Design
The study uses mixed methods. Qualitative data will help describe how admissions are understood and viewed from “the ground up”. Quantitative data will help determine the extent of the attitudes, perspectives and intentions we have gathered from the qualitative part of the study reflect patient and practitioners experiences. Together, these data will then be used to evaluate current services and inform policy.
Setting
The study focuses on two adjoining but separate administrative areas, with different demographic profiles relevant to place of death, including access to services. The first area is largely rural, with a university city and larger towns and areas of significant social deprivation and affluence. The second area is mainly urban, characterised by high levels of ethnic diversity and significant pockets of urban deprivation. The study area includes one large Primary Care Trust (PCT) which will shortly become a Clinical Commissioning Group (CCG), responsible for purchasing care for the local communities. There are two providers of Out of Hours (OOH) primary care and two Local Authorities responsible for social care. The study will focus on the two acute settings in the study areas.
Data collection
The study is organised in three data collection phases, encompassing five research tasks.
Phase 1: Patient case-studies
Phase 1 of the study adopts a qualitative case-study approach centred on patients aged 65 or older with advanced dementia, Chronic Obstructive Pulmonary Disease, or cancer who die within 72 hours of admission to of the two hospitals. Patients will be identified retrospectively with help from the hospitals’ Bereavement Services. These Services receive the hospital notes and organise the death certificates and other paperwork that needs to be completed following all deaths in hospital: the notes will be used to identify patients for the case studies. The Services also provide comprehensive bereavement support or onward referral as appropriate. The study aims to identify a maximum of 48 patient case-studies, half from each hospital, with approximately eight patients in each condition group. This is shown diagrammatically in Figure 1. Patients will be purposively sampled to ensure coverage of a range of variables known to predict place of death [4, 22].
Participants in Phase 1 are (1) community and hospital staff involved in the admission who will be interviewed shortly after the death and (2) patients’ next of kin, who will be interviewed three to six months after the patient died. Up to six people involved with each patient’s care will be invited for interview, including one next of kin if possible. Staff will be identified from patient medical records and then through a “chain” or snowball sampling (See Figure 2).
The recorded next of kin in the deceased’s medical record will be contacted by the hospitals’ bereavement offices and invited to take part in the study. Participants will be asked to return a reply-slip indicating interest to the university-based research team. In both cases, data will be collected through semi-structured interviews.
The interviews are designed to explore carers’ descriptions and explanations of what happened and why; the factors that influenced the decision to admit the patient to hospital; how they view the decision in hindsight and what, if anything in their view, could or should have been done differently to prevent admission. Interviews with the next of kin will cover similar issues as well participant’s thoughts on what the patient wanted or might have wanted; and what, if anything, they would like to have been different in the support given.
Phase 2: Non-clinical stakeholder perspectives
Phase 2 of the study involves vignette-based focus groups with patients and carers living with the three conditions of interest to gain a patient perspective. Focus groups will be organised by place and condition (n=6 groups of 6–8 participants). Participants will be identified through hospital services, support groups, the ACE Study user group and existing patient and public involvement networks. Interested participants who reply to the research team will be invited to take part in the study if they meet eligibility criteria.
The focus group will involve the researchers telling a brief vignette story about a patient admitted to hospital who dies shortly afterwards and, depending on time, a patient who dies at home. Participants will be asked to comment on issues such as: what the story tells them; what they might expect to happen; what the people in the story want to happen; what they expect from taking the patient to hospital; what reasons there are for not taking the patient to hospital; what the right thing to do was; what alternative options there were.
A qualitative vignette approach was chosen as being particularly well suited to supporting discussion around topics that are “difficult to broach” or potentially upsetting [23], by creating distance between the participant and the topic [24]. This approach is similar to that employed by Gott et al. in their study of older people's preferences for place of care at end of life [19].
We will use the same vignette approach with a group of senior policymakers and commissioners to compare responses. Participants will be identified through existing professional networks and snowball sampling.
Phase 3: Clinical stakeholder perspectives
In Phase 3, a vignette-style questionnaire for staff in hospitals and the community will validate and quantify earlier findings. The research team will compile a database of relevant professionals within the National Health Service (NHS) boundary of the area. This will include GPs, District Nurses, OOH staff, ambulance staff; palliative care specialists in hospitals and the community, and hospital staff who care for older people, in particular Registrars and Consultants in Geriatrics, Chest Medicine, Oncology, and Emergency Medicine. District Nurses will be identified though area managers, and hospital staff through a “snowball” technique working with departments and existing research team contacts. Details of GPs are in the public domain. A questionnaire will be sent to everyone on the database and it is anticipated that the sample size will be approximately 1000.
The focus of the questionnaire will include a set of patient vignettes with closed multiple-choice options about whether the patient should be admitted or referred, followed by open text boxes asking for reasons and comments. These vignettes will be developed from case-studies in the interview phase of the research, with further input from clinicians and the user group. Vignette questionnaires are considered superior to a traditional questionnaire format because the questions asked are less abstract and closer to real-life [25]. When used quantitatively they also allow the collection of data from large numbers of people [24, 26] and are particularly good for exploring ill-defined issues or dilemmas [27].
Analysis
Table 1 summarises how the data will be analysed to answer the research questions. Data analysis will happen in two phases. First the empirical data will be analysed. Secondly, the findings will be used to evaluate current policy around palliative care and make policy recommendations.
Analysis of empirical data
Data from interviews will be organised thematically using an adapted framework analysis approach [28]. This uses a thematic template [29] and a data recording matrix of theme against participant to identify and organise responses for further analysis and interpretation [30]. The focus of the analysis will be to ascertain participant’s perspective, including their opinions, values and assumptions. This data is likely to be gathered from indirect questioning about the participants experience.
The analysis approach will be guided by psychological models of behaviour and will use a social marketing framework to help highlight key factors in decision-making. Psychological models of behaviour change are useful for understanding behaviour as they can help direct analytical attention to issues that might be relevant to behaviour change [14, 15] and are well established in health psychology. Social cognition theories relate to how individuals make sense of social situations, recognising that individual behaviour is influenced by a number of factors such as a person’s belief and their self-efficacy. An example of these models of behaviour is shown in Figure 3.
Social marketing is used in public health to help bring about behaviour change [31]. It starts with generating “insights” into actors’ current practices, beliefs, and preferences which influence their behavioural choices. It assumes that individual behaviour is premised on many factors with trade-offs between them, and that current practice may have benefits which are forgone if the behaviour is changed, such as emotional well-being and social acceptance. In Public Health research these insights have been used to develop “attractive exchanges” [32], designed to support the adoption of new behaviours.
Social marketing assists analysis by focusing attention on the factors that influence behaviour from the perspective of the actors themselves, without attempting to attribute blame. The current palliative care policy directed towards reducing inpatient death does not explore why formal and informal caregivers decide to admit a patient and using an analysis approach that is focused on behaviour will provide useful insight. Applying the social marketing framework to the data will help order the analysis and generate future policy options for providing high quality Palliative care grounded in the empirical data.
The questionnaire data will be analysed quantitatively, requiring open questions to be coded. The sampling frame will be the entire population of healthcare professionals involved in decision concerning hospital admissions in the two hospitals studied and the surrounding community teams. We estimate this to be around 1000 professionals. The questionnaire is not designed for hypothesis testing and will generate primarily descriptive data. Basic descriptive analysis of categorical sets such as professional group or geographical area will be undertaken in SPSS, using methods including chi-square, logistical regression and cluster analysis as appropriate. Analysis categories will be derived from the literature and cluster analysis. Questionnaire results will be collated with the main qualitative analysis and used to evaluate current policy and make recommendations.
The second stage of the analysis will evaluate current policy and literature in the light of the empirical findings. We expect that combining the behaviour model and social marketing framework will provide powerful analytical leverage to offer a clear description of current behaviour and its associates, and a transparent means to evaluate current policy. This analysis is intended to be solution-focused; to provide “actionable insights” to inform effective policy.
Ethical issues
Palliative care is recognised as a sensitive area to research. As such, a considerable time has been spent discussing the protocol and instruments with clinician stakeholders and the user group, who have advised on feasibility and acceptability. A central issue was the identification of suitable patients and the individuals who had been involved in the decision which had led to their admission to hospital. Consent could not be gained from the deceased patient and neither was proxy consent from next of kin possible. This was both because a patient's notes remain confidential to the patient after death and because our user group advised against contacting bereaved relatives so soon after death. Instead we gained approval from the National Information Governance Board (NIGB) to view deceased patients’ records to collect relevant information and identify an initial contact. In addition we have collaborated with the hospital bereavement services who will initially approach next of kin and invite them to participate in the study.
Working with the bereavement services is beneficial to maintaining the confidentiality of participants. In particular, the invitation letter to the patient’s next of kin will come from the care provider, thus ensuring patient and personal data remains within the service. In turn, the service will not know who responded to the research team, which will help increase the anonymity of the responses.
It is usual when undertaking research with people who have been bereaved or who are very ill to inform participants that you will contact their GP on their behalf if they become worryingly distressed during the interview and refuse to seek help themselves. In this case, the participants are not known to the researchers as patients, so it will be necessary to gather information about the participants’ GP before they are interviewed. The research ethics committee reviewing the study also requested that the interviewer follow up next of kin participants a couple of days after interview to check on their well-being. In addition we were asked to prepare an information sheet on sources of bereavement support to leave with participants.
The topic of care at the end of life is potentially very emotionally charged for all participants: care professionals who may have regrets over decisions made, and informal carers who are recently bereaved. However, since care in last few days of life is the focus of the study, we have been explicit in our material to ensure that participants are not surprised by this topic being raised. While this may bias the sample, this approach we believe is more ethical and is unlikely to be detrimental to overall validity of the study.
The vignette design of the focus groups will allow participants to distance themselves from the topic if they wish. The focus groups with patients and carers will be attended by two researchers, so that one can support any participant should they become distressed. At their own suggestion, members of the project's user group have volunteered to attend the focus groups to help support and reassure participants. At the end of the focus group participants will be debriefed and asked if they have any concerns arising from the group session. They will be signposted to contacts from where they could gain support. We will seek to arrange for an appropriate specialist nurse to come at the end of the group session to take any questions and help debrief participants.
Approvals
The study is approved by the Hertfordshire Research Ethics Committee (England) (#11/EE/0491), and National Information Governance Board (NIGB) (ECC 1–05 (g)/2012).