Design
The study was delivered in two phases. Phase one: intervention development using focus groups. Phase two: testing the feasibility of an intervention within a cluster RCT methodology.
Phase one: intervention development
To build on the original idea for the personhood and citizenship training intervention, a series of four focus groups were conducted in three residential care homes. Topic guides were adapted to reflect the potential diversity of participants while retaining the focus on intervention development. Sequentially gathering focus group data and then revising the PERSONABLE exercises iteratively informed intervention adaptations. Focus group dynamics and discourse were observed by an independent non-participating nurse. Focus group data were analysed using framework analysis (Smith and Firth, 2011). Mapping data by job role informed changes to PERSONABLE tailored to potentially diverse staff experiences and work practices. To highlight practical areas in need of consideration, the core issues arising from the framework analysis were separated into those which enhanced or detracted from a personhood or citizenship approach to care, or factors which might affect the effective delivery of a training intervention. The framework was then mapped against the topic guide themes of: (1) staff characteristics; (2) current training; (3) personhood and citizenship; and (4) PERSONABLE feedback. Subsequent changes to the PERSONABLE exercises were made in conjunction with a working group of professionals familiar with dementia care and adaptations were carried into subsequent focus groups.
Once adaptations informed by focus group data were complete the PERSONABLE intervention was delivered to a group of care staff and subsequently further refinements were made. This field work exercise was observed by AA, a nurse experienced in the care of people with dementia.
Phase two: feasibiliity testing
Setting
The study took place in care homes in the East of England. A mixture of small and larger care homes located in urban and rural locations were approached.
Participants
Care home: Residential care homes in the East of England were eligible if they provided care for people with dementia. Resident: Residents were eligible if they had a diagnosis of dementia with either capacity to consent to participation in the study or who had a consultee as outlined within the terms of the Mental Capacity Act 2005 (Department of Health, 2005). Residents were excluded if they were receiving palliative care or had an acute illness. Staff: Staff were eligible if they were employed on a full or part-time basis, working in any role such as care worker, ancillary, maintenance or administrative positions. Senior management were excluded because of their potential involvement in resident participant selection.
Baseline measurements
All baseline measurements were collected by JC prior to randomisation in all six sites. Descriptive information about each care home was obtained from the care home manager, including care home ownership, number of resident bedrooms and number of staff.
Residential care home
The quality of the usual care environment was measured using the Therapeutic Environment Screening Survey for Residential Care (TESS-RC) (Sloane et al., 2002), which consists of 31 items. In this study only the 23 items relating to public areas of the care home were used.
Dementia Care Mapping™
Dementia Care Mapping™ (DCM™) was originally developed (Kitwood and Bredin, 1992a, Kitwood and Bredin, 1992b) as a practice development tool. In DCM™ an observer attempts to interpret the experience of those who have dementia and then reflectively discuss the observation data with staff (Buetow, 2019, Whitlatch, 2001, Whitlatch, 2013, Keller et al., 2015). Because of the inter-relatedness of data collection, reflective staff discussion, staff adaptions to care and further data collection, DCM™ has rarely been used, in trials, solely as an outcome measure. Even when trials have stated DCM™ was used as an outcome measure, the observational data remains intrinsically linked to the intervention of reflective discussion with staff (Chenoweth and Jeon, 2007).
DCM™ comprises three elements. For every resident participant, each of these three elements was recorded in five-minute intervals over a four-hour period at baseline and at follow-up:
1) Mood and Engagement score: refers to two concepts ‘mood’ and ‘engagement’ recorded as one value, the one with the most potential for resident wellbeing. Often trials use a DCM™ wellbeing/illbeing ‘WIB’ score as the primary outcome measure, which is the mean of the aggregate mood and engagement score for each participant.
2) Behaviour Category Code: there are twenty-three different behaviour category codes designed to intricately describe the behaviours of people with dementia.
3) Personal Enhancing or Personal Detracting: DCM™ attempts to describe the nature of interactions which occur between the staff and those they care for, using thirty-four codes split into detracting or enhancing.
Residential care home staff
Residential care home staff knowledge about personhood was measured using the Personhood in Dementia Questionnaire (PDQ) (Hunter et al., 2013), consisting of twenty statements that a member of staff can rate, on a seven-point Likert scale, between agree strongly (7) and disagree strongly (1). How able the staff felt to deal with residents who had dementia was measured using a zero to 100 visual analogue scale (VAS).
Randomisation
Cluster randomisation occurred at the level of care home. Block randomisation was used with a block size of two. Once two sites had been recruited and baseline measures complete, they were randomised by AA, using the ralloc command in Stata version 14, on a 1:1 basis (PERSONABLE intervention or training as usual).
Intervention
The PERSONABLE dementia workshop was facilitated face-to-face in a private room by JC across all intervention sites. In response to phase one focus group data, PERSONABLE was designed to be brief, lasting no longer than one hour and fitting into the potentially quiet period after staff handover. The workshop comprised of five reflective exercises, described in phase one results below. PERSONABLE required minimal resources other than a simple and predominantly illustrative workbook which assisted the facilitator when guiding staff participants through the PERSONABLE workshop. Staff participants in the intervention arm were offered the PERSONABLE intervention once per participant. This was a waitlist trial and those randomised to the training as usual group were offered PERSONABLE once all follow-up data had been collected.
Training as usual
The control group received training as usual. During the phase one focus groups key data was collected about the usual type, content, length and frequency of training as usual. These data were collected for staff working across multiple roles and indicated that residential care homes typically provide (i) mandatory training with limited dementia specific content; (ii) occasional dementia specific training usually attended by senior staff and subsequently informally disseminated to junior care staff; (iii) usually no dementia specific training for ancillary, maintenance and administration staff; and (iv) when dementia training is delivered, it does not usually explore principles relating to personhood or citizenship.
Follow-up measurements
All follow-up measurements were assessed ten weeks post-randomisation to allow time for intervention delivery and for the staff to complete their pledges, a component of the intervention. All measurements taken at baseline, at the level of staff and residents, were repeated. DCM™ inter-rater agreement was assessed at follow-up by JC, and a DCM™ mapper independent of the study, at one intervention and one training as usual site. The second rater was experienced in the application of DCM™ and involved in the training of DCM™ practitioners. The observers took regular breaks to discuss the observation and refer to the DCM™ handbook (University of Bradford, 2010). To ensure a true representation of inter-rater agreement, even if a consensus was reached after discussion, the initial observation data was not altered.
Blinding
The person responsible for randomisation (AA) was blind to care home identity but blinding of allocation was not possible for JC or care home managers and staff. At follow-up, attempts were made to blind the second person undertaking DCM™ to study arm allocation. Staff and resident participants were assigned a code and JC remained unaware of which code represented which study arm during analysis.
Sample size
As the study was not a definitive randomised controlled trial the aim of the analysis was not to determine evidence of effect and therefore not powered as such. To help ensure smooth running of trial procedures, recruitment was limited to forty residents across the six care home sites, this number was derived from a four-hour scoping exercise of DCM™ conducted in a care home separate from phase two. There was no limit on staff recruitment within the participating homes. Care home recruitment was limited to six sites to help ensure meaningful feasibility data could be collected.
Analysis
DCM™ mood and engagement is measured on a -5 to +5 ordinal scale. DCM™ denotes a neutral state of mood or engagement with a +1 score. Frequency of the 23 behaviour category codes in each trial arm was explored by grouping the individual codes by their potential (high, moderate, low or none) for mood or engagement, as described in the DCM™ handbook (University of Bradford, 2010). The no potential for wellbeing behaviour code is ‘N’ when a participant is asleep, or ‘Land of Nod’. The use of personal enhancers and detractors is descriptive of the nature of the staff interaction rather than a numerical measure of interaction quality. There are seventeen personal enhancing staff interaction categories and seventeen detracting staff interaction categories. Usually, these codes are only used when an interaction of interest has taken place. However, for this study to capture the breadth of interactions across all sites, all interactions occurring between staff and resident were recorded. Because of the high number of interactions anticipated from taking this approach the rating of the thirty-four different interactions was condensed to enhancing (+1) and detracting (-1). DCM™ data can be converted into a concordance co-efficient to describe interrater agreement. The concordance co-efficient is percentage agreement between different observers for the behaviour category codes and mood and engagement scores, across the five-minute time frames for which both observers have data.
The PDQ uses a seven-point scale for each of the 20 items. This scale was adjusted to 0-6 to include a zero value. The VAS and PDQ data were reported as a mean for each group and data captured for missing cases and outcome concordance.
Study data was analysed using the Statistical Package for the Social Sciences (version 25). Analysis was predominantly descriptive in accordance with the feasibility design and the study objectives. Staff flow through the study was mapped at baseline, intervention delivery and follow-up to inform decisions on how future trials might weight recruitment.