Study design

This was an exploratory study, which was carried out within a constructivist qualitative tradition of research. Qualitative research is concerned with eliciting how participants understand and experience the world through the cultural filters internalised through developmental processes of socialisation and through interaction with others in real-world situations and contexts.¹⁰⁷ Constructivism seeks to ‘see what happens’ in real-world settings and to illuminate the relevance of social context and processes for resilience, self-reliance, and the opportunities and constraints that govern access to resources for individuals and connected groups. This research approach aims to elicit and understand how research participants construct, negotiate and share meanings around the phenomenon of interest. In this study the phenomenon is deterioration, dying and medical treatments at the end of life.

The study was designed to undertake an in-depth exploration of the diversity and complexity of experience within a limited number of cases. There were three WSs:

•

WS1: longitudinal patient-centred case studies comprising an ethnically diverse patient with advanced illness, one or two of their FCGs and a HCP nominated by the patient

•

WS2: single interviews with bereaved FCGs who have experienced the loss of an ethnically diverse family member due to advanced illness

•

WS3: public and professional stakeholder virtual workshops to discuss fictionalised, EOLCP scenarios, derived from real-life experiences reported by participants.

Qualitative methods of data collection and analysis are particularly suitable for use when little is known about the issue and in the study of sensitive and complex topics. Semistructured interviews enable the exploration of key topics, while concurrently allowing the flexibility to follow up on new issues of interest and importance.^107–109 Recruiting severely ill patients to take part in research projects presents challenges, which required flexibility and sensitivity in engaging FCGs and patients in a way and at a level that they could comfortably accommodate. We adopted the case study approach because the issues are socially and culturally complex. Case studies aim to explore complexity through an intensive, holistic focus on the components of each case, rather than obtain limited data from a larger number of single types of unconnected respondents.^110–112 Case studies are suited to the investigation of complex real-world situations in which a diversity of perspectives are at play.^113,114 Engaging different participants within each case across longitudinal follow-up enables the comparison of different perspectives and the exploration of processes and experiences emerging over time, rather than the cross-sectional snapshot provided by one-off interviews. Interview data were supplemented when possible by additional data from a focused review of medical records. Each case was followed up, when possible, for a period of approximately 6 months.

Setting/context

The geographical setting for the study was the East Midlands across Leicester, Leicestershire, Nottingham and Nottinghamshire. Throughout the report, we shall refer to the settings as ‘Leicester’ and ‘Nottingham’. Patient participants were under the care of services in these areas and health professionals for WS1 and WS2 worked in these areas. FCGs for WS1 and WS2 did not have to be resident in these areas. Participants for WS3 were largely from these areas, but some key stakeholders worked elsewhere in the midlands or UK.

All research activity took place in this geographical area, and Leicestershire and Rutland Organisation for the Relief of Suffering (LOROS) Hospice was the single recruiting research site. For WS1 and WS2, interviews with FCGs who were not resident in these areas took place in a venue within a reasonable distance of LOROS Hospice or, after a protocol amendment in relation to COVID-19, virtually, as highlighted in the following section.

Ethics and governance approval

NHS Research Ethics Committee approval was obtained in December 2018 (reference number 18/WM/0310), and then two substantial amendments followed. Substantial amendment 1 (1 May 2019): we did not give copies of audio-recorded consent to the participants. This related to potential difficulties with participants having access to appropriate technology to listen to it. When interpreters were used, we informed participants that consent would be audio-recorded. The recording was then retained as source data. Paper copies of the consent forms were filed and a copy given to the participant. For practical and resource reasons, copies of audio-recordings were not distributed to participants. To enable greater flexibility for patient participation, we increased the time of maximum involvement from 4 months to 6 months, to enable our engagement with them to be at a slower pace should patients need this.

In the light of government and NIHR guidance on COVID-19, substantial amendment 2 (9 March 2020) was made to the study protocol to ensure the safe delivery of the remainder of the study. The first priority was the safety of participants, many of whom were very vulnerable to the virus. The second consideration was to minimise the burden on health services and professionals. In WS1, we completed the remaining interviews for existing participants by telephone, and also amended the timeline for these in the protocol from 6 months to 6–9 months. We recognised that, in these challenging times, it was not possible for us to review all medical records as originally planned, so we amended the relevant text in the protocol to capture this. In WS2, we obtained consent by telephone for the remaining participants and completed the single interview for these participants by telephone. In WS3, we amended the protocol to allow the workshops to be conducted face to face, via webinar or by individually completing a workbook via e-mail.

Although COVID-19 did not unduly affect the scientific quality of the project, it affected the progress of analysis and development of WS3. The chief investigator (CI) was redeployed clinically for 5 months from April to the end of August 2020. This had a knock-on effect on the timeline for study outputs. Hence, a case was put forward, and subsequently granted, for a 5-month costed extension.

As a research team, we believe that not offering vulnerable people the choice to take part in research because of assumptions made about their experiences and preferences is inequitable and exclusionary.^111–114 We also recognise that any research engaging with severely ill patients and bereaved FCGs presents challenges and needs to be broached with sensitivity and flexibility throughout the research process.

Prior to the decision to take part, participants were asked to contemplate how they would feel about deliberating their experience of illness and dying. They were assured that participation was voluntary and that they could withdraw from the study at any time. To further mitigate the potential risks for participants and researchers, an information sheet with contacts for further information or support was provided, and a follow-up telephone call with the study CI, if desired, was also offered. We developed a ‘pyramid of support needs’ to manage distress within the study, which comprised assessment by the interviewer, resulting in (1) no support required, (2) signposting to a general practitioner (GP), (3) written information provided about local third-sector organisations or (4) formal referral initiated for counselling (via study CI). Patients’ GPs were also informed that the patients were participants in this study to enable the GP to document this in their medical records as evidence of participation.

Eligibility

Sample size and structure

There are no hard and fast criteria for establishing the sample size required in qualitative research.¹¹⁵ Rather, this is determined by the circumstances and context of each study. Morse¹¹⁶ and Malterud et al.¹¹⁰ propose that the number of participants required depends on the range and depth of information collected about each participant or case (information power): the greater this is, the fewer participants required. A qualitative sample size of between 20 and 40 participants is likely to include the majority of views and experiences to be found within the target population and is in line with previous longitudinal studies in EOLC adopting a similar design and method.^{72,106,115,117}

Purposive sampling, which involves participant selection being guided by strategic choices regarding the individuals who or groups that can yield the most valuable and relevant information for the study, is a strategy that can optimise the depth and breadth of data in a sample and reduce the sample size required to achieve thematic saturation (i.e. when no additional themes about the research topic are emerging from the participant interviews). Stake¹¹³ cautions against increasing the number of cases much beyond 15 because the number of qualitative data generated by a larger number of cases will become unmanageable. However, a guiding principle of qualitative research is that the nature and adequacy of the final sample must be kept continuously under review, and adjustments made if necessary, to enable the study to achieve its aims.¹¹⁰ Above all, it is the relevance of the participants and the quality of the data that are important, rather than the number of participants per se.

In WS1 and WS2, we purposively sampled participants across three elements of diversity:

•

religious/faith group

•

ethnic backgrounds (ethnically diverse groups)

•

disease/illness group.

Findings from our previous work suggested that cultural–religious customs and mores are one of the key factors that increases the complexity of navigating EOLC and achieving patient preferences.^41,97 It was our intention that we achieve the greatest heterogeneity of religion/faith and ethnicity in the samples of WS1 and WS2.

This concern with cultural–religious diversity was complemented by selection criteria for recruitment across illness contexts and we also purposively sampled to achieve heterogeneity across a number of disease groups (including cancer, frailty in old age, and heart and renal disease).

In addition, demographic data were recorded, including gender, age and migration generation, and we aimed to include a diversity of these characteristics in the sample.

A sampling frame was utilised to target recruitment to achieve these purposive sampling aims and to construct a matrix that may be important for attribution of themes/subthemes and in our search for examples of variance within the data. The nationally agreed guidelines to characterise self-reported ethnicity into a set of 16 codes were used.¹¹⁸

Recruitment

Taking on board best practice in recruitment for research involving ethnically diverse participants, an extensive programme of awareness-raising about the project took place within community groups and organisations.^119,120 This work was informed by the networks and experiences of our PPI lead (IM) and PPI consultees [named public, carers and bereaved relatives (PCBR) research consultees] and of the co-applicants and advisors to the project. The work involved communicating the project objectives and the opportunity to be involved in research through social and traditional media; the placement of flyers in organisations and services including, but not limited to, general practices, hospice services, patient and carer support groups, community and faith groups, and libraries; and attendance at meetings and events of patient and carer support groups and community and faith groups by members of the project team.

Workstream 1: longitudinal patient case studies

The Clinical Research Network East Midlands (in which co-applicant Simon Royal and the CI have clinical leadership roles); Clinical Commissioning Group research networks and the well-established clinical and research links of the CI and co-applicants Simon Royal, Simon Conroy and Alison Pilsworth; and collaborators including specialist nurses and research nurses within participant identification centre site services were engaged to promote awareness among HCPs of the study, to promote understanding of the criteria of case eligibility and to seek identification of patients within the participant identification centre sites, which were as follows:

•

five general practices with a substantial population of ethnically diverse patients and the linked community nursing services

•

community and primary health-care services that support a population with particularly harder-to-reach ethnically diverse communities (e.g. Somali)

•

palliative care services in community, hospice and hospital sectors

•

community and secondary care services for heart failure, elderly and psychogeriatric care, renal medicine and oncology.

In total we engaged with 110 community groups in Leicester and 67 community groups in Nottingham and attended 15 community events. Our engagement with a variety of community groups provided a platform to discuss the study and how it aimed to explore experiences and improve EOLC, thus achieving a high level of community engagement in the study. In small part this was about widening recruitment; in large part it was about underpinning dissemination of findings in the future and maximising the potential for impact on patient and family care, and outcomes.

We liaised and worked with research and EOLC leads, practice managers, nurses and other workers and primary care patient participation groups to promote understanding of the study. Presentations were delivered by the CI, the principal investigator (PI), a research fellow and a research associate at a number of meetings, including those of cardiology and renal teams, general practices and local hospices within the area, and at a number of community events in which the recruitment process was discussed (see Figure 1).

FIGURE 1

Workstream 1: patient recruitment flow chart.

Health-care professionals were asked to identify eligible patients through a variety of means, depending on their role and the systems in place within their service. HCPs were provided with the Supportive and Palliative Care Indicators Tool (SPICT™, University of Edinburgh, Edinburgh, UK)¹²¹ as a means of identifying patients at risk of deteriorating and dying with one or more advanced, progressive and life-threatening conditions and the Gold Standards Framework ‘surprise’ question: ‘would you be surprised if this patient were to die in the next 12 months?’.^80,122

A HCP known to the patient was the first to approach a patient and provide them with the participant information leaflets for patients and FCGs. Patients, or someone acting on their behalf, then contacted the research team directly about their interest in the study by telephone, e-mail or returning a reply slip in a freepost envelope given to them by the HCP. Alternatively, if they preferred, they could give permission to the HCP who first approached them to pass their contact details to the research team, requesting that the team contact them.

If a patient did not speak English, the HCP would direct the patient (or FCG, if they were interpreting on their behalf) to the appropriate information video on the LOROS project website and/or ask if a researcher with shared language skills may contact them.

If potential patient participants lacked the capacity to give any or full consent owing to the nature of their illness (e.g. dementia), the study was introduced to them and to their FCG by a known HCP. If the FCG considered that the patient would want to consider involvement, then information about the study was shared with both parties in the ways described previously. For such patients, their involvement could be limited to permission to review their medical records and to approach a FCG and a HCP to seek their involvement. However, patients were not excluded from contributing to the study through interviews if the consultee considered that this would be appropriate and acceptable.

The research team contacted patients, or their consultees, who indicated that they were interested in the study and provided them with full participant information, including in audio/video format in their first language if needed. This was available in Gujarati and Hindi translations via the LOROS website link to the study. Audio-/video-recorded methods of facilitating informed consent are regarded as acceptable alternatives to written consent for study populations in which literacy skills are variable.¹²³

Workstreams 1 and 2: consent

The process for obtaining participant-informed consent at the outset was undertaken with Research Ethics Committee guidance, and in accordance with good clinical practice and General Data Protection Regulation (2018)¹²⁴ regulatory requirements.

The participants in WS1 and WS2 were given a copy of the original signed and dated consent form. The original signed and dated consent form was retained in the trial master file.

For patients who lacked capacity, a consultee was asked to complete the consultee declaration form in accordance with the process of the Mental Capacity Act 2005² and the Health Research Authority guidance.

The decision regarding participation in the study was entirely voluntary. The researcher emphasised to all participants that consent regarding study participation could be withdrawn at any time without penalty, without affecting the quality or quantity of their future medical care, and without loss of benefits to which the participant was otherwise entitled.

•

Written consent was taken.

•

If the participant’s first language was a language other than English, consent was taken verbally via a digital recorder, as well as the participant completing a written consent form in English. The interpreter was also asked to sign the written consent form under ‘witness signature’.

•

Informed consent took place at least 24 hours after the participant had been given the full participant information, or as long as they required to decide.

Consent from patients and FCGs (WS1) and BFCGS (WS2) was undertaken face to face, and via a professional interpreter accompanying the researcher when needed. Patients were seen in the LOROS Hospice, in their own home or at a suitable location of their choice. For those individuals thought not to have capacity to give fully informed consent, we used the provisions of the Mental Capacity Act 2005² and the consultee provided information regarding the patient’s wishes about participation in research.

For HCPs, consent was taken immediately before the interview. For telephone interviews, consent was taken verbally via a digital recorder and consent forms were sent to the HCP and they signed and sent the original back to the researcher by post, fax or e-mail. Further details of the consent process are detailed in Ethics and governance approval.

Data collection

Case studies: workstream 1

Longitudinal case studies comprised sequential interviews over a period of up to 9 months with a patient, FCGs and a HCP nominated by the patient. Participants took part in a semistructured, audio-recorded interview arranged at their convenience. Interviews with the patient and FCG participants were mostly face to face, but, post COVID-19, some were via telephone. If a patient died during the follow-up period, a bereavement interview was requested with each FCG at a minimum of 8 weeks after the death. When possible a review of each patient’s medical records was also undertaken.

We intended to be as inclusive as possible and each case study was explored on its own terms. It is not necessary for case studies to conform to a ‘standard’ composition and patients were given flexibility and options in how they wanted to be part of the study. Consequently, the data set includes the following:

•

patients who lacked FCGs or who were unwilling/unable to nominate specific FCGs or a HCP

•

patients who nominated more than one FCG to take part in their case study.

In some case studies, the greater part of the interview data were drawn from interviews with FCGs and HCPs. This occurred in the following situations:

•

Patients did not wish to participate in some or all interviews, but were willing for their nominated FCG(s) and HCP to do so and to have their notes reviewed.

•

Patients lacked capacity to give full consent from the start (e.g. patients with dementia or a brain tumour).

The number of interviews conducted with each participant was determined by the circumstances of each case study. A minimum of two interviews was needed for each case study. A maximum of two follow-up interviews over 6–9 months were undertaken with each patient and FCG. Interviews were with patients and FCGs separately, jointly or a combination, according to participant preferences and convenience.

A nominated HCP was invited to take part in at least one interview, and, at most, two interviews, to discuss their involvement in the case (see Figure 2).

FIGURE 2

Patient case study composition.

Experienced researchers conducted all interviews. The same researcher conducted all the interviews with each participant in the patient’s case study.

The content discussed in each interview developed over time, but, in general, interviews with patients and FCGs included exploration of the following themes:

•

experience of living with serious illness

•

understanding of the illness and the prognosis

•

goals and values with regard to their future care

•

anticipation of and thinking about the future, about decision-making and treatment preferences, and the significant factors influencing these

•

preferences in how such decisions are made

•

expectations and experiences of professional support and communication about deterioration, EOLC and EOLCP.

We were mindful that the experience of taking part in the interviews may have had an influence on the individuals.¹²⁵ Participation in research has the potential to change the nature of the patient and their family’s way of living and coping with their illness and their future, and this was to be incorporated as a topic in the interviews. The potential impact would be beneficial in that the patient/carer may communicate more effectively with care providers. The listening skills of the researcher and the facilitation of reflection and discussion may potentially change things for the participant.

In subsequent interviews, we asked the following questions:

•

How are things?

•

What’s changed in the illness?

•

Have they been thinking about the future?

•

Have they had any conversations with family/HCPs about their illness/wishes and future death (if the patient brings this into the conversation)?

•

Do they think that they have done anything differently as a result of having these conversations/being part of the study?

Such actions were explored as an outcome of the research; importantly, such observations may provide information about potential facilitators of EOLCP, as discussed in the findings and discussion in Chapter 5.

The interview with HCPs also included the following themes:

•

their experience of providing care and support for the patient and family

•

barriers to and enablers of their care of the patient and family in:

∘

understanding the illness and prognosis

∘

thinking about the future

∘

decision-making

∘

EOLCP.

Health-care professionals were also asked for their assessment of whether or not they felt that being part of this study (either they themselves or the patient or FCG) had an impact on their relationship with the patient/carer or any other research-related outcomes. This will be further discussed in Chapter 5.

Medical record review

Patients were asked to give permission for access to relevant parts of their medical records and related documents recording future care preferences.

At the end of patient involvement in the study (after 6–9 months or at death), when possible, the medical records, including nursing and allied health professional records, were sourced. These included palliative care services, primary care, community services and hospital services. The records were scrutinised for data concerning the following:

•

discussions and information-sharing about prognosis, deterioration, dying, EOLC and EOLCP

•

records relevant to recording patient views and wishes, and EOLCP, including advance care plan and DNACPR documentation

•

preferences regarding place of death.

Data from electronic and paper case notes were extracted verbatim and entered into the electronic case report form for subsequent analysis. The date, place of discussion and role of the HCP were noted for each data extracted. The researcher also added an explanation of the circumstances of the discussion to provide context for the narrative.

Completed case report forms were incorporated within the project database and form part of the data set relating to each case. This is discussed further in Chapter 5.

Workstream 2: data collection

Interviews with recently bereaved FCGs of ethnically diverse patients who died following a period of advanced disease and deterioration were conducted, either face to face or by telephone, by experienced researchers in a participant’s primary language (using translation strategies described in WS1). The interviews explored participants’ perspectives of the following:

•

living with serious illness and EOLC

•

thinking ahead about deterioration and dying

•

information-sharing, communication and decision-making in illness deterioration with the patient, family members and HCPs involved in providing care

•

the role and value of EOLCP

•

the role of the HCP in helping patients and families prepare for deterioration and dying.

Interpreters and transcription of interviews

Our research team have considerable language skills to conduct interviews in the language preferred by a participant, and one of the BFCG interviews was conducted by the bilingual PI (ZI) in Punjabi. Zoebia Islam translated the audio-recorded interview into an English-recorded version before a professional transcriber transcribed the interview. This method has previously been employed by Zoebia Islam.¹²⁶ Professional interpreters were used for patient participants in four of the case studies.

There is limited guidance available for how to appropriately brief and debrief interpreters before/after each interview. Therefore, the research team developed its own to ensure that the interpreters felt suitably prepared for the types of questions that would be posed and the sensitive nature of these. Zoebia Islam also checked that each interpreter was able to pose the questions and obtain responses in debrief meetings, especially when there may not have been any literal translations for certain words. The debrief meeting also added further explanation and understanding to the data collected. This was particularly pertinent in one case for which the interpreter was able to explain the reasoning behind certain cultural beliefs as they held the same beliefs as the participant.

All interviews were audio-recorded. Interviews in English were transcribed verbatim. Interviews in other languages recorded Zoebia Islam asking the question in English and the interpreter then repeating this in the participant’s first language. After the participant responded in their first language, the interpreter would then repeat the response in English. The English parts of the audio-recording were transcribed verbatim. Again, this method has previously been employed by Zoebia Islam.¹²⁶

All transcripts were checked for accuracy against the audio file and anonymised. Turns of phrase, idioms and metaphors were sometimes complex to translate and a note was made on the transcript of literal translation, as well as the meaning constructed by the translation during the interviews conducted by Zoebia Islam. These notes were an important aspect of quality assurance and data accuracy.

Workstream 3: public and professional virtual workshops

Workstream 3 involved conducting virtual stakeholder workshops, with participants joining a virtual workshop or participants completing a workbook. Based on the findings from WS1 and WS2, we developed topic guides for workshops and workbooks. These had commonalities, but also included materials that were specific to each of the different participant groups; they iteratively evolved during the course of the workshops. Stakeholders were categorised as members of the public from diverse ethnic backgrounds (lay), community and faith leaders, academics, educators or HCPs.

Workstream 3: sampling and recruitment

For WS3, we purposively recruited and sampled across the range of stakeholders. The recruitment of stakeholder participants was via a process of snowballing. The study team and all collaborators, including members of the PCBR group, identified potential participants via local small-world networks. Study flyers and e-mails were circulated. Through this process, > 140 potential participants were contacted. Potential participants were contacted by e-mail, were sent a participant information leaflet and reply slip, and were offered the opportunity to attend a workshop or compete a workbook in their own time.

Workstream 3: data collection

Based on the findings from WS1 and WS2, eight fictionalised stories were developed for use in the workshops. Each of the stories highlighted key themes identified in WS1 and WS2. A topic guide for workshops and workbooks for those unable to attend workshops was drafted for each of the different stakeholder groups. These draft stories, topic guides and accompanying workbooks were piloted with the PPI lead (IM) and members of the PCBR group, and were refined based on their comments and after further discussion within the core research team. The HCP topic guide and accompanying stories were also piloted with professionals, with subsequent refinement.

The workbooks and topic guides had some commonalities; for instance, all participants were asked to identify challenges within each story and best practice for HCPs in supporting patients in EOLC discussions and planning, and to comment on the implication of these best practices for training and for service delivery. Key questions were as follows: what are the ‘generic’ issues? Are there any specific ethnically diverse issues? How do ethnically diverse backgrounds overlay/increase complexity of the generic issues? How do professionals make ‘best-interests’ decisions in the cross-cultural context?

Each stakeholder workshop also included material that was tailored to each of the different groups and that iteratively evolved during the course of the workshops. The stories and the key themes highlighted in each will be further discussed in Chapter 6 and are presented in Appendix 1.

Each virtual workshop had a maximum of five participants. Data resulting from the workbook tasks, contemporaneous field notes, audio-recordings of workshop interactions and semistructured questionnaires captured the participants’ views and are discussed in Chapter 6. Demographic information was also gathered, together with feedback on the workshop or workbook materials to inform the iterative development of WS3 and the educational outputs from the study.

Workstream 3: process and consent

On the day of the workshop, the chairperson of the session, the CI (CF), welcomed everyone joining the workshop; led the introduction of the team; and went through each item on the consent form, asking participants to initial and sign accordingly and return by post or e-mail. On receipt of a signed participant consent form, the study research associate (MH) also signed and dated the form and allocated each participant an identity number before returning a copy to the participant. Participants completing workbooks were also required to read the information leaflet before completing their consent form and the workbooks. On receipt of a completed and signed participant consent form and completed workbooks, for both participants who attended virtual workshops and participants who completed individual workbooks, Matilda Hanjari sent each participant thank-you vouchers.

Analysis of workstreams 1, 2 and 3: multiple datasets

There were several data sets for analysis.

•

WS1: 18 patient-based case studies comprising 93 interviews across three stakeholder groups (i.e. patients, FCGs and HCPs). Review of medical records for 13 cases.

•

WS2: 19 interviews with BFCGs.

•

WS3: workshop reflection notes, recordings and workbook responses.

In WS1, each data set was subjected to both separate and integrated thematic analysis through constant comparison.^108,127 The qualitative software program NVivo (QSR International, Warrington, UK) facilitated the management and coding of data sets and allowed for cross-stakeholder and cross-WS coding and queries. Analysis proceeded through a process of open coding of initial manuscripts by at least two researchers independently, and then development of the coding framework through discussion of the themes between core research team members (AP, MH, ZI, CF and KP). Coding has involved ‘coding’ segments of interview transcripts to one or more broad ‘nodes’ (or themes) within NVivo to capture all text relating to an idea or topic. The coding framework was developed through an iterative process of reading, coding and discussing data elements within the research team to identify, compare and link ‘themes’ occurring within and across the interviews. A more refined and selective process of coding from individual nodes was undertaken largely by Anne Patterson and by Kristian Pollock (co-applicant) to explore, differentiate, reorganise and relate the themes identified as of greatest relevance to the study objectives.

In WS1, patient-centred narrative case studies were developed for each of the 18 cases individually by the researcher undertaking the interviews in each case and one other core researcher, and then further refined after discussion between core researchers. These triangulate different stakeholder perspectives and data sources, including a review of medical records. Each case study involved a holistic in-depth analysis of each case to enable an understanding of how ‘thinking ahead’ and EOLCP evolve over time, and also a systematic cross-case comparison to discover patterns in the characteristics of the cases and to identify and examine any relationship between themes within a co-ordinated explanatory model of study findings.^113,127 Serial follow-up interview data go beyond cross-sectional and static accounts of specific participants and groups of stakeholders and enable an understanding of how EOLCP is managed and experienced over time within a complex network of care. The additional comparative dimension afforded by thematic analysis of the interviews of each participant group completed the triangulation of the different data sets.¹¹⁴

Analysis of interview data for WS2 was similar to that for WS1: this was coded in terms of overall themes and subthemes occurring in the data through filtering in NVivo. Further exploration, by comparing themes between WS1 and WS2 and triangulating findings, was undertaken to enable an understanding of the key issues relating to EOLCP when bereaved relatives reflect on their experiences, and the degree of difference, overlap and mutual understanding that exists between them and FCGs involved in current, active care of a patient with advanced illness; patients; and HCPs.

In NVivo, data were initially coded for each of the stakeholder groups using a coding framework developed by (at least) double-coding of a range of interviews from each data set. Common threads were then drawn together from across the stakeholder groups. Relevant ‘nodes’ in NVivo were then drawn together around these common threads (e.g. communication, professional support). These were then interrogated further, for example enablers of and barriers to good communication; enablers of and barriers to effectively ‘knowing’ a patient and family. These were further coded to arrive at enablers and barriers for several aspects. For other sets of data, it was more appropriate to interrogate and code these further, using categories of ‘what worked’ and ‘what did not work’ (e.g. when scrutinising positive experiences and negative experiences of receiving health-care support). In this way, the coding framework looked both within and across stakeholder groups, which facilitated all stakeholders’ perspectives on particular emerging themes/issues.

Workstream 3: analysis

Analysis of WS3 involved a separate pragmatic thematic analysis of each workshop’s field notes (by ZI and MH). Field notes from each of the workshops were coded to themes and subthemes through a process of constant comparison, similar to that employed in WS1 and WS2.^108,127 In this, researchers focused on identifying themes and subthemes relating to the specific questions addressed in each workshop. Emerging themes were compared and grouped in relation to each category of stakeholder workshop and workbook (by ZI and MH), and then further refined after discussion with the rest of the core research team. The qualitative software program NVivo facilitated the management of this pragmatic thematic approach.

One of the co-applicants (LW) undertook separate analyses of the HCP workshop field notes. Analyses used two implementation science approaches. The first used a psychological approach to identify behaviour change techniques (BCTs) in the discussion of behaviour change by practitioners and family members. A BCT is an observable, replicable and irreducible component of an intervention designed to alter or redirect causal processes that regulate behaviour, that is a technique is proposed to be an ‘active ingredient’.¹²⁸ The 91-item taxonomy was applied to the transcripts of the workshops. The second is an organisation-level approach based in sociology: normalisation process theory (NPT).¹²⁹ It is a mid-range theory of the implementation, embedding and integration of organisational innovations, as result of individual and collective agency. This agency is developed through four processes: sense-making (coherence) work, engagement work (cognitive participation), enacting a practice (collective action) and how the intervention has an impact and on whom (reflexive action). Each of these four processes has four further components, making a 16-item framework applicable to any complex intervention or innovative practice involving the adoption of new practices. This approach was used to ascertain how the HCPs discussed the practices around ‘thinking ahead’ and EOLCP as applied to the stories and discussions in the workshops. These analyses informed the content of the learning resource.