliative care. Existing classifications (i.e. DRGs and HRGs) are based on diagnoses, but for patients receiving palliative care the priority is not on diagnosis but enhancement of well-being and quality of life, and the maintenance or maximisation of current health status in the face of advanced incurable illness. 52 A palliative care casemix classification must reflect these different goals. 53,54,57

Australia is the only country to have developed such a casemix classification for palliative care. 57 Diagnosis and procedure criteria were found to be ineffective in classifying the complexity of needs for those receiving palliative care. 53 Instead, palliative Phase of Illness and ‘problem severity’ were better indicators of increased complexity and consequent greater resource use. 56,59 Despite these advances, key questions remain unanswered:

1. How can the wide range of patient needs in palliative care provision best be classified across conditions and settings, so that services can be resourced to meet individual needs and deliver best outcomes? Without such a classification, it is difficult to ensure that sufficient resources can be matched to the right patient at the right time.

2. Within such a classification, how is complexity best understood/captured so that commissioners can commission, and providers can deliver, the optimal mix of services?

3. What are the best criteria for palliative care casemix? It is internationally acknowledged that diagnosis and procedures are not useful criteria in palliative care casemix categorisation53,57 and UK specialist palliative care providers seek a model more attuned to patient needs across wide-ranging levels of complexity.

4. What are the most useful patient-level data in palliative care? There is a lack of specific patient-level data in palliative care with which to model casemix and resource use. 14 This is especially true in community settings, where provision cuts across the NHS and voluntary sectors. 60,61

5. What evidence do we have to build on? Some work has been undertaken in recent palliative care funding pilots,58 but no evidence has been published from these to help determine which casemix criteria are optimal for use in England.

Rationale for our approach

This programme responds to these challenges by developing a palliative care casemix classification across conditions and settings. Recent UK initiatives to improve the provision of high-quality palliative care40,62 have not addressed the challenges of classifying needs and costs.

Worldwide, different palliative care funding models exist,63 but a formal palliative care casemix classification has been developed only in Australia. This Australian casemix classification incorporates functional status/dependency with problem severity and palliative Phase of Illness,57 and adopts a blended funding model. 64 There is no existing palliative care classification easily transferable to the UK, but the Australian model is a valuable starting point.

This programme leads directly to patient benefit through the improved matching of resources to needs at the individual patient-level, with corresponding outcome measurement, along with more effective and cost-effective use of resources through the following work:

1. the development of better patient-centred outcome measures that are relevant and meaningful for patients with advanced illness and their families

2. the development of a ‘standard’ way to assess, capture, and report the palliative care needs of individual people with advanced illness

3. the development of a casemix classification – casemix ‘classes’ based on individual-level criteria – which can predict the resources needed to address the symptoms/concerns of people with advanced illness needing specialist palliative care.

This approach has been directly recommended by the Palliative Care Funding Review,14 and we have liaised closely with NHS England in progressing this work.

Aims

The aims of the C-CHANGE programme were to report the costs of specialist palliative care; develop and test a person-centred, nationally applicable casemix classification for adult specialist palliative care provision in England; accurately capture the complex needs of patients with advanced disease in last year of life; better quantify those needs; and support more equitable allocation of resources to meet them.

The programme also aimed to identify ways to measure the improvements in health status and well-being which patients and families experience following specialist palliative care so that the casemix classification could be developed, but also so that quality and effectiveness of services can be more readily demonstrated to patients, families, commissioners and services.

Objectives

The C-CHANGE programme had five objectives:

1. to refine, validate or test new and existing person-centred outcome measures to assess the main health status and symptoms/concerns of, and services received by, patients and families receiving specialist palliative care

2. to utilise the perspectives of key stakeholders (i.e. patients, families, professional caregivers, commissioners and policy-makers) on complexity in palliative care to inform subsequent casemix development

3. to understand the criteria which distinguish different models of palliative care to help inform how a casemix classification and per-patient funding models can best be utilised across different models of specialist palliative care

4. to develop a person-centred palliative care casemix classification, based on individual patient and family needs and costs of care, for adults with both cancer and non-cancer conditions in the last year of life

5. to test this person-centred palliative care casemix classification in terms of its ability to predict resource use in last year of life and to better understand transitions between services in order to improve care.

Validation of the Integrated Palliative care Outcome Scale– cognitive testing

This work has also been published in Schildmann et al. 72 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Validation of the Integrated Palliative care Outcome Scale: full validation study

This work has also been published in Murtagh et al. 73

To assess how palliative Phase of Illness related to the other measures

This work has also been published in Mather et al. 74

A patient experience measure: Views on Care

This work has also been published in Pinto et al. 75

Aim

Our aim was to understand patient experiences of uncertainty in advanced illness and develop a typology of patients’ responses and preferences to inform subsequent research in this programme.

Design

We performed a secondary analysis of qualitative interview transcripts. 78 Studies were assessed for inclusion and interviews were sampled using maximum variation sampling. Analysis used a thematic approach with 10% of coding cross-checked to enhance reliability. Qualitative interviews from six studies were analysed, comprising patients with advanced heart failure, end-stage chronic obstructive pulmonary disease, end-stage renal disease, advanced cancer and advanced liver failure. 78

Results

A total of 30 transcripts were analysed. The median patient age was 75 years (range 43–95 years) and 12 patients were women. The impact of uncertainty was frequently discussed: the main related themes were engagement with illness, information preferences, patient priorities and the period of time that patients focused their attention on (temporal focus). A typology of patient responses to uncertainty was developed from these themes (see Figure 2). 78

FIGURE 2.

Two different examples of patient responses to uncertainty, dependent on level of engagement, information preferences and temporal focus. Reproduced with permission from Etkind et al. 78 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY-NC 3.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by-nc/3.0. The figure includes minor additions and formatting changes to the original figure.

Copyright © 2023 Murtagh et al.

Design

We designed a qualitative study using semistructured interviews with key stakeholders in specialist palliative care.

Recruitment and consent

We undertook audio-recorded, individual interviews with various stakeholders across specialist palliative care. Participants were either recruited from one of the C-CHANGE participating sites for workstreams 3 and 4 or – for policy and national leads – sought out by the wider research team at the Cicely Saunders Institute (the leading UK research institute for palliative care). Within this frame, participants were sampled purposively, by personal and/or professional background, geographical location and experiences of settings of care (hospital, hospice and community).

Data collection

A topic guide was developed from a review of evidence on complexity, potential criteria for casemix already used46 or proposed,58 existing casemix classifications in palliative care57 and predictors of resource use in the last year of life. It was refined by our PPI group, by the research team, and through discussion with and feedback from the Programme Steering Committee.

Face-to-face, semistructured interviews were conducted by C-CHANGE researchers in the participant’s preferred setting. To increase the credibility of the data, interviewers summarised the interview back to each respondent, to allow the participant to verify the data and clarify any misconceptions or add additional information. All interviews were digitally audio-recorded, anonymised and transcribed verbatim to ensure confidentiality.

Analysis

Interviews were analysed independently by two researchers from the C-CHANGE team using the five analytical steps of framework analysis: (1) familiarisation, (2) identifying a thematic framework, (3) indexing, (4) charting and (5) mapping and interpretation. Framework analysis was considered the optimal approach to allow both an inductive and deductive approach, which would facilitate comparison across stakeholder groups and support the service delivery and policy focus of this research. Emerging themes were discussed with the whole C-CHANGE research team to improve the confirmability and dependability of the findings. Charts were created for each theme, grouped by stakeholder type, and were used to explore stakeholder assonance and dissonance among perspectives on each theme and subtheme. Analysis was managed using NVivo version 10 (QSR International, Warrington, UK). The framework was presented to the PPI advisory group, the Project Steering Committee and other qualitative research experts to refine and improve the presentation of the developed framework.

Ethics approval

Ethics approval was gained from the King’s College London REC (BDM/14/15–2).

Design

We developed a mixed-methods study with (1) semistructured interviews to identify criteria for models of care, (2) a two-round Delphi study to rank/refine these criteria and (3) structured interviews to test the acceptability and feasibility of these criteria. 81

Semistructured interviews

A rapid scoping review was conducted to identify literature related to models of palliative care. Original papers and reviews were examined for possible criteria which could help define models of specialist palliative care and a topic guide was created covering the 28 preliminary criteria identified from this literature. Semistructured interviews using a pre-specified topic guide were conducted with a range of palliative care service leads across organisations. 81

Delphi study

We selected Delphi survey methods for this second stage because it enabled us to present potential criteria derived from the semistructured interviews to all respondents, allowed them time to absorb this complex information at their own pace and enabled us to sample a wide range of views in a way that gave all opinions equal weight. A two-round Delphi survey of UK clinical, policy or PPI leads were invited from the Outcome Assessment and Complexity Collaborative network (a multidisciplinary network of professionals engaged in the implementation of outcome measures in specialist palliative care in England), and the national C-CHANGE sites. Participants were advised that we were aiming to establish a list of key criteria to describe and compare models of care. The Delphi survey was conducted to refine the criteria from the semistructured interviews, identify any additional criteria, achieve consensus on how each criterion was defined and rank the criteria in terms of importance. 81 CREDES (Conducting and REporting DElphi Studies in palliative care) guidelines were followed. 82

An online survey was developed using Bristol Online Survey (BOS) v1.0 Bristol, UK. 83 The survey was piloted for face validity prior to going live.

In the first Delphi round, panel members were presented with a list of 34 criteria. Participants were asked to state whether or not they agreed with the inclusion of each criterion as an important criterion for describing and comparing models of specialist palliative care (answering ‘yes’, ‘no’ or ‘don’t know’) and their reasons for this. They were also asked to comment on the phrasing and clarity of the criterion, as well as the answer options listed. Finally, participants were asked to suggest any additional criteria they thought should be included. 81

Each criterion was retained if at least 75% participants answered ‘yes’. Free-text comments were analysed using content analysis and used to refine and expand the set of criteria.

In round 2 of the Delphi process, participants received anonymised feedback from round 1 and the amended list of criteria for further refinement and ranking. Participants were asked to rate the importance of each criterion for characterising and comparing different models of care on a five-point Likert scale (1 = not at all important; 2 = not very important; 3 = important; 4 = very important; 5 = extremely important). In addition to the rating scales, participants were also given the opportunity to add additional free-text comments to help refine criteria and answer options. 81

Responses were analysed to capture both central tendency (median rating) and dispersion [interquartile range (IQR)]. Consensus was deemed to have been reached for criteria that received aggregated responses with an IQR of ≤ 1 and a median of 4 or 5. Both methods are considered to offer robust measurements for Delphi surveys. 84,85 Criteria reaching this consensus were then included in the final set.

Ranking responses were collated and analysed using IBM SPSS Statistics version 22. Free-text responses underwent content analysis and were used to refine the criteria and response options.

Structured interviews to test for acceptability and feasibility of the criteria

The criteria developed from the Delphi survey component were then tested with clinical leads from three different specialist palliative care settings (hospice inpatient care, hospital advisory teams and community-based care) using structured interviews. Participants consented to be interviewed and audio-recorded. The data from these interviews were classified according to the criteria used, and entered into Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) to identify whether or not the criteria could discriminate between services. 81

Ethics approval

Prior ethics approval for all three components was gained from the King’s College London REC (LRS-15/16–2449).

Semistructured interviews

Semistructured interviews were conducted with 14 service leads from eight organisations, discussing 12 settings of care (five hospice inpatient units, two hospital advisory teams and five community teams).

An early finding was that the clinical leads struggled to know at which level within the organisation to describe their models of care. 81 It was often confusing when an organisation covered multiple settings of care (i.e. hospice inpatient, community, hospital inpatient and day services) and also provided multiple services within each setting, which often overlapped. For example, a hospice may have inpatient hospice, home care and ambulatory settings. Within any one of these settings, multiple services or teams were often operating. Within the day services there may be a physiotherapy clinic, a lymphoedema service and a day service, all operating with different models of care.

After all interviews were completed, out of the 28 criteria in the topic guide, 11 were not reported as useful and were removed; 17 criteria were refined; and a further 17 criteria were created. This resulted in 34 criteria to take forward to the Delphi survey. 81

Delphi survey

A total of 190 participants were invited to take part in the Delphi survey. Of the 190 clinical, policy and PPI leads contacted, 54 agreed to participate (response rate 28.4%).

Structured interviews

Interviews were conducted with 21 service leads from 19 different services (six hospice inpatient, four hospital advisory and nine community settings). The responses to each criterion were compared to see if the criteria could distinguish and discriminate effectively between services. A further four criteria relating to context were also added (see Box 1); these were reported by the clinical leads as providing important context for the practical application of the criteria. These four contextual criteria were the purpose of the team, who funds/manages the team, the ability to self-refer and the discharging of patients. 81

Design

We designed a multicentre prospective cohort study, following patients during episodes of specialist palliative care and reported according to the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) statement. 86

Definitions

We defined an episode of care as a ‘period of contact between a patient and palliative care service provider or team of providers that occurs in one setting’. 87 A new referral into palliative care or change of setting (e.g. re-location into an inpatient setting from home/care home, or vice versa) signalled the start of an episode of palliative care. Discharge from palliative care, change of setting or death signalled the end of an episode of care. We defined the setting of care as one of three types: hospital advisory (hospital-based specialist palliative care teams providing an advisory or consultant service), inpatient hospice (where patients are admitted to a hospice or specialist palliative care unit for an overnight stay of one or more days) or community-based specialist palliative care (where the patient receives care in their usual place of residence, either at home or in a care home).

Population and settings

Patients were recruited from 14 organisations providing specialist palliative care services in England: four hospital advisory services, five inpatient hospice services and seven community-based services (some organisations provided more than one setting of care). Sites were selected to ensure our study sample was representative of UK palliative care patients and services in terms of participant age, ethnic background, service size, proportion of cancer/non-cancer patients and urban–rural balance. We included more community services than hospital or inpatient hospital services.

Inclusion/exclusion criteria

We included consecutive adult patients (≥ 18 years) receiving specialist palliative care at all participating sites. Exclusion criteria comprised patients aged < 18 years, those who declined participation and/or those who previously expressed a wish not to participate in research.

Data collection and primary outcome

Data were collected from clinicians between July 2015 and October 2016, with completion of follow-up at the end of January 2017; no data were collected directly from patients. Collected data included demographic and clinical variables, episode start and end dates, potential casemix variables, and data on patient-level and other costs of providing specialist palliative care. All participants received the usual specialist palliative care at that site, including a multidisciplinary team with specialist training delivering holistic care focused on physical and psychological symptom management, social/family support, planning ahead around priorities and preferences, and care into the dying phase including post-death care of the family, where relevant.

Potential casemix variables were selected based on (1) being patient-level attributes and (2) existing evidence of association with casemix/complexity. 88 The key casemix variables included were age, sex, ethnicity, living circumstances, need for interpreter, primary diagnosis, palliative Phase of Illness, functional status, dependency and symptoms/problem severity.

Details of the measures used are presented in Appendix 1.

Our primary outcome was the cost of specialist palliative care per day. We adopted a (palliative) provider perspective for costs. The costs of acute hospital care, primary care, generic end-of-life care (i.e. provided by non-specialist teams), and informal care costs were excluded, not because these are unimportant, but because we sought casemix criteria relevant to specialist palliative care. A broader perspective on costs is planned for future work.

Palliative Phase of Illness was assessed daily for people receiving inpatient (hospital or inpatient unit) care and at every contact for those receiving community-based care. Each change in Phase of Illness (or end of episode) triggered the collection of the AKPS, IPOS or Palliative Care Problem Severity Score (PCPSS), and the Barthel Index. All staff involved in patient care recorded the time spent delivering care to participants at the patient level using the staff activity matrix.

We collected data from participating sites on the costs of delivering their services and patient-level resource use data from the staff activity matrix to derive actual patient-level costs according to a standard costing methodology based on current NHS costing principles. 89 Note that costs captured for the hospital advisory and community-based settings represented the additional or ‘top-up’ costs for adding palliative care support to the hospital or community setting (and thus are reasonably compared). In contrast, the costs for inpatient hospices represented all the costs of inpatient care (and so are more reasonably compared with the costs of acute hospital admission). The full costing methodology (how costs were collected, classified and compiled) is available from the corresponding author on request.

Sample size

Based on standard recommendations for fitting multivariate models, a minimum of 50 + 8 × m cases for testing multiple correlation (where m is the number of predictors) are required to test the null hypothesis that the population multiple correlation equals zero with a power of 80%, α = 5% and a medium effect size for the regression analysis (R² = 0.13). 90,91 The unit of analysis was episodes within sites; therefore, 10 predictors required 130 episodes per site. Allowing an additional 15% for episodes with missing data and 20% for cost outliers, we estimated that a target of 2674 episodes of care (191 episodes × 14 sites) was required.

Data handling

Data were collected prospectively at each site, recorded on an electronic database, cleaned and checked. Checked data were transferred to statistical software [Stata^® standard edition (SE) V.12 (StataCorp LP, College Station, TX, USA) and MATLAB^® 8.2 (The MathWorks, Inc., Natick, MA, USA)] for analysis.

Analysis

An exploratory data analysis was undertaken first, examining variables of interest one at a time. Descriptive statistics (means, standard deviations, medians, ranges and correlations) were calculated. Comparative box plots were constructed to investigate the differences between sites, episodes and phases. Cost of care was used as the response variable, measured as the total cost per episode.

The aim of the analysis was to form distinct groups within the data, such that patients within each group were similar to each other, but different from patients in the other groups.

To discover which baseline casemix variables could best predict the cost of a particular episode of care, the following nine steps were then undertaken:

1. We removed incomplete episodes, retaining only the complete episodes of care.

2. Following a previously adopted approach,46 high- and low-cost outliers were identified and removed using a trimming algorithm based on the IQR with the upper trim point at Q3 + 1.5 IQR and the lower trim point at Q1 – 1.5 IQR (where Q1 is first quartile and Q3 is third quartile). 46 The trimming algorithm was applied to each setting separately.

3. We examined the distribution of costs of specialist palliative care, by setting.

4. Then, following the same approach as the development of the Australian casemix classification,88 we developed and validated a cost-predictive model using classification and regression tree (CART) analysis, which constructs decision rules in a hierarchical manner to form a branching classification.

5. We used CART analysis to enable the more complex interactions between the predictor variables (both categorical and continuous) to be explored. CART has the advantage of being non-parametric and is not significantly impacted by outliers in the input variables. It enables the use of each variable more than once, if required for the optimal regression tree.

6. Explanatory variables were compared to find the one which could best split the data into two homogeneous groups that were as different from one another as possible. These two groups would then be further split, using the same or another explanatory variable. Successive binary splits were performed on the data until there was no further improvement to be made and the best possible classification solution was reached.

7. The best CART was deemed to be that which accounted for the largest proportion of variation in the cost of care (the response variable). The criterion used to compare the different ‘trees’ was the proportion of the variance of the response variable that could be explained by the selected groups.

8. Costs were log-transformed for better modelling and back-transformed for providing mean costs per class or at each terminal node. Decisions about rules for splitting were informed by clinical utility (for instance, allowing branches that made clinical – as well as statistical – sense), as well as statistical performance as outlined in 7. We selected a maximum of four branches and a minimum of 30 cases per branch, for reasons of clinical utility. 10–fold cross-validation was used to prevent overfitting of the developed classification. No recalibration was undertaken. The analysis was done in R 3.5 (The R Foundation for Statistical Computing, Vienna, Austria) using the rpart, caret and (for bagging the trees) RWeka packages.

9. For each setting, we reported the variance explained by the CART model and the root-mean-squared error (RMSE) (i.e. the square root of the variance of the residuals). The RMSE indicates the absolute (rather than relative, as with R²) fit of the model to the data.

Ethics approval

The trial registration number is ISRCTN90752212. Ethics approval was received from the Camberwell St Giles National Research Ethics Service REC on 2 July 2015 [REC Reference: 15/LO/0887, Integrated Research Application System (IRAS) Project ID:172938].

Subject characteristics

A total of 2469 patients were recruited, providing data on 2968 complete episodes of specialist palliative care (12 incomplete episodes were removed prior to analysis); 2087 participants contributed one episode of care, 283 participants contributed two episodes and 99 contributed three or more episodes. Demographic and clinical characteristics for the 2469 participants are reported in Table 1. No participants withdrew after recruitment. Further details are available in Appendix 2.

Episode characteristics

Details of the 2968 complete episodes of care and related casemix variables are reported in Table 2.

Outliers

Using the trimming algorithm described in Methods, Analysis, 123 (16.0%) hospital advisory episodes, 185 (24.2%) inpatient hospice episodes and 305 (21.2%) community episodes were removed to ensure that the principal cost and classification reporting was not based on outliers (a common challenge in costing studies).

Costs of specialist palliative care

The distribution of the total cost of specialist palliative care episodes, derived from the trimmed data set, is shown in Table 3; costs are shown (1) by day, (2) by day, broken down by Phase of Illness, and (3) by episode of care, together with details of length of episodes.

Classification and regression tree analysis

Figures 4–6 show the CARTs for each setting. The per cent variance explained (and RMSE) were 20% (RMSE = 0.30), 51% (RMSE = 0.51) and 27% (RMSE = 0.36), for hospital advisory, inpatient hospice and community episodes, respectively.

FIGURE 4.

Classification tree of casemix criteria for hospital advisory episodes of specialist palliative care (costs per day reported for each class).

FIGURE 5.

Classification tree of casemix criteria for inpatient hospice episodes of specialist palliative care (costs per day reported for each class).

FIGURE 6.

Classification tree of casemix criteria for community episodes of specialist palliative care (costs per day reported for each class).

Seven different casemix variables provide the optimal combination to develop classes for each of the settings. Table 4 shows which variables were used and how they were combined to constitute the casemix classes, including cost weights.

Design

We designed a multicentre prospective cohort study, following patients during episodes of specialist palliative care, with a qualitative nested component (interviews with a subsample of participants to better understand the experience of transitions between care settings).

Definitions

We defined both an episode of care and the setting of care as in Workstream 3: development of the casemix classification, Definitions.

Population and settings

Patients were recruited from 12 organisations providing specialist palliative care services in England, comprising three hospital advisory services, eight inpatient hospice services and five community-based services (some organisations provided more than one setting of care). Sites were selected for diversity in terms of participant age, ethnic background, service size, proportion of cancer/non-cancer patients and urban–rural balance

Inclusion/exclusion criteria

The inclusion criteria comprised adult patients aged ≥ 18 years who were able to consent and receiving specialist palliative care at any of the participating sites. The exclusion criteria were patients aged < 18 years and those unable to consent.

Data collection and primary outcome

Data were collected from patient participants and clinicians between December 2016 and May 2018. Collected data included demographic and clinical variables, episode start and end dates, casemix variables as required for the casemix classification developed in workstream 3, and data on patient-level and other costs of providing specialist palliative care. All participants received the usual specialist palliative care at that site, including a multidisciplinary team with specialist training delivering holistic care focused on physical and psychological symptom management, social/family support, planning ahead around priorities and preferences, and care into the dying phase including post-death care of the family, where relevant. Details of the measures used are presented in Appendix 1.

Palliative Phase of Illness was assessed daily for people receiving inpatient (hospital or inpatient unit) care and at every contact for those receiving community-based care. Each change in Phase of Illness (or end of episode) triggered the collection of the AKPS, IPOS or PCPSS and Barthel Index score. All staff involved in patient care recorded the time spent delivering care to patient participants at a patient level using the staff activity matrix.

We also collected data from participating sites on the costs of delivering their services, plus patient-level resource use data, as in workstream 3.

A subsample of participants who had experienced at least two transitions between care settings were invited for interview. A purposive sampling approach was used to include participants from a range of age groups, sex, diagnoses, types of transitions in either direction and geographical areas.

Data handling

Quantitative data were collected prospectively at each site, recorded on an electronic database, cleaned and checked. Checked data were transferred to statistical software (Stata SE V.12) for analysis. Qualitative data were recorded, transcribed verbatim and handled using NVivo version 12.

Analysis

For the quantitative data, the casemix classes developed in Workstream 3 were applied to predict costs for episodes of care and this was contrasted with the actual costs captured for each episode of care. For the qualitative data we adopted a similar approach to Pinnock et al.’s qualitative study,93 undertaking a thematic94 and narrative analysis of interviews, exploring how perspectives on transitions evolve over time, with detailed attention to patient and family perspectives on their experience of care in each setting and during transitions, including their experience of interventions that potentially influenced changes in settings of care.

Ethics

The trial registration number is ISRCTN90752212. Written or oral witnessed consent was taken and documented for each participant, and continuing consent was confirmed at follow-up. Ethics approval was received from Bromley REC on 5 September 2016 (REC Reference: 16/LO/1021, IRAS Project ID: 204926).

Subject characteristics

A total of 309 patients were recruited, providing data on 751 episodes of specialist palliative care. Of these participants, 309 contributed one episode of care, 177 (57%) contributed a second episode of care and 119 (39%) contributed a third episode of care. Only 63 (20%) participants contributed four or more episodes of care. Demographic and clinical characteristics for the 309 participants are reported in Table 5. Just over three-quarters (76%) of participants had cancer, but we were able to recruit one-fifth with a range of different non-cancer conditions.

Episode characteristics

Details of the episodes of care are reported in Table 6. Seventy-seven episodes of care (10% of all episodes) occurred at sites not participating in the study; although we had endeavoured to include all specialist palliative care sites in each geographical area, inevitably some episodes of care (especially in hospital) were outside of our C-CHANGE sites. Tables 7 and 8 report the duration of episodes of care, by setting and by episode number, respectively. As expected, the median duration of episode was shortest for hospital advisory episodes (10 days) and longest for community episodes (26 days). Note that the absence of the dying Phase of Illness strongly affects these data; those first seen in the dying phase had the shortest length of stay yet are effectively excluded from these data because of the requirement for consent.

Casemix variables

Details of the casemix variables are reported in Table 9. Apart from the dying Phase of Illness, the casemix variables extended across the full range of categories or scores.

Actual and predicted costs per day

Table 10 presents the actual costs per day, whereas Table 11 compares the predicted class costs with the actual costs per day. The actual cost of inpatient hospice care are consistently higher than the predicted class cost, perhaps because of the 2-year interval between the collection of costs for workstreams 3 and 4 (no uplift was applied) or possibly because of improvements in the application of our costing methodology (the same costing methodology was used in both workstreams, but the teams were more familiar with how to apply it).

Theme 1: uncertainty about the new care setting

Lack of information about the new setting of care and patients’ uncertainty regarding their discharge plan predominated in interviews.

Many participants commented on the ‘bad’ timing of a transition in care setting (e.g. emergency admissions during out of hours) and uncertainty around access to care.

Participants highlighted the fact that clear and effective communication is as important as the quality of care delivered; receiving the right information in the right place at the right time helped them to gain a better understanding of their health status and care plan, which reduced their anxiety and worries.

Theme 2: biographical disruption

Sense of identity was considered important but participants felt that transitions often compromised one’s sense of self, which linked to an inevitable loss of control and independence.

Theme 3: importance of continuity of care

Both family and patients praised clinicians who provided continuity and got to know them as individuals. They appreciated not having to retell their stories and recognised the difference in care when staff knew them and their stories.

Participants highlighted the importance of the correct and timely transfer of patient information, medical notes and medication prescriptions; when things were disjointed, patients and their families felt unsafe and that lack of continuity could lead to medical mistakes.

Theme 4: need for emotional and practical support

Family members often act as advocates in the healthcare setting and the majority of participants specified a great need for support from family and friends not only in practical matters, such as arranging transport or coordinating care, but also in the emotional aspect of transitions in care.

Engagement with the Consumer Panel

We also worked with a more diverse group, the Consumer Panel (a network of patients, families and the public developed and sustained at the Cicely Saunders Institute; see Brighton et al. 95 for further details), which enabled engagement with a wider and more diverse group constituency. The following individuals and organisations especially contributed in this respect: Carolyn Morris and Kirstie Newson (who worked with us on our Dissemination and Engagement Group at the Cicely Saunders Institute); our Macmillan Information & Support Centre (which was located in the Cicely Saunders Institute, provided information and support to patients and public, and hosted several patient/family support groups who were able to contribute); and Hospice UK (London, UK) (a national organisation which supports palliative care providers).

We engaged with the Consumer Panel throughout the programme, largely by phone, Skype and e-mail but also through several smaller face-to-face group meetings (according to need) to discuss issues that emerged. The focus of our work with the Consumer Panel was on considering and discussing individual measures, considering which measures could be used together, how the measures met the priorities and domains of concern of patients/families with advanced illness, how the measures work both individually and together for this population, and how a casemix classification might best support good quality care and reduce inequities in access and care. Through the Consumer Panel we were able to test the training materials for the use of measures, and review and refine participant and other materials for workstreams 3 and 4. The Consumer Panel was kept informed through our quarterly newsletter, although we also met with a number of individuals on the Consumer Panel for specific pieces of work, as the programme required.

Patient participation in the Programme Steering Committee

Mr Jonathan Hope and Mrs Sue Farr (co-leads for PPI) regularly attended the meetings of the C-CHANGE Programme Steering Committee, where they were updated on the programme’s progress, provided feedback from the PPI group, informed the planning of next steps, addressed challenges that arose and helped the committee to better consider patient and family perspectives. We owe them both a great debt for their wisdom, their contributions, their ability to challenge us, and the insight they brought from extensive experience of both serious illness and providing care.

Innovation and evolution in our patient and public involvement and engagement

Patient and public involvement in palliative care is challenging because of the severity of illness and short survival time of those affected (some of our PPI partners, for example, were very seriously ill), the substantial caring demands on families, and the impact of bereavement often experienced. However, we linked to Collaborations for Leadership in Applied Health Research and Care (CLAHRC) South London to help us develop and sustain a sufficient ‘critical mass’ within the C-CHANGE PPI group to support effective PPI engagement over the years of the programme.

In addition to the PPI described above, over the course of the programme we held three PPI workshops across the Cicely Saunders Institute (14/06/2016, 15/11/2016 and 10/04/2017) to improve the quality and relevance of the research and maximise its impact on improving care. Both the C-CHANGE researchers and our PPI C-CHANGE group members were actively involved and engaged in this work. These workshops focused on increasing dialogue and collaboration, listening to our PPI partners, facing challenges, and seeking ideas and solutions together.

Impact of patient and public involvement to date

Dissemination

As a way to disseminate the C-CHANGE project and the PPI work, our PPI group developed a video diary, taking a ‘reflections and responses’ approach. The aim was to capture PPI perspectives on video and enable the research team to consider and respond to the evolution of the research. In total, our PPI group, alongside a C-CHANGE researcher, produced six videos (see Table 12). The process for creating the videos was as follows:

1. The PPI group suggested a topic for a video log.

2. A PPI member co-wrote a script with a C-CHANGE researcher.

3. A PPI member and C-CHANGE researcher edited and refined the script and filmed the video.

4. The PPI member and/or group provided feedback on how best to edit the video.

The main topics raised in the videos comprised: (1) identifying different cultural needs, (2) the individual concerns of people with advanced illness, (3) capturing emotional distress, (4) thinking about goal-based outcomes, (5) the experiences of patients alone and (6) the safety of patients and their family members.

Views on Care measure

The VoC measure is an assessment of patient quality of life and experience of care (for the full measure, see https://pos-pal.org).

Views on Care focuses on the patient’s quality of life and perceived impact of the palliative care service, and provides staff with an indication of whether or not they are having a positive impact on patients’ lives. 125

Design

We designed a multicentre prospective cohort study, following patients during episodes of specialist palliative care and reported according to the TRIPOD statement. 86 The study was registered with a International Standard Randomised Controlled Trial Number on 2 March 2017 (https://doi.org/10.1186/ISRCTN90752212).

Definitions

We defined an ‘episode of care’ as a ‘period of contact between a patient and palliative care service provider or team of providers that occurs in one setting’. 87 We defined the ‘setting of care’ as: hospital advisory (specialist palliative care teams providing an advisory or consultation service within hospitals), inpatient hospice (patients admitted to a hospice or specialist palliative care unit for an overnight stay of one or more days) or community-based specialist palliative care (when the patient receives care in their usual place of residence, i.e. at home or in a care home).

Population and settings

Patients were recruited from 14 organisations providing specialist palliative care services in England: four hospital advisory services, five inpatient hospice services and seven community-based services (some organisations provided more than one setting of care). We included consecutive adult patients (≥ 18 years) receiving specialist palliative care at all participating sites. The exclusion criteria were being aged < 18 years, declining participation and/or previously expressing a wish not to participate in research.

Data collection and primary outcome

Collected data included demographic and clinical variables, episode start and end dates, potential casemix variables, and data on patient-level and other costs in providing specialist palliative care. Potential casemix variables were selected based on (1) being patient-level attributes and (2) prior evidence of association with casemix/complexity. 88 The key casemix variables included were age, sex, ethnicity, living circumstances, need for interpreter, primary diagnosis, palliative Phase of Illness, functional status, dependency and symptoms/problem severity. Our primary outcome was the cost of specialist palliative care per day, measured over an episode of care.

Sample size

Based on standard recommendations for fitting multivariate models, a minimum of 50 + 8 × m cases for testing multiple correlation (where m is the number of predictors) were required to test the null hypothesis that the population multiple correlation equals zero with a power of 80%, α = 5% and a medium effect size for the regression analysis (R² = 0.13). 90,91 The unit of analysis was episodes within sites, therefore, 10 predictors required 130 episodes per site. Allowing an additional 15% for episodes with missing data and 20% for cost outliers, we estimated a target of 2674 episodes of care (191 episodes × 14 sites) was required.

Analysis

To discover which baseline casemix variables could predict the cost of that episode of care, the following steps were undertaken:

1. We removed incomplete episodes, retaining only the complete episodes of care.

2. Following a previously adopted approach,88 high and low cost outliers were identified and removed using a trimming algorithm based on the interquartile range (IQR) with the upper trim point at Q3 + 1.5 IQR and the lower trim point at Q1–1.5 IQR (where Q1 is first quartile and Q3 is third quartile). 57

3. We examined the distribution of costs of specialist palliative care by setting.

4. Following the same approach as in the development of the Australian casemix classification,88 we developed and validated a cost-predictive model using CART analysis, which constructs decision rules in a hierarchical manner to form a branching classification.

Subject characteristics

A total of 2469 patients were recruited, providing data on 2968 complete episodes of specialist palliative care (12 incomplete episodes were removed prior to analysis). Demographic and clinical characteristics of the 2469 participants, and details of the 2968 episodes of care and related casemix variables are reported in Workstream 3: development of the casemix classification, Results, with 767 episodes in the hospital advisory setting, 764 episodes in the inpatient hospice setting and 1437 episodes in the community setting.

Costs of specialist palliative care

The distribution of the total cost of specialist palliative care episodes, derived from the trimmed data set, is shown in Table 3; costs are reported (1) per day, (2) per day, broken down by Phase of Illness, and (3) by episode of care, together with details of length of episodes.

Classification and regression tree analysis

Seven different casemix variables provided the optimal combination to deliver classes in each of the settings. Table 14 shows these variables and how they are combined to constitute the casemix classes, including cost weights.