Self-management toolkit and delivery strategy for end-of-life pain: the mixed-methods feasibility study

Phase I: development objectives

• Establish a patient and public involvement (PPI) panel.

• Establish the content of a prototype intervention and a manualisation strategy that includes a protocol to standardise (1) the training of HCPs and (2) the delivery of the intervention by HCPs to patients and carers.

• Understand self-management needs and capabilities of patients and carers related to strong opioid medication.

• Define usual care.

Phase II: modelling objectives

• Refine the prototype intervention and manualisation strategy.

Phase III: feasibility assessment objectives

• Assess acceptability and up-take of the intervention in a mixed-methods observational study involving patients, informal carers and HCPs from four palliative care services.

• Assess the feasibility of obtaining outcome data for a larger trial.

Phase I

• Establishment of a PPI panel and assessment of members’ support and training requirements.

• Development of a usual-care protocol based on literature review and clinical practice observations.

• Development of prototype intervention materials, manualisation strategy and usual care protocol.

Phase II

• Establish members of focus groups.

• Development of refined intervention materials and manualisation strategy.

Phase III

• Sampling strategy: recruit three patients per month at each site within 4 months.

• Feasibility of data collection: key clinical and health economic measures, and health-care resource measures, have sufficient complete data to estimate primary study end points.

• Trial experience: patients and carers reporting acceptable and sustained use of intervention materials.

Scoping of the literature

Scoping of the literature was undertaken following the preliminary work described above. The objective of this exercise was to examine:

1. the content and form of previous interventions effective in improving pain management by patients (see Appendix 2)

2. key systematic reviews of what can be done to support self-management across a range of long-term conditions of relevance to this particular application at the end of life

3. key literature and reviews that identify factors that enable HCPs to support patient self-management

4. existing public guidance based on NICE clinical guideline 140 – Opioids in Palliative Care (Box 1)

5. description and potential components of self-management at the end of life (Box 2).

Theoretical underpinning of the intervention

The literature-scoping work informed the selection of theory focused on self-efficacy and behaviour change to best suit the developing intervention. Self-efficacy is a key component of Bandura’s social cognitive theory. 54,55 The theory of social cognition states that knowledge acquisition can be directly related to observing others within the context of interactions and experiences. According to this theory, self-efficacy is the belief in an individual’s capabilities to organise and carry out courses of action to manage situations. Bandura states that behavioural techniques can be used to target the four sources of self-efficacy: mastery experience, role modelling, verbal persuasion and the regulation of physiological and affective states. Michie et al. 39 undertook a systematic search and consultation with behaviour change experts to identify frameworks of behaviour change interventions. These were then evaluated for comprehensiveness, coherence and a clear link to a model of behaviour. A new framework was subsequently developed to fully meet all these criteria: the behaviour change wheel. The wheel characterises behaviour change interventions around nine intervention functions aimed at addressing deficits in one or more of three essential conditions for behaviour change: Capability, Opportunity and Motivation (termed the COM-B system). The behaviour change wheel identifies three main target constructs (sources of behaviour), which are capability (physical and psychological), motivation (automatic and reflective) and opportunity (social and physical). Around these target constructs are nine intervention functions (education, persuasion, incentivisation, coercion, training, enablement, modelling, environmental restructuring and restrictions), which represent ways to address deficits in one or more of the target conditions.

Key learning from studies conducted by the team

The intervention development was also informed by key learning from two studies undertaken by members of the research team. These were Cancer Carer Medicines Management (CCMM),56 funded by Dimbleby Marie Curie Cancer Care, and Improving the Management of Pain from Advanced Cancer in the Community (IMPACCT), a programme grant funded by the NIHR (RP-PG-0610-10114).

The SMART intervention drew on carer needs for medicines management at the end of life identified from the CCMM study (see Appendix 3), including needs for information about pain medicines and side effects, changing beliefs about opioids and providing skills and opportunities for self-evaluation. The results of the CCMM study also showed that a conversation-driven intervention, delivered by nurses in routine practice and supported with a toolkit of resources, was acceptable to carers and to nurses and was compatible with their existing practice. CCMM showed some evidence of benefit in influencing carers’ knowledge, beliefs and behaviours related to management of pain medicines. Principles of the CCMM conversational process (assessment, education and review), as well as elements of the toolkit (information leaflets, medication chart, pain diary and contact details for local and national services), informed the development of the SMART intervention.

Learning from the IMPACCT study was derived from a review of the optimal components of educational interventions for advanced cancer pain, which was carried out as part of the study (see Appendix 4). In this meta-review,57 the authors used Michie et al. ’s39 behaviour change wheel as theoretical underpinning. Mapping findings from six reviews and two papers led to identification of five out of the nine behaviour change wheel intervention functions. These were considered essential for successful interventions:

1. education, for example providing written information about pain management, including analgesic and non-pharmacological approaches

2. training, for example providing instruction, demonstration and coaching of new skills (techniques for managing daily drug regimes, relaxation techniques)

3. enablement and persuasion, for example overcoming cognitive and emotional barriers to pain management through addressing concerns about tolerance or addiction

4. environmental restructuring and resources, for example incorporating the delivery of education for self-management into the usual care provided by specific health professionals, such as specialist nurses, primary care practice nurses and community pharmacists

5. modelling, for example patients talking to other patients about their successful use of various pain management strategies.

Aim

The aim of this phase of the study was to refine and detail an intervention for self-management of analgesia (opioids) and related treatments (for nausea, constipation and drowsiness) at the end of life.

Objectives

The literature scoping informed the objectives and these were to:

• explore views regarding the nature and components of supported self-management regarding analgesia and related treatments at the end of life, and test our preliminary model of self-management in this context

• introduce participants to, and ask for their views on, aspects of self-management in this context – in particular our selected definition of supported self-management in palliative care (Johnston et al. 51) and practical difficulties regarding supply and medicines taking encountered by patients (Schumacher et al. 52,53) (see Table 2).

For HCPs, the objectives were to:

• reveal the self-management promoting activities and behaviours already used by specialist palliative care HCPs and gain views on the professional supportive self-management roles proposed by Johnston et al. 51

• rehearse previous examples of interventions to test possible delivery modes and how these could link to existing patterns of practice.

For patients and carers, the objectives were also to:

• understand what supported self-management at the end of life was for them

• explore patient and carer needs using the framework of Schumacher et al. ’s52,53 commonly encountered practical difficulties regarding supply and the taking of medicines

• gain views regarding potential options for the content, form and delivery of the intervention.

Approach to data collection

Focus groups and interviews were held in two geographical regions: Hampshire and Yorkshire. They were conducted with patients, their carers and specialist palliative care health professionals (including service managers and commissioners).

Inclusion criteria

Patients were included if they:

• were aged ≥ 25 years and considered to be in the last year of life

• were experiencing pain

• were being treated with, or starting, opioid analgesia

• were experiencing, or anticipating, adverse effects of nausea, constipation and drowsiness

• were living at home

• were being cared for by specialist community-based palliative care services in Hampshire or Yorkshire

• had capacity to consent.

Carers were included if:

• they were the primary carer of a patient meeting the above inclusion criteria and

• the patient gave consent to their involvement.

Health-care professionals were included if they were:

• CNSs and doctors who were part of specialist palliative care teams or

• service providers or managers of specialist palliative care services or

• local commissioners of palliative care services.

Sampling strategy and recruitment

To access a range of individuals (HCPs, patients and carers), we aimed to recruit 35 participants via various strategies across four hospices and two acute NHS trusts. The sampling and recruitment strategy is detailed in Table 1.

Focus group and interview guides

Topic/interview guides (as well as supporting slide/card packs) were developed to meet the study objectives (see Appendices 5–8). A number of tools were used to manage the interview and focus group discussions:

• a definition of supported self-management in the context of palliative care (Johnston et al. 51)

• eight proposed professional roles to support self-management in the context of palliative care (Johnston et al. 48)

• the practical difficulties around supply and medicine taking encountered by patients and carers in the work of Schumacher et al. 52,53

• the content and form of previous interventions to improve pain management by patients. 58–61

The interviews and focus groups were conducted by the study’s researchers. The focus groups were conducted with a co-facilitator another researcher with expertise in the field, present to aid moderation.

Data analysis

The audio files from the interviews and focus groups were professionally transcribed. They were listened to by the researchers alongside the transcripts to check for complete accuracy. Both study researchers familiarised themselves with the data by reading and rereading the transcripts and identifying key issues, concepts and themes. Initial coding took the form of indexing on the transcripts, and each research fellow summarised the key themes from the data related to the sites in their regional area. The themes were discussed for comparative purposes. Natasha Campling then coded the entire data set for all issues, aspects and themes that were relevant to supported self-management in this field. Coding was performed in NVivo (version 11; QSR International, Warrington, UK) utilising framework analysis. 62

The development process

The process of intervention development is illustrated in Figure 1. The contextual work and literature scoping informed a preliminary model of supported SMART. This model, a definition of self-management,51 practical difficulties related to supply and medicine-taking52,53 and the content and form of previous interventions58–61 were used as tools within the focus groups and interviews to gain participants' views. The result was development of a concept of self-management, to inform the required components (including the content and form) of the intervention: a four-step educational approach and toolkit, plus a training package to enable nurses to deliver the intervention within a feasibility study.

FIGURE 1.

Intervention development process.

Findings from the intervention development interviews and focus groups

Concept of supported self-management in analgesia and related treatments at the end of life

The development process, informed by the varied sources of learning outlined in Evidence synthesis, enabled the generation of a preliminary model of supported self-management in analgesia and related treatments at the end of life. The model was tested, within this specific context, through the interviews and focus groups with patients, carers and HCPs. The model included the self-management definition, professional roles outlined by Johnston et al. 51 and the practical difficulties regarding medicines access, supply and taking outlined by Schumacher et al. 52,53

This process revealed new components of supported self-management within the end-of-life context. It displayed an ever-changing process enacted on a continuum of behaviours dependent on the responsibility taken by the patient, carer and specialist nurse. This was a complex web of behaviours, varying day by day, if not hour by hour, within this context. With continual disease progression, there were frequent changes in symptoms and side effects from both medication and palliative treatments, with behaviours profoundly affected. This context was complicated by the surrounding ‘swirl’ of what individuals and their families were already striving to deal with, the wider context of psychological distress and high levels of carer strain. Individuals in this context could be struggling to cope with a palliative care diagnosis and there was anxiety and clinical depression of both patients and/or carers. Consequently, the capabilities of the patient and carer fluctuated greatly, influencing supportive self-management roles and the required behaviours of the specialist nurse.

The concept is presented here as it is key to understanding the development and form of the intervention. The key components of the concept, the issues of responsibility and the supported self-management roles of the patient, carer and CNS are outlined. Figure 2 is a diagrammatic representation of the concept.

FIGURE 2.

A concept of supported SMART.

The four-step educational approach

The four-step educational approach was maintained from the study protocol, with data confirming that such an approach was an appropriate way to proceed. Our contextual work had identified that needs assessment, information provision and regular review were important components of a supported self-management intervention that would potentially fit well into nurses’ usual practice processes. Goal-setting (with associated action plans) was proposed following expert educational psychologist input (SM) as the mechanism by which information provision could lead to behaviour change. In addition, a meta-review of self-management interventions for long-term conditions identified that action plans for deteriorating conditions were a key component of successful interventions. 21 The approach was designed to sit alongside the everyday practice of specialist nurses, with four cyclical steps to occur at each nurse–patient visit.

The self-management support toolkit

The evidence from the data was mapped to components of the intervention (both the educational approach and proposed SMST resources), resulting in the framework outlined in Table 6. The framework also linked the intervention components to the target participant understanding and behaviour in relation to self-management, including the behaviour source,39 the target supportive self-management roles of the specialist nurse,51 the self-efficacy techniques to be used by the nurse54,55 and the intervention function,39 and served to ensure a theoretically modelled intervention.

As a result of this framework, the SMST resources were developed to address the evidence from the data. The resources comprised eight factsheets, two podcasts, a pain diary, medication charts and goal-setting sheets. The factsheets were drafted to encompass all the core themes and issues outlined by the patients, carers and palliative care HCPs. They went through iterative cycles in the development process. The study PPI group was asked to review the content of all the factsheets, as were specialist HCPs from two different palliative care teams who were not going to be involved in the feasibility study (in order to prevent potential issues of bias and/or contamination). The final stage of development was review of the factsheets by an expert in health literacy (HB). The presentation and structure of the facts were recast with reference to the evidence on how people make sense of illness and decisions about treatment. 63,64 Factsheet content did not change, but readability was improved (assessed using Simple Measure of Gobbledygook readability formula65) by shortening sentences and eliminating some longer words (except drug names). The text was subdivided in all the factsheets into small titled sections to facilitate review by patients, with spaces provided for patients and carers to write notes or questions. These changes improved the factsheets’ health literacy and utility to support patients’ reasoning about treatment in the context of their experience of illness and lifestyle.

Patient and public involvement

Specific engagement for this feasibility study took place via regular study meetings and individual correspondence with PPI panel members. PPI panel members have been closely involved in the design and delivery of this feasibility study. PPI representatives were involved with:

• reviewing the content and format of the SMART intervention materials and the nurse delivery strategy

• reviewing and feeding back to the study team on the draft patient study materials, including the self-reported outcome measures included in the patient questionnaire, which led to the removal of one questionnaire [Beliefs about Medicine Questionnaire(3)] as panel members felt the wording to be inappropriate for the end-of-life context.

Amendments to protocol

There was one substantial amendment, which outlined three changes to the protocol. This was approved on 23 November 2015 and the changes to protocol are summarised below.

1. Changes to patient-reported outcome measures: replacement of the Self-Management Ability Scale68 with the Self-Efficacy for Managing Chronic Disease Scale (SES)69 and removal of the Beliefs about Medicines Questionnaire,70 because it was felt to be too burdensome for participants to complete.

2. Change to assessing fidelity of delivering the SMART intervention: study nurses would be asked to audio-record the nurse–patient consultations when the SMART intervention was used.

3. Change to wording of ‘weekly’ to ‘regular’ review of patients’ progress with goal-setting review so that the protocol reflects study nurses’ usual practice.

See Appendix 15 for communication with the REC.

Eligibility criteria

Patients were eligible for participation if they:

• were aged ≥ 18 years

• had been prescribed strong opioid analgesia

• were living at home

• were being cared for by specialist community palliative care services

• were considered by the clinical team likely to survive beyond 6 weeks of follow-up

• had the capacity to provide informed consent.

Patients were ineligible for participation if they:

• had insufficient literacy or proficiency in English to contribute to the data collection that was required for the research.

Carers were eligible to take part in an end-of-study interview if:

• they were the primary carer of a patient meeting the above inclusion criteria

• the patient whom they cared for had consented to their involvement.

Sites

Study nurses and patient participants (and carers, when appropriate) were identified from community palliative care services at four hospices: two in Hampshire (site codes HANTS1 and HANTS2) and two in Yorkshire and the Humber (site codes YKHB1 and YKHB2). Identifying patients from community palliative care services offered the most efficient access to patients approaching the end of life who were living at home. Hospital-based palliative care services were used as recruitment sites for the focus groups and interviews described in the previous chapter. However, patients recruited from hospital-based palliative care services were either inpatients or in transition between inpatient and community health-care services and, therefore, not living in their own homes. Therefore, it was decided not to recruit from hospital-based palliative care services for the feasibility study as the focus was to identify patients who were managing pain in their own homes.

Nurse recruitment: identification and consent

At each recruitment site, between two and four community-based palliative care CNSs were identified by contacting service team leaders; 12 CNSs (hereafter referred to as study nurses) were identified from the four recruitment sites. An invitation was sent to all CNSs in the four community teams to attend a brief presentation by the SMART study research team. Following this, CNSs interested in the study were invited to attend a half-day workshop at which they received further information and training on the trial procedures and delivery of the intervention.

All study nurses who were trained to use the SMART intervention were also invited to take part in a one-off face-to-face interview at the end of the trial. Study nurses were given a recruitment pack (see Appendix 16) by a researcher, who explained the purpose of the qualitative interview. Study nurses were asked to provide written informed consent to take part in an interview with the research no less than 24 hours after receiving a recruitment pack and having had sufficient opportunity to ask any questions about their participation.

Non-study nurses working with research active sites were identified by contacting community service-lead CNSs. An e-mail was sent to lead CNSs with a link to an online survey and a request to circulate among all palliative care CNSs within their teams.

Participant recruitment: eligibility, approach and informed consent

Potentially eligible patients were identified by screening all new referrals and existing patients on study nurses’ caseloads against the eligibility criteria. Screening caseloads was done by a clinical research nurse (CRN) or researchers with the study nurses. This activity was recorded on a screening log (see Appendix 17), kept at each site. Before eligible patients were approached, their participation was discussed among the study nurse, researchers and wider clinical team to consider patients’ capacity to participate.

Eligible patients were first approached face to face by their CNS, who gave a verbal explanation of the study and provided them with a recruitment pack, consisting of an invitation letter, a patient information sheet and a consent form (see Appendix 17). At this point, patients not interested in participating were thanked for their time and asked if they would say briefly why they refused participation and, if willing, whether or not they thought the SMART intervention was acceptable in principle.

To identify carer participants, patients who were interested in the study were asked if they had a main informal carer (family member or friend) who may also be interested in participating in an interview with a researcher. If a main informal carer was identified, they were provided with a recruitment pack for carers (see Appendix 17).

Patients (and carers) interested in knowing more about participating in the SMART study were asked to provide their contact details and told that a researcher would contact them to discuss the study in more detail and answer any questions that they may have about participation. At this point, all subsequent recruitment and consent activities were completed by a researcher or CRN.

Patients were then contacted by telephone or face to face by a researcher to discuss the SMART study and answer any questions they had about participation. At this point, patients who refused participation were thanked for their time and asked if they would say briefly why they did not wish to participate.

Following provision of the recruitment pack, patients and their carers were given at least 24 hours to consider their participation and were encouraged to discuss the study with their family/friends and other HCPs.

Patients (and carers) willing to participate were asked to provide full written informed consent, following which they were recruited into the trial, and hereafter are referred to as participants.

This study nurse-led approach to identification and introduction and researcher/CRN-led approach to recruitment was the most efficient way to identifying eligible interested patients while keeping the burden of recruitment to a minimum for the study nurses and separate from their clinical practice.

Recruitment schedule

We estimated that within a 4-month recruitment period approximately 450 new patients would be referred to the four recruitment sites included in our study. We assumed that two-thirds of patients would have pain (n = 300), of which 50% would be eligible (n = 150). Based on these assumptions, we conservatively estimated that 20% of eligible patients would agree to participant (n = 30), which gave a predicted recruitment rate of eight consented participants per month across all sites for 4 months.

Training of the clinical nurse specialists

A bespoke training package was designed (outlined in Chapter 2, Training development), comprising a half-day workshop session. Following the training session, additional resource packs were developed for the study researchers to deploy with the nurses over the course of the trial to reinforce their engagement in the study and support their delivery of the intervention.

All 12 study nurses attended an initial SMART training workshop. The workshops were run in Hampshire and Yorkshire, with nurses from the two hospices in each region attending the same workshop. Sessions were facilitated by an expert Nursing and Midwifery Council-registered nurse educator who leads postgraduate palliative and end-of-life care education provision. In attendance, and acting as co-facilitators, were two study researchers. Sessions were experiential in approach, requiring reflection and demonstration of practice, and were run using the SMART training plan (Table 8) and supported with resources developed by the research fellows for the study nurses (see Appendix 3).

During the workshop session, it became apparent that the nurses wanted further detail on the research process, such as eligibility criteria, screening, approach and consent, before they felt able to give their attention to intervention delivery itself. As a result, these processes had to be outlined by the co-facilitators. This took time during the workshops, leaving no time to model the conversational process for and with the nurses, or the goal-setting, in particular the process of effective action-planning once goals are set. To respond to this, additional resource packs were created for the study nurses following the workshops. It was the intention that these resources would be given to the nurses via weekly opportunities to reflect on delivery of the intervention through face-to-face visits with the study researchers.

Additional resource packs comprised (1) a 30-minute nurse–patient modelled self-management-focused conversation between two skilled nurse educators (one of whom was the workshop session facilitator), which was supplied as an mp3 file on a memory sticks with an expectation that nurses would listen to the ‘model’ conversation; (2) self-management conversation prompts (see Appendix 18), outlining potential questions mapped to the four steps of the educational approach; and (3) a ‘making action plans’ document (see Appendix 19), outlining specific actions needed to help a patient meet their self-management focused goal(s). The use of these resources enabled the researchers to provide ongoing coaching and support to the study nurses during the trial and additionally helped to promote ongoing engagement of the nurses.

In order to help formalise and aid standardisation of the approach to ongoing nurse engagement, an aide memoire for the researchers to use with the study nurses at site visits was developed, which prompted questions related to the nurses’ experience of each step of the educational approach (see Appendix 20). A related framework to standardise the documentation of these discussions via researcher field notes was also developed (see Appendix 21).

Interventions details and schedule of delivery

Study nurses were asked to begin using the intervention with participants within 7 days of consent and baseline data collection. As described in Chapter 2, The SMART intervention, the SMART intervention comprised both a four-step educational approach and a SMST. In brief, the four-step educational approach was designed to reflect the everyday practice of specialist nurses and included an assessment of participants’ needs, provision of information, goal-setting and review and coaching of self-management progress. The SMST comprised eight factsheets, two podcast films, a pain diary, a medication chart and goal-setting sheets.

It was intended that the intervention would be delivered by study nurses each time they visited their participants (and a carer when appropriate) over the 6-week study period (each visit was referred to as a ‘SMART visit’). Study nurses were asked to visit participants a minimum of three times during this 6-week period (i.e. at least once a fortnight). During each visit, study nurses were required to use a conversational approach to go through the four steps of the educational approach and provide the resources from the SMST as required. As a minimum, study nurses were required to provide the factsheet ‘Contacts and further information’ and the goal-setting sheets as core resources at their first SMART intervention visit.

Screening data collection

Screening data were collected by a researcher or CRNs on all the patients on the study nurse’s caseload. Data were collected on sex, age, referral status (existing patient or new referral) and eligibility.

Baseline data collection

The researcher or CRN completed the baseline case report form (CRF), which captured data on participants’ medical history (e.g. type of advanced disease, date of diagnosis), reason for referral to palliative care, palliative treatments received (within the past month) and current medications for pain, nausea and constipation. The baseline CRF also asked participants whether or not, in a future study, they would take part if they were randomised to receive either the intervention or standard care as usual (yes/no response). This question was included to inform the acceptability of randomisation processes in a future definitive trial.

Participants were also asked to complete an outcome measure pack comprising five validated self-reported outcome measures on pain, self-efficacy, common symptoms, quality of life and satisfaction with medicines information. These outcome measures are described in Participant self-reported outcome measures.

Follow-up schedule

Participants were followed up for 6 weeks from the date of baseline assessment. Six-week follow-up was chosen because the risk of short survival in this population meant that demonstrating early and sustained improvements in self-management within a few weeks would be particularly important in this context. During the 6-week follow-up period, participants were contacted by a researcher at three time points post baseline:

1. week 2 (day 14)

2. week 4 (day 28)

3. week 6 (day 42).

The exact timing of follow-up visits was flexible to fit around participants’ medical and other appointments, although efforts were made to keep follow-up visits within ± 2 days of the scheduled appointment date. Follow-up visits were usually conducted face to face between the participant, their carer and a researcher. Participants were offered a telephone follow-up if this was more convenient.

At each follow-up visit the researcher completed a follow-up CRF and asked the participant to complete the outcome measure pack (as described above). The follow-up CRF captured data on any of the following activities since the last follow-up (or baseline) visit:

1. which factsheets had been given to the participant

2. whether or not any self-management goals had been set (or reviewed)

3. whether or not the participant and/or carer had watched either of the video podcasts.

Participant and carer interview

As part of the final follow-up visit (week 6) participants (and carers, when appropriate) were asked to take part in an audio-recorded face-to-face semistructured interview together with a study researcher. Participants and carers were interviewed together. The topic guide for these interviews focused on two broad areas. First, participants and their carers were asked what they thought about taking part in the trial and how they were managing their medicines. These questions were intended to explore participants’ and carers’ acceptability of trial procedures, what they thought of the intervention itself and how it had been delivered to them by their CNS. These questions were designed to understand the extent to which participants and their carers were aware of the formalised educational process, the acceptability of the SMST resources (i.e. what they liked and did not like about it), their continued use of the intervention and whether or not their participation had any effect on their confidence with managing medicines. Interview guides are reproduced in Appendix 22.

Study nurse interview

In addition to participant and carer interviews, all study nurses were invited to take part in an audio-recorded semistructured interview after follow-up had closed for all participants at their site. These interviews covered a range of topics designed to capture their experiences of participating in the research process (including the training workshop), as well as their views on the acceptability of the intervention and whether or not they felt that they could integrate it into their clinical practice. The topic guides for the study nurse interviews are reproduced in Appendix 22.

Final data collection

At each site, when the last participant had completed 6 weeks of follow-up (or had withdrawn from the trial) a final data collection CRF was completed to capture the following data on all patients at that site:

• date of death or date last known to be alive

• place of death and preferred place of death (if known)

• health-care resource use over the 6-week follow-up period (these data are reported in Chapter 4 as they relate to the health economic evaluation of the overall intervention)

• current medication prescribed for pain, nausea and constipation.

Adverse events

The intervention was evaluated within a patient population with advanced disease and who are approaching the end of life. Thus, it was expected that episodes of acute illness or infection, new medical problems and deterioration of existing medical problems would occur and could result in prolonged hospitalisation, hospital readmission, significant or permanent disability or incapacity, or death. In recognition of this, events fulfilling the definition of an adverse event or serious adverse event were not reported in this study unless the event resulted from administration of any research procedure.

Non-study nurse data collection: assessing contamination

To inform the design of a future definitive trial, it was considered appropriate to evaluate whether or not within individual sites the practice of non-study nurses was influenced as a consequence of working in a team where their colleagues were using the SMART intervention. To assess contamination, an online survey was sent to all non-study nurses working in the community palliative care teams at the four recruitment sites. The survey captured data on:

• demographics (age, sex, grade/band)

• duration of palliative care experience

• whether or not they were an independent prescriber

• whether or not they were aware of the SMART study and what it was about

• whether they had discussed the SMART study with colleagues who were using the intervention

• if they were aware of the study or had discussed it with a colleague, whether or not this had influenced their own practice supporting medicines management with their patients.

A copy of the online survey questions to non-study CNSs is reproduced in Appendix 23.

Participant self-reported outcome measures

Pain was measured using the short-form Brief Pain Inventory (BPI). 71 Using a 0–10 numerical rating scale (anchored 0, ‘no pain’, and 10, ‘pain as bad as you can imagine’), the BPI allows respondents to rate their worst, least and average pain intensity during the past 24 hours, as well as present pain. The responses to the four pain intensity items are reported separately. Using a 0–10 numerical rating scale (anchored 0, ‘does not interfere’, and 10, ‘completely interferes’), respondents are also asked to rate the extent to which pain interferes with general activity, mood, walking ability, normal work, relations with other people, sleep and enjoyment of life.

Self-efficacy was measured using the SES (adapted for palliative care). 69 This scale contains six items that assess an individual’s confidence with managing symptoms of illness and was adapted for a palliative care context. Each item is measured on a 1–10 numerical rating scale, anchored 1, ‘not confident at all’, and 10, ‘totally confident’.

Symptom burden was measured using the Edmonton Symptom Assessment Scale (ESAS). 72 The ESAS is a 10-item tool designed to assess common symptoms in palliative care patients. Each item is measured using a 0–10 numerical rating scale anchored 0, ‘no (symptom)’, and 10, ‘worst (symptom)’. The ESAS was originally developed in cancer patients but has been extensively used at end of life. The final item, ‘other problems’, was modified to represent drowsiness.

Health-related quality of life was measured using the EuroQol-5 Dimensions, five-level version (EQ-5D-5L),73 which is a standardised, generic measure of health-related quality of life. It provides a single index value for describing and valuing health status calculated from a simple descriptive profile consisting of the following five dimensions: usual activities, self-care, mobility, pain/discomfort and anxiety/depression. The EQ-5D-5L index is reported not in this chapter, but in Chapter 4, as it relates to the health economic evaluation of the overall SMART intervention. The EQ-5D-5L also provides an overall measure of health-related quality of life by asking respondents to rate their present health state from 0, ‘worst health you can imagine’, to 100, ‘best health you can imagine’, which is reported in this chapter.

Satisfaction with medicines information was measured using the Satisfaction with Information about Medicines Scale (SIMS),74 which consists of 17 items about the types of information required to facilitate safe self-management. For each item, respondents are asked to rate the amount of information they have received using the following response scale: too much, about right, too little, none received or none needed.

Feasibility of conducting trial procedures

Process outcomes providing measures of the feasibility of study procedures included eligibility rates, recruitment rates and follow-up rates at 2, 4 and 6 weeks; the fidelity of delivery (i.e. number of SMART intervention visits delivered per participant); completion and acceptability of participant self-reported outcome measures; completion of patient health-care records to collect outcomes; estimates of variability in patient-reported outcome measures; the extent of contamination of non-study nurses working in research-active teams; and level of carer involvement based on the number of carers willing to take part in a face-to-face interview.

The feasibility of conducting the study procedures was also assessed through semistructured face-to-face interviews with participants, carers and study nurses. Interview guides were developed (see Appendix 22) to explore the feasibility of the trial processes.

Choosing a primary outcome measure for a definitive trial

The primary aim of a definitive trial would be to observe a reduction in average pain intensity (measured using the BPI average pain intensity item) and is therefore a candidate primary outcome for a future definitive trial. However, the complexity of symptom control within an end-of-life population may mean that the average pain intensity scores do not improve over the course of the study period despite the participant having received some benefit from using the intervention with their study nurse. It is recognised that changes in pain interference (measured as a composite score of the seven interference items on the BPI) and self-efficacy (measured on the SES), may be more responsive outcomes. As outlined in the logic model presented in Chapter 2 (see Table 7), the SMART intervention is aimed at improving medicines self-management behaviours; therefore, it is possible that participants may have experienced improvements in self-efficacy and interference from pain (mediated via information provision, goal-setting and regular review and couching) without any direct improvement on pain severity score. Indeed, stability in pain severity score over the study period may indicate improvements in overall medicine self-management in the context of declining health. Therefore, the BPI pain intensity, pain interference measure and the SES self-efficacy score were assessed as candidate primary outcomes for a definitive trial.

Fidelity of intervention delivery

Intervention delivery ‘as planned’ was defined as:

• initiation within 7 days of baseline

• a minimum of three SMART study nurse visits over the 6-week follow-up period

• assessment of participants’ self-management needs/requirements (step 1)

• tailored information resource provision from the SMST (step 2)

• self-management goal-setting and action-planning (step 3)

• regular review and coaching of self-management goals (step 4).

To assess whether or not the SMART intervention was delivered as planned, nurses were asked to record the details of each SMART visit on a standardised CRF. This form captured information on:

• self-management needs identified and discussed

• intervention factsheets provided to participants

• self-management goals set or reviewed

• changes made to participants’ analgesic medication and who made the change

• details of any additional contact with participants between SMART visits.

The number of CRFs completed per participant was used as a proxy indicator of the number of SMART study nurse visits each participant has received. When study nurse CRFs were missing, participants’ clinical records were searched for evidence of a SMART visit having taken placed. To capture evidence of goal-setting, the goal-setting sheets were printed on carbon copy paper: the top copy was kept by the participants in their SMST folder, one copy was kept by the study nurses in the participant’s notes and one copy was collected by the researchers (either from the study nurse or from the participant at each follow-up visit) as evidence of goal-setting having taken place. When goal-setting sheets were missing, participants’ clinical records were searched for evidence of goal-setting.

Fidelity of the intervention delivery was also assessed by the researchers. At each follow-up visit the researchers completed a CRF capturing data on which SMST resources had been received by participants since the previous follow-up, whether or not the participants had set or reviewed their self-management goals, whether or not they had watched the video podcasts (if yes, which ones) and whether or not they were still using the SMART intervention.

In addition to the above quantitative assessment of intervention delivery, fidelity was also assessed during interviews with participants (and carers) after they completed 6 weeks of follow-up and with the study nurses at the end of the trial. Interviews included questions focusing on the extent to which the intervention had been delivered as intended as well as any barriers to or facilitators of delivering the intervention or using the intervention.

Finally, in order to develop an ‘intervention delivery fidelity checklist’ to be used in a future national multicentre definitive trial, study nurses were asked if they would be willing to audio-record their patient consultations when they used the SMART intervention. It was intended that these audio transcripts would be judged against a checklist of key elements required for intervention delivery.

Acceptability of the SMART intervention

Acceptability of the SMART intervention was assessed during recruitment by asking all patients who received an information pack whether or not they thought the intervention was acceptable in principle (yes/no response).

During the study period participants’ and carers’ acceptability of the intervention was assessed quantitatively by evaluating the number of participants agreeing to use the intervention compared with the number indicating that they were still using the intervention at each follow-up. Participant and carer engagement with the SMST was assessed at each follow-up visit by the question ‘Are you still using the SMART toolkit?’.

The end-of-study interviews with participants, carers and study nurses (described above) also assessed acceptability and usage of the SMART intervention. Interview guides (see Appendix 22) were developed to explore participants’, carers’ and study nurses’ experience of using the intervention.

Qualitative data: analysis of interview data

The audio files from the interviews were professionally transcribed. They were listened to alongside the transcripts by the researchers (NC and MM) to check for complete accuracy and ensure data familiarity. The data were coded utilising framework analysis62 by indexing transcripts for all issues relevant to the feasibility of conducting the trial processes (i.e. deliverability), acceptability and usage of the intervention, perceived benefits and any potential disadvantages of both the research design and the intervention.

The interview data were analysed within a framework designed for the study based on the recommendations of Moore et al. 76 regarding process evaluation of complex interventions. The analysis framework, and the ultimate higher level of analysis, focuses on the mechanisms of action, the participant, carer or study nurse responses to the research design or intervention, the mediating factors and the consequences.

The analyses of qualitative findings are presented in full in Appendix 24. These analyses focus on the feasibility and deliverability of the trial processes, acceptability and usage of the intervention, perceived benefits and possible disadvantages of both the research design and the intervention itself, including the four-step educational approach and the SMST.

Quantitative data

Unless otherwise stated, all percentages were calculated using the total number of participants (or forms completed) within the relevant population, which was the denominator (i.e. excluding all participants with missing data for that variable). All percentages, means, medians, interquartile ranges (IQRs), ranges, SDs, standard errors and 95% CIs will be rounded to one decimal place (or two significant figures for numbers < 0.1). To account for the variation in the amount of intervention received by participants, all calculations and analyses were performed on an intention-to-treat basis (i.e. based on all consented participants). This pragmatic approach to include all consented participants was taken as it more closely reflects the real-life situation in which the amount of palliative care support received by patients varies. All calculations and analyses were carried out using Stata^® version 12 (StataCorp LP, College Station, TX, USA).

Assessing feasibility of conducting trial procedures

Fidelity and acceptability of the SMART intervention

Data accuracy

To evaluate the accuracy of data input, a random check of 20% of the data entered into the trial database was carried out prior to data cleaning and analysis. This process identified that data entry accuracy was very high: > 99% across all the questionnaire and study CRFs.

Participant flow

Figure 3 presents the Consolidated Standards of Reporting Trials (CONSORT) flow diagram of participant recruitment from screening through consent and intervention delivery to follow-up completion. In total, 417 patients were screened against the eligibility criteria (Figure 4), of whom 103 (24%) were eligible. Of the eligible patients, 37 (36%) were approached, of whom 22 (59%) were interested and 19 (51%) consented and were recruited and are hereafter referred to as participants. Of the 19 consented participants, 15 (79%) completed 6 weeks of follow-up. A total of 17 participants (89%) received the intervention.

FIGURE 3.

The CONSORT flow diagram of recruitment. a, Calculated as the proportion of consented participants.

FIGURE 4.

Weekly screening rates by referral status.

Over the 4 months of the trial, the recruitment rate was 4.75 consented participants per month. Figure 5 presents the weekly recruitment figures for each site (bars) and the weekly cumulative accrual rate against target accrual rate (dotted line). The median (range) days between approach and consent was 7 (4–39) days.

FIGURE 5.

Weekly recruitment figures by site.

Recruitment sites

HANTS1 and HANTS2 were the two community palliative care services in Hampshire; YKHB1 and YKHB2 were the two community palliative care services in Yorkshire and the Humber. Stacked columns in Figure 5 show the weekly recruitment rate by site. The dots and lines show target cumulative accrual rate (black line) and the actual cumulative accrual rate (blue line).

Eligibility criteria

Table 9 summarises the screening activity undertaken across all sites and presents basic demographics for patients screened. Based on previous studies undertaken by the research team, we had assumed that 66% of screened patients would be ineligible for participation; however, 314 out of the 417 patients (75%) screened were ineligible. The primary reason was not having been prescribed strong opioid analgesia (Table 10). Our screening procedure did not stipulate rescreening of ineligible patients, which may have identified patients who had subsequently been prescribed opioid analgesia. The researcher field notes taken during the screening process identified that many patients were ineligible because they were not prescribed strong opioids but were prescribed weak opioids for pain.

The interviews with study nurses revealed that the eligibility criteria were initially seen as acceptable by being relatively broad. However, the study nurses were surprised by the lack of individuals who met the eligibility criteria. Moreover, those who did meet the eligibility criteria often had complex end-of-life needs and would have been ‘on their reserves to do it [very fatigued and only just able to participate]’ (H2CNS002). One study nurse responded that ‘It’s interesting that . . . there’s quite a lot of patients that aren’t even on opioids’ (H1CNS002). Another study nurse indicated that ‘I was surprised in a way that I didn’t have any patients that would be applicable for the study . . . maybe, that shows how complex the patients are that come to us, and how quickly people who may possibly have been suitable deteriorated’ (H1CNS004).

A mediating factor in the low eligibility rate was the number of study nurses involved in the study. Across the four sites there were 37 full-time equivalent CNSs, of whom 12 (33%) volunteered to take part in the SMART study. Consequently, the pool of patients available to screen was smaller than expected and those who were ‘eligible’ often had very complex needs that were not captured by the broad eligibility criteria.

Patient referral rate

We had assumed that approximately 450 new patients would be referred to the four community palliative care services during the recruitment period (a rate of 112 per month). As we recruited only 12 nurses to take part in the study, the actual number of new referrals from these 12 study nurses over 4 months of recruitment was 202 (a rate of 50 per month).

Screening and initial approach processes

The initial screen for eligible patients included screening all existing patients on the study nurses’ caseloads. Subsequent caseload screening was of new referrals (see Figure 4). Consequently, 54% (n = 227) of all patients were screened within the first 3 weeks of recruitment, resulting in 70% (n = 72) of all eligible patients being identified within the same 3-week period. Given that study nurses’ appointments with patients were usually weekly, fortnightly or ‘as required’, they had to prioritise which patients they approached first based on the next appointment date. This meant that many eligible patients identified in the first 3 weeks of the trial were not approached (reasons described below), or by the time the study nurses were able to see them had become ineligible because of declining health (i.e. they were not anticipated to survive more than 6 weeks).

Just over one-third of eligible patients were approached and invited to participate in SMART (see Figure 3). Reasons for the low approach rate were largely missing because of low CRF completion rates (71% missing; Table 11); however, end-of-study interviews and researcher field notes provided more explanatory evidence. Study nurses indicated during interviews that approaching eligible patients was not a problem in itself (i.e. patients were not intrinsically opposed to finding out about the research). However, the often complicated and rapidly fluctuating circumstances of the end-of-life context meant that it was often not appropriate to approach eligible participants about participating. In addition, some patients had already been approached about participating in other drug trials and study nurses felt that they ‘couldn’t burden them with something else at that time’ (H1CNS002). Many eligible patients, as a result of their complex situation, were frequently admitted to alternative care settings (hospital, hospice or nursing home) for symptom control or respite care; therefore, study nurses found it challenging to find a relatively stable period during which to approach them about participating.

Another factor affecting the screening and approach processes was the contact time between study nurses and researchers or CRNs to undertake screening activity, which was not consistent across all sites due to pressure on study nurses’ time and the size of their caseloads. The number of patients screened per CNS varied from 17 to 68 (median 36). Maintaining weekly appointments with each CNS to screen new referrals and review recruitment of eligible patients was challenging and, although the screening process was seen as deliverable and acceptable by the study nurses, they were very aware that they (and sometimes the CRNs) were not available when they had said they would be for screening appointments. One that responded: ‘It was only tricky because of time’ (H1CNS002). Existing pressures on study nurses and their often high caseloads meant that screening appointments frequently needed to be rearranged/reattempted, with the result that screening was missed on some weeks. One study nurse stated she found the screening onerous and two said they would have preferred to screen with just the researcher (rather than screening with two people, e.g. the CRN and research fellow). Consequently, screening was undertaken regularly, but not always on a weekly basis.

Participant recruitment

Despite the low approach rate, the proportion of patients interested in participating after reading the information sheet was higher than expected (60%; see Figure 3). The majority of those who refused to participate did so because they felt that they were too unwell at the time of being asked. Participant interviews revealed two primary motivations for taking part:

1. It was an opportunity to give something back and to help others.

2. It was a way for them (and their carer) to learn something new that might help them to manage their pain.

Of the 37 patients who were given an information sheet, 19 (51.4%) consented to participate (hereafter referred to as participants), of whom nine (47%) also had a carer who consented.

Nurse recruitment

Across the four sites, 12 CNSs were trained to deliver the SMART intervention with their patients. The demographic details of the 12 study nurses are summarised in Table 12. Eleven of the 12 nurses completed an interview with a researcher at the end of the trial about their experiences of being part of the trial and delivering the intervention.

Three study nurses (25%, two in Hampshire and one in Yorkshire and Humber) recruited no participants. Three study nurses (25%) recruited one participant each, three (25%) recruited two participants each, two (17%) recruited three participants each and one (8%) recruited four participants.

Study nurses were asked whether or not they would be willing to audio-record their consultations with participants when using the SMART intervention to enable the research team to develop an intervention delivery checklist to assess fidelity. This was generally met with apprehension by the study nurses, although the majority agreed to do it. To allow study nurses time to get used to using the intervention it was decided that they would start recording SMART intervention consultations with their second participant. However, the completion rate for this part of the study was poor, primarily because half of the study nurses recruited one participant or none. In the end, only two nurses (in Hampshire) audio-recorded their SMART intervention consultations; however, this provided insufficient data to test an intervention delivery checklist. For a future definitive trial, alternative methods should be considered.

Baseline participant characteristics

Participants’ baseline demographic factors and clinical characteristics were broadly similar across the four sites (Table 13). Participants recruited from the two Hampshire sites had slightly lower (worse) Australia-modified Karnofsky Performance Scale score than the two northern hospices. There was a range of disease types, with the most common being breast cancer (26%), followed by liver or pancreatic cancer (16%). One patient had a non-cancer diagnosis (liver cirrhosis). Although the study was open to patients with any type of advanced disease, we found that most of those on strong opioids, and, therefore, eligible for the study, were patients with cancer.

There were differences in the median time from diagnosis of advanced disease to referral to palliative care services between the Hampshire and Yorkshire recruitment sites: the median (IQR) time between diagnosis and referral in Hampshire was 4 (IQR 2–61) weeks compared with 38.5 (IQR 9–59) weeks in Yorkshire and the Humber. The end-of-study interview data with the study nurses from both Hampshire sites emphasised that late presentation at diagnosis (i.e. more advanced disease) meant that many patients referred to palliative care services were close to the end of life and subsequently had limited exposure to specialist palliative care services.

Across all four sites, the most common reasons for referral to palliative care services were for pain control and psychological support. Just over half of participants (n = 10) were undergoing palliative treatment at the time of enrolment. Six patients were receiving palliative chemotherapy and three were receiving palliative chemotherapy plus radiotherapy.

At baseline, all participants were prescribed at least one strong opioid for pain relief (Table 14). Two participants (10%) had one strong opioid prescription, 15 participants (79%) had two strong opioid prescriptions (for background and breakthrough pain), one participant (5%) had three strong opioid prescriptions [two background (tablets + patch) and one breakthrough] and one participant (5%) had five strong opioid prescriptions. Of the 15 participants who received two strong opioid prescriptions, three (16%) were also prescribed a weak opioid. Co prescribing rates of laxatives, antiemetics, neuropathic analgesics and non-opioid analgesics were 74% (n = 14), 79% (n = 15), 68% (n = 13) and 74% (n = 14), respectively.

Randomisation acceptability

During the baseline visit, participants’ general willingness and acceptability of randomisation was assessed by asking the question, ‘If the study had been designed so that those taking part would be randomly selected to receive either the intervention or standard care would you have taken part?’.

A total of 17 out of 19 participants (89%) responded to this question by saying that they would. One participant did not understand the concept of randomisation and, therefore, preferred not to comment either way. One participant–carer dyad indicated anxiety about the relinquishing of control associated with randomisation and that maintaining control of ‘critical decisions in palliative care’ was important to them and this was the reason why they would not have agreed to randomisation.

Researcher follow-up

All follow-up visits were conducted by one of two researchers, face to face at participants’ homes. Overall, participant retention was high: of the 19 participants who consented, 15 (79%) completed 6 weeks of follow-up (see Figure 3). One participant completed all follow-up visits with a researcher but did not receive the intervention because of complex pain management issues that prevented the study nurse from devoting time within the consultation to initiating the use of the SMART intervention. Therefore, of the 19 consented participants, 14 (74%) received the intervention and completed 6 weeks of follow-up. However, an intention-to-treat analysis approach was taken to account for the variation in the amount of SMART intervention received by participants; therefore, all consented participants were included in the analysis. As such, the denominator at baseline was 19 and at all follow-up time points was 15.

Withdrawals

The end-of-life context meant that many patients experienced uncontrolled symptoms (not just pain) and infections during the course of the 6-week study period. Despite this, relatively few participants were lost to follow-up. There were four dropouts, all associated with a rapid unexpected decline in health:

• One participant died 1 week after consenting, having received no SMART study nurse visits (only baseline data obtained).

• One participant died between baseline and the 2-week follow-up, having received two SMART study nurse visits (only baseline data obtained).

• One participant withdrew from researcher follow-up visits because of uncontrolled pain in week 2 (prior to week 2 follow-up), but continued to use the goal-setting element of the SMART intervention with the study nurse on four more occasions (baseline data obtained + evidence of four goal-setting sheets from the study nurse).

• One participant withdrew between baseline and the 2-week follow-up because of declining health and was subsequently moved to a nursing home where they were no longer managing their medicines; this participant received one SMART study nurse visit (only baseline data obtained).

In order to capture accurately why, when and from what participants withdraw, a future trial should explicitly record what parts of the trial participants are withdrawing from (i.e. the research elements or the intervention elements).

Acceptability of study length and frequency of follow-up visits

The 6-week study period was universally seen as ‘about right’ or ‘just right’ by the participants and their carers: ‘I think it’s just about right actually, you probably need that length of time to get any results’ (H2Pt019) and ‘It seems to have gone quick’ (H1Pt001-C). The study nurses also appeared to agree on the acceptability and deliverability of a 6-week study period.

In terms of the deliverability of follow-up appointments with researchers, Figure 6 shows that in the majority of cases it was feasible to conduct three follow-up visits within a tight time frame of 2 days either side of the scheduled fortnightly follow-ups. Nevertheless, it was not always possible to conduct follow-up visits within this tight time frame (see Figure 6). This is not surprising given the complex nature of participants’ homes.

FIGURE 6.

Time elapsed between baseline and researcher follow-up visits.

Interview data revealed that participants and carers all stated that the frequency of researcher visits (fortnightly) was acceptable, and some looked forward to these visits. Participants responded, ‘It’s been very nice you coming in’ (H2Pt019) and ‘It’s been lovely. I enjoy you coming . . . you’re so easy to talk to, you ask the right questions’ (H1Pt009). These responses indicate that people approaching the end of life valued taking part in a research study and that they saw the researcher as external to their care team and were able to develop a positive relationship with them over the 6-week duration of the study.

Participant and carer interviews

Thirteen participants completed an interview with a researcher (Table 15). Although nine carers consented at baseline to participate in an interview, only seven were available to complete an interview with a researcher.

Survival following study entry

At the close of the study, three participants were known to have died. Two participants died during the follow-up period at 21 days and 14 days following baseline assessment, respectively. One patient died after the follow-up period, 71 days following baseline assessment. Two participants withdrew from the study follow-up data collection for reasons of declining health; the number of days between study entry and withdrawal was 14 days in one case and 27 days in the other. These data are summarised in Figure 7.

FIGURE 7.

Survival (days) for each participant.

Intervention delivery

To what extent was study nurse training provided as planned?

During the interviews, study nurses were asked to comment on the training workshop (described in SMART intervention). Responses to the experiential, reflective style of the training workshops were mixed; they varied widely from generally or overtly positive through to neutral and negative responses. The reflective nature of the session appealed to some, whereas others found it challenging and preferred alternative approaches (e.g. along the lines of advanced communications training, accompanied by a video of a modelled conversation and then subsequent group discussion). However, critical reflection on existing practice is necessary to stimulate change, which is fundamental to the intervention delivery. 78,79 Nevertheless, there was a general view that the sessions covered what the nurses needed to know in anticipation of their study involvement.

There was a mixed reception to the underpinning self-management ethos of the sessions. During the workshops, the Johnston et al. 51 definition of self-management support in palliative nursing was explored, as well as the related eight professional roles from advocate to reporter (see Appendix 25). Some of the study nurses appeared to readily understand the supported self-management ethos, whereas others were more challenged by it. The self-management ethos of the study and the training sessions challenged the professional behaviours and identity of some of the specialist palliative care nurses, whose therapeutic role is often measured by their effectiveness to ameliorate pain. The nurse interviews also revealed that it was harder for some nurses to adapt their practice from imposing their views (i.e. telling patients how to use opioids) to collaborative discussion. Such discussion focused on the individual patient using and developing their own self-management strategies as result of the information provided by the nurse.

There was feedback that the four-step educational approach was viewed as akin to usual specialist practice, but some nurses felt that the workshop sessions made the four steps seem more complicated than it actually was. The nurses generally felt that the four-step approach mirrored normal practice and, consequently, was not viewed as being distinctive or novel. The study nurses reported that they valued the training materials supplied during and after the training workshop (see Appendices 18 and 19, and Figure 8). In particular, the nurses liked the figure illustrating the self-management conversation prompts mapped to the four steps of the educational approach (see Figure 8). Despite this, there was variation in the extent to which nursers referred to the training resources over the course of the study, with the majority of use being at the start of the trial.

FIGURE 8.

Diagram of the SMART four-step educational process.

There was insufficient opportunity during the training workshop for the study nurses to become familiar with the research process, which led to some anxiety during the workshop. This had an impact on time available for role modelling a conversation in practice. Therefore, a future definitive trial should consider providing an opportunity to receive information and ask questions about the research process prior to any training.

Based on our finding of a mixed reception to the self-management ethos, it appears important that nurses process the intervention and rehearse how they deliver self-management-focused conversations that fit their own practice. There are likely to be different timelines regarding adoption of such a focus, and nurses vary in how much they need to practise this before it becomes embedded. Therefore, some form of further training and support for those involved in delivering the intervention is desirable and could be combined with efforts to assure intervention fidelity.

The busy and time-pressured reality of clinical practice for these specialist nurses, who often managed large caseloads, meant that the delivery of ongoing training and support during the course of the trial was often problematic. These issues were captured in the researchers’ field notes and included difficulty making appointments to meet the study nurses. Visits usually had to be made at the start of their working day. However, not all the study nurses visited their office before going out to visit patients, and none wanted to make appointments at the end of the day. Furthermore, visits to the study nurses were complicated by nurses covering weekend working and, therefore, having days off in the week. Once appointments were made, the study nurses were not always subsequently available (because of extended/unexpected patient visits, over-running meetings, their own illness, etc.). Nonetheless, when they were available, they were usually open to discussing their experiences of delivering the intervention (utilising the four-step educational approach), but time devoted to this was always pressured or limited. Over the course of trial, the study nurses were usually visited by the study researchers once a fortnight depending on their availability, which was less often than the weekly frequency originally intended.

To what extent was the intervention delivered by study nurses as planned?

Table 16 summarises the pattern of the intervention delivery for each participant. Overall, there were 52 SMART study nurse visits across all participants, evidenced by completion of study nurse CRFs. When study nurse CRFs were missing, evidence of a SMART intervention visit having taken place was gathered from participants’ clinical records, the presence of a goal-setting sheet having been completed (with date) or from researchers’ field notes. The final column in Table 16 groups the participants based on whether they received the intervention as planned (group A), partially (groups B and C) or not at all (groups D and E).

TABLE 16 - Pattern of SMART intervention delivery: who got what when

Data from researcher follow-up CRFs: 1, managing pain with opioids; 2, contacts and further information; 3, getting hold of prescriptions; 4, organising medicines; 5, fitting opioids into daily routine; 6, checking opioids are managing pain; 7, common concerns; 8, keeping on top of side effects; 9, pain diary; 10, medication chart; and 11, goal-setting sheet.

A, received the intervention as planned, i.e. started within 7 days of baseline data collection, received at least a minimum of three SMART study nurse visits, received tailored staged information provision, goal-setting and regular review and coaching; B, received the intervention as planned (minimum of three SMART study nurse visits + goal-setting + review and coaching), but unclear if information provision was tailored to participants needs; C, received goal-setting, review and coaching as planned, but did not receive minimum of three SMART study nurse visits or staged information provision; D, did not received any elements of the intervention.

Participant withdrew from researcher follow-up after 14 days, but continued to use the goal-setting element of the SMART intervention with study nurse.

Participant died after receiving two SMART study nurse visits.

Participant died after receiving one SMART study nurse visit.

Participant did not receive the intervention from their study nurse, but did complete all follow-up visits with a researcher.

Participant died before receiving first SMART study nurse visit.

ID – site	Start 7 days?	Number of SMART visits	Average visit time (minutes)	New resources at each follow-upa			Number of factsheets received	Number of goal-setting sheets received	Podcast films watched?	Intervention delivered as plannedb	Complete all follow-up visits?
				Week 2	Week 4	Week 6
1 – YKHB2	Yes	3	62	1, 5, 7, 10, 11	6, 11	2, 3, 4, 8, 9, 11	10	3	No	A	Yes
2 – YKHB2	Yes	3	60	1, 7, 8, 11	4, 9, 10, 11	2, 3, 5, 6, 11	10	3	Yes	A	Yes
3 – HANTS1	Yes	4	47	2, 3, 7, 11	10, 11	11	4	3	No	A	Yes
4 – HANTS1	Yes	5	44	2, 3, 8, 11	5, 11	11	4	5	No	A	Yes
5 – HANTS1	Yes	4	71	1, 2, 4, 8, 10, 11	3, 7, 11	11	7	4	No	A	Yes
6 – HANTS1	Yes	3	47	1, 2, 11	7, 8, 9, 11	11	5	3	No	A	Yes
7 – HANTS1	Yes	4	55	1, 2, 10, 11	8, 11	11	4	3	No	A	Yes
8 – HANTS2	Yes	3	62	2, 3, 5, 11	11	9, 10, 11	3	3	Yes	A	Yes
9 – HANTS2	Yes	4	52	2, 7, 9, 10, 11	11	11	4	4	Yes	A	Yes
10 – YKHB1	Yes	2	60	11	11	11	0	2	No	A	Yes
11 – YKHB2	Yes	3	45	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11	11	11	10	3	No	B	Yes
12 – YKHB1	Yes	3	60	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11	11	11	10	3	Yes	B	Yes
13 – YKHB1	No	3	60	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11	11	11	10	3	No	B	Yes
14 – YKHB1c	Yes	4	22	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11	–	–	10	4	No	B	No
15 – YKHB1	No	1	60	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11	–	–	10	1	Yes	C	Yes
16 – YKHB2d	Yes	2	70	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11	–	–	10	1	No	C	No
17 – HANTS2e	Yes	1	95	7, 8, 10, 11	–	–	3	1	No	C	No
18 – YKHB1f	No	0	0	–	–	–	0	0	–	D	Yes
19 – HANTS1g	No	0	0	–	–	–	0	0	–	D	No

Of the 19 participants who consented, 17 (89%) had a first SMART study nurse visit and started using the intervention (see Table 16). From this point, 10 participants (53%) received the intervention as planned (group A). One participant in group A decided not to have the factsheets after reading through them, but did engage with the goal-setting. As the information provision was designed to be tailored to the participants’ needs, this still met the criteria for this element of the intervention. Four participants (21%, group B) received the minimum number of SMART study nurse visits, goal-setting, and review and coaching, but the information provision was not staged as they received all the factsheet on their first SMART study nurse visit. This does not necessarily represent inappropriate delivery of the intervention resources; however, it was not clear whether all the factsheets were delivered on the first visit at the participant’s request (which would satisfy the criteria for tailored provision of information) or whether the study nurses handed them all over in one go. One participant in group B [identification (ID) 14 – YKHB1; see Table 16] withdrew from researcher follow-ups because of uncontrolled pain but continued to engage with the goal-setting and reviewing (steps 3 and 4 of the intervention). Three participants (16%, group C) did receive the intervention materials and the goal-setting, but did not receive the minimum number of SMART study nurse visits. However, in group C one participant died during the first 2 weeks of the study and one participant was lost to follow-up because of rapidly declining health (see Table 16). Finally, two participants (10.5%, group D) did not receive any elements of the intervention: one participant died within a week of giving consent and the other did not receive the intervention (because of complex pain management issues that prevented the study nurse from starting the SMART intervention), but did complete all researcher follow-up visits.

The timing of study nurse visits is summarised in Figure 9, which shows that participants generally received a SMART study nurse visit once a week or once a fortnight. The average duration of SMART study nurse visits was 57.2 minutes (SD 15.1 minutes), but this covers the whole consultation and not just time spent using the SMART intervention. Interview data with study nurses revealed that they were able to successfully deliver the SMART intervention during their visits and the pattern of delivery was acceptable as it matched their normal visiting pattern. However, in relation to the acceptability of the length of the SMART study nurse visits, the nurses were very conscious of the extra time required for study visits (approximately 30 minutes for the first visit and 15 minutes for each further visit) and the impact this had on their workload. One study nurse responded:

We [one study nurse and another study nurse] talked about timekeeping a lot, because in my first couple of SMART study appointments they were really long, partly because the lady I had, it was quite difficult to keep to time with them anyway.

H2CNS001

FIGURE 9.

Timing of SMART study nurse visits.

The study nurses managed to successfully accommodate this additional workload without changes to their usual care patterns, except for one (part-time) study nurse who asked a colleague to follow up some patients on her caseload by telephone on a single day because of a SMART study visit. The extra time required for SMART study nurse visits led the study nurses at the individual sites to discuss this impact with one another.

Acceptability of the intervention

Acceptability of the four-step educational approach

The deliverability of the four-step approach was seen as acceptable to the study nurses, who did not perceive that they needed any additional skills to deliver it; it was viewed as a normal part of the specialist role. Still, study nurses perceived value in the explicit nature of the four-step educational approach as it formalised and gave structure to supporting participants and their carers to self-manage medicines within the complexity of the end-of-life context with frequent deterioration of health and depression.

All participants stated that they derived a benefit from the intervention, but the extent to whch the participants were aware of, or acknowledged, the four-step approach varied. The acceptability of the delivery of the intervention and resulting benefit were increased for the participants and their carers because of the almost universal value that they placed on contact with their study nurse (particularly when face to face and in their own homes).

The second part of the four-step approach was provision of information tailored to patients’ needs. Frequently, delivery of the SMST resources (i.e. the factsheets) by the study nurses was less than ideal, which, as a result, had an impact on the acceptability and benefit of the resources, particularly when they were not discussed or talked through, or when all the factsheets were given all together as a large file. There was also difficulty in providing the intervention at the most appropriate time for participants (and their carers) while they were well enough to engage in it, given the unpredictability of end-of-life context.

Overall, the four-step educational approach helped to stimulate suggestions for self-management strategies and then enabled patients, carers and study nurses to determine which strategies worked for the individual. In addition, the intervention as a whole (see Box 3) stimulated appropriate questioning by the patient or carer to the CNS; for example, ‘it’s been easy to ask questions’ (H1Pt001). Participants who were already effective self-managers prior to the trial felt that they would have used the materials more if they had received them earlier (at their first contact with their CNS). For example:

I’ve been doing this for over a year. And a lot of the things in here I knew. And frankly, I’m a very organised person, so I’ve got all the diaries, and I’ve got all the prescriptions, and they’re all online. Whereas it takes a while to get to that stage, to figure out what you’re meant to be doing, how you’re meant to be doing it.

H3Pt002

Completion of study nurse case report forms

Study nurses were asked to complete a brief CRF after each SMART visit, documenting what had been delivered. Completion rates were high; study nurses’ CRFs were completed in 43 out of 52 SMART intervention visits that were delivered across the whole feasibility study (83% completion rate). For SMART study nurse visits for which study nurse CRFs were missing, information on the date of visit was gathered from participant clinical records and evidence of the four-step educational process and the use of the SMST tool resources (including whether or not goal-setting sheets had been completed) was gathered by researchers at follow-up visits. For four participants [all from the same site in Yorkshire and the Humber (YKHB1)] no study nurse CRFs were completed.

Exploratory analysis of participant self-reported outcomes

Descriptive statistics (mean and 95% CIs) of the participant self-reported outcome measures at each time point and at 6 weeks compared with baseline (difference) are presented in Table 18 and summarised overall by time point. Owing to the small number of participants taking part, further outcome summaries by potential confounders such as disease state, age, sex, level of support and recruitment site, were not undertaken.

This study was not powered to detect any changes in outcome measure score, and there was no change in average pain scores; however, there was a slight reduction in interference from pain (–1.6, 95% CI –2.8 to –0.4) and a modest increase (0.7, 95% CI 0.3 to 1.2) in self-efficacy scores (see Table 18). There was no overall change in the intensity of common end-of-life symptoms (ESAS), health-related quality of life [EuroQol-5 Dimensions (EQ-5D)] or satisfaction with information about medicines (SIMS).

Figures 10–12 present histograms of the change in average pain, pain interference and self-efficacy scores, respectively. These histograms show that, for the majority of participants, average pain intensity worsened over the study period (i.e. scores of > 0), whereas there was greater stability or improvement across the board on BPI pain interference (i.e. scores of ≥ 0) and, largely, there were improvements for all participants on the self-efficacy scale (i.e. scores of > 0). Table 19 presents the variability (SD) with 95% CI for candidate primary outcome measures, along with the estimated effect size for the change in average pain, pain intensity and self-efficacy at 6 weeks compared with baseline.

FIGURE 10.

Histogram of change in BPI average pain intensity scores. N.B. change scores calculated week 6 score minus baseline score.

FIGURE 11.

Histogram of change in BPI pain interference score. N.B. change scores calculated week 6 score minus baseline score.

FIGURE 12.

Histogram of change in SES score. N.B. change scores calculated week 6 score minus baseline score; > 0 indicates improvement.

The number, proportion and 95% CI around the proportion of participants with clinically meaningful reduction in average pain and pain interference are summarised in Table 20. These data show that at follow-up weeks 2 and 6 there were more responders based on pain interference than average pain intensity.

Participant acceptability of self-reported outcomes

Generally, participants’ experience of completing the self-reported outcome measures was acceptable; however, there were some limitations related to the wording of some questions given the end-of-life context (e.g. ‘normal work’ and ‘enjoyment of life’ on the interference subscale of the BPI). A small number of participants criticised the ‘duplicity’ (H4Pt001) of some questions, given the combination of five different measures. Overall, there was dislike for the SIMS as many participants experienced difficulty remembering specific information that they had received about their medicines over the preceding 2 weeks.

Overall, completion rates for the questionnaire packs were high: all questionnaires were completed by participants with a researcher at each time point. The proportion of missing data at individual item level was very low (Table 21) and did not prevent any summary scores from being calculated at any time point. The proportion of missing data was < 3% for all participant self-reported outcome measures at all time points, except for BPI at the week 4 follow-up (3.9%). The most commonly missing item on the BPI was the final item about the extent to which pain interferes with enjoyment of life (missing in three cases from the same participant); a number of participants responded that they felt that this question was inappropriate for people approaching the end of life. Similarly, for the EQ-5D, one response to the final item asking respondents to rate their overall health from best to worst was missing at all time points (from the same participant in each case). The field notes kept by the researchers identified a general lack of acceptability of this question by participants. One participant responded during the end-of-study interview, ‘it’s a stupid question to ask people in palliative care’ (H1Pt011-C), when asked specifically about this response item. The SIMS was least liked by the participants; nevertheless the proportion of missing data was extremely low.

Resource use

Data collection forms to identify the health and social care services that individuals used were completed by researchers using clinical records. The form, which was adapted from one used in the IMPACCT study, is included in Appendix 3.

Quality of life (utility)

In order to calculate quality-adjusted life-years (QALYs), it is necessary to collect data on health state utility. The EQ-5D is NICE’s preferred measure of health state. The feasibility of using the new EQ-5D-5L was explored. 81 The EQ-5D-5L formed part of the interview-administered questionnaire and was scored using a newly developed UK tariff. 82

The small sample size and absence of a control group meant that an economic evaluation based on patient-level analysis of the data was not possible. Thus, estimates of cost-effectiveness were based on a decision-analytic modelling (DAM) approach.

Model structure, cycle length and time horizon

The SMART DAM (Figure 13) was adapted from the IMPACCT DAM and is a Markov model with weekly cycles that runs for a time horizon of 52 weeks. The model structure, time horizon, cycle length and parameters were developed using the findings from a model literature review and expert and patient opinion. The model was developed in line with current best-practice standards. 84,85

FIGURE 13.

Diagram of the SMART DAM.

The DAM comprises health states based on level of pain severity (no/mild pain, moderate pain, severe pain) and death (see Figure 13). Within each of the pain health states, the occurrence of side effects (i.e. constipation, drowsiness and nausea) was permitted.

In the DAM, a cohort of hypothetical patients (mean age 72.4 years) who have advanced cancer and pain move (or transit) through the health states in accordance with specified transition probabilities. Each health state has a mean cost and utility value associated with it so that the cohort members accrue QALYs and costs as the 52 weeks pass. Patients can move between pain states and between pain states and death. Side effects are not represented as separate health states.

Model parameters

The model parameter values are described in Tables 19 and 21. They were derived from a number of sources, including the IMPACCT patient survey data, data from the literature, analysis of palliative care patient survival data and analysis of a previous advanced cancer trial. 86

Cost and utility parameters

The model uses resource use and utility data collected from participants during the NIHR-funded IMPACCT study. Between August 2013 and June 2014, 248 patients were recruited to the study and completed a survey. Community-based patients with pain from advanced cancer who were aged ≥ 18 years were eligible for the study. Patients with advanced cancer were defined as those patients with metastatic cancer (histological, cytological or radiological evidence) and/or those receiving anticancer therapy with palliative intent. Patients with pain were defined as those receiving analgesic treatment for cancer symptom-related and/or therapy-related pain. Patients had to be able to complete the questionnaires and provide informed consent to participate. Thirteen palliative care services across England recruited patients to the study. The participants completed a resource use questionnaire capturing health-care use, pain rating scales, the EuroQol-5 Dimensions, three-level version, and additional utility measures: the European Organisation for Research and Treatment of Cancer 8 Domains (EORTC-8D)87 and ICEpop CAPability measure for Adults (ICECAP-A). 88

The resource use questionnaire asked participants to recall use of primary and community care (e.g. GP visits and nurse contact) and secondary or hospital care (e.g. visits to accident and emergency and hospice stays) in the previous 4 weeks. Unit costs were assigned to the data in order to estimate the average cost of health and social care service use for the sample. Unit costs were obtained from national sources including the Personal Social Services Research Unit (PSSRU) Unit Costs of Health and Social Care 2015,89 NHS Reference Costs 2014–201580 and the British National Formulary. 90

Mean cost and utility estimates (with variance) were estimated for each of the model pain health states (Table 24). Individuals were classified into pain health state using a 0–10 pain severity rating scale, where:

• 0–4 = no/mild pain

• 5–6 = moderate pain

• 7–10 = severe pain.

The impact of side effects (nausea, constipation and drowsiness) were estimated using a previous trial data set that included EQ-5D, cost and side effect data. COUGAR II86 was a trial of chemotherapy compared with active symptom control for those with refractory oesophagogastric adenocarcinoma. 92 These data were thought appropriate as the participants were at the end of life and had similar characteristics to those in the survival estimation and IMPACCT survey samples. 91 A regression model was run predicting, in turn, EQ-5D and costs, and using side effect identifiers based on European Organisation for Research and Treatment of Cancer – Quality of Life Questionnaire C30 (EORTC-QLQ C30) questions as predictors. The beta coefficients on each of the predictors denote the utility decrement or cost impact associated with each side effect. These were applied in an additive way in the model.

The development and implementation of the new tool were costed following consultation with the SMART study researchers. In addition to the participant-completed data, the resources associated with development and delivery of the SMART intervention were recorded based on routine data, such as administrative records and participant records, as well as a detailed description of the intervention costs. The costs of material development, printing and nurse training are included in Table 25.

Survival parameter

Although it was assumed that neither of the interventions compared in this evaluation influences mortality, it was necessary still to estimate background survival in this population. The survival of the hypothetical model cohort was estimated using a parametric regression, which was fitted to data gathered in another IMPACCT workstream. The data (n = 4638, of whom 84% of patients had a cancer diagnosis and 16% a non-cancer diagnosis) were retrospectively collected on all patient referrals to specialist palliative care services in the city of Leeds, West Yorkshire, over a 2-year period (2012–14) and contained variables on date of referral to palliative care, age, sex and date of death. The characteristics of the sample are described in Table 26. There was no censoring of death and a Kaplan–Meier (K–M) curve was estimated for the time between referral and death. Negative numbers were assumed to be errors and, where they occurred, resulted in the cases being dropped.

A number of models were applied to the data, including exponential, Weibull and Gompertz. Visual comparison of the modelled survival curve with the K–M curve and the Akaike information criterion were used to judge model quality. Weibull had the lowest Akaike information criterion (16,891.5 vs. 17,760.15 for exponential and 16,891.5 vs. 17,138.03 for Gompertz functions) and had good fit with the observed K–M curve. Age and sex covariates were tested in the models but only age was found to be significant. The results of the Weibull regression are shown in Table 27 and Figure 14. As the gamma factor was significant, the use of the Weibull model is justified as this indicates a non-constant (and declining) hazard function. The same risk estimates from this analysis were applied to all health states and the Markov model was relaxed to allow these risk estimates to vary over time. The survival model estimates were permitted to vary in the probabilistic sensitivity analysis following Cholesky decomposition for correlated regression parameters. During the 52-week model time horizon, 97% of the cohort were expected to have died. Survival was assumed to be unrelated to pain, and thus the mortality rate was the same for all pain health states.

FIGURE 14.

Weibull estimated survival (weeks after referral to palliative care).

Transition probability parameters

Table 28 summarises the transition probabilities used in the model. The starting proportions of those in each pain health state and of those with each of the three side effects were taken from the IMPACCT patient survey, but allowed to vary in sensitivity analyses. Data from the feasibility study could not reliably inform on the effectiveness of SMART as the study was not powered to do so. In the absence of effectiveness data on the impact of the SMART intervention on pain and side effect management, we relied on expert opinion on the likely effectiveness of the intervention (Professor Michael Bennett, University of Leeds) and on a review of the literature. Given the uncertainty over the intervention effectiveness, we explored scenarios in the model which assumed, for example, that the SMART intervention led to an overall reduction in the proportion of people with side effects of 5% and a weekly reduction in pain status (from moderate and severe to no/mild) of 1% per week. The baseline pain and side effect levels and pain progression were assumed equivalent to those estimated in the standard care arm. To help inform model parameter estimation, literature searches were conducted in June and July 2016 across the databases MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations, EMBASE and The Cochrane Library to identify economic models of behavioural and educational self-management interventions for patients to manage pain and side effects of medications at the end of life.

The change in pain status in standard care was modelled using the COUGAR II trial data set. 86 In COUGAR II, patients completed the EQ-5D measure at 3- and 6-weekly follow-ups during the trial, meaning that a substantial number of longitudinal data were available for this group. The EQ-5D pain and discomfort item response options were considered to be equivalent to the health states in the model defined by the numeric pain scale categories. Thus, the EQ-5D responses ‘I have no pain or discomfort’, ‘I have moderate pain or discomfort’ and ‘I have extreme pain or discomfort’ were assumed roughly equivalent to the pain rating categories of 0–4, 5–6 and 7–10, respectively. Hence, by observing changes in EQ-5D pain item responses over time, we were able to estimate the change in pain status over time at the end of life.

The EQ-5D pain item response was predicted in a multinomial regression with study week number and survival included as covariates. Thus, the coefficient on the week covariate indicates the likelihood of change in pain item response as time progresses, after controlling for survival. The results of the regression are shown in Table 29 and indicate an increase in pain level over time, albeit a small one. Marginal effects were used to estimate the transition probabilities for pain progression using the multinomial results. These results inform only on the change in proportions over time (and not all the possible transitions between health states), and the COUGAR data were insufficient to inform on all possible pain health state transitions. Therefore, we assumed that transitions occurred only from the ‘no pain’ group to moderate and severe pain groups. This background pain progression at the end of life was assumed to be the same in both arms, but the scenario analyses allowed for improvements in health status in the SMART intervention arm.

Cost-effectiveness analysis

The economic evaluation follows the NICE reference case and hence a cost–utility analysis was conducted with a cost per incremental QALY presented from UK NHS and PSS perspectives. The costs are reported in 2015 prices and patient health is measured in terms of QALYs. Cost-effectiveness was assessed using the incremental cost-effectiveness ratio (ICER) and net monetary benefit (NMB) values, and a range of sensitivity analyses were conducted to explore the impact of key model assumptions and parameter uncertainty on the results.

The ICER is calculated by dividing the difference in mean costs between two arms by the difference in mean QALYs between the two arms:

where C_SMART and E_SMART are the expected cost and effectiveness of the intervention (i.e. SMART) and C_UC and E_UC are the expected cost and effectiveness of the usual care arm. The incremental cost and effect of the SMART arm compared with usual care arm are represented by ΔC and ΔE, respectively. The NICE willingness to pay per incremental QALY threshold [(λ) = £20,000] was used to define cost-effectiveness. ICERs < £20,000 are usually indicative of cost-effectiveness. It was assumed that the NICE end-of-life criteria were not met as these require an intervention to deliver an average increase in survival of 3 months over usual care.

We account for parameter uncertainty in non-linear models by assigning probability distributions to each of the input parameters and randomly drawing from these probabilities over the 10,000 Monte Carlo simulations. This probabilistic sensitivity analysis allows the calculation of 10,000 ICERs and informs on the level of uncertainty in the model. The probabilistic sensitivity analysis results were plotted on the cost-effectiveness plane and NMB estimates used to generate the cost-effectiveness acceptability curve (CEAC). 92 The CEAC illustrates the probability that each intervention would be cost-effective given a range of willingness-to-pay thresholds per incremental QALY.

The NMB was derived thus:

Discounting was not required as all costs and benefits were experienced within 1 year. A half-cycle correction was applied to account for the likelihood that model health state transitions occur half-way through the model cycles. All analyses were conducted in Stata software (version 14) and Excel^® (2013; Microsoft Corporation, Redmond, WA, USA).

Resource use

Full resource use data for 6 weeks were collected for 15 of the 19 participants (78.9%). Four participants (21.1%) were lost to follow-up and, therefore, their patient records were not accessed. It was of note that use of health-care services outside the GP practice, for example by the community team, was often not recorded. Further investigation of how these services may be captured is required. Despite these challenges, the researcher-completed outcome measure packs were completed.

There were 164 recorded instances of health-care service use for 15 participants over the 6-week study period. As seen in Table 30, community nurse and palliative care CNSs were the services most frequently accessed.

In order to calculate mean cost of care for each participant over the 6-week period, we assigned unit costs to each service use (see Appendix 26 for unit costs). Cost data were completed for 15 participants. As seen in Table 31, with unit costs applied to service use it can be observed that the main drivers of cost are GP, CNS and district nurse home visits.

The questionnaire also included space to record prescribed medications. These data were collected for the final 2 weeks for 15 patients and included only prescribed medications and not over-the-counter medications and treatments the patient may have paid for themselves. These data have not been included in the descriptive analysis.

Quality of life

In respect of assessment of quality of life, of the 19 participants in the study, 15 had complete information for calculation of EQ-5D-5L. The four participants who were lost to follow-up had missing data for this part of the questionnaire. Mean change from baseline to week 6 was 0.1079. Table 32 provides a summary of the EQ-5D-5L.

Intervention costs

Intervention costs were calculated and included for SMART. The total intervention cost is estimated at £320.25 per patient (Table 33). Development costs were also estimated; this included material development (printing SMST resources), training sessions for nurses and researchers and time to deliver intervention (Table 34).

Developed a construct of supported self-management in palliative care in relation to analgesic medicines

This was derived through a synthesis of our existing research and literature reviews of patient needs, behaviour change theory, existing interventions and ways to optimise the context in which HCPs can provide support. The development of the supportive self-management concept included key contextual factors in end-of-life care pain management, patients’ concerns about analgesia and the roles and responsibilities of patients, carers and professionals that we represented on a continuum in relation to self-management behaviours.

Developed and refined an intervention consisting of a self-management toolkit and a four-step education approach

We explored and refined our concept of supported self-management with 11 patients, eight carers and 19 HCPs though interviews and focus groups, informing the development of our intervention. From these data, we developed a four-step education approach that consisted of a needs assessment, including capacity for self-management, provision of information, goal-setting and review and coaching. We supported this approach with a self-management toolkit that comprised information factsheets, a pain chart and medication diary, goal-setting sheets and two 5- to 6-minute podcasts. We mapped data from our literature review and interviews against target behaviours and techniques to enhance self-efficacy to ensure that the toolkit and education approach would address the needs of patients, would be based on sound theoretical principles of enhancing self-management and would therefore be more likely to be effective in practice.

Developed and delivered a brief training programme for clinical nurse specialists in palliative care

The intervention was designed to be delivered to patients by community-based CNSs in palliative care. We engaged an expert nurse educator to design and deliver the training, which modelled the four steps within a clinical encounter and was based on a therapeutic conversational process between the specialist nurse and patient. This training included reflection, experiential learning and the development of a modelled self-management-focused conversation between two nurse educators.

Conducted a feasibility study of the intervention to inform the design of a future randomised controlled trial

This tested the recruitment and follow-up rates, fidelity of treatment delivery and suitability of outcome measures. We trained 12 CNSs from four UK hospices in the delivery of the intervention. During the 4-month study period, we identified 103 eligible patients from 417 who were screened. The most common reasons for ineligibility were patients not treated with a strong opioid (53%) and expected survival of < 6 weeks (19%). Of the 103 eligible patients, 37 (36%) were approached, of whom 19 (51% of those approached) agreed to participate and 15 completed the 6-week follow-up period. We found that 13 out of these 15 patients received all components of the intervention. Most patients (13/15) received a minimum of three visits during the 6-week study period. Rates of missing data for our outcome measures were very low. Although we observed no changes in our measures of pain intensity, we did observe improvements in our measures of interference from pain and in enhancing self-efficacy, which our evidence synthesis and interview data highlighted as being more important outcomes to patients than pain intensity alone.

Acceptability of intervention

Through qualitative interviews with patients and CNSs who participated in our feasibility study, we were able to understand the acceptability of the intervention and potential challenges within a large RCT. Patients and carers perceived that they all derived some benefit to them from the SMST, but the degree and nature of this benefit was variable and dependent on individual circumstances and preferences. The goal-setting sheets were the most frequently valued element; however, there were often other elements that were liked (e.g. the pain diary). The value of the factsheets to patients and carers appeared to be in reinforcing information that they had already been provided with by their CNS. Nurses reported that, although some patients found the concept of supported self-management more difficult to grasp, in general they felt that the educational approach was in keeping with their usual practice. All of them valued the training and materials. The busy and time-pressured reality of clinical practice for these specialist nurses, who often managed large caseloads, meant that the delivery of the intervention per protocol during the course of the trial varied. Patients perceived that the intervention was most effective when nurses delivered the factsheets according to need, completed contact information sheets, reviewed patients’ diaries and set goals. We assessed the potential for contamination of nurses not involved in the study by those nurses who were. We found that, although there was a general awareness of the study, there was no evidence that practice had changed or that the intervention was used by nurses who were not involved in the study.

Health economic analysis

We estimated cost-effectiveness of the SMART intervention within this feasibility study using a DAM approach because of the small number of patients and the lack of a control arm. We demonstrated that it was feasible to collect information to inform a full economic evaluation within a definitive RCT. The cost–utility analysis suggested that the SMART intervention appears to be cost-effective compared with standard care alone and could lead to cost savings. The SMART intervention yielded QALY gains and cost savings in our base case and many of the deterministic sensitivity analyses.

Phase I

Phase II

Phase III

Recruitment

Intervention content

Intervention delivery

Trial support

Trial processes

Outcomes and outcome measures

Trial design

The Phase I systematic review and Phase II patient and carer interviews

Conversational process

Parts of a personal action plan

1. Something the individual wants to do.

2. Reasonable/achievable (something that the individual could expect to be able to achieve within the week).

3. Action/behaviour specific.

4. Answer the questions: what, how much, when, how often.

Consider

• The specific steps needed to achieve the goal (include what, when, how, where and how often).

• The things that could make it difficult to achieve the goal.

• The plan for overcoming these challenges.

• Supports and resources needed to achieve the goal.

Needs assessment

• What is your experience of asking patients how they are managing their pain medicines?

• How are you finding using the four-step conversational process?

Information provision

• Thinking about your experience of providing the intervention, are there parts of it that feel easier/more appropriate to do than others?

• Could we do anything additional to support you in delivering the process? Are there any extra resources/materials that might be of help to you?

Goal-setting

• Thinking about the process of goal-setting with study patients, what are the kinds of things patients are prioritising? Do you feel that this prioritisation process is helpful in supporting self-management?

• Are you managing to use the goal-setting sheet at each visit? How are you finding this? What sort of action plans are you developing with the patients (are they practical steps, action orientated, time specified, barrier focused with strategies for overcoming these, etc.)?

Review and coaching

• I’ll be here next week; can we talk over these issues again to see if there are any changes or additional things that would help you?

Needs assessment

• Response to how she/he is finding asking patients about their needs (beliefs, behaviours and knowledge), resulting discussion, any researcher recommendations.

Information provision

• Response to how she/he is finding providing information, issues raised and researcher’s response, when appropriate.

Goal-setting

• Response to how she/he is finding using the goal-setting sheet, resulting discussion and recommendations.

Coaching

• Response to how she is finding the reviewing with the study patient(s), resulting discussion and coaching of CNS by researcher.

Delivery and discussion of further training resources:

• For example, Johnston et al. ’s51 self-management definition, self-management conversational prompts, audio file of modelled self-management conversation, action-planning sheet.

Response of the CNS to the above process and trial delivery.

Any confounding factors?

• Lack of CNS availability/time?

• Lack of receptiveness? Is there an apparent reason for this?

Researcher reflexivity

• Should I have handled anything differently?

• What lessons can be learnt for the other CNSs or other patients in terms of delivery of the intervention by the respective CNS?