The investigators developed the theory-driven 'Compassion Intervention', an integrated, interdisciplinary approach to address existing gaps in end-of-life care for people with advanced dementia. The Intervention consists of two core components: facilitation of an integrated, multi-disciplinary approach to assessment, treatment and care; and education, training and support for formal and informal carers. The intervention is implemented by an Interdisciplinary Care Leader. The primary aim is to understand how the Intervention operates in two care homes (with nursing support) in two different health and social care economies; one in the Camden Commissioning Group and one in the Barnet Commissioning Group, both in London, UK. The secondary aim is to collect preliminary outcome data and estimate the cost of employing an Interdisciplinary Care Leader to inform further evaluative studies. The final aim is to check that the Intervention causes no harm to residents and their family carers.

Within a three-year research programme the investigators used the RAND/UCLA Appropriateness Method (Fitch, 2001) to achieve national consensus on the components of Compassion ('the Intervention'), a complex intervention for a model of end of life care for people with advanced dementia. The Intervention consists of two core components: facilitation of an integrated, multi-disciplinary approach to assessment, treatment and care; and education, training and support for formal and informal carers. The development is described in detail in a manual and a publication (Jones, 2015). The Intervention is coordinated by an Advanced Dementia Interdisciplinary Care Leader (ICL) The ICL role requires extensive clinical experience in care of frail older people and those with dementia, particularly towards end of life. Skills may be drawn from the fields of nursing, social work or a profession allied to medicine. This theory-driven approach to improving end of life care in advanced dementia has not previously been tested in a real-world setting. The protocol is published (Elliott, 2014).

The study will be an exploratory, naturalistic implementation study of the Compassion Intervention following the principles of dynamic sustainability, recognising that implementing protocols in real-life settings requires adaptations, and that rigid adherence to guidelines tested in controlled settings may not be suitable or effective in broader contexts (Chambers, 2013). The investigators structured their approach using the phases of implementation described by Grol and colleagues (2007) including orientation, insight, acceptance, change and maintenance. A full time ICL with a social care background and considerable experience working with people with dementia in care homes was employed. The ICL will be supervised by clinicians with palliative and dementia expertise within the research team and two external clinicians.

Two care homes, both predominately nursing home level of care, were invited to participate having been involved in an earlier phase of the research programme; a longitudinal (9 months) cohort study to examine the needs of people with advanced dementia and their family carers (Jones, 2012). Care home managers gave written consent for their site to participate, and permission for the ICL to carry out clinical assessments of eligible residents and have access to their files. Managers will identify eligible residents. Implementation will occur over 15 months. In month 1, in each care home, the ICL will meet with care home managers, deputy managers and key external healthcare professionals who visit the care homes. Posters promoting the intervention will be displayed in each care home. The Intervention was launched in Care Home 1 in May 2014 and Care Home 2 in June 2014. The ICL will be actively engaged in each care home for six months to co-ordinate activities. After six months the ICL will cease active engagement. To assess maintenance of activities, interviews with relevant stakeholders will be conducted after the ICL withdraws at months 7, 11 and 15.

The ICL will keep a reflective diary to document observations of practice, liaison with staff, family and residents, examples of improvements in care and personal responses to the role. The diary will influence implementation by enabling the ICL to reflect on her role in situ and modify her approach if required. The ICL will keep a daily log of time spent on tasks related to implementation to enable estimation of costs. It was expected that care home and external healthcare professional time to be involved in Intervention meetings and training will be consistent with their usual job descriptions and so will not be considered an additional cost of the Intervention and any opportunity costs will be offset by the training skills acquired. During the six months at the care homes, the ICL will collect monthly care home-wide data on the number of residents with: documented resuscitation status; a pain management plan; preferred place of death recorded; and hospital admissions. Data on emergency phone calls and location of deaths will be collected. Resident assessments undertaken by the ICL will not be used as research data. These data will be maintained within the care home as clinical information according to their governance polices. Findings from assessments, however, can be reflected on in the anonymised ICL diary and used to inform other Intervention activities such as training. Formal training sessions will be evaluated using evaluation forms.

Monthly individual resident outcome data will be collected by researchers (not the ICL) from residents assessed by the ICL and their family carers. As residents are unlikely to have capacity to make an informed decision to participate in research, care home managers will invite next of kin/primary contact to give consent for this monthly outcome data to be collected. If a next of kin is not available, a professional consultee will provide agreement according to UK's Mental Capacity Act (2005). Residents will be recruited during the first four months of implementation to ensure collection of at least three months of outcome data.

Process data will reported as total number of activities undertaken and total ICL hours spent on different activities. ICL hours spent on activities associated with the implementation will be costed using the UK Department of Health and Health Education England tariffs to estimate the cost of engaging the ICL. Training evaluations and outcomes (facility wide and individual) will be analysed using descriptive statistics using statistical package IBM SPSS Version 22 (2013). Outcome data will be used for monitoring potential harm and to examine the feasibility of collecting measures in future trials, hence a sample size calculation will not performed. Using descriptive statistics we will compare individual outcomes with data from our earlier cohort study but the sample size will be too small for statistical analyses.

Semi-structured interviews will be conducted with a purposive sample of care home staff and external healthcare professionals after the ICL has completed six months at each site and also to assess sustainability after the ICL has left. Interviews will be conducted at months 7, 11 and 15. Family carers who agree for a resident to have monthly individual data collection will be invited to take part in an interview at month 7. Participants will provide written informed consent prior to each interview. Interviews will be audio-recorded and transcribed verbatim. The interviews aim to: assess participants' views of the strengths and weaknesses of the Intervention; identify whether any changes in practice to improve the care of residents with advanced dementia and their family carers were implemented due to the Intervention; and explore whether these changes were maintained after the ICL left. Transcripts will be checked against the audio-recording. One researcher involved in interviewing and transcribing will code all transcripts using QSR International Pty Ltd NVivo V10 software (2012). Framework analysis will be used (Gale, 2013), based on Grol's (2007) phases of implementation. Small chunks of text will be extracted and coded, summarising their content. Coded text will be grouped into larger themes within each of the five phases. Themes will be reviewed and checked by a second researcher.

Inclusion Criteria:

• residents residing in one of the two participating care homes;
• memory problems indicating a diagnosis of dementia according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV);
• Functional Assessment Staging scale (FAST) grade 6a (difficulty putting on clothing) and above (unable to hold head up) (Reisberg, 1988);
• Comorbidities or unmanaged symptoms such as agitation, recurrent infections, pain and pressure ulcers.

Exclusion Criteria:

• Residents who indicate either verbally or non-verbally that they do not wish to participate.
• Residents who are moribund, in a coma, or those where there are clinical concerns that may preclude them being approached.