A Naturalistic Two-year Cohort Study of Agitation and Quality of Life in Care Homes

Research questions What is the relationship between paid carer coping, and agitation and quality of life in people with dementia living in care homes? What are the economic implications? What carer, resident and environmental characteristics predict better quality of life in this group, and should therefore be incorporated into our draft care home intervention before it is evaluated? Do agitation, quality of life and quality of care predict death? Primary hypothesis For people with dementia living in care homes, paid carer use of dysfunctional coping strategies predicts lower quality of life in residents with dementia and that this is more so at higher levels of agitation.

Methods Setting and sampling We will recruit care homes from across England. Our sampling frame will encompass all care homes types where people with dementia reside to ensure external validity and generalisability. We will recruit homes to ensure a representation of each provider type (voluntary, state and private) care provision (nursing, residential) and of urban/suburban and rural locations. We already carry out research with: Jewish Care, Dementia UK, Camden Adult Social Care, Springdene Care Homes group, BUPA Care Homes, Care UK and Abbey Care. We will also work with DeNDRoN ENRICH project (CC has been involved in North Thames pilot) and our project care home partners (Jewish Care, BUPA and Barchester) to recruit care homes. We have discussed with DeNDRoN ENRICH; the network has 389 research-ready care homes and is growing rapidly; our target is achievable by targeting larger homes including our collaborators; BUPA, Jewish Care and Barchester. There are around 18,000 English care homes and 80% of residents have dementia (Alzheimer's Society).

Procedures We will seek care home managers' agreement for their care home's inclusion. We will note refusal, in order to consider external validity and use its predictors in a sensitivity analysis. In included homes, we will identify through the home care staff which people cared for have dementia using a carer proxy measure, the Noticeable Problems Checklist (Levin 1989) if they do not have a known dementia diagnosis. We will do this rather than screening residents using a cognitive measure, as it will not cause distress and is independent of culture and education. We will invite all people with dementia to participate. Some will be able to give informed consent to take part, but the majority are unlikely to have capacity to do so. We will follow the Mental Capacity Act (2005) provisions when deciding whether to include people with dementia who do not have capacity, and in appointing professional consultees for those without a personal consultee as we have in previous studies(Jones et al. 2012). We will also interview the consenting primary family carer of each resident included, if they see their relative at least monthly, and all consenting care team members, who provide hands-on care. This is to understand the coping strategies used by all who care for them and the impact on residents. We will ask the paid carer working most closely with each resident with dementia to complete proxy measures. For each home we will record care home characteristics, including number of residents (in total and with dementia), staffing: resident ratios, turnover and details of regular home activities. All assessments will take place at the care home in a private room, for paid carers and residents; family carers may choose to be interviewed in the care home, in their own home or at UCL.

Measures Clinical research assistants will conduct interviews at baseline, 4, 8, 12 \& 16 months.

Analysis We will test our hypothesis that paid care teams using more dysfunctional coping strategies, is associated with people with dementia having a worse subsequent quality of life (measured using paid carer DEMQOL) and that is more so when their agitation levels are higher. We will take into account the influence of agitation and other predefined carer and resident environmental characteristics (e.g. specialist dementia home; nursing staff; dementia severity; resident sex and age). Hierarchical multiple regression models will be used to allow for the repeated measurements of quality of life over time and clustering by carer team. Interaction terms will be included in the models to consider differential effects of baseline agitation levels. We will conduct a cost-consequences analysis, reporting by different coping strategies the mean per patient resource use and costs, agitation and quality adjusted life years (QALYs) as measured by the DEMQoL. The net monetary benefit, calculated as mean per patient costs and QALYs multiplied by the amount a decision maker is willing to pay for a QALY. Analysis will also be reported by coping strategy. Descriptive statistics of mean per patient costs and resource use by agitation will also be included and will follow a predefined Statistical Analysis Plan (SAP). We will obtain date and cause of death from the ONS mortality data set; the Health and Social Care Information Centre will obtain this data for us by matching NHS number and date of birth against this database.

Sample size calculation In our START study the correlation between dysfunctional coping and quality of life was -0.31. To detect this magnitude of correlation with 90% power \& 5% significance requires 105 people with dementia(Machin et al. 2009). Adjustments to allow for clustering by care team (estimated average team size: 40 people with dementia; Intra cluster correlation (ICC: 0.075)(Fossey et al. 2006), impact of confounding (Variance Inflation Factor = 2) [8] \& an expected average 2.5 repeated measurements/person (based on 30% drop out/year) and correlation between repeated quality of life measurements of 0.75 (from START data) results in a total required sample size of 700. To investigate the interaction between coping strategy and high and low agitation groups will inflate sample size to 2800. This sample size will be reduced if cluster size is less than 40 people with dementia; to 1734 people with dementia based on 87 clusters with 20 people with dementia per cluster, or to 2000 people with dementia based on 80 clusters with 25 per cluster, or to 2537 people with dementia based on 72 clusters with 35 per cluster. Although not powered to show outcome we will investigate the responses of family carers and people with dementia in sensitivity analyses.