Pain self-management interventions for community-based patients with advanced cancer: a research programme including the IMPACCT RCT

Hackett et al.8

We interviewed patients and carers at two time points to explore their experiences of pain in advanced cancer and the processes in which they engaged to manage pain. Patients were drawn from oncology outpatient clinics in a tertiary cancer centre and a hospice palliative care service. In total, 35 patients (29 patient–caregiver dyads and six patients on their own), of whom 21 were first interview participants and 14 were second interview participants, were interviewed. Four open-ended audio diaries were completed. A grounded theory analytic approach was employed. We used the method of constant comparison within cases over time and across cases to explore patterns of variation in the experience of pain and the strategies employed to manage it, as these were shaped by the course of illness, disease process, treatment regimen and access to supportive care, including for pain.

We found that pain in advanced cancer is complex, multidimensional and dynamic: patients and caregivers reported varying types and sources of pain, experienced both singly and together, that altered over the advanced illness course. For patients, pain management is an active, dynamic process, involving considerable ongoing work, so that developing expertise in pain management is not a one-off achievement. The amelioration of pain is uncertain and involves much trial and error (including by professionals) in the combinations of pharmacological and other therapeutic options, particularly because bone pain and neuropathic pain, common here, are more difficult to treat. Thus, a medication regimen often involves different types and combinations of drugs alongside prophylaxis medication to manage side effects. For most patients, managing pain is about securing ‘good enough’ relief that is consistent with balancing the side effects of medication and sustaining what they value in their lives. Three distinct patterns of pain were identified, varying in their degree of complexity, severity, transience/persistence and perceived control over pain, which are of practical and theoretical significance. This research has been published. 8

Bennett et al.20

We sought to determine when referrals to palliative care occur by exploring the length of time between an initial referral and death. A retrospective cohort study was undertaken that sought to determine the timing of referrals to three specialist palliative care services: two hospice-based community palliative care services and one hospital-based inpatient palliative care service. For each patient who had been referred to any of the three services, we identified the date on which they were first referred to the service and then calculated the median number of days between the first referral and the patient’s death. We also examined variation in the timing of referral before death across a range of characteristics of the patient (age, sex and diagnosis) and the service to which they were referred. Data were included for 4650 patients referred to specialist palliative care services in Leeds between April 2012 and March 2014. The median age of the sample was 75 years, and 3903 (84.0%) of the patients had a diagnosis of cancer. Overall, the median duration from referral to death for all patients was 34 days. This varied by hospital-based (20 days) and community-based palliative care (46 days). Age, diagnosis and place of referral were significant predictors of the duration of palliative care before death. We found that age was independently associated with the duration of palliative care, regardless of diagnosis. For patients aged > 75 years, 29 fewer days of palliative care were received than for patients aged < 50 years. Furthermore, patients with conditions other than cancer were found to receive 13 fewer days of palliative care than patients with cancer. When looking at services, we found that patients referred to hospital palliative care were receiving 24.5 fewer days of palliative care than patients being referred to community palliative care services. The study suggests that the current timing of referral to palliative care may limit the benefits to patients in terms of improvements in end-of-life care, particularly for older patients and patients with conditions other than cancer. Given that retrospective routinely collected clinical data were used, we do not have data on date of diagnosis, which prevents us from relating the duration of palliative care services to the duration of clinical awareness of disease. Late diagnosis may be a key factor in late referral, although the almost universally short duration of palliative care in our cohort suggests that this factor is unlikely to account for the observed referral pattern. This research has been published. 20

Allsop et al.22

Having identified that patients in Leeds were receiving a median of 34 days of palliative care prior to death [when community (46 days) and hospital (20 days) service data were combined], we sought to understand the national picture. Working in partnership with Hospice UK (London, UK), the research team conducted a national retrospective cohort study of the timing of referral to hospice-based specialist palliative care. The aim of the study was to identify patient and organisational factors that influence the duration of hospice-based palliative care in the UK prior to death. Overall, 64 UK hospices (i.e. one-third of all UK hospices) providing specialist palliative care inpatient beds and community services extracted data for all adult decedents (aged > 17 years) with progressive advanced disease who had received a prior referral (e.g. inpatient, community teams and outpatient), who died between 1 January 2015 and 31 December 2015. Data were requested for factors relating to both the patient and the hospice site. In total, data for 42,758 decedents were included in the analysis. The overall median time from referral to death was 48 days. Significant differences in referral to death days were found for those with cancer (53 days) and for conditions other than cancer (27 days). As age increases, the median number of days from referral to death decreases (those aged > 50 years, 78 days; those aged 50–74 years, 59 days; and those aged ≥ 75 years, 39 days). An adjusted multivariable negative binomial model demonstrated that increasing age was a significant predictor of fewer days of hospice care, as was being male, having a missing ethnicity classification and having a non-cancer diagnosis. This work provided a broader context for earlier findings on the timing of referral, with more than half of all patients being referred to palliative care < 7 weeks before death. This provided the national context in which to understand the Leeds data. The median of 46 days between referral and death for community-based patient in Leeds aligns with the 48 days found in national data relating to referrals to hospices before death.

The national survey highlights that, despite increasing rhetoric around the need for early referral to palliative care, patients with advanced disease across the UK receive referrals to hospice specialist palliative care very late in their illness trajectory. Age and diagnosis persist as determinants of duration of hospice specialist palliative care before death. Recent projections of the demand for palliative care by 2040 suggest that health-care systems need to adapt to the age-related growth in deaths from chronic illness, with dementia and cancer likely to be the main drivers of increased need. 23 Findings from this study suggest that there may be a need for reorientation of services to both older age groups and non-cancer conditions; both groups are associated with limited duration of hospice-based palliative care prior to death. The remit of palliative care has expanded to have relevance for any patient at an early stage in the disease process whose death can be medically anticipated. 24 Such a broad remit, when people are living longer with an increased illness burden, suggests that many patients will need primary, secondary and specialist palliative care. With increasing calls to reduce hospital bed-days in the last months of life and to support patient preferences for home death, it is particularly important that community-based specialist palliative care can consider how provision might match demand. This research has been published. 22

Hackett et al.25

Alongside patients’, caregivers’ and oncologists’ perspectives, we engaged with primary health-care teams, which are a key component of care delivery to patients with advanced cancer during the last year of life. The Gold Standards Framework is proposed as a mechanism for co-ordinating and guiding identification, assessment and support. There are still considerable variations in practice despite its introduction. The aim of this qualitative study was to improve understanding of variations in practice by exploring the perspectives and experiences of members of primary health-care teams involved in the care of patients with advanced cancer. Qualitative, semistructured interviews, focus groups and non-participatory observations involving 67 members of primary health-care teams providing palliative care were undertaken. Data were analysed using a grounded theory approach. We identified distinct differences in the drivers of and barriers to community advanced cancer care co-ordination, which relate to identification and management, and access to effective pain management, and go some way to explaining variations in practice. These include proactive identification processes, time and resource pressures, unclear roles and responsibilities, poor multidisciplinary working, and inflexible models for referral and prescribing. These provide valuable insights into how health professionals work together and independently within an infrastructure that can both support and hinder the provision of effective community palliative care. Although the Gold Standards Framework is a guide for good practice, alone it is not a mechanism for change. Rather, it provides a framework for describing quality of practice that was already occurring. Consequently, there will continue to be variations in practice. This research has been published. 25

Taylor et al.26

Increasing evidence suggests that, for patients with a range of advanced cancers, earlier integration of palliative care should be an essential component of their care. 27–31 Communication is an important element of the provision of advanced cancer care32 and may play a role in facilitating awareness of and access to palliative care services. However, a high level of unmet information needs exists among patients whose care has become palliative in focus. 33 We conducted a study to understand the current evidence on the information provision for patients with advanced cancer while exploring where deficits in provision may exist. The study was conducted in three distinct phases: (1) a literature review, (2) a regional audit of patient information and (3) a critique of patient information. The literature review was conducted as part of a scoping exercise to focus the direction of future research projects. The review highlighted particular issues of interest around patient information that were investigated further in the audit. The literature review identified patient-related barriers to earlier integration of palliative care, misconceptions of what palliative care is and a lack of understanding of the role and breadth of services available. 34 The lack of information available to patients about palliative care and their unmet information needs was evident in the literature, which led us to conduct an audit of patient palliative care information in our geographical area. Our regional audit of patient information resources found that patient information relating to palliative care is not widely available to cancer patients. Our audit showed that < 13% of all inpatient units, 7% of outpatient departments and 25% of chemotherapy day units had written information available for patients regarding palliative care. Despite this, > 90% of palliative care teams that were surveyed said that information leaflets had been produced but were not in routine circulation. Our audit shows that, although information resources exist within the trusts, these are often distributed to patients only after they have received a palliative care referral. Our review of the content of patient information resources, where these exist, shows considerable variation. The majority of the resources fail to describe the referral process and what patients can expect once they have been referred. None of the information leaflets explain that palliative care can be integrated alongside oncology care. Patient and health professionals’ understanding of this concept is key to achieving an integrated service in which patients can receive appropriate palliative care input alongside cancer treatment. 35 More research is required to explore ways to disseminate information about palliative care effectively and sensitively. This research has been published. 26

Flemming et al.37

A review of evidence drawn from systematic reviews of complex interventions was combined with a synthesis of qualitative research to identify the key components of a self-management intervention for advanced cancer pain. We identified four systematic reviews examining interventions for the self-management of advanced cancer pain. Although attributes of a pain management intervention were recommended in each of the reviews, the essential key components could not be determined. As part of subsequent qualitative evidence syntheses, a further three systematic reviews were identified and integrated with the effectiveness reviews. The key components of a self-management intervention following the integration included the importance of addressing patients’ knowledge, skills, individualised approaches to care and attitudes towards pain management, alongside the significance of the role of team approaches and interdisciplinary working in the management of pain. Mapping the findings of each paper onto a behaviour change framework led us to identify the contextual and intervention components that are essential for the development of successful educational programmes. We concluded that educational interventions to promote the self-management of advanced cancer pain by patients and that their carers should seek to include content that addresses how individuals manage their pain in the context of their situation and that of their health professionals, while also focusing on those intervention functions that are known to influence behaviour. This research has been published. 37

Hughes et al.38

This study was undertaken to ascertain the views of specialist palliative care health professionals on patient self-management of cancer pain to inform the development of a new educational intervention. Focus group interviews were conducted with 17 health professionals [community CNSs, n = 6; complementary therapists, n = 3; hospice nurses, n = 5; hospice social worker, n = 1; hospice spiritual care co-ordinator, n = 1; and palliative care consultant physician, n = 1). The aim of the focus groups was to elicit experiential perspectives that would help us to extend and refine an educational intervention that focused on the use of strong opioids for control of cancer pain, which was produced and tested during earlier work by the team. 39 Participants viewed self-management of cancer pain as desirable and achievable, but also as something that could be problematic. Challenges to self-management were perceived in patient attitudes and behaviours, health professionals’ own beliefs and actions and the wider social system. Practitioners showed awareness of potential tension between their espoused views (the desirability that patients manage pain autonomously) and their tacit views (the undesirability of patients managing pain in ways that conflict with health professionals’ knowledge and identity). This research has been published. 38

Allsop et al.52

Our aim in this systematic review was to review existing ICT systems developed for pain management in palliative care. The majority of the literature identified in our systematic review was of a non-experimental research design. ICT systems included in the review were at various stages of development (planning, analysis, design and evaluation), with no systems implemented into routine care. Most ICT systems measured pain as part of a quality-of-life measurement and there was wide variation in the approaches used to assess pain. We concluded that future development of ICT systems needs to increase the quality and scale of development work, consider how recommendations for pain measurement can be integrated, and explore how to effectively use system feedback with patients. This research has been published. 52

Taylor et al.53

We recruited 34 participants with chronic pain who were asked to complete twice-daily pain assessments over two consecutive days. Participants alternated their mode (personal computer/laptop, smartphone, short messaging service or tablet computer) of accessing the ICT platform. Most participants completed all four assessments. The system was reported to be useful for reporting pain. Participants suggested that the platform could be useful to improve recall and monitoring of pain, promote self-management and improve links with health professionals. Perceived benefits for health professionals included increased understanding of pain experiences, improvements in pain monitoring and management, and enhanced communication. This research has been published. 53

Taylor et al.54

We undertook usability testing of the ICT system with 29 participants (advanced cancer patients, n = 13; GPs, n = 4; CNSs, n = 4; district nurses, n = 3; palliative care doctors, n = 5). Patients and health professionals were quickly able to understand and use the ICT system with very limited information about what the system was designed to do or how it worked. Both patients and health professionals were generally positive about PainCheck and found it easy to understand. Participants did have some concerns about how PainCheck might work in clinical practice. Their concerns were largely related to process issues, such as whether or not health professionals would have time to use the ICT system and how a lack of response may have an impact on patient care. Patients’ and health professionals’ ability to engage and use the technology was also mentioned as a potential barrier to ICT system integration. This research has been published. 54

Allsop et al.55

Face-to-face interviews were conducted with 13 patients with advanced cancer who were receiving palliative care. Patients described technology as peripheral to existing processes of care. Simple approaches that employ well-established technologies may be a preferred starting point. For system content to have relevance for a patient with advanced cancer, it needs to take account of the complexity of pain experiences and existing relationships with health professionals. Future research is required to understand how ICT systems can be positioned flexibly within existing delivery models of palliative care. This research has been published. 55

Taylor et al.56

We carried out face-to-face interviews with 15 health professionals managing the palliative care of patients living in the community. Participants included GPs, CNSs, district/community nurses and palliative care doctors. Within our work, even the most sceptical of health professionals acknowledged the potential benefits of implementing an electronic patient-reported pain monitoring system. Health professionals had reservations about how PainCheck would work in practice. For optimal use, an ICT system would need to be embedded within existing electronic health records. Electronic pain monitoring systems were reported to have the potential to enable health professionals to support patients’ pain management more effectively but only when barriers to implementation are appropriately identified and addressed. This research has been published. 56

Allsop et al.57

We adopted a research-led development process that sought to develop an ICT system that met the needs of patients and health professionals in real clinical settings and that was fit for clinical trial. Agile software development methods were combined with health science research methods and ‘participatory design,’ including diary studies, face-to-face interviews, questionnaires, observations of clinical settings, prototyping, think-aloud, agile sprints, process reviews, requirements clustering and pilots. Three iterations of ICT system development were necessary to prepare the tool for the feasibility study.

PainCheck, our information and communication intervention for the management of pain, was built by a private company, X-Lab (Leeds, UK). X-Lab used a University of Leeds platform called QTool. QTool is an electronic online questionnaire management software suite. QTool is used by health-care practitioners and researchers to build and schedule complex questionnaires that can be completed by patients and clinical staff. QTool has been used for patient-reported outcomes in cancer survivors and self-report and management of adverse events during cancer treatment. QTool was selected as a starting point for the development of PainCheck. This research has been published. 57

PainCheck was accessed by patients and health professionals through a website hosted by the University of Leeds. PainCheck allows patients to record their pain and gives them access to personalised pain management advice. Patients are asked to respond to various questions, including providing a description of their pain (Figure 4a); providing a rating of the intensity of current pain and the intensity of pain in the last 12 hours (see Figure 4b); and giving the perception of control of their pain and how pain has interfered with daily activities and their sleep. The majority of items seeking this information were taken from the BPI.

FIGURE 4.

Examples of PainCheck user interface. (a) Qualitative description of pain; (b) numerical sliding scales rating pain and interference; (c) questions exploring what patients have undertaken to control their pain and if they were helpful (the second page included self-help measure such as use of hot/cold, having a bath, exercise); and (d) patient feedback.

Patients are asked about pain management techniques, which of these techniques were helpful and how likely they are to try them in the future (see Figure 4c). Various question response options are used, including multiple choice, numerical slider scales (0–10 or 0–6) and free text. After completion, patients are provided with a summary of their results and suggestions of pain management techniques that they may want to try in the future (see Figure 4d). Health professionals can log in to PainCheck and view all patients registered on PainCheck and see who has completed reports (Figure 5a). They are then able to select a patient and view their responses to individual questions (see Figure 5b). Health professionals are presented with a graph that tracks patients’ current pain and their pain in the last 12 hours over time (see Figure 5c). Patients are given a ‘red flag’ in the health professional system if they reach certain thresholds for current pain and pain control. After reading the patient report, health professionals can decide what action, if any, they would like to take (see Figure 5d). Health professionals have the option of contacting patients through PainCheck to provide information and advice. All data entered into PainCheck can be exported into Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) in comma-separated value format.

FIGURE 5.

Screenshots of health professional dashboard. (a) Health professional dashboard; (b) individual patient report; (c) patient scores over time; and (d) record of health professional action.

Mapping pathways of prescriptions among cancer patients during their last year of life

We began by seeking to understand existing patterns of prescriptions for strong opioid treatment in patients with cancer and to identify current practice. We investigated this through a regional analysis of prescribing data on patients who had died from cancer.

In parallel, we undertook a systematic review to synthesise the research evidence on the relationship between strong opioid analgesia and survival in patients with advanced cancer. We hypothesised that one of the reasons that patients with advanced cancer pain are often undertreated may be that the patient or their clinician perceive that strong opioids used in the terminally ill population can hasten death. By undertaking this systematic review, we hoped that it would be possible to substantiate or refute this perception.

Opioid prescribing for patients with cancer in their last year of life

Ziegler et al.58

We originally set out to understand the prescription pathways in a population of 400 deceased cancer patients during their last year of life, including the analgesics prescribed (e.g. paracetamol, codeine, strong opioids) and the adjuvant analgesics, such as antidepressants and antiepileptics, often used for cancer neuropathic pain. We were able to substantially expand the original scope and increased the study population to 6800 patients, accordingly increasing the power of our study to detect and test associations. We used an innovative data linkage system (i.e. Openpseudonymiser) to link data derived from the primary care electronic patient record system (SystmOne) within the electronic system within a large acute NHS trust (Patient Pathway Manager) and data from the Northern and Yorkshire Cancer Registry. Comprehensive linked data on 6080 patients who died from cancer over the 7-year period (2005–12) were retrieved and analysed. To the best of our knowledge, this is the most comprehensive data set of its type and is of international significance to the field.

For all patients included in the linked cohort, all prescriptions for analgesics were identified for the period 12 months prior to their death. These prescriptions were analysed to determine to what extent and for what duration strong opioids were provided, exploring any differences that arose when analysing prescriptions by clinical and patient characteristics. Strong opioids were prescribed for 48% of patients in the last year of life (Figure 7). The median interval between the first prescription of a strong opioid and death was 9 weeks (interquartile range 3–23 weeks). Prescribing was not influenced by cancer type, duration of illness or sex, but we found a strong association with patients’ age, with older patients much less likely to be prescribed a strong opioid. Patients who died in a hospital were 60% less likely to have received a strong opioid in primary care before admission than patients who died in a hospice [relative risk ratio 0.4, 95% confidence interval (CI) 0.3 to 0.5; p < 0.01].

FIGURE 7.

Cumulative proportion of patients (%) prescribed analgesics week by week for the last year of life.

The study provides the first detailed analysis of the relatively late onset and short duration of strong opioid treatment in patients with cancer before death in a representative UK cohort. This pattern of prescribing does not match epidemiological data, which point to the earlier onset of pain. 5 Although the persistent under-treatment of cancer pain is well documented, this study suggests that strategies for earlier pain assessment and the initiation of strong opioid treatment in community-based patients with cancer could help to improve pain outcomes. In addition, our finding of unexplained variation in the prescribing of opioids for younger and older patients indicates a previously unknown inequity in access to pain relief. This research has been published. 58

Association between regular systemic opioid analgesia and survival in adult patients with cancer

Surveys of non-medical prescribing in palliative care

Analysis of growth and impact of non-medical prescribing

Survey of health professionals’ views on community pharmacy services for patients with cancer pain

Evaluating recruitment methods used in a pharmacist-led cancer pain medicines consultation study

Patient views of pharmacist medicines consultation

Systematic review of pharmacist educational interventions for cancer pain management

Feasibility and acceptability of pharmacist medicines consultation

Understanding patient preferences

We conducted a DCE with patients to understand their preferences for various aspects of pain management services. This also provided information on patient preferences for alternative pain management service types.

Health economic model development

Summary of trial design

IMPACCT is a pragmatic multicentre RCT in patients with active and incurable cancer that aimed to evaluate the delivery and implementation of evidence-based interventions into routine clinical practice, developed and piloted as part of the IMPACCT programme grant.

We aimed to recruit 160 participants at the point of identification for referral into palliative care. Participants were randomised on a 1 : 1 basis to receive either the intervention [UC plus SSM delivered within the oncology clinic and palliative care services by locally assigned community palliative care nurses (health professional), consisting of self-management/educational support and pain monitoring], or UC. Participants were identified and recruited in the oncology (or related) clinic by the research nurse in consultation with the patients’ clinician/treating team. All participants were referred to their local palliative care team for pain management. The recruiting team were not involved in subsequent intervention delivery.

The follow-up period of the trial was 12 weeks, with patient-reported outcome measures collected by post at 6 and 12 weeks following randomisation. When required, the blinded IMPACCT trial researcher collected these data by telephone. Medical data were obtained directly from patient hospital and palliative care records by the research nurse and palliative care team, respectively. All participants were followed up for overall survival until the final participants’ 12-week follow-up.

A subsample of consenting participants and health professionals were invited to take part in an interview at approximately 6 or 12 weeks post randomisation (see Process evaluation).

Protocol amendment

A substantial amendment to the protocol and trial design was implemented early on in the trial, 6 months after opening to recruitment and after 13 participants had been recruited (Table 1). The original protocol (v3.0) had an additional aim to evaluate early screening and referral. However, early screening and recruitment methods were scrutinised and it was not feasible to implement an early referral pathway into the existing cancer pain pathways in the RCT. Once our screening procedures had identified cancer patients with significant pain (≥ 4 on a rating scale of 0–10), oncology staff were reluctant to allow patients to be randomised, potentially to UC, and felt obliged to make a clinical referral to palliative care. With the support of the TSC, we agreed an amendment to recruit at the point of referral so that all patients eligible for community-based palliative care could be recruited and randomised, with community palliative care referral in both trial arms, essentially removing the early referral pathway of the intervention.

Primary objectives

• Delivery and implementation of the intervention: the primary implementation objective is to evaluate adherence in terms of the uptake and retention rate of each intervention.

• Potential effectiveness of the intervention: the primary effectiveness objective is to assess the effectiveness of the intervention compared with UC, as measured by pain severity on the BPI (the mean and the proportion of responders with a ≥ 30% reduction in the BPI pain score) 6 and 12 weeks after randomisation.

Secondary objectives

• To assess differences in patient-rated pain, as measured using the BPI pain interference and the 7-point global rating of change in pain at 6 and 12 weeks.

• To assess health-care use in each arm, in particular hospital admissions within 12 weeks of randomisation.

• To assess patients’ pain knowledge and experience, as measured using the PPQ at 6 and 12 weeks.

• To assess the differences in patients’ general and cancer-specific QoL as measured using the EQ-5D and EORTC QLQ-C30 at 6 and 12 weeks.

• To document the cost-effectiveness of the interventions.

Furthermore, we aimed to report on the delivery and implementation of the intervention through a process evaluation using qualitative data to explore participant and palliative care nurse views.

Sample size

To assess the effectiveness of the intervention, the power calculation was based on the difference in mean pain severity on the BPI at 12 weeks. With 80% power and a two-sided type I error rate of 0.05, we estimated that 128 patients (64 per arm) were required to detect a moderate intervention effect size of 0.580 between the intervention and the control arms.

In addition, a relative reduction of ≥ 30% in pain severity (BPI) is an accepted threshold for clinically significant improvement in pain trials. 81 Results from a previous study involving automated symptom modelling found a 27% difference in such improvement rates at 12 weeks. 82 Our estimated sample size also provided 80% power to detect a similar difference in rate.

As our patient sample was drawn from a generally frail population, we allowed for an attrition rate of 20% and, therefore, aimed to recruit 160 participants (80 per arm).

Withdrawals

Nine (5.6%) participants withdrew from trial processes: four (5%) who were allocated to the SSM arm and five who were (6.2%) allocated to the UC arm (see Report Supplementary Material 1, Table 20). All withdrew from questionnaire completion via researcher contact and all but one withdrew from postal questionnaire completion. All remained willing for further data to be collected from their medical records. The majority of withdrawals were because the participant was too unwell and were clustered around the 6- and 12-week follow-ups.

Participant questionnaires

A total of 115 (71.4%) participants had at least one successful follow-up, with questionnaires completed at 6 or 12 weeks post randomisation: 56 (70.0%) in the SSM arm and 59 (72.8%) in the UC arm (see Report Supplementary Material 1, Table 21).

Baseline questionnaires were completed face to face with the recruiting researcher and were returned for all but one participant (see Report Supplementary Material 1, Table 21, and Table 6). The majority of participants completed follow-up questionnaires via the post; over one-quarter had help completing the questionnaires at 6 weeks and just under 20% had help at 12 weeks, with help provided primarily by the participants’ partner or child (see Report Supplementary Material 1, Table 22). Questionnaire packs were completed for 107 (66.5%) participants at the 6-week follow-up and 87 (54.4%) participants at the 12-week follow-up (Table 7). Questionnaires were not returned at 12 weeks owing to participant death (n = 29, 39.2%), illness (n = 11, 14.9%), and withdrawal (n = 6, 8.1%). Questionnaires were also not returned when these had been sent by post and the participant was contacted by telephone to offer telephone completion (n = 11, 14.9%), when there was difficulty contacting the participant by telephone (n = 15, 20.3%) or when the participant was not contacted by telephone (n = 2, 2.7%).

The majority of questionnaires were completed within a 2-week time window: 94.4% at 6 weeks and 93.1% at 12 weeks (see Table 6; and Report Supplementary Material 1, Figure 17).

Comparison of baseline characteristics between participants with and participants without follow-up (primary outcome completion; see Report Supplementary Material 1, Table 23) indicated that those not completing follow-up at 6 or 12 weeks were more likely to be male, younger and taking a strong opioid and to have a higher ECOG performance status; had more recently received their original diagnosis; had lower baseline QoL scores (EORTC-8D, EQ-5D-3L); and had worse baseline measures on the BPI (worst pain, pain severity and interference), PPQ (Patient Pain Questionnaire) (experience scale) and QLQc30 (Quality of Life Questionnaire c30) (global health status, physical, social and role functioning).

The completeness of questionnaire data, including missing item level data, outcome scores and multiple item responses, is summarised in Report Supplementary Material 1, Tables 24 and 25.

Researcher unblinding

Researcher unblinding occurred for three participants by their 6-week follow-up; all of the participants were allocated to the SSM arm and were recruited under the original v3 protocol. Two instances occurred as the participant discussed the intervention during the telephone follow-up (follow-up completed) and one occurred as the researcher became aware of palliative care contact following nurse contact (follow-up not completed).

Safety

No related and unexpected serious adverse events (RUSAEs) were reported during the trial.

Deaths were recorded throughout the trial up to the last participant’s 12-week follow-up. A total of 92 (57.1%) participants were reported to have died (see Table 26, Figure 11), with an overall median survival of 53.3 weeks (95% CI 40.9 to 59.6 weeks) and similar estimates across trial arms. A total of eight (5%, 95% CI 1.6% to 21.1%) participants had died by 6 weeks post randomisation, and 25 (15.5%, 95% CI 9.9% to 21.1%) participants had died by 12 weeks post randomisation. Almost half of the participants who died did so in a hospice (44.6%), 28.3% died at home and 21.7% died in hospital.

FIGURE 11.

Kaplan–Meier survival estimates.

Protocol amendment

In accordance with the original protocol (v3.0), six participants allocated to the UC arm were not referred to community palliative care, whereas the 75 (92.6%) participants allocated to UC in accordance with the amended protocol (v4.0) were referred to community palliative care as part of UC (see Report Supplementary Material 1, Table 16).

Intervention training

Intervention training in each hospice took place face to face prior to the hospital site opening to recruitment, and ranged in duration from 20 minutes to 2 hours 20 minutes (median 1 hour 15 minutes) (see Report Supplementary Material 1, Table 27). Refresher training was offered to all sites during the trial; 5 out of 11 hospice sites received further face-to-face training, with a maximum of four face-to-face refresher training sessions taking place for a site. The number of health professionals (palliative care nurses, health professionals) trained across the hospices (with recruited participants) ranged from 5 to 16.

Usual care

An initial palliative care visit took place for 78 (97.5%) participants in the SSM arm and 71 (87.7%) participants in the UC arm (see Report Supplementary Material 1, Table 28). The visits mainly took place in participants’ own homes (95.3%) a median of 1 week following randomisation, and lasted a median of 75 minutes in both arms.

During the initial visit, the following palliative care needs were addressed with similar rates across trial arms (see Report Supplementary Material 1, Table 29): pain in 94.6% of participants, additional symptoms in 57.5%, psychological needs in 47.7%, carer concerns in 20.1% and other care needs in 14.1%. A recommended change to pain medication was made for 67.1% of participants: 65.4% in the SSM arm and 69.0% in the UC arm.

A total of 58 health professionals were involved in the 149 participants’ initial palliative care visits (Figure 12); 46 health professionals for 78 participants in the SSM arm and 38 health professionals for 71 participants in the UC arm. Of these, 26 health professionals saw participants across trial arms for 44 SSM and 56 UC participants.

FIGURE 12.

Health professional delivery of the initial palliative care visit.

For further details of palliative care receipt during the trial, see Health-care use (see Table 13; see Report Supplementary Material 1, Tables 40 and 41).

Primary intervention implementation outcome

Of the 80 participants allocated to the SSM arm, the Tackling Cancer Pain booklet and DVD were introduced and accepted by 72 (90.0%, 95% CI 81.2% to 95.6%) participants (see Figure 8).

The PainCheck online monitoring system was introduced to 47 (58.8%, 95% CI 47.2% to 69.7%) participants (see Figure 8). The main reasons that PainCheck was not introduced were that the participant did not have a computer (n = 19), the health professional felt that it was not appropriate (n = 5) and the participant chose not to use the computer (n = 3). A total of 32 (40%, 95% CI 29.2% to 51.6%) participants logged on to PainCheck and 23 (28.8%, 95% CI 19.2% to 40%) were found to have engaged with the intervention, having used PainCheck on three or more occasions.

Of the 32 participants using PainCheck, the median total time spent on PainCheck over the 12-week trial period was just under 1 hour, with times ranging from 6.5 minutes to > 10 hours. Participants logged on a median of five times, ranging from one time to 82 times when the participant used PainCheck daily (see Figures 8 and 9). A health professional also used PainCheck for 21 (65.6%) of the 32 participants using PainCheck (see Report Supplementary Material 1, Table 30). The median number of times that a health professional logged on was one time, ranging from one time to 55 times (see Figures 8 and 9). Of the health professionals using PainCheck, the median total time spent on PainCheck per participant (n = 21) was 25.7 minutes, ranging from 3.1 minutes to > 7 hours.

Contamination

Contamination was reported for one participant allocated to UC for whom the Tackling Cancer Pain booklet and DVD were provided to the participant in error.

Process evaluation during implementation of multicomponent intervention for pain management

To understand the process of implementation surrounding the pain self-management intervention, we conducted semistructured interviews with 12 patients and 12 hospice-based palliative care nurses. Participants were interviewed at their home and at their place of work. Interviews were guided conversations to elicit accounts of participants’ experiences in their own words of taking part in the trial using a topic guide. Key themes that emerged from the process evaluation are outlined below.

Feasibility of eHealth intervention

Health professionals felt that there was a place for PainCheck in current practice if its use was streamlined (Figures 13 and 14). Some felt that it easily supported their current way of working by adding in another layer of detail, which they could use to monitor patients’ pain. Others felt that it enhanced the care that they provided because it enabled them to think about other aspects of pain management. Although digital technologies were viewed as becoming more pervasive within health care, health professionals believed that these would not replace their current way of working. This research is now published and available at http://eprints.whiterose.ac.uk/155107/.

FIGURE 13.

Intervention uptake.

FIGURE 14.

Participant and health professional PainCheck use.

Missing data

The primary outcome was available for all time points for 78 (48.4%) participants, for at least one follow-up time point for 37 (23.0%) participants and for baseline only for 45 (28.0%) participants, and was not available at any time point for one (0.6%) participant (Table 8). Forward selection identified BPI worst pain, sex, age, centre, opioid use, health today and QoL from the EuroQol-5 Dimensions, three-level version (EQ-5D-3L), as key participant characteristics predictive of missing data at 6 or 12 weeks in multivariable models (see Report Supplementary Material 1, Tables 31 and 32). Multiple imputation for primary analysis, therefore, used the above predictive variables (worst pain on the BPI replaced by the relevant baseline value for secondary outcomes) in addition to randomisation strata (dichotomised average pain response and centre) and allocation. A total of 52 imputations were made for all models, corresponding to the overall percentage of missing data (i.e. 52% missing at least one time point).

Pain severity: worst pain (Brief Pain Inventory)

The mean worst pain was 7.7 points (SD 1.76 points, n = 160) at baseline, representing a severe level of pain. This was 5.8 points (SD 2.35 points, n = 107) at the 6-week follow-up and 4.7 points (SD 3.09 points, n = 86) at the 12-week follow-up, each representing a reduced moderate level of pain and an increase in variability given the high level of loss to follow-up (see Report Supplementary Material 1, Table 33 and Figure 18). Three-quarters of participants reported a severe level of worst pain at baseline (7–10 points), reducing to 43% of participants at 6 weeks and 34% of participants at 12 weeks. Mean scores were slightly higher in the SSM arm than in the UC arm and an increased proportion of participants reported severe pain at 6 and 12 weeks.

Based on participants’ change in worst pain compared with baseline, there was a mean reduction of 1.5 points (SD 2.62 points, n = 107) at 6 weeks and 2.7 points (SD 3.54 points, n = 86) at 12 weeks (see Report Supplementary Material 1, Table 34). A total of 41.1% of participants were classed as a responder at 6 weeks and 57% at 12 weeks, having had a reduction in worst pain score of ≥ 2 points or ≥ 30%, with similar rates across trial arms.

The primary and sensitivity analyses found no significant treatment differences for the primary outcome or for other secondary outcomes of pain severity items on the BPI (Table 9, and see Figure 10 and Report Supplementary Material 1, Tables 35 and 36). The mean adjusted difference between the trial arms (SSM – UC) at 12 weeks was 0.5 (95% CI –0.7 to 1.8; p = 0.4102), representing no significant difference between the trial arms. There was similarly no significant difference in the odds of a response, with an odds ratio of 0.84 (95% CI 0.39 to 1.82; p = 0.6558) (Table 10).

Exploratory subgroup analysis

Based on summary statistics, participants in the intervention arm who engaged in the use of PainCheck (n = 23) had slightly increased worse pain scores at baseline, 6 weeks and 12 weeks compared with those who did not engage (n = 57); there was no evidence of a significant difference (Figure 15; see Report Supplementary Material 1, Figure 19).

FIGURE 15.

Estimated mean worst pain scores with 95% CIs for unadjusted, adjusted primary and adjusted sensitivity analyses. Primary analysis – multiple imputation; sensitivity analysis – to availability of data.

Global rating of change

An increase in pain on the 7-point global rating of change in pain was reported for 31 (29.5%) participants at 6 weeks and 27 (31.8%) participants at 12 weeks, and a reduction in pain was reported for 59 (56.2%) participants at 6 weeks and 43 (50.6%) participants at 12 weeks (see Report Supplementary Material 1, Table 39). A higher rate of increase in pain and a lower rate of reduction in pain was observed in the SSM arm than in the UC arm at 6 weeks. However, this difference was not observed at 12 weeks; instead, a higher rate of reduction in pain was observed in the SSM arm than in the UC arm.

Hospital attendances

Hospital outpatient appointments were reported during the 12-week trial period for the majority of participants (95%), with a similar rate across trial arms (Table 12). The number of visits per participant was larger in the UC arm, with a mean of 7.6 visits (SD 5.42 visits) compared with 5.6 visits (SD 4.56 visits) in the SSM arm. Over three-quarters of appointments were for cancer treatment and routine follow-up. Accident and emergency (A&E) attendances were reported for 14.9% of participants overall, with similar rates across the trial arms. A&E attendances were most frequently for disease complications (66.7%), and a higher proportion were owing to cancer-related pain (22.2%) than outpatient appointments (6.3%). Hospital admissions were reported for almost half of participants: 38.8% in the SSM arm and 48.1% in the UC arm. A maximum of four admissions were reported per participant. The most frequent reasons for admission were disease complications (46.3%) and cancer-related pain (35.2%).

Palliative care contacts

All participants in the SSM arm and the majority (92.6%) in the UC arm received some palliative care contact, including telephone contact, home visits, day or outpatient hospice attendances and inpatient stays, during the 12-week trial period (see Report Supplementary Material 1, Table 40; Table 13).

Over half of participants continued within palliative care at the end of the trial, with a slightly larger number of SSM participants (62.5%) than UC participants (53.1%). One-fifth were discharged from palliative care at the end of the trial: 17.5% in the SSM arm and 22.2% in the UC arm. The remaining participants were discharged before 12 weeks, largely owing to the participants’ death (see Report Supplementary Material 1, Table 40).

Almost all participants (93.2%) received telephone palliative care contact, with a median of four contacts (see Table 34). A median of two home visits took place in both arms, ranging up to 10 in the SSM arm and eight in the UC arm. The reason for home visits were mainly because of the initial palliative care visit, cancer-related pain or cancer-related pain with other reasons.

The number of participants with a day/outpatient visit and an inpatient stay and the number of visits/stays was similar between arms. Just under one-fifth of participants had a day/outpatient visit: 17.5% in the SSM arm attended up to four times and 21% in the UC arm attended up to nine times. A total of 10.6% of participants had an inpatient stay: 12.5% of participants in the SSM arm attended on up to two occasions and 8.6% of participants in the UC arm attended once. Inpatient stays were on average a median of 11.5 days, ranging up to 50 days.

Pain medications

All participants but one were receiving pain medication at baseline and all but four participants (alive, with available data) were receiving pain medication at 6 and 12 weeks (see Report Supplementary Material 1, Table 41). Participants were taking a mean of 2.4 different medications (ranging from 0 to 5 medications) at baseline, increasing to a mean of 2.8 and 2.9 medications (ranging from 0 to 7 medications) at 6 and 12 weeks, respectively. The proportion of participants on a strong opioid increased from a baseline rate of 59% to 72.4% at 6 weeks and 71.8% at 12 weeks. Medication use was similar across trial arms, albeit with a slightly higher rate for all classes of medication in the SSM arm than in the UC arm.

Cost-effectiveness analysis of the IMPACCT trial

The cost-effectiveness analysis of the IMPACCT trial was based on within-trial outcomes, with costs and HRQoL data collected at baseline, 6 weeks and 12 weeks. The decision model presented in WP 4.1 was not used here to extrapolate the outcomes of the trial over a longer time horizon, principally because the statistical analysis of the trial data found no difference between the trial arms in terms of the primary outcome measure (pain). Without a differential in pain outcome, there could be no differential in the costs and effects in the model. Thus, the time horizon for the evaluation was 12 weeks. Where possible, we followed the National Institute for Health and Care Excellence’s reference case, reporting cost per incremental QALY [based on EuroQol-5 Dimensions, 5-level version (EQ-5D-5L), and EORTC-8D].

Our primary analysis was based on the EQ-5D-5L direct valuation84 with supplementary analyses presented based on the EQ–5D-5L crosswalk to the EQ-5D-3L85 and the EORTC-8D, derived from the QLQ-C30. 86

The QALY calculation combined HRQoL and survival information and, thus, was based on a linear interpolation, area-under-curve approach. Given that some patients were observed to die between follow-up waves, some further assumptions were required for such cases. For patients who died with their last HRQoL observation as positive, we calculated the aggregate HRQoL measure on the basis that such patients had a linear fall in this measure from this point to death, imposing the assumption that HRQoL deteriorates in a linear fashion until the value zero is reached at death. For patients who died with their last EQ-5D observation as negative, we calculated the aggregate HRQoL observation measure on the basis that such patients had this constant level of HRQoL according to this measure from this point until death, imposing the assumption that HRQoL does not improve from this level until death.

Costs

We used a combination of the IMPACCT patient-completed questionnaires and administratively collected records to estimate the health resource use of each patient within the trial. We prioritised data from the latter source because patient-completed questionnaires rely on both accurate recall and the individual being willing and able to provide information at the time of the 6- and 12-week follow-ups.

Our cost estimates were formed of four components: primary care costs, hospital and hospice secondary care costs, prescribed medication costs and programme costs. For all but primary care costs and programme costs, we derived activity estimates for each of these from administratively completed records. The administratively completed records specified whether or not hospital visits and admissions were pain related. Although the a priori assumption would be that data on the former would be more specific and reduce noise in the results, discussions with nurses collating administrative data, and examination of the text responses made alongside these records, indicated that these data were potentially confounded. For example, frequently, although resource use was indicated to have been pain related, text responses indicated that there were multiple, including non-pain-related, reasons for admission. That being the case, we elected to use all reported resource use and not just those indicated to be pain related. Prescribed medication details (both specific type/brand of medication and dosage) were also derived from administrative records.

Health-care resource use (primary and secondary care) was costed using the most recently available data from national resource reports. The unit costs employed, sources and the year from which these costs were taken are documented in Table 14. We inflated all such costs to 2018/19 prices using the NHS Pay & Prices index for years up to 2015/1691 and Monitor’s published Economic Assumptions92 for 2016/17 onwards, with the inflator used for data from each calendar year documented in Table 15. Prescribed medication costs were assigned from the August 2018 NHS Electronic Drugs Tariff,93 with adjustments made for dosage as applicable, with the modal dosage assumed in cases for which this was missing. Programme costs, arising from staff costs and the production of digital and physical information materials to support pain self-management, amounted to a one-off per-patient cost of £63.35 and a weekly per-patient cost of £15.17.