Carer administration of as-needed subcutaneous medication for breakthrough symptoms in people dying at home: the CARiAD feasibility RCT

How does the existing literature support this study?

Carers prioritise rapid symptom control and are willing and able to administer injectable drugs, including controlled drugs such as morphine: a narrative literature review of family carer perspectives on supporting a dying person at home illustrated the desire of families to be able to provide immediate symptom relief. 23 Our review found that caregivers are willing to learn to overcome their reservations about administering SC medications. 5,15–22 The ability to alleviate their loved one’s symptoms and support them to stay at home was paramount.

There is an existing evidence-based education package and medication resources; a Brisbane group developed and evaluated an educational package and a randomised trial was completed of who prepares the SC injections (carer, nurse or pharmacist). 5,14,24 In Singapore, a colour-coded ‘Comfort Care Kit’ is in use,25 with oral and non-oral as-needed medication provided for caregiver administration. A telephone survey of 49 family carers showed that 67% of them used this kit, all family members found it easy to use and 98% of them found it effective for symptom management. All except one patient died at home. In Canberra, the provision of an emergency medical kit (including for use by lay carers) was largely viewed as an effective strategy to achieve timely symptom control and to prevent inpatient admissions. 19

There is growing UK evidence around the carer role for patients who are in the last months/year of their life, but few studies focus on the last days of life (as reiterated by the Neuberger Review into the Liverpool Care Pathway26). The evidence that is beginning to accumulate mostly focuses on patients who have capacity in the last year of their life. UK and Australian research includes ‘Unpacking the home’,27,28 the Cancer Carers Medication Management work,29,30 the SMART (Self-Management of Analgesia and Related Treatments) study31 and ImPaCCT (Improving Palliative Care through Clinical Trials). 32 Our study, by contrast, focuses on the last days of life at home, when the individual’s capacity is likely to be limited or absent, with very different implications and issues for carer administration.

Community receptivity

The UK was ready for testing this extended lay carer role:

• Primary care teams and families are used to similar practices in other areas of medicine (e.g. insulin for diabetes mellitus and intravenous antibiotics for children with cystic fibrosis).

• The PeolcPSP report incorporated the views of 1403 people across the UK and placed great emphasis on empowerment of family carers and symptom management during the last days of life. 11

• Ambitions for Palliative and End of Life Care: A Framework for Local Action was published in September 2015. 33 It was jointly developed and published by the National Partnership for Palliative and End of Life Care (27 national organisations) and has widespread support in the UK, especially as the partnership included the Patients’ Association and charities with large public contributor groups. They identified eight foundations for their six ambitions; one of these foundations relates to ‘Involving, supporting and caring for those important to the dying person’, acknowledging the importance of lay carers in the caring team. Each ambition has a set of building blocks; the building block of ‘practical support’ in ambition 6 is particularly applicable to the CARiAD trial as it calls for ‘new ways to give the practical support, information and training that enables families . . . to help’. There has been a strong positive response to the framework, and many localities are using it to consider their local strategies. The message about shared ownership and responsibility is particularly pertinent.

• In the UK, this practice is not widely accepted as usual care; however, over the past few years, we have identified a small number of geographically distinct sites where the practice occurs (< 10 sites). Recently, the Lincolnshire project was showcased on national radio as part of a series of talks on dying. 21,34,35 Since the conception of the CARiAD trial, at least seven other areas have, unsolicited, expressed interest in exploring the practice, including joining as potential recruitment sites in a future definitive trial.

Pressure on health and care services in the UK

Health-care professionals in all three sites were universally positive about testing the intervention; if found to be beneficial, it could make their patients more comfortable and their jobs more manageable. In the longer term, this innovation could relieve some pressure on emergency departments by reducing inappropriate emergency (crisis) admissions due to uncontrolled symptoms. 36,37

Pressure on DN time could also be relieved as extra visits (in addition to the daily check) to administer as-needed medication would decrease, which would contribute to the sustainability of services.

Choice of design

Our team was aware of the challenges that are associated with research in the last days to weeks of life in general, including recruitment and ethics considerations. Although we recognised the benefits of conducting an internal pilot trial with progression rules to a full trial, we felt that an external pilot and feasibility study was more appropriate. Regarding recruitment, an external pilot trial required three sites recruiting 50 patients. A full trial would require at least 30 sites recruiting 520 patients. Although our three sites were confident about recruitment, and a number of other areas had already expressed interest in participating in a future trial, we could not disregard the effect of the other complex factors affecting a broader roll-out.

These specific additional considerations added to the decision to propose a stand-alone (external) pilot trial:

• The current UK context (post Shipman,38 post Liverpool Care Pathway26 and with the ongoing euthanasia public debate) – this called for careful attention to its affect on consent mechanisms and attitudes of carers, patients and clinicians to this innovation.

• The lack of clear UK-wide guidance on carer administration of as-needed SC medication to patients dying at home – the practice is legal but current guidance was not detailed or specific enough for wide adoption (see Appendix 1).

• The lack of a clear and widely accepted training package for lay carers, adapted for the UK context.

• The uncertainty about the primary outcome measure for a definitive trial.

These were unpredictable barriers until we began to introduce the reworked Australian manualised intervention and tested the trial processes. If the intervention was proven to be feasible and acceptable, we anticipated a phase of ensuring that new guidance is developed and put in place at national level in UK health systems to enable the practice, prior to rolling out a full trial.

We aimed to demonstrate a clear path towards a definitive randomised controlled trial (RCT), as per Medical Research Council Framework, for the evaluation of complex intervention principles, further informed by the MORECare guidance developed for palliative care research. 39,40

Introduction

From discussion with our Australian co-investigators, we developed a good understanding of the Caring Safely at Home resources, which consist of carer and HCP resources. When the CARiAD study commenced, carer resources that were part of the Caring@Home study comprised:

• Illustrated step-by-step guides that provided a simple structured approach to the preparation and administration of SC injections –

  • opening and drawing up from an ampoule

  • 10-step plan – blunt needle technique

  • 10-step plan – no-needle technique

  • injection via cannula – blunt needle technique

  • injection via cannula – no-needle technique.

• A practice demonstration kit, including a cannula and other equipment that are required for the preparation and delivery of SC injections.

• Colour-coded medication labels for labelling prepared syringes.

• A fridge magnet that was colour coded to help the clinician or caregiver match the relevant medication with the patient’s symptoms.

• A daily diary to document aspects of medication administration.

• A series of videos, ‘Palliative subcutaneous medication administration: a guide for carers’, that demonstrated aspects of SC medication preparation and administration.

• A booklet, ‘Subcutaneous medication and palliative care: a guide for caregivers’, that covered topics including frequently asked questions, the importance of symptom control, management of common palliative symptoms, commonly used SC medications and injecting processes.

The HCP resources comprised:

• the Caring Safely at Home Standardised Education Framework

• detailed instruction guidelines for the Caring Safely at Home resources and education framework

• a competency checklist that provides the clinician with a mechanism to ensure that the caregiver is competent to safely inject SC medications

• a clinician lanyard that provides a useful overview of the educational framework and the colour-coding scheme utilised for the medications.

The Queensland materials continue to be freely available on the project website (www.caringathomeproject.com.au/; accessed 5 January 2020) and are updated frequently (most recent update 12 June 2019). The materials are now available for use in all Australian jurisdictions.

In this chapter, we describe the development of the manualised training package that was used in the CARiAD trial. It does not cover all of the supporting trial documents, for example the participant information sheets (PISs), screening log or daily DN checklist.

Methods

We developed the manualised training package materials in three distinct stages:

1. clarifying focus and scope

2. developing the content

3. deciding the design, colour scheme and layout.

The final materials would be used in the trial and feedback would be obtained on their utility, acceptability and feasibility via qualitative interviews and informal feedback processes.

A more detailed description of each of the three stages is as follows:

1. Clarifying focus and scope.

   After detailed review of the suite of Australian materials (including any updates during the time the CARiAD trial materials were developed), and after accounting for practice differences between Australia and the UK, the materials were adapted for a UK trial setting. All references to the Australian practice of drawing as-needed medication up in advance, labelling with colour-coded labels and storing in the fridge (with supporting aide memoires such as fridge magnets and lanyards) were removed. All references to the Queensland-wide framework Guidelines for the Handling of Medication in Community-Based Palliative Care Services in Queensland (March 2015) were removed. Wording was scrutinised and, if needed, aligned with the UK practice of carer administration being tested in the trial. Details in the Australian framework pertinent to the CARiAD trial were included in the trial protocol.

   Permission was granted by our Australian co-applicants for us to use the graphics in their materials as appropriate.

   The Quality of Life in Life-Threatening Illness--Family Carer Version (QOLLTI-F) questionnaire was added to the diary, to be completed every 48 hours following the first administration of SC as-needed medication (see Appendix 2). 41

2. Developing the content.

   This stage required detailed, word-for-word consideration of the Australian materials and adjustments as appropriate. In the first instance this was carried out by the core CARiAD team, who also cross-checked the content against other available documents (specifically the Lincolnshire policy and supporting documentation and the risk assessment tools for self-medicating). 35,42 This was reviewed by the co-applicants who are clinicians. As a final check prior to discussion at the expert consensus workshops, it was considered by our public contributor co-applicants. Overall, minimal adaptation was needed, including of terminology and drug names. Detail included in the Australian Standard Education Framework was incorporated into the training that HCPs received on the trial.

3. Deciding the final design, colour scheme and layout.

   The final design stage was an iterative process between the core team and the design company, with invaluable input from the public contributor co-applicants, especially on colour scheme choice (distinctive enough but not intrusive, i.e. not using too strong colours and using different colours for different trial booklets to support ease of recognition for carers and HCPs) and cover graphics.

Results

The final set of documents included in the CARiAD trial manualised training pack for carers comprised:

• ‘Subcutaneous medication for breakthrough symptoms in the last days of life: a guide for carers’ (see Report Supplementary Material 1).

  This was adapted from the booklet ‘Subcutaneous medication and palliative care: a guide for caregivers’ and covered similar topic areas. In addition, it included information on how to know which medication to give, how long to wait for medication to work, how often to give medication, which symptoms medication can be given for in the trial, how many doses could be given, what to do if too much medication is given and what to do if the carer is not available to give medications. A section was also included broaching the concept of the ‘last injection’. After careful discussion with our public contributor co-applicants, the following paragraph was added –

  There may come a time when you are giving the injections when the person you are caring for is very ill and will soon die. This might mean that the time when they die is near to when you have last given them medication. It is very important for you to know that these two things are not related and the medication has not ended their life. The nurse training you to give injections will discuss any worries or concerns you may have about this.

Rather than leaving it to individual clinicians to find their own way of explaining this, the text was intended to provide a script on which the HCP’s explanation to the carer could be based.

• A carer diary (intervention or usual care), which was A5 size (see Report Supplementary Material 1). This was divided into the following sections –

  • introduction, explaining provenance and focus

  • contact details of the teams involved in the dying person’s care and space to record medical appointments

  • instructions for use of the diary when recording symptom management

  • information on the prescribed as-needed medication (intervention group diary only)

  • record of symptoms and subsequent management (one page per entry, 15 entries per booklet)

  • QOLLTI-F questionnaire (three copies per diary, additional diaries provided by DN teams as required).

Changes to the carer diary included assigning a full A5 page to each administration entry (rather than five administration entries per page, as in the Australian carer diary) and amending the medication information table to align with commonly used medication in the UK. The context of a RCT necessitated the need for carer diaries with different content for the two trial groups; therefore, two versions of the diary were produced.

• Step-by-step guides with images illustrating the required actions (see Report Supplementary Material 1) –

  • step-by-step guide to opening and drawing-up medications from an ampoule

  • 10-step plan for preparing and giving as-needed SC injections using a blunt needle technique

  • 10-step plan for preparing and giving as-needed SC injections using a no-needle technique.

An injection training pack was also included so that the carer could practise giving needle-less injections (i.e. via a cannula). We listed the suggested content to be assembled by the HCP from their stocks (see Report Supplementary Material 1). Given that ampoule openers are not in widespread use, these were provided to the HCPs for inclusion in the training packs.

In relation to carer resources, we were advised that the video resources were not often used and that carers preferred the printed resources; therefore, video resources were not developed.

More extensive changes were required to the HCP resources, as it was outside the scope of the trial to generate a standardised education framework for a UK context. The framework was therefore removed from the HCP resources along with the detailed instruction guidelines, as HCPs delivering the intervention would be comprehensively trained by the core CARiAD trial team. Additional resources were developed to support HCPs and ensure the safety of patients and carers (see Report Supplementary Material 2) and comprised:

• risk assessment (RA) tool

• information for prescribers (intervention group only)

• daily checklist for DNs – this document was an aide memoire of all the trial-related tasks to be carried out at DN visits

• adverse event (AE) record – to be kept in the patient’s notes and returned to the trial team at the end of the study. It defined AEs, serious adverse events (SAEs) and related events, included a table to record AEs and provided a prompt of the circumstances in which a separate SAE form should be recorded.

The competency checklist was amended to include the following questions about the nominated carer:

• Is aware of the symptoms of pain, restlessness/anxiety, nausea/vomiting and noisy breathing and which medications to use for each of these symptoms?

• Can reconstitute drugs if required?

• Understands the importance of completing all associated study paperwork?

• Is aware of how many as-needed doses can be administered of each drug in 24 hours?

• Understands the importance of contacting the health-care team immediately if an error is made with medications or unusual symptoms develop?

The date, signature and name of the HCP were also included in the competency checklist. In addition, two columns were added so that ongoing checks of competency could be recorded if needed.

The intervention under scrutiny in the CARiAD trial was carer administration of as-needed SC medication and, for this reason, the training to enable carers in this practice was key. The printed documents described supported carer training and, therefore, there was ongoing review of these materials throughout the trial, resulting in minor changes.

Introduction

The aim of the expert consensus workshops was to refine trial processes, guide the adaptation of the existing Australian training materials for use in a UK context and inform the development of other supporting documents. The chosen format was based on the successful model used in the Early Lung Cancer Identification and Diagnosis (ELCID) trial. 43

Methods

Two workshops were planned with invited UK experts: one in North Wales and one in the southern part of the UK (either south Wales or Gloucestershire). The invited experts from the three recruitment sites were to include public contributors, primary care clinicians [general practitioners (GPs) and DNs], specialist palliative care (SPC) clinicians, research nurses (RNs), pharmacists and representatives of other support services [including hospice at home and Marie Curie (London, UK)]. Workshops were facilitated by two members of the core CARiAD trial team, one of whom made notes throughout the workshop. Workshop attendees were given background to the trial, including detail on how the question for the research was developed and the existing Australian training package that was to be adapted for the CARiAD trial. The facilitators then led discussions on the following areas:

• identifying and approaching participants

• consent processes

• prescription, supply and storage of drugs

• delivery of the intervention

• monitoring and accountability

• follow-up measures

• post-bereavement interviews

• ethics concerns

• resource pack.

Attendees were given the opportunity to freely discuss, ask for clarification on or raise any queries or concerns about these areas. In addition, the core CARiAD team raised specific questions that required input.

Workshops were audio-recorded and a summary of discussions and resultant actions from each workshop was produced.

Results

Following the first workshop in North Wales, the decision was taken to have two separate workshops in the southern part of the UK (i.e. for both south Wales and Gloucestershire recruitment sites), as it became apparent that the specific details of the local ways of working would be vital in understanding how to deliver the trial intervention. Discussions at each workshop built on and refined those from previous workshops, interpreting them in a local context. There were 10–12 attendees at each of the three workshops, with at least one representative of each of the invited groups present at each meeting. A co-chief investigator for the study (MP) facilitated two of the workshops, supported by the trial manager (JR). The third workshop was facilitated by the principal investigator (PI) for the site (AB) and the qualitative lead for the study (AN).

Discussion

The consensus workshops provided rich, site-specific information from a range of expert stakeholders to inform the trial processes and the development of carer training materials. Attendees were positive and supportive about the trial and its aims, and the majority of concerns that were expressed were addressed by the trial team or other attendees by sharing their experiences.

Suggestions made in the workshops were considered by the core CARiAD team and were incorporated into training materials and other trial documents accordingly. It was apparent that clear guidelines for prescribing for intervention patients would be required and that an information sheet for prescribers should be compiled to enable prescribers to adhere to these. This summarised the issues raised in the workshop and suggested dose regimens for as-needed medications. As reconstitution was not likely to be common, it was felt that no step-by-step guide for this would be needed. Specific changes that were required to the RA tool were removing reference to the risk of transmissible disease, adding a check that there was a suitable place to store medications and amending it to allow it to be used as an ongoing document. Carer burden concerns were addressed as much as possible in the carer handbook and through signposting to OOH services as well as by ensuring that during training carers were aware that there was no obligation to administer as-needed medications, even when trained.

The importance of GP support for the study, particularly regarding the prescribing requests for intervention patients, required all general practices in the areas covered by the chosen DN teams to indicate that they had no objections to patients being approached to take part. Although some DN teams planned to involve the GP in the initial RA, this was unlikely to happen in every case, and for this reason the general practices were targeted by the core CARiAD team prior to opening trial recruitment.

The issue of whether or not the RNs could be unblinded was not resolved, and further discussion of this with the core CARiAD team and trial steering committee was required.

After DN teams were identified at each site, they were trained by the trial manager in how to recruit participants and how to train carers in the intervention group. They were provided with all of the materials and information, and the purpose of the documents was clearly explained, particularly the RA tool and the competency checklists. When usual-care processes already existed and there had been no strong views in the consensus workshops about the need for any changes to these, and as long as there was no impact on safety, DN teams were encouraged to continue with these processes. For this reason, no trial paperwork was provided for drug stock checks because these checks were already being carried out as part of current practice. When DN teams felt that it was appropriate and feasible for the carer, the carer was trained to complete the Medication Administration Record along with the carer diaries.

The General Medical Council (GMC) and Nursing and Midwifery Council (NMC) were contacted to establish their stance on recruiting carers to the CARiAD study who were also HCPs. Although they were unable to offer complete clarity, they did not object to HCPs being included in the study. As all carer participants were to be trained as lay carers, regardless of their background, they were not excluded from taking part.

Trial setting/context

The trial was carried out in community settings in North Wales {Betsi Cadwaladr University Health Board [BCUHB], the Vale of Glamorgan [i.e. Cardiff and surrounding areas, Cardiff & Vale University Health Board (CVUHB)] and Gloucestershire [Gloucester Care Services (GCS)]}, where patients were likely to die at home in accordance with their wishes and without the provision of round-the-clock paid care. The three pilot study sites were chosen as they are representative of the range of sites for a future definitive study.

Eligibility criteria

Recruitment

Informed consent

A RN sought consent at a time that was judged to be suitable by the attending HCP. This gave the patient and carer as much time as they needed to understand the nature of the research, ask questions and make their feelings clear on trial participation. 49,50 If the patient was unable to consent or they had lost the capacity to do so after they had previously given consent, the assent of a personal consultee was sought (as required by the Mental Capacity Act 200551) to the patient’s participation in the trial. 49,50 As the RA excluded dyads for whom there were concerns about relational issues between the patient and the carer, the carer could act as personal consultee.

If the carer did not wish to act as the personal consultee and there was no additional family member or close friend to take on this role, we appointed a nominated consultee (e.g. a HCP not associated with the research) who would act for all patients in this situation in the trial.

The PI retained overall responsibility for the informed consent of participants at their site and was mandated to ensure that any person who was delegated responsibility to participate in the informed consent process was duly authorised, trained and competent to participate according to the ethically approved protocol, principles of Good Clinical Practice and Declaration of Helsinki. 52,53

Method of implementing the allocation sequence

Once the dyad had consented and baseline data collection had been completed, the dyad was randomised to one of the trial arms. Secure online randomisation hosted by the North Wales Organisation for Randomised Trials in Health (NWORTH) Clinical Trials Unit was carried out by the researcher who obtained consent or the trial manager. The system used a dynamic adaptive method of randomisation in which it stratified for recruitment centre and diagnosis (cancer/non-cancer). 54 Confirmation of allocation was sent only to those members of study staff who needed to be aware of the result.

Blinding

The CARiAD trial was an open-label trial in which blinded-outcome assessment was not feasible; therefore, it was important that outcomes were as robust as possible in the light of the lack of blinding. See Appendix 2 and Table 27 for details of the primary outcome contenders, methods of assessment, strategies to reduce bias (by increasing subjectivity) and criteria for assessing feasibility as a primary outcome measure for a future definitive trial. Outcome assessors were experienced RNs.

Data entry was completed unblinded; the trial statistician who carried out the data analysis was the only individual who was blinded to randomisation allocation. The analysis was unblinded at a combined Trial Steering Committee (TSC) and Data Monitoring and Ethics Committee (DMEC) meeting, with independent members present.

Withdrawal criteria

Participants were free to withdraw from the trial at any time without giving reasons and without prejudicing their further treatment. This was made clear to all potential participants at the time that they consented and throughout their time in the trial. Non-completion of the follow-up questionnaires did not constitute formal withdrawal from the trial; unless the participant requested withdrawal of their data completely, all data collected were used for analysis. The RA was reviewed at intervals based on the HCP’s judgement, and if the criteria were not met the dyad was withdrawn from the trial.

Schedule of study procedures

Table 1 details the planned study procedures.

Health technologies being assessed

The technology under scrutiny was the extended role of lay carers to administer as-needed SC medication for common symptoms to a person who was dying at home. Lay carers were trained in this practice, and their training was supported by a manualised training package that was based on the Australian package Caring Safely at Home (see Chapter 2 for details of the development of the intervention).

Of note, it is usual practice in the UK to ensure that there is provision of as-needed medicines for breakthrough symptoms in the patient’s home for administration by the attending HCP. 55–57 The difference in this technology is that lay carers were trained and, therefore, had the option to administer these medicines (instead of and/or in addition to HCP administration).

Recruitment measurements

Recruitment measurements were the number of eligible patients who fulfilled the inclusion criteria and were willing to be randomised, which was expressed as a percentage of the number of patients screened, the number who withdrew after baseline assessment and randomisation, the number who completed the various outcome measurements at baseline and at later time points, and reasons for any non-completion.

Patient measurements

Patient measurements were baseline information (including demographic information, medical history, capacity assessment, preferred place of care in the last days of their life and current drug management) and a daily carer diary during the study that was related to the presence and treatment of breakthrough symptoms (for use in both trial groups). Data points in each medication administration diary entry included the initial time when the breakthrough symptom triggered a perceived need for an additional SC dose; whether this was noted by the patient or the lay carer; medication and dose, and time given; reason for medication (e.g. pain, nausea, restlessness or noisy breathing); symptom score before and 30 minutes after medication administration; and when symptom control/reduction of the symptom to an acceptable level was achieved. Hospital or hospice admissions during last illness and the actual place of death were also recorded. Approximately 6–8 weeks post bereavement, carers were asked to complete the Family Memorial Symptom Assessment Score – General Distress Index (MSAS-GDI). 10,59–61

Carer measurements

Carer measurements were demographic information at baseline; QOLLTI-F questionnaire (at baseline, after the first as-needed SC medication, then every 48 hours thereafter until the patient’s death); whether or not HCP support was sought; and Family MSAS-GDI (a measure of the patient’s symptom distress in the last 7 days of life) at the 6–8 week post-bereavement visit. In the intervention group, the confidence in administering SC medication and competence at intervals after training were recorded.

Health-care professional measurements

Health-care professional measurements were the baseline measurements of attending team structure, primary prescriber and carer trainer.

Safety

The CARiAD study contained a number of safety outcome measures at different stages of the clinical journey taken by the patient, carer and HCPs. Safety outcome measures include the RA tool, competency checklist and significant event reporting (see Report Supplementary Material 2). Significant event reporting included the following: the appropriateness of administration (is administration accompanied by evidence of need?); proportionality (has the correct dose been administered?); side effects both anticipated and not anticipated; drug accountability (do stocks tally?); and carer events (e.g. distress, injury, accidental or purposeful self-administration).

An AE was defined as any untoward medical occurrence in a trial participant (either the patient or the carer) and included incidents that were not necessarily caused by or related to the trial. A SAE was any untoward occurrence that resulted in death, was life-threatening, required inpatient hospitalisation or prolonged existing hospitalisation, resulted in persistent or significant disability/incapacity or was otherwise medically significant.

All AEs and SAEs were captured via the significant event form. SAEs were reported to the PI and sponsor within 24 hours. As this was a study involving patients who were close to the end of their lives, death was an expected outcome. It was recorded and reported to the sponsor but was not considered a SAE if, in the opinion of the PI, it was a natural conclusion to a patient’s life-limiting illness. Owing to the nature of the study, events of death did not require immediate reporting to the DMEC.

Criteria for assessing feasibility as primary outcome measure

All outcome measures were assessed on the same criteria for consistency.

Applicability:

• This was assessed by an independent expert panel and was based on feedback from participants (HCPs and carers).

• Each measure was assessed by the panel with regard to its relevance and applicability to the population, based on the outcomes of the pilot data collection phase.

• The panel will recommend an ‘accept’ or ‘not accept’ status for each outcome based on the criteria below and their expert opinions, and taking into account the RATIONALE (Risk of Bias Justification Table) statements on outcome measures and assessment of bias risks in ultimate reporting. 62,63

Acceptability:

• This was assessed by participants and HCPs during the qualitative aspects of the feedback interviews.

Level of completeness:

• This was assessed by the frequency of missing data during the data collection phase. This would require potential primary outcome measures to have > 70% completeness.

• An assessment will also be made of the reasons for missing data to establish whether or not anything systematic in the trial design could be adjusted to mitigate for the missing data.

Once the feasibility of the outcomes is established, the design of the definitive trial will consider whether a single or a combined primary outcome of interest is appropriate.

Secondary outcome measures

The potential suitability of the secondary outcomes will be considered, as detailed below.

Detailed considerations

The 2013 Office for National Statistics data66 showed that 8.6% of all deaths in those aged > 15 years are home deaths caused by neoplasms. Deaths caused by neoplasms were seen as a useful proxy for expected deaths. Therefore, the three recruitment sites had the following numbers of home deaths caused by neoplasms available per annum: BCUHB, 653; CVUHB, 349; and GCS, 517.

Lay carers of patients who are cared for at home in their last days of life

Drawing on the collective clinical experience of the study team, we knew that it is rare for a patient to successfully fulfil their wish to die at home if they did not have the support of a lay carer. This was because most health and care services (and specifically in the sites that we propose to recruit from) could not provide 24/7 care in a patient’s own home. That said, the best data available on this topic came from a recent quasi-experimental study examining palliative care interventions at home. It noted that of 953 patients who expressed a preference to die at home, 72.2% had an informal carer. 67 It is worth noting that the study did not record where patients who were unable to identify an informal carer actually died, or if previously unidentified lay carers in the patient’s circle stepped in to support their wish to die at home.

The pilot trial

For the pilot trial we proposed to approach 200 dyads (66 or 67 at each site). Therefore, even if only 72.2% of the patients had an identified carer, the numbers of dyads in the three sites that could be eligible were as follows: BCUHB, 471; CVUHB, 251; and GCS, 373. It is also worth noting that all of the participants in the focus group indicated they would take part in a trial if invited (caution: selection bias). Of these, on very conservative calculation, 100 dyads would consent to participate (33 or 34 at each site). The percentage recruited in the Australian study was much higher: 97.6% of those approached consented to participate. In total, 50 of those dyads would be in the intervention group (receiving training) and 50 dyads would be in the usual-care group (i.e. 16 or 17 per group per site), with, again on very conservative calculation, 25 in each group (50 in total) completing the trial (i.e. 8 or 9 per group per site). In total, 50 completers were needed for the analysis (i.e. 25% of those initially approached for participation).

Statistical analysis

The primary analysis was concentrated on the feasibility metrics and adherence outcomes based on defined thresholds (Table 2). There was limited preliminary analysis of intervention outcomes. Point and 95% confidence interval estimates were calculated and used to estimate the variability and direction of effect to further inform the sample size calculation for a definitive study.

Summary statistics of all outcomes were used to inform the approximate models of analysis that would be used in a full trial. The results of the feasibility trial were used to inform the most appropriate analysis models (e.g. the number of episodes in which as-needed medication was used and proportion of participants who never required as-needed medication). A preliminary analysis of the outcomes was completed using an intention-to-treat approach. All analysis that was undertaken was prespecified in a statistical analysis plan that was written and agreed before data collection was completed.

As this was a feasibility trial, there was no imputation of missing data. Missing data were considered as a criterion for assessing the suitability of measures. Descriptive statistics were produced for each of the outcome measures to evaluate the appropriateness of the measures for inclusion in a definitive RCT.

Progression to a full trial

Clear progression rules were defined to determine whether or not an application for a future substantive trial that was powered to study clinical effectiveness and cost-effectiveness should proceed. Our progression rules related to the following measure that we considered important to feasibility: reaching our target (16–17) for the number of patients recruited per site.

We also established clear assessment criteria for establishing the acceptability of the potential primary outcome measures.

Table 2 explains sample size calculations and Table 3 summarises the objectives, action plan and criteria for progression to a full trial.

Summary of any changes to the study protocol

Version 1 was approved by the funder (October 2016) and detailed the whole CARiAD study (including the work-up phase/expert consensus workshops). The workshops had concluded and the results were incorporated in the trial processes and materials. Version 2 focused on the trial (with embedded qualitative study), which was due to commence in October 2017. Further amendments were made in response to REC comments (July 2017) to clarify details of the qualitative methods (June 2018) to reflect the decision to unblind RNs and the affect on trial processes (August 2018) and to make general corrections to trial processes and reflect the extended trial period (November 2018). The current protocol is version 6 (27 November 2018).

Aim

In this study, the aim of public contribution was to collaboratively involve patients as co-applicants and partners at all stages of the study, with the intention of improving the design and conduct of the study and ensuring that it was informed by experiences of giving injections to loved ones dying at home, and to involve public contributors with appropriate experience to ensure that there was full understanding of the needs of research participants.

Methods

A range of methods were used for public contribution at different stages throughout the study. These included the following:

• The initial decision to develop the CARiAD trial resulted from the Palliative and End of Life Care Priority Setting Partnership report, which incorporated the views of 1403 people across the UK, which placed great emphasis on empowerment of family carers and symptom management during the last days of life. 11

• The study design was heavily informed by the work-up stages, which comprised two group consultations with carer groups (suggestions on consent mechanisms, drug safety, training and ongoing support were incorporated into the study design), one telephone consultation with a bereaved carer (JO’C) who had given SC medications to her late husband and one telephone consultation with a nurse who had trained a carer to give SC medication to a relative and had supported them in doing so.

• The CARiAD trial had two public contributors as co-applicants to the study (BF and JO’C) who contributed to the submitted application, including providing the lay summary. Following this, one of the public contributors (JO’C) attended and contributed significantly to the CARiAD REC meeting.

• Trial materials, including the topic guides for the qualitative interviews, were produced with the input of the public contributors. Public contributors also took part in expert consensus workshops and formed part of the TSC.

• The public contributors strongly supported the study dissemination. One of the public contributor co-applicants (JO’C) presented the results of the CARiAD study at a national palliative care conference.

Public contribution was of an extremely high standard in this study and public contributors formed an integral part of the research team. When public contributors were able to attend meetings in person, measures were taken for considerate inclusion, such as access, briefing, refreshments and travel provision and expenses. The two public contributor team members/co-applicants were supported in the role directly both by the CARiAD team and by their respective groups (Sue Ryder Hospice and North Wales Cancer Network).

Results

The outcomes of public contribution in the early stages of the study included a better-designed study, particularly based on input during the study work-up phase. This added to, altered or confirmed a number of study design elements. Throughout the study, the study team were better informed through the sustained and active public contribution in meetings and developed a stronger sense of bereaved carers’ perspectives on the appropriateness of the approaches considered.

In more practical ways, it was an enormous advantage to have the presence and participation of Julie O’Connor at the NHS REC meeting. This gave the REC a sense of the meaning of the study and how it may be perceived by bereaved carers, and reassured them that the study intervention could be welcomed by carers. The public contribution improved the quality of a number of the public-facing documents, such as the qualitative topic guide and the study close-out documentation for carers. We identified no negative outcomes from public contribution.

A more indirect outcome that may have been influenced by participating in this study is that one of the public contributors (JO’C) has decided to embark on a new career in end-of-life care.

Reflections

Overall, public contribution influenced this study to a great extent. At all stages, the study was kept on track by ongoing and active public participation. The study team learnt a lot and changed a lot. We were guided through a very sensitive and potentially distressing study and steered towards appropriate and acceptable approaches. Not only was the instructive, beneficial and reassuring for the team but it also resulted in more of the right approaches being used to ensure that the study proceeded with respect and consideration for the study participants at a very difficult, and important, time of their lives.

Site initiation

Final REC approval was granted on 3 August 2017 and excess treatment costs were approved on 26 September 2017, both ahead of schedule (Table 5).

Participant flow

Trial recruitment is presented in Figure 2 and Table 6. There were 189 potential dyads identified at pre-screening in which the patient was in the last weeks of their life and was likely to lose the oral route of medication administration. Of these, 169 dyads were screened.

FIGURE 2.

Trial recruitment overall and by site (BCUHB, CVUHB and GCS).

In total, 68 out of 169 dyads were ineligible. The main reasons for ineligibility were that only paid or formal care was in place (n = 18, 26.5%), the patient did not wish to die at home (n = 13, 19.1%), the carer was disorientated, confused or forgetful (n = 5, 7.4%) and there were relationship issues between the patient and the carer (n = 5, 7.4%).

For 11 out of the 169 dyads eligibility was confirmed, but for eight no further information was provided and in three dyads the patient died before approach.

The remaining 90 dyads were approached to participate in the study. Of these, 25 declined to take part. From the patients’ perspective, the reasons given for declining were not wanting to be randomised (n = 1, 4%) and not wanting the carer to administer medications (n = 5, 20%), and two (8%) declined for reasons not specified. From the carers’ perspective, three (12%) carers were concerned about burden, two (8%) did not want to be randomised, six (24%) did not want to administer medications, one (4%) did not want to complete paperwork and five (20%) declined for reasons not specified.

There were 40 dyads (40/101, 39.6% of the eligible population; 40/90, 44.4% of those approached) who completed the baseline visit and were randomised. These 40 dyads were allocated evenly: 20 to the usual-care group and 20 to the intervention group. A total of 22 carers (55% of those randomised) completed the follow-up visit: 16 (80% of those randomised) from the intervention group and six (30% of those randomised) from the usual-care group.

Recruitment site context

Betsi Cadwaladr University Health Board serves a population of 678,000 people in North Wales, providing community and inpatient services. BCUHB employs 16,500 people. Seven DN teams were part of the CARiAD trial from the outset and were chosen after discussion with DN matrons based on their capacity and capability, and were located across a mix of urban and rural areas. One additional team was invited to join the study and opened to recruitment 6 months after the first seven teams.

Cardiff & Vale University Health Board serves a population of 445,000 people in the Vale of Glamorgan and provides community and inpatient services. CVUHB employs 14,500 people. Two DN teams participated and recruitment was also opened via the SPC team, covering the acute hospital and local hospice.

Gloucester Care Services serves a population of 600,000 people in Gloucestershire, providing community services only. GCS employs 2700 people. Five DN teams and one SPC team recruited to the study.

In all sites, the number of DN teams involved in the CARiAD trial was a small proportion of the total number of DN teams across the health authority. This had an impact on the number of potential patients screened compared with our initial assumptions. However, DN matrons and other senior staff members who were involved in the assessment of capacity and capability of the teams were confident that our required numbers could be met from a smaller number of teams, and felt that this was an acceptable compromise in managing the aims of the trial and the capacity of the NHS services.

Measures to support recruitment

Affect of the measures to support recruitment

A number of measures were taken throughout the recruitment period to try to improve the number of dyads who were recruited and retained (Figure 3). These included the following: refresher training for DN teams in May (three teams), June, July and October; feedback enhancements to all recruiting teams in addition to the regular planned feedback; increased involvement of SPC nursing colleagues; identification of the difficulty for RNs who were partially blinded; the substantial amendment and unblinding; the ability of teams to implement unblinding; opening of new sites in BCUHB and GCS; and the overall no-cost extension to gain more recruitment time.

FIGURE 3.

Measures to support recruitment. SPCN, specialist palliative care nurse.

Patient demographics

There were 40 dyads who were randomised to the study. Among these 40 dyads, 22 carers completed the follow-up and it was possible to obtain some information from medical notes for a further three patients; therefore, partial data are available for 25 patients at follow-up.

The mean age of the patients who were randomised was 68.3 years (SD 12.6 years); 23 out of the 40 (57.5%) patients were male and 17 (42.5%) were female. Information on rurality was available for only three patients (two rural and one semirural) from the quantitative data, although rurality was revealed for a subset who took part in the qualitative study.

There were 18 (out of 25 patients with data, 72%) patients who died at home, with other patients dying in either a hospital (n = 2, 8%) or a hospice (n = 5, 20%). Seven patients (28%) had at least one unscheduled admission to hospital. Of these seven patients, six died during admission and one died at home following admission.

Carer demographics

The mean age of the carers who were randomised to the study was 56.6 years (SD 12.0 years); 35 out of the 40 (87.5%) carers were female and five (12.5%) were male. Half of the 40 patients were cared for by their spouse/partner (n = 20, 50%), 15 (37.5%) were cared for by their child, three (7.5%) were cared for by another relative and two (5%) were cared for by a friend. There were 32 (80%) carers who lived with the patient and eight (20%) who did not live with the patient. Of the 40 carers, 26 (65%) declared that they lived in their usual place of residence and 14 (35%) did not live in their usual place of residence. There were 31 (77.5%) carers who had a role in managing the patient’s medications prior to involvement in the trial and eight (20%) who did not have this role prior to the trial; data were missing for one carer. Half of the carers were the sole carer (n =20, 50%) and 18 (45%) were not the sole carer; data were missing for two carers.

Medications

At baseline, 33 out of the 40 (82.5%) patients had anticipatory prescribing in place and seven (17.5%) did not. Medication was prescribed by a medical prescriber from the primary care team (13/32, 40.6%), a medical (9/32, 28.1%) or non-medical (8/32, 25%) prescriber from the SPC team or a non-medical prescriber from the secondary care team (2/32, 6.3%). At death, all patients (n =25) had anticipatory prescribing in place. Of these patients, nine (36%) had this prescribed by a medical prescriber from the SPC team, six (24%) had this prescribed by a non-medical prescriber in the SPC team, seven (28%) had this prescribed by a medical prescriber in the primary care team, one (4%) had this prescribed by a non-medical prescriber in the primary care team and one (4%) had this prescribed by a medical prescriber in the secondary care team. Information on who prescribed the medication was missing for one patient. Of the 25 patients who had anticipatory prescribing in place at death, four patients did not have this in place at the baseline visit.

At baseline, 27 out of the 40 (67.5%) patients did not have a continuous SC infusion set up and 13 (32.5%) did. Of those who did have a continuous SC infusion, six (46.2%) were prescribed it by a medical prescriber in the primary care team, two (15.4%) were prescribed it by a medical prescriber from the secondary care team, one (7.7%) was prescribed it by a non-medical prescriber in the secondary care team, two (15.4%) were prescribed it by a medical prescriber in the SPC team and one (7.7%) was prescribed it by a non-medical prescriber in the SPC team. Information regarding who prescribed the medication was missing for one patient. In total, 21 out of the 25 (84%) patients had a continuous SC infusion set up at death and four (16%) did not. Among patients who had a continuous SC infusion, eight (38.1%) were prescribed it by a medical prescriber in the primary care team, six (28.6%) were prescribed it by a medical prescriber in the SPC team, three (14.3%) were prescribed it by a non-medical prescriber in the SPC team, two (9.5%) were prescribed it by a non-medical prescriber in the primary care team and one (4.8%) was prescribed it by a medical prescriber in the secondary care team. Information regarding who prescribed the medication was missing for one patient.

At baseline, the numbers of patients who had been prescribed anticipatory SC medication were 36 (out of 40, 90%) for pain, 25 (62.5%) for anxiety/restlessness, 31 (77.5%) for nausea/sickness and 21 (52.5%) for noisy breathing.

At the time of death, of the 25 patients for whom there were data, the numbers of patients who had anticipatory SC medication prescribed were 18 (72%) patients for pain, 19 (76%) for anxiety/restlessness, 14 (56%) for nausea/sickness and 15 (60%) for noisy breathing.

Services

In total, 21 out of the 25 (84%) dyads had accessed GP services and four (16%) had not. DN services were accessed by 21 (84%) dyads and four (16%) had not accessed them. There were 20 (80%) dyads who had accessed nursing SPC and four (16%) who had not; data were missing for one dyad. There were 13 (52%) dyads who did not access medical SPC and 10 (40%) who did; data were missing for two (8%) dyads. Other nursing was not accessed by 14 (56%) dyads and was accessed by 11 (44%); data were missing for one dyad. For health-care support, 16 (64%) dyads did not access this and seven (28%) did access this; data were missing for two (8%) dyads. Other allied health professionals were not accessed by 16 (64%) dyads but were accessed by six (24%), with missing data for three dyads (12%).

Intervention

Usual care

Dyads randomised to usual care were provided with a carer diary by the DN or RN and were instructed to complete a medication entry when the patient received medication from the DN, and a QOLLTI-F every 48 hours. The HCPs continued usual-care duties with these patients and a daily checklist was included in the notes to prompt DNs of necessary trial processes for these patients. However, there appeared to be minimal trial-related contact with the usual-care group participants.

Carer diary returns

It was intended that the DN would collect the diaries post bereavement and return them to the trial team, but diaries were also collected by RNs post bereavement or by the qualitative researcher during the carer interview. Carers indicated that they would have appreciated feedback on the diaries for reassurance that these had been completed correctly. Table 9 details reasons why diaries were not returned.

General feedback on the carer diaries included that some of the text was too small and that clarification was needed on confidence and symptom scales in the medication administration tables. On the prescribed medication table, additional lines would support dose changes and the option to add more than one medication per symptom when drugs are used in combination (e.g. cyclizine and haloperidol). The table could also be prepopulated with ‘3’ in the maximum number of doses per 24-hour period. One carer was unsure whether this was a rolling 24-hour period or from 00.01 a.m. to midnight. This could be changed in a future version along with the Medication Guide for Carers.

Symptoms resulting in medication administration

The number of occurrences per patient of pain, nausea or vomiting, anxiety/restlessness and noisy breathing that resulted in medication administration is presented in Figure 4. Three dyads in the usual-care group completed 20 medication administration entries and 11 in the intervention group completed 147 medication administration entries. These dyads are labelled UC1 to UC3 for the usual-care group and I1 to I11 for the intervention group. Most patients have zero, one or two occurrences of each symptom resulting in medication administration, with some patients having six or seven occurrences. One patient in the intervention group had 28 occurrences of pain and 43 occurrences of anxiety/restlessness recorded.

FIGURE 4.

Number of medication administrations per dyad.

The carer diaries were assessed for completion by considering the medication administration entries only (i.e. documentation regarding administering medications).

In the intervention group, among the 147 medication administration entries, when the carer administered the medication (n = 131, 89.1%) the medication administration documentation was:

• 100% completed in 10 instances (7.6%)

• 70–99% completed in 116 instances (88.5%)

• 50–69% completed in five instances (3.8%).

In the intervention group, when the HCP administered the medication (n = 10, 6.8%) the medication administration documentation was:

• 70–99% completed in nine instances (90%)

• 50–69% completed in one instance (10%).

In the intervention group, when it was not specified who administered the medication (n = 6/147, 4.1%) the medication administration documentation was:

• 70–99% completed in four instances (66.7%)

• 50–69% completed in one instance (16.7%)

• 1–49% completed in one instance (16.7%).

In the usual-care group (n = 20), when the HCP always administered the medication the medication administration documentation was:

• 100% completed in one instance (5%)

• 70–99% completed in 17 instances (85%)

• 50–69% completed in two instances (10%).

Therefore, the medication administration documentation was at least 70% complete for 94.5% of medication administration instances in the intervention group (139/147) and 90% of instances in the usual-care group (18/20). Overall, across both groups, documentation was at least 70% complete for 94% of all medication administration instances (157/167).

Time to symptom relief

Time to symptom relief refers to the time from the onset of symptom to medication administration and the time from medication administration to symptom resolution.

Time from onset of symptom to medication administration

One important item of information was the time period from the onset of symptom to the time of medication administration. For this to be calculated, the time of symptom onset and the time of medication administration were required. The time of symptom onset was available for 131 out of the 147 (89.1%) medication administration entries in the intervention group and 19 out of the 20 (95%) in the usual-care group. The time of medication administration was available for 136 out of 147 (92.5%) medication administration entries in the intervention group and 16 out of 20 (80%) in the usual-care group.

The intervention group had a considerably shorter time from onset of symptom to medication administration than the usual-care group: the median number of minutes [and associated interquartile range (IQR)] from onset of symptom to medication administration in the intervention group was 5 minutes (IQR 0–18 minutes) and in the usual-care group it was 105 minutes (IQR 74–330 minutes). In the intervention group, the median time from onset of symptom to medication administration was 10 minutes (IQR 0–20 minutes) for pain, 0 minutes (IQR 0–7 minutes) for nausea/vomiting, 10 minutes (IQR 0–30 minutes) for anxiety/restlessness and 20 minutes (IQR 15–77.5 minutes) for noisy breathing. In the usual-care group, the median time from onset of symptom to medication administration was 120 minutes (IQR 55–285 minutes) for pain, 120 minutes (IQR not applicable as only one instance) for nausea/vomiting, 92.5 minutes (IQR 70.5–461.25 minutes) for anxiety/restlessness and 80 minutes (IQR not applicable as only two instances) for noisy breathing. The time from onset of symptom to medication administration is presented in Figure 5. This graph presents the median time with whiskers from the lower quartile to the upper quartile.

FIGURE 5.

Median time between symptom onset and medication administration (minutes).

In the usual-care group, it was possible to calculate the time from onset of symptom to medication administration for 15 out of the 20 (75%) medication administration entries. Of these 15 entries:

• in three instances (20.0%) medication was administered within 1 hour

• in seven instances (46.7%) medication was administered within 2 hours

• in one instance (6.7%) medication was administered within 3 hours

• in one instance (6.7%) medication was administered within 6 hours

• in one instance (6.7%) medication was administered within 7 hours

• in two instances (13.3%) medication was administered within 15 hours.

In the intervention group, it was possible to calculate time from onset of symptom to medication administration for 127 out of the 147 (86.4%) medication administration entries. Of these 127 entries, medication was administered by the carer in 114 instances (89.8%) and by a HCP in eight instances (6.3%). In five (3.9%) instances it was not specified who administered the medication.

For the entries in which the carer administered medication (n = 114), it was administered:

• within 1 hour in 104 instances (91.2%)

• within 2 hours in five instances (4.4%)

• within 3 hours in one instance (0.9%)

• within 4 hours in three instances (2.6%)

• within 5 hours in one instance (0.9%).

When the HCP administered the medication (8/127, 6.3%) or when it was not specified who administered the medication (5/127, 3.9%), it was administered within 1 hour.

In the intervention group, in 56 out of the 147 (38.1%) medication administration entries the time of medication administration was the same as the time of symptom development. Of these entries, 50 were carer administered, four were HCP administered and for two entries it was not specified who administered the medication.

Symptom scores

The symptom scores before and 30 minutes after treatment are presented in Figure 6. This graph presents the median with whiskers from the lower quartile to the upper quartile. Scores are from 0 to 10, with 0 being no symptom and 10 being the worst the symptom can possibly be.

FIGURE 6.

Median symptom scores before and 30 minutes after treatment.

The symptom scores in the intervention group were completed for 81 out of the 147 (55.1%) medication administration entries at the time of symptom development and for 69 (46.9%) 30 minutes after medication administration. The symptom scores in the usual-care group were completed for 19 out of the 20 (95%) medication administration entries at the time of symptom development and for 19 out of the 20 (95%) entries 30 minutes after medication administration.

Both groups had an improvement in symptoms demonstrated by a decrease in the symptom scores.

In the intervention group, the symptom scores were as follows [initial median score (IQR), median score 30 minutes after treatment (IQR)]:

• pain – 7 (6–8), 0 (0–2)

• nausea/vomiting – 8.5 (7–10), 1 (0–2)

• anxiety/restlessness – 8 (6–8), 3.5 (0.25–6)

• noisy breathing – 7 (7–9), 2 (0–7).

In the usual-care group, the symptom scores were as follows [initial median score (IQR), median score 30 minutes after treatment (IQR)]:

• pain – 8 (7–10), 4 (1–10)

• nausea/vomiting – 8.5 (not applicable as only one instance), 1 (not applicable as only one instance)

• anxiety/restlessness – 8 (7.25–9), 4.5 (2.25–8.25)

• noisy breathing – 8 (not applicable as only three instances), 5 (not applicable as only three instances).

In the intervention group, the symptoms were noted by the carer in 124 out of the 147 (84.4%) medication administration entries. In the usual-care group, the symptoms were noted by the carer initially in eight out of the 20 (40%) medication administration entries and by the HCP in six out of the 20 (30%) entries.

Carer confidence (intervention group)

Carers in the intervention group were asked to note how confident they felt about administering the medication. This was a scale from 1 to 7, with a higher score indicating higher confidence. Of the 11 carers in the intervention group, nine (81.8%) administered medications. There were some entries in which the HCP had administered the medication but a confidence score had been recorded; we report the confidence scores only when the carer administered the medication. Of the 131 instances in which the carer administered the medication, the confidence score was recorded in 128 (97.7%). For pain, nausea or vomiting and noisy breathing the median score was 7, the highest level of confidence, and the median score was 6 for anxiety/restlessness. The confidence scores over time for the nine carers are presented in Figure 7. The final score is 6 or 7 for all carers, but there is fluctuation over time.

FIGURE 7.

Carer confidence scores in the intervention group over time by individual. (a) I1; (b) I2; (c) I3; (d) I4; (e) I5; (f) I6; (g) I7; (h) I8; and (i) I9.

Health-care professional visits

In 24 out of the 147 (16.3%) medication administration entries in the intervention group it was recorded that HCP support was sought. In 119 (81%) instances it was indicated that HCP support was not sought and this question was not answered in four (2.7%) of cases.

In 24 (16.3%) instances in the intervention group it was recorded that HCP support was sought. These were not clustered towards the earlier part of the care episode when the carer administered medication. In 10 instances, medication was administered by a HCP. We did not have access to the patient’s usual medication administration record nor the community/DN notes to comment on the exact amount of times a HCP attended for these 24 instances.

In the usual-care group, a HCP visit was required for all 20 cases.

Adverse event reporting

In total, there were seven AEs and 20 SAEs that we are aware of. Breakdown by site and randomisation allocation can be found in Table 11. Overall, reporting of AEs and SAEs by DNs was poor, particularly in North Wales. In most cases, the trial manager reported the event; therefore, it is possible there were other AEs/SAEs that the trial team were unaware of and that went unreported.

Of interest in Table 11 is that the total number of AEs/SAEs per site is approximately proportional to the recruitment at each site [23/40 (57.5%) at BCUHB, 3/40 (7.5%) at CVUHB and 14/40 (35%) at GCS].

Table 12 documents the causes of AEs. All were reported according to the BCUHB/Bangor University standard operating procedure (SOP) and, where necessary, further action was taken (e.g. additional team training).

All 20 SAEs were inpatient admissions to either a hospital or a hospice. Three patients were admitted once or more and one patient in the usual-care group was admitted eight times. It is understood that six patients died during their admission.

Carers were asked to return their diaries to the trial office as soon as practicable after the patient’s death. However, delays were sometimes unavoidable; for example, one diary was locked in the house of the deceased patient and their CARiAD carer lived elsewhere. Some of the SAEs/deviations were identified only many months post bereavement.

Protocol deviations

In total, 13 protocol deviations were identified. These were related to the recording of the RA process prior to approach (n = 5 deviations across four DN teams), sharing of patient-identifiable information with the central team (n = 2), having more than one carer involved in the trial processes (n = 2), errors on the medication table (n = 2), mismatch between trial documentation and MACRO (n = 1) and declining the use of a continuous SC infusion despite clinical advice (n = 1) (see Appendix 4).

Two were categorised as serious (RA recording = 1, mismatch between trial documentation and MACRO = 1) and all were dealt with according to local BCUHB/Bangor University SOP, and corrective action was taken as appropriate to each case.

The study team noted that the majority of protocol deviations reported were related to recording RAs. Although only one of the deviations related to RA recording was classed as serious, the fact that these issues occurred in two recruitment sites (BCUHB and GCS) and across different DN teams has clear implications for future work.

Survival per site

The survival times for the three sites are presented in Figure 9. The points on the plot represent the survival time for each individual patient. The box and whisker plot shows the median, quartiles and ranges of survival times.

FIGURE 9.

Box-and-whisker plot of survival (from consent to death) in days.

In BCUHB, one patient died on the day of consent, the longest survival time was 153 days and the median survival time for 15 patients (for whom the date of death was available) was 14 days. In CVUHB, two dyads completed the study: one patient survived for 23 days after consent and one patient survived for 58 days after consent. In GCS, one patient died on the day of consent, the longest survival time was 90 days and the median survival time for eight patients (for whom the date of death was available) was 36 days. The cumulative survival times are presented in Table 14. Within 28 days, 17 out of the 25 (68%) patients for whom there are data had died.

Trial close-out

The CARiAD trial recruitment ended on 15 March 2019. As of 15 May 2019, three patients of dyads recruited to the trial were still alive. Detailed close-out arrangements were made for these dyads and were approved by the REC (see Appendix 5).

For those dyads for whom survival was longer than anticipated and post-dated the trial end, careful consideration was required on an individual basis to balance the expectations and communication needs of the dyad, burden of trial procedures and safety mechanisms built into the trial. There were also specific considerations based on whether the surviving dyads were in the usual care or intervention group; therefore, in essence, each case was considered separately. To inform these decisions, a detailed document was generated to propose ‘rules’ that differed for the intervention and usual-care group, and informed trial procedures. Specific information sheets for both dyads and HCPs were also generated to ensure that best communication took place.

Family MSAS-GDI

Quality of Life in Life-Threatening Illness – Family Carer Version

The QOLLTI-F was self-completed by the carer during the baseline visit. It was then anticipated to be completed when the first as-needed medication was administered, and every 48 hours thereafter until the patient’s death.

Purposive sampling

Purposive sampling was used for the qualitative interviews with carers. The purposive sampling criteria used were geographical location, gender, age and rurality. As patients and carers were recruited to the trial they were added to sample monitoring tables, which were used to monitor participant characteristics and create a purposive sample for interviews. Using this system, carers were purposively identified to provide a range in terms of the purposive sampling criteria. Those carers were then invited to participate. These systems were in place and were followed; however, owing to study attrition, particularly in the usual-care group, carer characteristics (fewer men than women were recruited overall) and difficulty in contacting some carers, there was a need to adopt a more flexible approach to sampling, particularly as the study drew closer to closure. The final sample obtained is outlined in Table 20. Purposive sampling continued until data saturation was reached in the case of the intervention group, but it was not possible to reach this in the usual-care group owing to low numbers resulting from study attrition in the usual-care group.

Conducting the interviews

A sample of 12 informal carers was contacted by telephone 2–4 months post bereavement. The qualitative researcher telephoned the carers and explained the purpose of the interview and what would happen should they choose to take part. The carers were provided with a PIS and informed consent was taken at the time of interview.

Interviews lasted 30–60 minutes and were carried out in participants’ homes, at the research centre if requested by the carer or by telephone if circumstances necessitated. Face-to-face interviews were used in the study. However, in the case of three carers, interviews were conducted by telephone owing to the carer living overseas and having travelled back home by the time of the interview, or owing to carer request as a result of working unsocial hours. In these three cases, consent was given by post prior to the interview.

All interviews followed a topic guide (see Appendix 6), were audio-recorded and were transcribed verbatim.

The rationale for using interpretive phenomenological analysis

The carer data were analysed using interpretive phenomenological analysis (IPA). 70 The carer aspect of the embedded qualitative component of this study was designed to explore, in depth, the views of carers and the meaning that administering the medication had for them. The intention was not to test a predetermined hypothesis; research questions of this flexible nature are suited to IPA. 71 IPA is an appropriate analysis method for research such as this and seeks to understand how a particular set of circumstances are perceived by an individual and how the person is making sense of those circumstances in the context of their own social world; this was exactly what we were aiming to do in the carer interviews. Semistructured interviews are the best method for IPA because they allow the researcher and participant to engage in dialogue, during which participants are able to give their accounts as they perceive them and the researcher is able to probe areas that provide depth and richness to the data. 71 IPA was chosen as the ideal method for analysis in this study, as it sought to explore personal accounts, views and experiences. 70

The process of interpretive phenomenological analysis

In accordance with the principles of IPA, we followed a process in which transcripts were read thoroughly, case by case, and annotated with notes that were made of observations, insights and points of significance. We also made sure that notice was taken of participants’ use of language and the meanings behind their words and phrases. Each transcript was then re-read and potential emergent themes were noted. IPA follows an idiographic approach, starting with singular examples and slowly building up to wider, more comprehensive categories. 72 In this study, this was achieved by the case-by-case approach to the transcripts and the careful noting of examples. Transcripts were then re-read in more detail and themes organised into groups. The final phase was an in-depth defining and developing of themes and examining relationships between them, which included team involvement. Statements were then made to create a meaningful account. 70

Rigour and reflexivity

During the carer interview phase it was important to consider rigour at all times, which includes reliability and validity that may be translated to trustworthiness in qualitative research. 73 We believe that the methods have been well documented and could be repeated by another researcher, if trustworthiness were to be assessed. The interpretations given within this report are in participants’ own words and illustrate that interpretations reached are grounded in the raw data. 74,75 For this qualitative work, it was also important to consider the role of the researcher in creating the data and care was taken to ensure that no bias was introduced into the data. In this study, rigour and reflexivity were also considered. These were addressed by undertaking several team interpretation workshops to ensure that interpretations were true to the transcripts and analysis was careful and thorough.

Ethics considerations

The interviews with carers were likely to involve emotive topics. It was therefore of utmost importance for the interviewer to carry out the interviews in an appropriately sensitive manner. If a carer became distressed, they were given the opportunity to pause or discontinue the interview. It was also felt important to bear in mind that emotion does not always equal distress,76 so participants were offered the opportunity to continue the interview even if they became emotional, allowing them to decide for themselves whether or not they wanted to continue. Information on signposting to relevant HCPs was also available if necessary.

In the carer interviews, some participants did indeed become emotional; however, no carer decided to discontinue the interview. Rather, they seemed to find the experience somewhat cathartic and described taking comfort in the knowledge that contributing to research may help others in the future.

Sample

The characteristics of the carers interviewed are shown in Table 20.

Summary of key themes

The findings from the carer interviews are presented thematically below. The analysis process revealed key themes and subthemes that held meaning and importance for carers. The initial approach for the study, timing and first impressions were found to be instrumental in influencing decision-making regarding taking part. Carers also expressed their views and experiences of the training provided and of completing the carer diaries. Many of them found that the diaries were a useful tool for reflecting on their experiences and, for those in the usual-care arm, they added a sense of participation. The use of medications and equipment was also discussed during interviews, in terms of both the practical and the emotional aspects of delivery.

Further emotional themes emerged from the analysis, with carers showing concerns regarding the management of symptoms, particularly pain, and their ability to recognise when a patient was in need of medication. This was also bound with concerns about waiting times and delays, as no one wanted prolonged periods of pain or suffering.

The potential of the intervention to empower patients and carers was also of salience in the interviews, as was the deep desire for carers to fulfil the wishes of their loved ones and to keep them at home. However, this also came with the notion of potential carer burden and the possibility of the needs of the patient being prioritised over their own needs as the carer.

First impressions and recruitment

Carers were asked to discuss their first thoughts when approached for recruitment to the trial. Some carers reported initial hesitancy; however, this diminished with further explanation of how the intervention would work. Several carers described feelings of surprise, as they had never before considered the possibility that carers could administer these kinds of medications. There was also relief at the realisation that there was something more that they could do to help keep loved ones at home.

Randomisation

Carers’ accounts indicated a less than complete understanding of the randomisation process. Carers talked of randomisation in terms of being ‘selected’, ‘chosen’ or ‘accepted’ for the trial itself, rather than having been allocated to either group as a result of the randomisation process.

Carers who were allocated to the intervention group reported being pleased with their allocation, although there was misunderstanding of the randomisation process. Carers expressed relief at having been ‘chosen’, as if this had been done on purpose. Some questioned the randomisation and were doubtful of how the trial could be random at all, confusing the different common usages of the word ‘random’. One carer described that, although it had been explained to him that this was a randomised trial, he felt that he knew that he had already been chosen because it would not be possible to do something of this nature at random:

And I’d said from the start, ‘Well, I think it’s random but I bet you’ll choose me’. And then straightaway. So I always felt as though everybody said it was random, but when I were asked, you’d already picked me to do it anyway, sort of thing.

Tom, urban, intervention

Carers who were allocated to the usual-care group expressed disappointment with their allocation and reported feeling that they had been offered an opportunity to do something beneficial that had then been withdrawn. One carer in the usual-care group described her initial disappointment and that the patient had been ‘angry’ about the allocation; however, the carer later felt relieved that she could not administer the drugs because she had the sense that she was being manipulated by the patient. Carers in the intervention group who were asked how they would have felt had they been allocated to the usual-care group explained that they would have been very disappointed at having had their hopes raised by the prospect of the intervention and then being let down, as they would not be in the administering group. The ways in which carers and patients view randomisation is important, as allocation to the usual-care group has the potential to cause distress if participants believe that they have somehow failed to be accepted to the intervention group for reasons other than it being a random process:

The way it was presented to us, it was more that this was something that we could do, rather than something that we might not be able to do. And I think it was quite hard on the district nurse who had to tell to tell [the patient], as I say, she was still an angry person and that made her angry.

Kathy, urban, usual care

Trial materials and training

Medications and equipment

Symptom management

All of the carers interviewed wanted to help their loved one avoid uncontrolled symptoms, particularly pain, in the last weeks and days of their life. A key factor in deciding to participate was the fear of uncontrolled symptoms and the idea that, with the intervention, they would be able to alleviate that. Some patients had already experienced periods of uncontrolled breakthrough symptoms prior to recruitment. This may also have been an influencing factor for study participants, who were hoping to avoid those experiences happening again:

That you weren’t scared of the uncontrollable pain or severe sickness or something that might’ve been there. That-that gave us comfort – comfort – and confidence, I would say, to keep my wife at home, rather than, you know feeling that you had to take her to a hospice to get her symptoms relieved.

George, rural, intervention

One carer in the intervention group reported being upset during times when symptoms were uncontrolled, as he felt that HCPs were blaming him for being unable to manage the symptoms with oral medication. Carers in the usual-care group expressed disappointment, as they had wanted to help patients by managing their symptoms and they then felt that they would not be able to. A usual-care group participant, Paula, had taken part because she wanted to ease uncontrolled pain and was relieved that in the end her loved one did not experience uncontrolled symptoms.

The desire to be able to manage and control symptoms as they occurred was a major contributor to wanting to take part in the trial:

So, yeah, it was just pain relief, you know, in my mind if anything I was doing was gonna relieve it then it was worth doing so I didn’t think of anything else than it was gonna alleviate any pain or suffering for my mum.

Tom, urban, intervention

Concern about delays and rurality

Closely linked to the fear of uncontrolled symptoms is the fear of long waiting times for HCPs to arrive at the home to relieve symptoms. Long periods of waiting for help to arrive, particularly at night, were a source of concern for all carers in the study, particularly those who lived in rural locations. There was a common feeling throughout the accounts that waiting times seem even longer at night and that these long waits during sleepless nights contributed to feelings of powerlessness and isolation. Waiting times were also one of the reasons that carers gave for being disappointed about being allocated to the usual-care group, as they felt that now they were likely to experience delays when others would not:

As I said before, it’s just positive. I was not reliant on waiting for the team to come out to give injections to relieve any symptoms she might have with the vomiting and the anxiety and the pain relief. So that was all positive, not having to wait. It was . . . I could give it as soon as it was required.

David, urban, intervention

Symptom recognition

The ability to recognise breakthrough symptoms, what signs to look out for and when to administer the drugs were of concern to some carers. As the management of symptoms depends on recognising the symptoms, carers worried about leaving loved ones in pain without realising. Recognising the difference between symptoms, such as pain and agitation, was reported as sometimes difficult, particularly if the patient was no longer able to communicate. Some carers in the intervention group described consulting with other relatives about whether or not to give medications. One carer, Josephine, reported having many conversations with her son about when to give medication; although she had a medical background and he did not, she still sought his reassurance. Carers seemed to gain confidence in recognising breakthrough symptoms once they had administered their first dose or once they had been reassured by either family members or HCPs that they were right in their observations:

I was glad that my son was here as well. He-he’s not medically trained at all but (. . .) I-I asked him, ‘Oh, do you think Dad should have something now?’ And-and I valued his input, really, because he said ‘Yes, definitely’. You know, ‘Yes, yes, yes. He’s getting a bit restless now’. Or, ‘His breathing is not so good, you know, and if we can relieve that’.

Josephine, urban, intervention

Administration error, last injections and hastening death

The issue of making mistakes with medication was raised throughout the carer interviews. All of the carers expressed some concern about the potential for errors in administration. Concerns about potentially causing pain were often alleviated by training and the use of no-needle injections. However, the biggest concern for carers was the possibility of causing or hastening death. Although carers knew that their loved ones were in the last stages of their life, there was a desire not to contribute to that in any way and there was a fear among participants of what could happen were they to give an injection and the patient die very soon after: they did not want to take what little life their loved ones had away. There was also a concern among carers that they could be held accountable or be blamed if the patient passed away after an injection that they had administered:

I was concerned if I was to administer morphine and he, say, passed away 10 minutes later, could I be held accountable um that I’d done something slightly wrong?

Claire, rural, intervention

Most carers worried about accidentally hastening the death of the patient; however, two carers in the study raised the concern that patients may ask to be given large doses to end their lives. One carer explained that she remembered a relative in the past who had requested a hospice doctor to assist him with dying; she wondered whether the intervention would increase the chances that patients would make requests like this of their carers. One carer had worried about this happening to her in the trial and was glad to have been allocated to the usual-care group, as she believed that otherwise the patient might have asked her to assist her death:

I think you’re-you’re in a position where you can quite easily be manipulated.

Kathy, urban, usual care

A home death and wish fulfilment

The desire of their loved one to have a home death was something that all carers reported. Carers presented the importance of a home death in a variety of ways by carers: the home was a more comfortable environment, a place where family and pets could be present at all times and a place over which the patient was able to have more control:

Thought of her being in her own bed, surrounded by her own pictures and books and music and, you know, the views. We live in a lovely place and the views and the noises and familiarity of home gave us comfort, really.

George, rural, intervention

Taking part in the trial, if assigned to the intervention group, was seen as an extra way of ensuring that a home death could happen. This was a driving force for many carers, who were seeking above all to fulfil the wishes of their loved one:

And they were all saying that [husband’s] wish was the most important . . . to remain in his home, they could see it was so important. And it’s that reason, you know, obviously I did it [took part in the trial]. Um the hardest thing I’ve ever done. Yeah.

Claire, rural, intervention

It may be that some carers put the wishes of their loved one to stay at home above their own concerns regarding administering medication and being the main carer. One intervention carer whose husband had been admitted to the hospice stated that she felt a great sense of guilt and regret that she had not been able to keep him at home, even though the circumstances of his admission had been out of her control.

Carer and patient empowerment

Carers in the intervention group found a sense of purpose in being able to manage and administer medications rather than have a more traditional, passive role. This led to increased feelings of empowerment and control. Some carers in the intervention group also reported a sense of personal peace in knowing that they were able to do all they could for their loved one during the last days of their life. One carer described how he missed his wife greatly, but was content in the knowledge that he had done his best for her. The intervention was also reported by carers to have been empowering for patients, in that it gave them more control over their own destiny rather than having decisions made for them:

So, those were two really important things for her and CARiAD was part of that, I think. ‘These are things, I can make decisions about. I can’t not have cancer. I can’t not die. But I can make decisions about how that happens’. And that, I think, was really important to her.

Laura, urban, intervention

Potential for carer burden

The potential for increased carer burden arose in several ways, which potentially added to the sense of responsibility for the intervention group carers, particularly in giving medications as part of the intervention. Some carers also expressed that they felt that this increased the time that they needed to spend at home with the patient, especially as only one carer could be trained for the study. It is possible that carers may believe that once they have been trained to give the medications they should try to call on the HCPs less often. Several carers in the intervention group reported feeling reassured by HCPs that they could still call for help whenever they needed to.

The issue of there being only one trained carer was a concern for some, especially those who were used to caring as a family. They worried about who should be the nominated carer and what would happen if they were not always around. Carers in the intervention group who were used to being the sole carer, such as David, were less concerned about being alone, as they were used to being in that situation and preferred it to having control taken away from them by relying on HCPs:

They wanted to put a health-care plan in with people coming in but I’ve always done the cleaning and the cooking and the mending and I just didn’t want people around me when I wanted to cope for myself. I didn’t want things to be taken away from me. To be given this help right at the end, when I did need it, and to be able to do it myself, it was great.

David, urban, intervention

As mentioned above, the possibility of the desire to fulfil a patient’s wish to die at home to be prioritised over the wishes of their carer was alluded to. This may increase burden on carers if they feel that they are obliged to agree to administer medication or to support a home death that they may not feel comfortable with. Carer burden may also increase if the patient wishes to resist, or avoid, contact with HCPs. Several carers in the interviews described their loved ones as being somewhat resistant to medication, particularly strong opioids, or to the more frequent visits from HCPs at the end of life. It may be that some patients are not ready to accept that they are at the end of life, simply prefer to remain in control of their own medications or do not like lots of people visiting. However, although being in the intervention group may reduce the number of visits by HCPs to deal with breakthrough symptoms, it does not mean that visits will stop altogether, although this possibility may be what attracts some patients to the intervention. One carer, Kathy, was glad that she had been allocated to the usual-care group because she realised how resistant the patient was to receiving help from professionals, and how much this would have increased her burden had she and the patient been allocated to the intervention group:

She was very reluctant to accept any external help at all from the district nurses or our more local hospice at home. I think if I had been in the position where I could give her medication, she would have wanted to shut the district nurses out, which would have been very difficult.

Kathy, urban, usual care

Support needed from health-care professionals

The amount and the nature of support from HCPs in both groups of the trial varied according to the carer, the needs of the patients and the home situation. The majority of HCP support, by home visits or telephone, was reported to have been provided by DNs and SPC nurses. Some carers, such as David and George, also said that their GP had been particularly supportive and had made house calls OOH to make sure that things were OK. The support from HCPs was praised by many of the carers, who explained that they felt reassured that they could contact their health-care teams whenever they needed support. Although carers acknowledged that sometimes health-care teams were overstretched and delays occurred, they believed that the teams were always doing everything they could. Carers reported being able to ask any questions that they wanted and that they felt well supported throughout the trial and the palliative care process:

That was why it was so nice for the nurses to say, ‘If you-you get a bit wobbly or you’re a bit worried, ring’. And I thought, ‘Yes, if it comes to it, I shall ring them and they shall talk me through it’. And that was like, ‘Oh, thank you! Sigh of relief’.

Myra, rural, intervention

Family dynamics and health-care professionals as carers

The home circumstances of each dyad in the trial were different. The carer was often the patient’s partner or child. However, that was not always the case, as there was also a neighbour, niece and sister-in-law. Carers reported different circumstances that had led to them being the nominated carer in the trial, and these had caused different dynamics within the home. Ffion, for example, was the niece of the patient and it seemed that her uncle, Phil, the patient’s husband, had been very involved with caring for his wife but the DNs had chosen Ffion to be the carer. During the interview, Ffion and her uncle (Phil) seemed to feel that Phil would have been just as good a choice, but because Ffion had a nursing background she had been selected:

You was in charge of all the medicines. You were helping –

I knew what was what and I did that all the time.

Exactly. I don’t think being a nurse, in that sense, helped. You don’t –

You don’t have to be a nurse.

A carer, Kathy, who was the patient’s neighbour, said that she had been deliberately chosen by the patient because she had a nursing background, even though the patient lived with her husband and had grown-up children. It may be that these carers were chosen because they had a health-care background over somebody else who may have been in a better position. Like Kathy, another carer, Laura, had been approached not by the nurses but by the patient. Laura lived in a large household with her husband and children, as well as her brother and his family. The patient was the wife of Laura’s brother and she and Laura had a close relationship. The patient expressed that she wanted Laura to be the carer who took on the intervention so that her husband (Laura’s brother) would be able to concentrate on providing emotional support and keeping their relationship as ‘normal’ as possible for them and their children.

Paula, another carer and a nurse, was caring for her mother. She described the situation among her brothers and sisters as somewhat tense, as her siblings had not all accepted that their mother was dying. Paula had been expected to move back to her hometown to care for their mother, as she was a palliative care nurse herself. Having a HCP background, however, did not mean that as carers they needed less support from their family and friends. Josephine, as previously mentioned, described consulting her son each time she gave medication, despite having been a palliative care professional in the past. It would seem that the difference between HCPs and lay carers takes on a different meaning when the caring is for a loved one.

A good death

Some of the carers in the study, particularly those with HCP backgrounds, had previous experience of caring for, or being with, people who were in the last days of their life. Previous experiences seemed to have influenced the desire to take part in the trial, especially if during those previous times they had experienced long delays in medication administration or uncontrolled symptoms. Throughout the interviews, carers explained that they did not want their loved one to be in pain or to suffer, as they had seen that happen to others in the past. Taking part in the trial seemed to be a way of trying to help and trying to ensure that a loved one was able to have a ‘good death’:

Well, I had seen people die before and I wanted him to have the best kind of death. I really did.

Josephine, urban, intervention

Implications for feasibility

The qualitative interviews revealed that the intervention was acceptable to carers. Implications for feasibility also arose from the carer interviews, and their accounts show several areas in which changes could be made to improve feasibility for future work.

Many carers in the study had previous health-care training themselves, which may have influenced their decision-making in taking part; however, as evidenced above, this did not mean that they were at an advantage during the study owing to the emotional difference between patients and loved ones. It may improve feasibility if this is taken in to account during recruitment.

One key area that has implications for feasibility is the ways in which the randomisation process is explained to carers. Many carers showed a lack of understanding of how randomisation works, which indicated that a better explanation is needed. This had implications for recruitment and retention to the study; therefore, if randomisation could be explained in a way that would be better understood by participants or if a different strategy could be used, feasibility may be improved. Carers may be less likely to withdraw on allocation to the usual-care group and may engage more with the study if in the usual-care group, which would, therefore, reduce non-completion rates.

Although carer diaries were generally acceptable to carers, the QOLLTI-F was reported to be hard to use, which led to lack of completion. Feasibility may be improved with more training regarding QOLLTI-F or use of an alternative measure.

Purposive sampling

A strategy was developed to recruit a purposive sample of HCPs for the qualitative interviews. The purposive sampling criteria were location, rurality, professional role and years in practice. Potential HCPs to interview were identified when their patients (and carers) were recruited to the trial (both groups). They were then added to sample monitoring tables that were used to monitor HCP characteristics and create a purposive sample for the HCP interviews. Using this system, HCPs were purposively identified to provide a range in terms of the above purposive sampling criteria. The identified HCPs were then invited to participate. These systems were in place and were followed; however, owing to lower rates of study recruitment than expected, there were less HCPs taking an active role in the trial than first anticipated. The sample was monitored throughout, but the strategy needed to be applied with more flexibility because of the availability of HCP participants. The final sample (see Table 21) provided a range of perspectives comprising a range of types of nurses and three GPs.

Health-care professionals were first recruited via e-mail and then by telephone. Interviews were conducted by telephone and lasted between 15 and 60 minutes. All of the interviews followed a topic guide (see Appendix 6), were audio-recorded, fully transcribed and analysed using the Framework approach. 77

Analysis

The Framework approach comprises a systematic, five-stage matrix method; this study followed the five stages. 77–81 During the first stages of data familiarisation, the researcher was immersed in the data by re-reading all of the transcripts thoroughly; this was then followed by the construction of a coding structure that was based on key themes emerging from the data. 81 This was used to code all transcripts thematically and charts were created to group and synthesise data according to the coding structure. 81 The final stage was an interpretive stage in which the researcher identified the patterns and links between the themes. 81 The research team also held several group interpretation workshops to discuss the key themes.

Sample

The characteristics of the 20 HCPs who were interviewed as part of the CARiAD trial across the three recruitment sites are shown in Table 21.

Summary of key themes

The findings from the 20 HCP interviews are presented thematically below. The HCP interviews revealed a set of key issues that were of importance to those supporting patients and carers throughout the trial. Initial participant identification was discussed by many HCPs who had concerns about screening, RA and choosing the ‘right people’ for approach. A sense of the need to approach the ‘right’ people may have influenced the selection of many carers, for example those who already had some HCP background themselves. HCPs as lay carers became an important topic in the interviews.

As well as considerations about who should be approached, HCPs described concerns about the timing of the approach. Prognostication was a key concern for HCPs who expressed the difficulties in pinpointing the last weeks of life.

Given that randomisation was an issue for carers it was also so for the HCPs, who discussed the disappointment that they saw in those randomised to the usual-care group. HCPs also discussed the training involved with the study, including their own training and the subsequent training of carers to administer medications. Training was often discussed in terms of the amount of time it could take and the needs and feelings of the carers.

Trial materials and carer diaries were also discussed and, although these were generally found to be useful, HCPs were able to give suggestions for delivery and improvements. In practical terms, use and storage of medications and equipment were discussed along with the potential for medication misuse; interestingly, there was less concern in this last area than perhaps expected, as the medications would be in the home anyway and RAs would always be conducted.

Health-care professionals often noted the symbolic difference between a no-needle injection and an injection that breaks the skin, as a no-needle injection may make the delivery of medications less daunting and ease worries about causing pain or injury. Among those HCPs supporting a dyad in the intervention group, there was discussion of family dynamics and household composition, especially when shared caring was happening. A key theme was also the potential for carer burden and distress, particularly in reference to ‘last injections’ and concerns about hastening death.

Finally, it was important to discuss the impact of the intervention on the time and workload of HCPs. Although views differed on this, the most salient point was that HCPs saw things in terms of patient benefit above all.

Participant identification

Recruitment

Randomisation

Following recruitment and consent, the dyads were randomised to receive either the intervention or usual care. During the interviews, HCPs were asked for their views on the randomisation process and the effects on patients and carers. The RNs were happier with the randomisation process following their unblinding, as they had found that being blinded to allocation was somewhat problematic. They explained randomisation to dyads carefully, ensuring that it was understood that there was no guarantee of being allocated to the intervention group. One BCUHB GP thought that patients enter trials hoping to gain something beneficial; therefore, if the patients are assigned to usual care they feel that they have missed out on an opportunity. HCPs described the disappointment felt by those allocated to usual care, including being ‘upset’ or ‘angry’ at not being ‘chosen’. The use of words such as ‘chosen’ and ‘selected’ by carers may also indicate that they did not understand randomisation despite the HCPs’ explanations, but may also indicate that HCPs were not always clear about how the process works:

The participants were bitterly disappointed that, having reviewed the information and looking forward to do the care, they weren’t actually randomised to receive it, so that was an issue in itself.

DN

It also appeared that the HCPs themselves were disappointed if they felt that they had identified someone who would benefit from the intervention and that person was allocated to usual care. They expressed sadness that these dyads would not be able to administer the medications. One SPC nurse described the disappointment of a family allocated to usual care as ‘heartbreaking’. Some participants also reported being aware of dyads who had withdrawn from the study following their allocation to usual care, and in one case a dyad had declined to take part at all on hearing that the study would involve randomisation:

The only person that we had that was the male carer, he was keen to do it because he actually was a health-care professional. Different sort of background, but he was in health. And because we couldn’t assure him that he’d be on the intervention arm, he opted out.

DN

Training

Trial materials

Medications and equipment

Supporting a dyad in the intervention group

Impact on health-care professional time and workload

The impact of the intervention on the workload and time of HCPs was discussed during the interviews. The main concern of the HCPs was timely symptom control and the benefit to the patient and carer, not the possibility of reducing their own workload. In fact, the HCPs did not report that the intervention saved them any time, as they would still visit the home just as often as before, and some of them needed to invest more time in training the carers. Participants reported that the pressures facing the NHS alongside the issues brought about by limited OOH staff and rural patients could mean that patients would face long waiting times during breakthrough symptoms and the key impact of the intervention was that they were reassured that the carer would be able to relieve those symptoms until the HCP arrived:

On one hand I’m saying it is feasible because it is feasible, but it’s a challenge because you do need some time to explain the intervention to the participants and select suitable people and then teaching.

DN

Introduction

Discrete choice experiments are a survey method to measure patients’ stated preferences for goods and services (e.g. medicines and health-care provision). 82 Respondents choose between hypothetical but realistic alternatives, which are described in terms of a number of attributes (e.g. waiting time) that are each characterised by specific levels (e.g. 1 hour). This allows for the estimation of the relative importance of each attribute, assessment of any trade-offs between attributes (e.g. waiting time vs. symptom severity) and estimation of respondents’ total satisfaction (utility) with the service. 82

The preferences of carers towards administering SC medications will have a bearing on their willingness to adopt this practice and the effectiveness of carer-administered medication. Evidence on the most important attributes of the intervention may be used to estimate the optimal configuration for service delivery and to explain why, if implemented, the option to administer SC medications is not adopted by all. Although the DCE, aiming to ascertain carers’ preferences for their administration of SC medications, will be conducted as part of a future main study, the formative work required to identify relevant attributes and levels was carried out as part of the embedded qualitative study component of the randomised pilot study.

Objectives of the formative study are to (1) identify and rank factors that are important to carers in guiding their choice between HCP and own administration of SC medications and (2) assess suitability for inclusion in a DCE.

Methods

The process of attribute selection was informed by best practice. 83 A starting list of potential factors was derived using the draft CARiAD study documents/guide for carers and previously published resources for caregivers from an Australian study. 14 This was supplemented with attributes that are commonly used (and found to be significant) in published DCEs of medicinal products. 82,84 The reduction process focused on identifying plausible, salient attributes for a future DCE. Factors from the original starting list that did not qualify as ‘attributes’ for the ranking exercises were discussed in the context of other aspects of the DCE, such as inclusion in the DCE scenario decision question (e.g. appropriate description of breakthrough symptoms) or as candidates for subgroup analysis (e.g. confidence to recognise breakthrough symptoms).

Owing to cognitive burden and the limited time for the ranking exercise (approximately 15 minutes), the study team reduced the starting list of 23 factors to 10 attributes; this was achieved using good practice criteria for what makes a plausible attribute83 and by combining attributes that represented parallel constructs (overlap). A convenience sample of nine university staff and two study team members completed the ranking exercise using the reduced list. Respondents completed the exercise and were asked to give comments/feedback. In response, the list was further reduced to nine attributes, with revised wording, to be used in the embedded qualitative study (Table 22).

A topic guide was developed around the final list of attributes (see Appendix 7). In the interviews, carers were presented with nine attributes that were printed onto show cards that contained the attribute heading and a brief descriptor. Interviewees had the opportunity to nominate additional attributes before selecting those that they considered important in the decision to choose own versus HCP (DN) administration of medications to people dying in their own home. They were then asked to place the selected cards in rank order, from most to least important. An in-depth interview preceded this exercise (see Chapter 5). The ranking was recorded and the corresponding dialogue was transcribed verbatim and analysed independently of the preceding in-depth interview. Interviews were conducted face to face in the respondent’s own home or by telephone (with modified topic guide). Ranking data were entered into Excel^® (Microsoft Corporation, Redmond, WA, USA) and analysed descriptively using counts and mean rank scores for each item. Rankings were converted into rank scores using:

where rank 1 = most important. The mean rank score was calculated as the sum of rank scores divided by the number of times that the attribute was selected. The second part of the topic guide was intended to pilot the presentation of the highest ranked attributes and was to be used for respondents 6–10 in the intervention or usual-care group.

Results

In total, eight out of 12 interview participants completed the attribute ranking exercise: seven from the intervention group and four over the telephone. The ranking exercise was the final task of the interview, which on average lasted 51 minutes. All of the interviews used part one of the topic guide, cognitive interviews on attribute presentation were not feasible due to sample size and there was a preference for telephone interview.

The average number of attributes selected for ranking was six (range 3–10). Symptom severity was the most commonly selected attribute. Waiting time for a HCP had the highest rank score (Table 23). Only one respondent selected cost to the NHS and they ranked it the least important. Three additional attributes were suggested, each only once: timing of introducing the intervention, privacy and ownership of decision-making about death and back-up/being able to call someone for support.

Discussion

All nine attributes were selected at least once and appeared valid for inclusion in a DCE. The potential for overlap was observed; however, this could potentially be reduced by further clarification within attribute labels and descriptions that could be obtained from training resources. Respondents were able to select what was important to them and when ‘thinking aloud’ demonstrated a tendency to trade between attributes, which would imply that the DCE would be acceptable. To reduce cognitive burden it is unlikely that all nine attributes would feature in the DCE. Instead it is likely that between four and six would be piloted (Figure 10).

FIGURE 10.

Example of a DCE choice question.

The key strengths of our study are the (1) systematic approach to attribute development and refinement, (2) mixed-methods approach to attribute selection, that enabled greater interpretation of the relationship between attributes and the decision-making process, and (3) conduct of formative work using good practice guidelines. The number of respondents and the proportion of telephone interviews limited the study.

Conclusion

By using a systematic approach to attribute selection, including literature review, expert opinion and ranking exercises, we demonstrated that waiting time for a HCP, symptom severity, frequency of additional HCP visits, frequency of symptoms and potential for administration error are important attributes of the decision of carer administration of medication. These attributes can be constructed into a DCE to quantitatively measure the strength and direction of preferences for each attribute, the willingness to trade between attributes and the total utility and probability of uptake for different service configurations.

Introduction

The CES is an index measure of the caring experience that focuses on six domains: (1) activities outside caring, (2) support from family and friends, (3) assistance from the government and other organisations, (4) fulfilment from caring, (5) control over caring and (6) getting on with the care recipient. The construct validity of the CES instrument has been demonstrated in a heterogeneous group of 730 carers in the UK,85 and specifically in the context of palliative care. 86 The CES benefits from having preference-based index values, based on 162 unpaid carers of older people from five geographical locations in the UK,87 that allow for calculation of utility for use in economic evaluations that focus on care-related (rather than health-related) quality of life. The CES was piloted to inform the choice of primary health economic outcome for a cost-effectiveness analysis alongside a definitive trial.

Methods

The CES was completed by the carer at baseline and post bereavement. To contextualise the questionnaire post bereavement, the carer was asked to complete the measure reflecting back to ‘during the last few weeks of life’. Completion rates were analysed as counts of expected responses. Missing data were summarised as a percentage of the partially completed measure. A decision threshold of what was considered feasible was not set a priori. A score for the CES may be calculated only if all questions are completed, given that there is no rule for calculating in the presence of missing values. Responses to each of the six CES domains were coded (from 1 for the top level to 3 for the bottom level) to produce a caring state (i.e. 21213) and were then assigned a utility value, which ranged from 0 (for CES state 333333) to 1 (for CES state 111111). The utility values applied represent the mean preferences of a sample of carers aged ≥ 65 years in England, and were derived using best–worse scaling. 87

Results

Table 24 summarises the CES domains by completion at baseline and post bereavement, and the associated utility scores. In total, 21 of the 22 carers who completed follow-up completed the CES. The majority of carers completed all of the CES (21/22) and one carer completed 50% of the measure.

At follow-up, the usual-care group had a mean score of 70.94, which was a decrease of 6.93 from baseline, and the intervention group a mean score of 72.51, which was an increase of 0.66 from baseline. A disutility (negative difference) indicates a decline in carers’ care-related quality of life during the period from baseline to bereavement, and a utility gain (positive difference) indicates an increase in carers’ care-related quality of life; however, it is difficult to make inferences about change from baseline to follow-up because of the small number of participants.

Conclusion

The CES offers a preference-based approach to incorporate the effects on carers in the economic evaluation, which focuses on care-related quality of life (rather than health-related quality of life). In this pilot study, the CES was completed by > 90% of the sample, which would indicate that this was a feasible measure of carer utility that could be used in a future economic evaluation alongside a clinical trial. A future economic analysis would need to consider baseline adjustment and the implications of completing the measure in retrospect at post-bereavement follow-up.

List of positive or acceptable findings

• The recruitment rate was above what was stated to be necessary for a definitive trial and was close to the 50% rate that was assumed.

• Retention was above the 50% that was stated to be necessary for a definitive trial.

• Outcome measures under contention for a main trial did assess main/general areas of interest: overall symptom control and carer self-efficacy/empowerment. When completed, measures were completed well.

• Acceptability of the intervention was overwhelmingly positive, adherence to the intervention was generally good and the intervention was safe in the study population.

• There was no evidence of contamination in the usual-care group.

In addition:

• Overall, eligibility based on clinical estimate of survival worked well.

• Recruitment appeared to be strongest where the DN team had close links with their local SPC team.

• Conversion to consent was close to the 50% rate that was assumed.

• Randomisation yielded equality in groups.

• The training materials were well received, including by the REC.

• Secondary outcome measures of carer confidence in administering medication and time to symptom control for patients were appropriate, and fully aligned with importance assigned by carers.

• Close-out procedures were planned and executed well.

• The CES provides a feasible preference-based approach to estimating carer utility, for use in an economic evaluation alongside clinical trial.

• Formative work on DCE attribute identification suggests that it would be feasible and remain very pertinent to quantitatively elicit carers’ preferences as part of the overall research question.

List of ‘problems’ to scrutinise using ADePT

• Not reaching the target recruitment.

• Differential retention between groups in the trial.

• Not able to do a sample size calculation for a main trial.

• Unclear which outcome measure(s) is/are best to use for clinical effectiveness.

• Suboptimal HCP adherence to the intervention.

In addition:

• Sites opened for recruitment later than planned.

• One site recruited significantly fewer participants.

• Survival time was short for many potential participants as well as for many participants.

• Although processes of randomisation were successful, acceptance of the randomisation results proved difficult.

• A high proportion of carers in the dyads had a HCP background.

• No consultee declaration forms were received in relation to patients who gave initial consent.

Not reaching target recruitment (with one site struggling to significantly to recruit)

These issues are likely to be a trial-specific problem, as recruitment numbers would not be relevant in a real-world context.

The reasons for these observations are mainly the lack of DN research experience and time to commit to the study, which were very marked in the site that struggled to recruit. In addition, we observed a high proportion of carers with a HCP background. This could have been because HCPs appeared to apply the RA only when they felt sure that the carer in a dyad would be able to engage in the intervention. It is, therefore, likely that DN teams did not identify or approach all potentially eligible participants, as suggested by the qualitative findings.

Despite similar levels of input from the central CARiAD research team, CVUHB started recruitment late and recruitment was poor. In addition to pressures on DN time and DN research naivety, the DN teams in CVUHB did not have the same close working relationship with SPC teams as was observed in the well-recruiting DN teams in BCUHB and GCS. Additional factors may have been the different recruitment systems in place in that site and a preference for doing the trial training in a different manner. This site was particularly clear that the partial blinding of RNs created barriers to study progress; however, recruitment did not increase after RNs were unblinded, mainly owing to a long-term sickness absence of the RN that coincided with unblinding. It is not completely clear why there were such marked differences in site recruitment, so it is possible that other issues affected this that are not addressed by the suggested solutions.

Potential solutions include changing aspects of the following:

• Trial design – rebalance research tasks away from DNs and towards RNs or study-specific SPC nurses (including the identification and recruitment of participants). In addition, teach DNs and SPC clinicians to apply RA to all eligible dyads and not only to those they already feel are competent, and emphasise to carers and HCPs alike that many people are fully capable of engaging in the intervention and not only those with a HCP background.

• Context – a specific, detailed review of capability and capacity at the site selection stage. Introducing a study-specific SPC nurse to identify and recruit participants could lessen the burden on DN teams and improve engagement.

Differential retention between groups

These issues are likely to be a trial-only problem, as trial retention is not relevant in a real-world context.

Only six out of the 20 (30%) dyads in the usual-care group completed the trial. Reasons for non-completion (n = 14) included carers not being contactable post bereavement (n = 6), withdrawals (n = 4: three because they were not randomised to the intervention and one initiated by the trial team as part of close-out arrangements), lost to follow-up on PI advice (n = 3) and one carer declining follow-up.

Differential retention (lower in the usual-care group) is a common observation in randomised trials. In the CARiAD study we observed that carers assigned to the usual-care group were less engaged; this is probably linked to the fact that, to be eligible for the intervention group, the carers had to be willing to be trained to give SC medication. They therefore may not have been in equipoise regarding the intervention and may have felt resentful at not having the opportunity to be trained in the intervention.

Potential solutions include changing aspects of the following:

• Intervention – if there is more evidence that the intervention is a ‘common good’, a future trial might consider randomising to different intensities of the intervention. Another possibility is to test two different interventions at once, both of which need testing in the end-of-life situation (A vs. B design).

• Trial design – increased RN attention and follow-up in the usual-care group might improve retention; this role could also be undertaken by a study-specific SPC nurse (see Potential solution: employ a study-specific specialist palliative care nurse). A cluster-randomised design could be considered; however, this might simply move attrition to the level of the cluster.

• Context – consider recruiting only in areas with explicit SPC nurse support.

Acceptance of randomisation proved difficult

These issues are likely to be a trial-only problem, as randomisation would not be relevant in a real-world context.

Processes for randomisation worked successfully. However, although participants were willing to be randomised, the randomisation results were not universally accepted.

Participants were required to wish to undertake the intervention as part of eligibility for the trial and, although randomisation would have been explained as part of the trial process, participants may not have been in true equipoise about the treatments on offer. Resentful demoralisation may form part of the explanation for disengagement from the usual-care group. In interviews, HCPs said that they believed that dyads allocated to the usual-care group were disappointed, and gave this as a reason for withdrawal and loss to follow-up.

Potential solutions include changing aspects of the following:

• Trial design – further explanation of the need for randomisation at the point of consent, use of a cluster-randomised design, possible use of Zelen’s randomisation90 and consideration of changing the level of system measurement.

• Context – assess the policy context to see whether or not this practice is a common good and generate background policy positions and documentation to support more widespread use of the practice prior to further testing.

Unclear which outcome measure(s) is/are best to use for clinical effectiveness

This issue is likely to be a problem in both a trial and the real world. The latter is because, even if the intervention is implemented, any future service evaluation will face this question.

The outcome measures under contention for a future definitive trial were not deemed appropriate for assessing the effect of the intervention. Although the completion rate of MSAS-GDI was acceptable, assessment of sensitivity to change was not possible with any reliability owing to the small number of data available at the follow-up time point. The completion rate of the QOLLTI-F was lower than anticipated, which may be a result of when this was supposed to be completed. Completion at baseline (when the measure was introduced by a RN) was good, but it may simply be too burdensome to ask carers who are looking after somebody in the last days of their life to complete this, or any other outcome measure, independently and at regular intervals. Feedback from carers was that the QOLLTI-F did not capture the nuances of their situation accurately; although they may not have had the time to do the things that it asked about, they did not necessarily feel burdened by this as they had known to expect this and it was acceptable to them during this time.

At the time of designing the CARiAD trial, there was no suite of COMET measures to apply to studies of the last days of life. In January 2017, development of an international core outcome set of the best care for the dying person commenced. 91 The outcome of this work is still awaited, and may affect any future studies during this care episode.

Potential solutions to the issue of unclear outcome measures include changing aspects of the following:

• Trial design – identify alternative outcome measures for clinical effectiveness, use ‘time to symptom control’ as a patient-level primary outcome measure. Alternatively, consider the effect of the intervention being measured at an alternative level, such as process- or system-driven outcomes. As this intervention is unlikely to replace any service provision within palliative care teams but become an adjunct to usual care, measuring the effect at this level may be more appropriate and enable those with a desire to support their loved ones in this way to do so.

Strengths

The study illuminated numerous aspects of research related to last-days-of-life research: it illustrated both dying people’s and carers’ willingness to be randomised, strong and successful public contributor involvement at all stages and the value of rich qualitative findings in explaining the quantitative results. It used a structured approach to assess the outcome of such feasibility work.

In addition, it generated a new UK training package and method for lay carer role extension that can be used in any further research or implementation, and clarified limitations of community-based nursing care in three areas in the UK. It also illuminated a number of further research questions, listed above, in recommendations for future research.

The CARiAD study has helped to move research methods for last-days-of-life research forward by providing much-needed evidence of conversion rates from approach to recruitment, which will help future researchers in estimating these rates for their studies. It also gives an example of how such trials can be introduced sensitively to patients and their families, as well as what works in terms of data collection time points and methods. The REC chairperson commended the study’s consideration of these matters and how these were reflected in the trial materials and processes. In particular, she commented on our careful consideration of the consultee process. The results from the trial have further illuminated the realities of this process in the context of last-days-of-life research, as the current legislation does not adequately account for how mental capacity can be assessed in this situation, who it should be assessed by and how consultee assent can be obtained in a way that is sensitive to the difficult emotional situation of patients and their families.

The CARiAD study, like the AMBER study,93 adds weight to the idea that vulnerable patient groups at the end of their life are ready and willing to engage in research of this type. The study showed the tension between the anxieties that carers may have had about taking an extended role and the relief that they felt when they were able to relieve the symptoms of a dying family member.

Limitations

The limitations of this feasibility trial have been fully explored in previous sections (see Chapter 8) and only the main points are summarised here. The key limitations can be summarised as follows:

• A series of key progression criteria were not met.

• The differential recruitment to the control group was of particular concern.

• Overall, the limitations led to an equivocal decision about progression to a full Phase III RCT.

The lack of a policy framework to underpin practice in the UK may have impeded the study in a number of ways; however, if such a policy framework existed, a randomised study would not have been appropriate.