Challenge Demcare: Management of challenging behaviour in dementia at home and in care homes

Aim

To describe the development and field testing of an interactive online training and decision support intervention, using functional analysis approaches for the management of CB in dementia.

Method

An e-learning course and two decision support e-tools were developed to help staff to use functional analysis-based interventions for up to 25 commonly reported CBs in dementia. The intervention was tested (2011–12) with 92 nominated ‘staff champions’ from 27 care homes and 26 community mental health practitioners from six NHS organisations across England.

Results

The course was well received and strongly recommended by care home staff champions who completed an evaluation sheet (n = 92), but only when this occurred at an external venue, with opportunity for facilitated discussion and practice. Although freely available within homes, e-learning take-up by other staff was limited. Staff selected as champions by their managers were, on average, younger [t(606) = 2.12; p = 0.032], had higher educational attainment (Fisher’s exact test p = 0.0448) and were more likely to have had dementia training (χ² = 4.38; p = 0.036) than others working in the care homes. E-tool-assisted action plans were developed for 199 residents with CB. Aggression was mostly selected by staff where 58 action plans (29%) were delivered. Immediately after training, staff appeared to have expanded the way in which they viewed some behaviour. They were less likely to perceive behaviour as challenging, with a significant reduction in ratings of CB following training [t(178) = 7.4; p < 0.001]. Community mental health practitioners, who tested the community decision support system for their patients with CB, valued its logical assessment framework and the ‘if–then’ algorithmic method for choosing potentially helpful case-specific interventions.

Conclusions

Worksite-based e-learning opportunities are not at present readily taken up by staff working in care homes in England. Computerised decision support for interventions for CB appears premature in care homes, but shows promise for training community dementia practitioners. However, usability will depend on successful collaboration between clinical experts, information technology (IT) advisors within NHS organisations and software engineers.

Development of the intervention

We conceived a multimedia interactive functional analysis-based intervention for CB in dementia. This comprised a training programme, together with two suites of decision support systems (one for staff in care homes and the other for staff supporting family carers in the community) for the targeting of individualised or person-centred interventions for CB in dementia. A range of options for the platform was considered, including our original plan to develop DVD and CD materials. We explored the potential strengths of an online solution, which were as follows: increased accessibility to information, where content could be standardised, easily updated and revised (see Ruiz et al. 87 for an overview of e-learning in education); options for self-pacing to overcome time pressures;88 immediate feedback with options for improvement, which is an important approach for adult learners;89 and fidelity of presentation with automated documentation, such as tracking and reporting of the learner’s activity,73 to allow monitoring of usage and problems with the program itself and thus facilitate improvements to meet the needs of the learner. In addition, feasibility studies in the USA have noted that direct care workers in nursing homes respond positively to internet-based multimedia training, including the management of aggression. 70,71,73 In family care settings, information and communication technology (ICT)-based solutions have been piloted74 and are also described as important uncharted territory in supporting families to cope with CB at home. 90 Given the huge growth of internet use in recent years, we considered that the online option would offer greater flexibility for those, including care home workers and community practitioners, who may wish to access the e-learning aspect of the resource from other venues, including potentially their own home.

The e-learning course was a multimedia, interactive, skills-based method, to encourage comprehensive assessment, such that knowledge about the person with dementia’s behaviour in their environment could be understood in order to provide solutions. These could include signposting or referral to other professionals when needed. Its aim was to encourage staff to systematically consider variables such as the person’s medical status, life story, communication and a host of other ‘unobservables’ that may give clues about the function of the behaviour. The information management systems (IMSs) for the two decision support e-tools were designed to follow on from a ‘functional analysis’ of the behaviour, in selecting approaches that were likely to ameliorate the behaviour, the emotional response to it or both. Three e-learning modules (outlined below) were developed, using actors to simulate common behaviours seen in people with dementia in context. 70,73,91 The learner is required to observe the potential function of the behaviour from observations and knowledge provided about the person’s past and present circumstances and then consider supportive actions within three domains (i.e. health, psychological and caregiving context) to meet their needs. The e-learning course was underpinned by a learning management system that allowed users to manage their learning and user input to be recorded and potentially assessed and used for targeted feedback. The e-tools that followed led to an assessment summary and algorithms for developing an action plan, in each of the three domains, with the third domain being relevant to the context of either a care home or family care setting. These were structured in line with the third e-learning module, to provide functional analysis-based interventions for up to 25 CBs in dementia. The three e-learning modules are as follows:

1. Module 1: an introductory module, introducing person-centred approaches in which CB is seen as a response to a frightening environment or unmet need. This introduced the notion of a shared sense of humanity and affinity with the person, in line with the philosophy underlying person-centred care. 92

2. Module 2: a skill development module, using interactive video-clips to practise observation and interviewing. This allowed staff to move away from the traditional ‘antecedent–behaviour–consequence’ observational approach to that of observing the relationship between emotion and communication through the behaviour,93 in order to enhance emotion-orientated care94 where relevant. The intention was to enable staff to appreciate the ‘language of behaviour’,95 that is, the feelings and intentions of the person with dementia that were being communicated by the behaviour, using observation of real-life practice in the care setting and thus recognise triggers and early warning signs that can then be acted on to prevent escalation of the behaviour into full-blown CB. 96

3. Module 3: this comprised nine case examples of graded complexity, incorporating video-clips, and an interactive procedure for accessing information about the person, relevant to a functional analysis of the behaviour. This was followed by a case-specific intervention summarised in an ‘action plan’ to address unmet need97 and thus reduce CB. Cases were derived from the clinical situation in which some cases have been documented in the literature. 1,13,68,69,98 Learners were guided through interactive video-clips, resources that are relevant to the person with dementia, such as life story information or case notes, audio-clips on views of others within the care context, and then provided with structured feedback to set their observations of the person’s CB in the context of ‘causation’. 49 The learning focused on guiding staff to use ‘why’ questions63 to consider the potentially multiple causes for what they had observed about the person and the behaviour in the vignette. 13 Through using information about the person’s current health and functional status, their life story and how others respond to the person with dementia during an episode of CB, key concepts for managing CB13,99 were systematically covered to address causation and possible methods of remediation. Three groups of ‘actions’ or interventions were designed and tested in the clinical situation, including those to address unmet somatic and psychological need97,100 in the person with dementia, as well as the caregiving environment, in which consideration of the particular needs of the staff group or family carer were considered. 86,101 These we refer to as functional analysis-based interventions for the management of CB in dementia. These groups were (1) ‘actions to support health needs’, for example signposting for help by the GP, to alleviate pain or discomfort due to constipation, or review of medications, such as antipsychotics or sedatives that may have been overlooked;100 (2) ‘actions to meet the psychological need of the person, such as how to support the person who may be surrounded by a sense of ‘disorder’, perhaps trying to escape from this, or feeling ‘trapped’ or ‘set aside’;102 and (3) ‘actions to support the caregiving context, environment or system’,101 such as accessing support for care home staff or psychological therapy for the family carer. Thus, in addition to the commonly advocated biopsychosocial approach for the management of CB in dementia care,103 our system included interventions to address contextual needs such as those of people providing care, which can be associated with CB,49,86 and are often overlooked (see Chapter 1).

The decision support e-tools were based on clinical practice guidelines76,104 with our assessment questions firmly set within Dementia: A NICE–SCIE Guideline on Supporting People with Dementia and their Carers in Health and Social Care for ‘behaviours that challenge’ in dementia. 12,105 The questions that were used to provide the ‘assessment summary’ are found in Appendix 1. We used case-based reasoning,106 in which information collected about the person with dementia and their setting is stored and can be used as a knowledge source to develop personalised interventions for particular problems. The case-based reasoning method then flows into the action-planning part, in which the individual’s assessment is collated and ‘if–then’ logic is applied to generate choices for the practitioner to consider. The stored information can be used to work on different problems, as a setting-specific (care homes and family care settings) CB checklist is supplied at the beginning. The information source can also be updated; for example, if the person has a change in their health or if they have recently experienced a bereavement or if new knowledge of a person’s life story or personal ways of managing life is discovered. Our e-tool started with the 25-item Challenging Behaviour Scale (CBS) of commonly reported behaviours in care homes107 for staff to select a behaviour to work on. This scale has been recommended as a helpful measure for use in care homes across the UK (see Appendix 2 in Brechin et al. 108). For community settings, we used the Problem Checklist,109 as this was based on the concerns of UK family carers and had been used effectively in our previous study. 65,109 This was seen as important, as many guidelines for the management of aggression, for example, have poor representation of some of the important individualised and contextual characteristics necessary for management. 110 Having chosen the person’s behaviour to work on, the staff member or practitioner is then asked to input selective information about the person with dementia and CB, as they had discovered was necessary in module 3 of the e-learning course. They are guided through the assessment process (see Appendix 1 for a detailed description), concluding with a printable summary of the individual’s assessment. This summary includes documentation of the potential cause(s) or function(s) of the behaviour that was based on the staff member’s or practitioner’s responses (see Appendix 1, Box 7). 65,109 The practitioner is thus provided with a systematic assessment framework where questions they considered covered important aspects of the person’s life story, health status and other factors13 (see also Appendix 1). The next part of the decision support system is the ‘algorithmic’ flow of the information into considering what interventions can be tried. The IMS was conceived to use algorithms (which we tested in the clinical situation using a paper workbook), using the aforementioned ‘if–then’ logic to provide options for treatment – in this case functional analysis-based interventions for given behaviour that is seen as challenging. The interventions in our system are structured within the three action groups that had been used in module 3. These are (1) support for health need, such as considering the effects of commonly encountered medical conditions or the effects and side effects of psychotropic and other medication; (2) support for psychological need, such as needs for reassurance, privacy, comfort and occupation or activity; and (3) support for contextual needs, such as advice on how to optimise the environment or system around the person. For each suggested action, tailored information is provided on what staff can do themselves and when and whom to access for further help. Many of the options for intervention for the first two groups of interventions are outlined in Dementia: A NICE–SCIE Guideline on Supporting People with Dementia and their Carers in Health and Social Care. 12 However, a structured approach to making decisions about treatment based on the particular cause(s) underlying CB (functional analysis) has been absent to date. The algorithms used for the decision support enable the care home staff or community mental health practitioner to apply a structured methodology to choose a set of actions that are possible for them to try within the resources available to them in their routine practice settings. The logic for the case-based reasoning, using tailored assessments, and the algorithms for action plans was developed by the co-chief investigator (BW). It was tested by the chief investigator (EM-C) using a paper workbook with 19 residents who had a score of 4 or more (eight residents) or 10 or more (11 residents) on the CBS. 107 These residents were drawn from eight care homes that were not involved in the CRT (see Chapter 3) and one inpatient dementia unit. All of the 25 items of CB were covered when testing the logic, and actions were refined at this stage. A separate paper workbook for family settings was tested with 15 cases of clinically significant CB determined by a score of 5 and above on a widely used research tool – the Revised Memory and Behaviour Problems Checklist (RMBPC). 111 The practitioner was provided with the Problem Checklist,109 as this was based on the concerns of UK family carers. Action suggestions were further refined to add opportunities that were available in NHS contexts, during our feasibility test with 26 community practitioners from six NHS organisations. The software engineers who were employed to develop the IMS were required to design the system to allow clinical experts (such as physicians, psychiatrists, psychologists and pharmacists) to add content for the functional analysis-based interventions, described within three action plan groupings. Thus, the utility of the action-planning component was that of flexibility for new actions that could be added on an ongoing basis to the system, as experience with the system grew.

Summary of the Challenge Demcare intervention

We used three modules of e-learning to introduce care staff to observational skills and the algorithmic approach to interventions comprising the first two components of the decision support e-tool. This required the practitioner, working with the care staff and the family, to collect important information on key contributory factors associated with CB (see Dementia: A NICE–SCIE Guideline on Supporting People with Dementia and their Carers in Health and Social Care, section 8.6.1.1 on p. 260 and section 8.6.3.1 on pp. 262–312), such as the person’s current health and functional status, their life story, interpersonal and communication style and how others respond to the person during an episode of CB. The decision support system comprised relevant assessment tools and systems to collect this information with algorithms to provide two sets of biopsychosocial groups of ‘action plans’ where practitioners could choose the most relevant way to meet the person’s health or psychosocial need. These are referred to as functional analysis-based interventions, as the approach was to assist practitioners to assess, analyse and then choose the most appropriate set of interventions (‘actions’) for a given episode of CB. Actions for these two components were extracted from the literature, including our overview in Chapter 1, the Cochrane review,82 guidance from the International Psychogeriatric Association (see Complete Guide to Behavioral and Psychological Symptoms of Dementia),36 the range of interventions for CB from Dementia: A NICE–SCIE Guideline on Supporting People with Dementia and their Carers in Health and Social Care12 and policy initiatives to reduce the use of antipsychotics in dementia. 112

Interventions included in the third component of the decision support tool were not related to the person’s unmet need, but arose from a new concept derived from our overview (see Chapter 1), relating to the needs of the caregiving system. These reflected, for example, the needs of the family carer for those living at home (such as counselling or skills training) and those of staff or the environment (such as training, skills enhancement or altering lighting or sound in a room). Algorithms for this component of the interventions were therefore bespoke to the care home or family care setting, where the focus was the need of the caregiving environment or system.

The e-learning course in care homes

The training, consisting of the three e-learning modules, was planned to occur within the care home, to facilitate ease of access by all staff who wished to engage with this. As noted by researchers in Canada113 and the Netherlands,104 we too found very early on in the process that the majority of the care homes did not have adequate computer equipment to offer online training. For example, most had just one or two computers in the home for sole use by the manager and/or administrator. Therefore, a lengthy process of installing the necessary additional equipment to enable access to the internet ensued. We supplied and installed computers in 19 of the 27 experimental homes (14 laptops and five desktop computers), and provided 16 printers. As noted by researchers in the USA when implementing internet-based training for nursing assistants, buildings were also not always designed to accommodate computers and the internet, in terms of either wiring or space. 73 We arranged for broadband installation in three homes and new telephone lines for broadband in two homes. On one occasion, a wireless internet booster had to be installed in a resident’s bedroom and, to activate the wireless access, the router (located underneath the manager’s desk) had to be switched on each time the staff required to access training. In some homes we provided memory upgrades to the existing computer, wireless access points, power extension cables and headphones to allow staff to access the interactive program without disturbing others. Apart from access to equipment and the internet, organisational obstacles had to be overcome to facilitate staff access to the e-learning. These included support from the research team when home managers were unable to ascertain whether or not they had internet access, obtaining permissions from organisation head offices where their policies would not allow for internet use in the home and negotiating with IT departments where technical support for care homes was outsourced.

Once access to the technology required was in place, as with the US and Canadian research teams,73,113 a specialist dementia care therapist needed to work with individual staff in the first three care homes to assist them with logging in and troubleshooting user errors or computer problems. We additionally offered homes money to backfill staff in order that others could complete the training. Thus, we overcame some of the impeding factors to e-learning in the care home setting, that is, a lack of ready access to computer equipment or internet services and low-speed connections that could detract from the utility of the programs, and staff confidence and motivation to use the training in their work environment. However, despite everyone’s best efforts, our strategy for in-home access to training was not successful because of interruptions, when staff felt that they were needed to help with care tasks or when they felt guilty because they thought that their colleagues might be busy or struggling. An alternative strategy was then used following discussion with care home managers, who nominated staff champions to undertake training, with the intention that they would then support other staff in delivery of support to residents with CB. As noted by others in this field,70,73 we needed to arrange training at an external venue where computer suites were available, to allow staff protected time to complete training within small facilitated group sessions. Funding for travel and time for backfill at the care home was made available. However, training facilities within small groups were still not used to capacity, as it was difficult at times for some homes to release staff to participate. Nonetheless, this worked much better than our attempts at offering in-home access to e-learning, as staff champions had their own uninterrupted learning time, with access to the specialist dementia care therapist and the opportunity to share and discuss with staff from other homes.

The decision support e-tool in care homes

The technological underpinning for, and functionality of, the decision support software proved unreliable, because of what may have been computer coding and software design problems with the IMS. The case-based aspect of the system was mostly, but not always, successful in producing case-specific assessment summaries, although these required some editing for presentation. The rule-based aspect was less successful and the range of actions built into the system remained smaller than desirable. Moreover, the IMS did not allow clinical experts to add management content on an ongoing basis. It had always been envisaged that input from a specialist dementia care therapist, such as a CMHN or a psychologist, would be required to assist with the care home intervention and support the action plans. With the limited range of actions available from the decision support e-tool, this was indeed the case. Therefore, a specialist dementia care therapist checked and, when necessary, edited the assessment summaries produced by the system using the rule-based logic we developed, and worked with staff champions to provide an action plan that the champions considered feasible to deliver in their home.

The decision support e-tool in the community

A second software company was employed to work on the community decision support e-tool. Community practitioners from six NHS organisations were selected by their managers to deliver the intervention. They had access to the e-learning course and were additionally trained in a small group setting by clinical experts from the research team, using relevant cases from module 3 of the e-learning course. This was supplemented by education about the rationale for case-specific functional analysis-based intervention, use of a training manual comprising tools and resources necessary for functional analysis in people with dementia and CB living at home (see Appendix 1, Box 9), and ‘hands-on’ practice with the community e-tool. For this case practice they used the e-tool, with anonymised cases from their own experiences, and engaged in facilitated group discussion with the clinical expert team and each other. These discussions focused on how to use the functional analysis for the anonymised case to develop action plans that were feasible to deliver in their own NHS context and its local resources.

Unlike the care home staff, all community mental health practitioners had access to computers and IT facilities within their NHS organisations. For them, the use of computers was a routine part of their job. Overall, the case-specific aspect leading to the assessment summary was superior to that of the care home e-tool, in terms of functionality, presentation and navigation; that is, the assessment summary was of the quality that was envisaged by the clinical expert research team. However, the software engineers did not deliver a system that was fully populated with actions or one that allowed flexibility for new actions to be added by the clinical experts. Therefore, the community mental health practitioners were provided with our manualised resources, including validated tools to assess common CBs in family settings, and other relevant contributory factors such as pain or discomfort in people with dementia or other family concerns (see Appendix 1, Box 9). This provided them with a workbook, to assist them in adopting a systematic functional analysis-based approach to individualised interventions for clinically significant CB in dementia within family care settings.

The e-learning course in care homes

Ninety-two staff champions across 27 care homes that constituted the experimental arm of the ResCare trial (see Chapter 3) were trained on the e-learning course: 10 managers/deputy managers, 36 senior care assistants, 45 care assistants and one administrator. Of these, seven staff completed the online training within their care home setting and 85 attended sessions over 1.5 days. These occurred at a training centre with a large computer suite. As with e-learning at the care home, this was facilitated by the specialist dementia care therapist, but it additionally offered staff the opportunity for group discussion and demonstration of the e-tool using anonymous cases of residents with dementia and CB from their own care home. In total, 11 training sessions were held outside the home over a 12-month time period (June 2011–June 2012). Attendance ranged from 5 to 10 staff for each cohort (average of eight staff per group).

In all homes the manager agreed to select at least two staff champions for training. Two of the homes were unable to meet this requirement. The number of staff who attended training from the homes varied, ranging from one to nine. Managers were asked to select ‘appropriate’ staff for the training, that is, staff who were involved in care planning for residents with CB and dementia and, to the best of the manager’s knowledge, were likely to be working in that home regularly for the foreseeable future. However, those who ultimately attended the training were not always the most appropriate, as was the case of an administrator who had no involvement in resident care and was not sure why they had been selected by the care home manager.

Characteristics of staff champions

The majority of participating care home staff across both arms of the ResCare trial were female (89.7%), and those designated as champions showed a similar gender balance to the overall pattern, with 89.1% being female. However, staff champions were slightly younger in age [mean 36.3 years, standard deviation (SD) 11.8 years] than other staff (mean 39.6 years, SD 13.3 years), that is, those not trained in experimental homes and all the staff in control homes. The difference in age was statistically significant [t(606) = 2.12; p = 0.032].

Figure 1 shows the age groups of the champions compared with the ‘other’ non-champion staff members in the ResCare trial, with differences particularly evident in the proportion of people aged ≥ 55 years and between 25 years and 34 years.

FIGURE 1.

Age distribution of care home staff selected as champions (n = 92) compared with other staff working in the homes.

All staff reported whether or not they had received previous training in dementia care or in the use of computers and provided details of their highest qualification. Tables 1 and 2 summarise these data, comparing staff selected as champions with the rest in the ResCare trial. The majority (n = 585) provided most of the details requested.

TABLE 1 - Dementia and computer training: champions compared with other care home staff

Previous training	Staff				Chi-squared test	p-value	All staff (N = 564)
	Champions (N = 77)		Other (N = 487)
	n	%	n	%			n	%
Previous dementia training
Yes	62	80.5	331	68.0	4.38	0.036	393	69.7
No	15	19.5	156	32.0			171	30.3
Previous computer training
Yes	32	41.6	171	35.1	0.94	0.333	203	36.0
No	45	58.4	316	64.9			361	64.0

TABLE 2 - Highest qualification: champions compared with other care home staff

A Level, Advanced Level; AS, Advanced Subsidiary; BA, Bachelor of Arts; BSc, Bachelor of Science; BTEC, Business and Technology Education Council; CSE, Certificate of Secondary Education; GCSE, General Certificate of Secondary Education; GNVQ, General National Vocational Qualification; HNC, Higher National Certificate; HND, Higher National Diploma; MA, Masters; NVQ, National Vocational Qualification; O Level, Ordinary Level; ONC, Ordinary National Certificate; OND, Ordinary National Diploma; PGDE, Postgraduate Diploma in Education; PhD, Doctor of Philosophy; RSA, The Royal Society of Arts; VCE, Vocational Certificate of Education.

Qualification level	Staff				All staff (N = 557)
	Champions (N = 76)		Other (N = 481)
	n	%	n	%	n	%
No qualifications	3	3.9	25	5.2	28	5.0
One to four O Levels/CSEs/GCSEs (any grades), entry level, Foundation Diploma, NVQ level 1, Foundation GNVQ, basic skills	6	7.9	68	14.1	74	13.3
Five or more O Levels (passes)/CSEs (grade 1)/GCSEs (grades A*–C), School Certificate, one A Level/two or three AS Levels/VCEs, Higher Diploma, NVQ level 2, Intermediate GNVQ, City & Guilds Craft, BTEC First/General Diploma, RSA Diploma	24	31.6	175	36.4	199	35.7
Apprenticeship, two or more A Levels/VCEs, four or more AS Levels, Higher School Certificate, Progression/Advanced Diploma, NVQ level 3, Advanced GNVQ, City & Guilds Advanced Craft, ONC, OND, BTEC National, RSA Advanced Diploma	29	38.2	139	28.9	168	30.2
Degree (e.g. BA, BSc), higher degree (e.g. MA, PhD, PGCE), NVQ level 4 or 5, HNC, HND, RSA Higher Diploma, BTEC Higher Level	10	13.2	36	7.5	46	8.3
Professional qualifications (e.g. teaching, nursing, accountancy)	3	3.9	26	5.4	29	5.2
Other	1	1.3	12	2.5	13	2.3

Champions were significantly more likely to have had previous dementia training (χ² = 4.38; p = 0.036), but were no more likely to have had computer training. As seen in Table 2, champions were more likely to have reached a higher educational level than other staff. Over half the champions (55.3%) had the equivalent of two Advanced (A) Levels or a National Vocational Qualification (NVQ) of level 3 or above, compared with 41.8% of the non-champions. This difference (higher vs. lower educational attainment) is statistically significant (Fisher’s exact test, p = 0.0448).

It may be that some managers’ decisions in selecting staff to receive training and to oversee the management of residents with CB may have been influenced by their perceptions of strengthening capability for future leaders within the home, by selecting younger staff, with higher educational qualifications and previous dementia care training.

Effects of staff training on reports of challenging behaviour in care home residents

Of the 92 staff trained on the e-learning course, 83 provided baseline CBS data for 78 residents with dementia and CB in the ResCare trial. Of these, 55 staff champions subsequently provided follow-up data for 112 residents with dementia and CB. Those staff champions who did not provide follow-up data had not left the care home: in three cases, the care home itself had withdrawn from the study and 25 champions were simply not available at the time of data collection (e.g. were on holiday, off sick or had switched to night shift).

Most CBs were reported to have declined between baseline and after staff training; for example, perseveration was recorded as a problem for 61.1% of residents at baseline, but for only 24.4% of residents following staff training, and levels of this behaviour remained lower at follow-up (reported in 46.1% of residents). Restlessness and lack of self-care also reduced following staff training, but this reduction was not maintained at follow-up. In contrast, small increases in certain behaviours, namely physical and verbal aggression, self-harm and inappropriate sexual behaviour, were reported (see Table 5).

All residents with action plans and for whom questionnaires were completed at all three time points were included in an analysis of change over time. Residents’ CB incidence scores, as measured on the CBS, decreased directly after staff training but returned to the baseline level by follow-up (Figure 2). The difference between baseline and post-staff training was significant [t(178) = 7.4; p < 0.001] and the difference between post-staff training and follow-up was also significant [t(178) = 7.6; p < 0.001].

FIGURE 2.

Average CBS incidence scores over three time points in care homes. Total incidence score = the total number of CBs (score range 0–25).

Uptake of the e-learning course after completion of the ResCare trial

Following completion of the ResCare trial, managers at 54 of the study care homes were sent a letter thanking them and their staff for their help. Homes were not sent a letter if they had dropped out of the study or had closed down. In this letter managers were offered free licences for the e-learning course; those in the control homes were offered up to a maximum of five free licences and the experimental homes were offered up to three. Only 30% (n = 16) of homes requested any free licences. In control homes, between three and five licences were requested, and in experimental homes this number ranged from one to three.

In total, we provided 60 free licences (April 2013) to care homes in the study. Sixteen weeks later, data extracted from the learning management system showed that only 12 staff from six homes had activated their licence and started training. Of these, only six accessed training for > 1 hour in total, and a further six accessed training for < 5 minutes. The majority of staff trained during the ResCare trial took around 5 hours in total to complete the three modules of the e-learning course. Therefore, staff who used the e-learning for < 1 hour are unlikely to have completed more than the first module. Once a licence is activated, training can be accessed online for 3 months. Users who successfully complete all three modules are issued with a certificate, which can be downloaded and printed off. Only two staff members completed the full e-learning course in the 16 weeks from the time the licences were provided. They were both from the same care home; one user took 3 hours 35 minutes and the other just under 5 hours to complete the course.

The low take-up rate of free e-learning for functional analysis-based training in dementia and CB is disappointing, particularly as the learning styles analysis showed that the type of training being offered through our e-learning course is compatible with the dominant preferred (i.e. visual) learning style in care home staff. Given that all those who completed an evaluation form following the e-learning said that they would recommend the training package, it was surprising that only 30% of the homes requested the free licences on offer. This perhaps confirms our initial experience that e-learning is hard to implement in care home environments, where staff have other duties to consider and practice development may not be the norm.

The decision support e-tool in care homes

Action plans

Following the training of staff champions using the e-learning course, three research therapists, including the specialist dementia care therapist, then worked with the care home staff on the development of functional analysis-based interventions for 199 residents with dementia and CB (defined as having a score of ≥ 4 on the 25-item CBS107 at baseline data collection). Residents were living in 26 intervention homes, in which an action plan was developed for an average of 7.65 residents (range 2–18 residents). Initially, there were 28 intervention homes; however, one home declined the intervention but remained in the study and another withdrew prior to the intervention. As seen in Figure 3, the main reasons for not providing action plans were that the resident had died or that the care home had either declined the intervention or withdrawn from the study.

FIGURE 3.

Challenging behaviour in residents within the intervention group. a, Two homes withdrew from the study (one before action-planning and one after); b, one intervention home declined the intervention, but contributed to follow-up assessment.

Research therapists gathered the information required about residents using the decision support e-tool. The specialist dementia care therapist was responsible for developing and checking the feasibility of the action plans to address CB for each resident with the relevant staff champion. The other two research therapists assisted with the gathering and inputting of information for populating the action plans in 64% of the 199 residents for whom care plans were developed, but they were not involved in communicating with staff champions in the final development of these. Twenty per cent of the resident action plans were checked and refined, if necessary, by one of the clinical experts from the research team.

Excluding travel time, the research therapists spent 158 hours with care home staff champions (n = 49) gathering information required for the development of action plans. Therapists spent a further 273 hours inputting the information into the e-tool, most of which was done away from the care home. The specialist dementia care therapist then spent an additional 464 hours enhancing the action plans that were produced via the system for 199 residents, and an additional 69 hours with care staff (n = 36) going through them. Additionally, a further 33 hours of resource was provided to homes (n = 17), in the form of ‘booster’ visits or telephone calls, whereby the specialist dementia care therapist provided extra support or information to staff in relation to action plans, for approximately 65% of residents (n = 139). Across all tasks involved in the development of the action plans, 101 hours of research therapist travel time was also incurred, with an additional 8 hours spent on visits to seven of the care homes at various points being unproductive because care home staff were not available for pre-arranged meetings. Excluding travel time and unproductive visits, research therapists spent a total of 997 hours on the generation and delivery of the action plans. This equates to an average of 5 hours per action plan produced. The action-planning aspect of the intervention may be described as ‘low intensity’,119 in terms of specialist input to the care home, particularly as significant time was spent interacting with the technology. Higher-intensity interventions were described in our functional analysis-based interventions review,82 in which just three studies in the care home setting were found.

The decision support e-tool in the community

This version of the decision support e-tool was extensively tested for acceptability with 26 community mental health practitioners between November and December 2011. They included CMHNs and others from occupational therapy, social work and clinical psychology disciplines, across six specialist community mental health NHS organisations in England. Prior to use of the e-tool, practitioners practised its logic using the e-learning. They used the system with ‘anonymised cases’, in that they adapted information from their current case to practice, but then ensured that for data protection reasons they did not include any personal identifiable details.

In most cases, but not all, training occurred at the host NHS sponsor site in Yorkshire, allowing practitioners from different parts of the country to discuss their experiences of providing care to families supporting people with CB and dementia at home. Connectivity fluctuated across training sessions, depending on internet speed at a given time or NHS location. For example, at one NHS trust, a thunderstorm delayed use and overall speed was significantly slower than at the sponsor’s NHS site; and, at the sponsor site, practitioners reported varying difficulties with connectivity and speed when using the e-learning course at their own team location. Many practitioners reported that computer and internet use in their routine practice was a predictable pressure on their work, stating connectivity and slow speeds as some common difficulties. Overall, these internet experiences with the e-tool were not perceived as an obstacle to the delivery of computer-assisted functional analysis-based interventions for their patients with dementia and CB, by community mental health practitioners.

One practitioner felt that use of the e-tool did not add anything to their own practice:

Feels like it’s what I do a lot of already without technology to help me – but it’s a different way of formulating. Glad it may help all therapists to work in the same way.

Occupational therapist 1

Overall, practitioners found the e-tool easy to use and understandable, and comments were positive:

Makes you think differently using functional analysis.

CMHN 1

This will help families to feel listened to and give therapists the tools to help decide what needs to be focussed on.

Psychologist

Like the assessment summary as it gives the key relevant information all in one place rather than trawling back through notes.

Occupational therapist 2

Glad it’s not lots of paperwork and glad it’s not just limited to medical interventions. Keen to reduce reliance on meds that are not needed.

CMHTOP practitioner

Think I can use the videos and audio to help families see that they are not alone and we can work together.

CMHN 2

Think the tool will fit in smoothly with work I do already.

CMHN 3

It’s better than I expected, pulls all the information together, action plan printable. I like the format and that the scoring is done for you.

CMHN 4

I like that it’s planning specific and evidenced.

CMHN 5

The behaviour checklist is helpful carer’s can complete whilst waiting and we can use it with the carer to decide how to help.

Psychologist

The behaviour checklist is better than what we use . . . and its evidenced.

CMHN 5

This version of the e-tool was subsequently adapted to take into account the practitioners’ suggestions for improvements, mainly aesthetics and clarifications on wording rather than content. A limitation to our evaluation was that, although participants were trained in a small group setting by clinical experts from the research team, using relevant cases from module 3, we did not take formal data on this aspect of the training. However, our previous participatory workshops with teams (see Chapter 5) to introduce the study suggested that practitioners were enthusiastic about the video material presented of real-life situations in dementia care. Another limitation was that usability of this computerised system could not be fully piloted in routine practice for two main reasons. First, we were impeded by the time required to make arrangements with clinical information governance systems within individual NHS organisations. They were under pressure to develop their own online clinical decision support tool,120 to train practitioners to cluster patients in groups in preparation for the Payment by Results (PbR) initiative (see Chapter 5), which is now mandated for all mental health NHS organisations in England. 121,122 In particular, as was noted by another study of online decision support software for mental health conditions,123 we had yet to resolve complications such as security and privacy concerns with each NHS organisation. Second, the software company was provided with a database of expert interventions, but there were delays in full population of the system. Because of these impeding factors, although we fully investigated acceptability, we were unable to conduct a feasibility study using the e-tool in the NHS within the time scale of our programme. As noted earlier (see Methods), it was not possible for our clinical expert team to update and manage the content of the e-tool, in order to reflect knowledge or resource updates relevant to action-planning.

However, some practitioners tested feasibility of our protocol to deliver functional analysis-based interventions (biopsychosocial) for dementia with CB, within their own work with their patients and families. During our consultations with them and their managers (March–May 2011, see Chapter 5) many had realised that they had been overlooking potential cases for the functional analysis–based intervention as a result of their misunderstandings about the nature of dementia and CB in family settings. This was at a time of significant redesign of specialist dementia services across the NHS in England (see Chapter 5). Consequently, some NHS organisations involved their memory clinic services in order to locate families that might be suitable for the FamCare study. All CMHTsOP and memory clinics involved in the FamCare study had access to clinical expertise with respect to medical (health) and psychosocial (the person with dementia and the family) needs that were identified by the trained practitioner, through use of the functional analysis (see Appendix 1, Box 9). Although practitioners did not have access to the e-tool, paper copies were included in the therapist manual that was supplied during training.

Limitations

A limitation to our approach is that, apart from the paper workbook-based clinical checks of our e-tool, it was not possible in the project time scale to validate the e-tool itself outside the CRT (see Chapter 3). Reasons for time scale delays are documented in this chapter (see Changes to protocol: the ResCare trial and Changes to protocol: the FamCare trial). In addition, it was also not possible to implement our original plan to build into the action-planning cycle an iterative process that would enable staff to refine and modify a resident’s action plan based on their findings from trying out the first set of actions. According to the principles of a functional analysis approach to understanding behaviour,1 and our modelling of this in clinical practice,13,68,69 an iterative ‘detective’ approach to problem-solving is usually required. Nor did our time scale allow staff to receive assistance from the specialist dementia care therapist to target the range of multiple behaviours, as often there are a number of behaviours that cause distress for the person living with dementia and those supporting them. Moreover, neither the e-tool nor the specialist dementia care therapist monitored detailed implementation of the action plan in the care homes. For example, although clear strategies on contextual issues in the care home, such as assigning responsibility across shifts to ensure that all staff took a consistent approach and time scales for specific aspects of the action plan were documented, these were not actively monitored. A further limitation is that if the success (or not) of the action plan is not reviewed then poor action plans may be implemented in the future. The assumption underlying the decision support e-tool was that encouraging staff to ask ‘why’, to seek fuller understanding of individuals and their behaviour, would lead to change in staff behaviour and perception. Although the training may have contributed to a change in perception, the next stage – of recording actions undertaken and the response – was not part of the e-tool as designed here. The intention was that this aspect should fit with systems already in place in the care home, but this may be a limitation of our approach.

Implications for the design of the care home (ResCare) and family care (FamCare) studies: major design changes

Aim

To evaluate the clinical effectiveness and cost-effectiveness relative to usual care of an online application to enable care home staff to understand the function of CB in people with dementia and support them accordingly.

Design and methods

A CRT that allocated 63 care homes in Yorkshire between intervention and usual care. The primary outcome was the well-validated NPI using frequency and severity scores taken at 4 months to address the primary research question, that is, whether or not the intervention reduces CB in dementia. Secondary outcome measures (n = 21) monitored both residents and staff and included resident quality of life using the EuroQol-5 Dimensions (EQ-5D) self-reported and proxy versions. The basic statistical model for effectiveness analysed follow-up scores by treatment group; corresponding baseline scores; and other covariates for both resident and care home. Resources used by residents with CB were costed by adapting the Client Service Receipt Inventory (CSRI) to focus on health and social care over 4 months; and assuming no marginal change in care home resources, as these are less likely to change and more difficult to cost.

Results

Eight hundred and thirty-two residents (555 with CB) and 609 care staff at baseline were reduced to 658 (79%) residents [428 (77%) with CB] and 436 staff (72%) at follow-up. No participants reported any serious adverse reactions (SARs) to the online intervention or to analogous activities in the control group. Our primary outcome measure, the NPI, showed that the intervention reduced the frequency of NPSs by 0.60 relative to treatment as usual; however, as the 95% confidence interval (CI) from –1.18 to 2.38 includes zero, this finding was not statistically significant. Although the intervention also reduced the severity of those symptoms by 0.45 (95% CI –1.03 to 1.93), this also lacked statistical significance. Although 14 of the 21 secondary outcome measures showed positive effects of the intervention, none reached statistical significance. Furthermore, the intervention generated little change in the prescription of drugs relevant to dementia – notably antipsychotics (χ² > 0.999), antidepressants (χ² = 0.635), hypnotics and anxiolytics (χ² = 0.215), anticonvulsants (χ² > 0.999) and the anti-dementia (such as the cholinesterase inhibitors) drugs (χ² > 0.999) or those for pain relief – that is, the opioids (χ² = 0.399) and non-opioids (χ² = 0.996). Hence, there is no evidence that the intervention changed the care of CB in dementia. Health- and social-care costs over 4 months did not differ significantly between groups (mean cost was £331 less in the intervention group, with the bootstrapped 95% CI from –£927 to £272) and staff reports of quality-adjusted life-years (QALYs) over 4 months differed little between groups. Hence, there is no evidence that the online intervention was cost-effective.

Conclusions

This computer-assisted intervention was neither clinically effective nor cost-effective. It comprised a comprehensive interactive e-learning program, delivered with group discussion to facilitate case-specific management of CB in dementia, and was followed by assisted decision support to provide case-specific interventions for residents with dementia and clinically significant CB. It was not enough to reduce CB in dementia in care homes.

Trial registration

The International Standard Randomised Controlled Trial Number (ISRCTN) is 02553381.

Research questions

• Does the experimental intervention described in Chapter 2 reduce CB in residents with dementia, as measured by the frequency and severity scores of the NPI?

• Does the experimental intervention improve the experience of staff and the quality of life of residents in care homes?

• Does the experimental intervention reduce the prescription of psychotropic medications and change the usage of medications, such as analgesia for management of pain and laxatives for the management of bowel movements?

• Is the experimental intervention cost-effective?

Design

This pragmatic CRT allocated participating care homes to two groups. The experimental (intervention) group received a functional analysis-based intervention and the control group delivered ‘treatment as usual’. Randomisation at care home level ensured that all recruited residents and care staff in each home were allocated to either experimental or control group. Following consent procedures at a care home, a Registered Clinical Trials Unit, the North Wales Organisation for Randomised Trials in Health (NWORTH), undertook computerised randomisation remotely. The researchers who assessed participants were blind to treatment allocation at baseline and follow-up.

Governance and study approvals

The York Research Ethics Committee (REC; reference number 09/H1311/29) approved this study in May 2009 (see Appendix 4). Details of amendments that were approved during the trial by the REC are also found in Appendix 4. In April 2009 the Research and Development (R&D) Department at Humber NHS Foundation Trust confirmed its sponsorship of the trial and gave local research governance approval. Appendix 5 describes the management arrangements for this study, including patient and public involvement.

Changes to the protocol

Chapter 2 outlined reasons for changes to the ResCare trial, namely the decision not to go ahead with a second follow-up assessment at 12 months.

The intervention

Ninety-two nominated champions from homes allocated to the experimental intervention received e-learning (functional analysis training), usually away from their care homes. The purpose was to teach them ‘detective’-like skills to improve their understanding of the ‘language’95 of common behaviours that were seen as challenging. By using simulated case studies of varying complexity, filmed in a care home setting, this course aimed to change or expand the way staff perceived and responded to resident behaviour in their own care home. Staff champions also worked with a specialist dementia care therapist, who used a decision support e-tool to develop action plans for a particular behaviour that was identified by staff as challenging (see Chapter 2, Table 4). The ‘usual’ long-term care services and interventions that were available to residents in control homes were also available to those in the experimental homes, as we were evaluating the additional benefit of our e-learning course and decision support e-tool-assisted intervention. Both control and experimental homes therefore had access to other training materials and courses, as they would normally.

Study population

In July 2010 we invited all care homes in East Yorkshire, Hull and York with 25 or more beds (including nursing care beds if available) listed on the Care Quality Commission (CQC) website (URL: www.cqc.org.uk/) as old-age care homes to take part in the study. Based on data from a previous study,107 homes with 25 beds or over were judged to be the smallest care home sample that could be included to achieve the numbers of people with dementia and CB required per home to meet the sample size calculations of this CRT (see Sample size). To ensure consistency in quality of care across the homes for this trial, we invited only those that the CQC had rated as ‘good’ or excellent’ at the time. All types of homes were invited, including those managed by local authorities, regional and national commercial chains and the not-for-profit sector and small, independently run private homes. Recruitment of care homes began in October 2010, with participants recruited between March 2011 and March 2012. Follow-up was completed in October 2012.

Eligibility criteria for residents

Sample size

Originally we planned to recruit 48 homes, expecting to recruit an average of 13 CB participants per home and 624 in total. Mozley et al. ,131 in a study carried out in the same locality as the ResCare trial, reported a 39% loss of care home participants over 12 months because of death, moving elsewhere or illness too severe for follow-up. So we expected to lose two recruited CB residents (15%) per home over 4 months, yielding a completed sample of 48 × 11 = 528 participants. Re-analysis of data from previous studies suggested that intracluster correlation coefficients (ICCs) rarely exceed 0.03. 131 We therefore judged that the ICC would not exceed 0.03, and calculated that 528 CB completers would provide the same power as a simple random sample of 528/[1 + 0.03 × (11 – 1)] = 406. Our original design had > 80% (85%) power using a significance level of 5% to detect a difference in NPI scores equal to 0.3 of their population SD. Though this difference is sometimes regarded as ‘small’, our research team and Programme Steering Committee (PSC) judged that it was both clinically important and feasible. However, as the number of CB residents per home was smaller than expected, we recruited 63 homes to the trial with the aim of achieving 63 × 7 = 441 CB completers, equivalent to a simple random sample of 441/[1 + 0.03 × (7 – 1)] = 374. Using the increased number of care homes with the decreased number of residents in each care home still achieves the > 80% (82%) power of the initial design to detect an effect size of 0.3.

Recruitment procedures

When appropriate, we approached the organisation owning the home for support and then contacted the home manager to discuss the trial, including the implications of participating. We needed to recruit care staff and residents through participating care homes. At this stage, 63 homes provided us with anonymised data on CB, using the 25-item CBS incidence domain,107 for all their residents aged > 65 years (n = 2185). This was in order for us to establish homes where CB was highly prevalent; these cross-sectional data are referred to as the ‘screening data’. Next we began the process of seeking consent for all homes simultaneously in order to seek permissions for residents who lacked capacity to consent to participate in the study. Care home managers introduced research team members to staff, who introduced them to residents. If residents were happy to talk, the researcher explained the study to potential participants, provided information sheets and answered their questions.

Care homes then sent letters advising each resident’s main relative about the research, asking them to inform the home manager or research team if they had any objections to being contacted by us. For residents without capacity to consent, we explained the study to the relative, asked them to act as personal consultees for the resident’s participation and answered their questions. We then asked those willing to act as a consultee whether or not their relative would have wished to take part in this research and to sign a consultee form. For residents with no key relative we approached an alternative consultee (e.g. a solicitor or social worker).

We gave each resident notified to the research team a trial identification number, whether or not they subsequently joined the study. This enabled us to document the flow of participants through the trial, as required by the Consolidated Standards of Reporting Trials (CONSORT). 132

Informed consent

In line with current guidance on the use of the Mental Capacity Act 2005,133 produced by the British Psychological Society,134 we assumed that each person had the capacity to consent, and explained the research in an accessible manner, increasing their likelihood of being able to consent for themselves. We encouraged them to discuss the research with their family when appropriate and gave them as much time as they needed before reaching a decision. Only after residents had been assessed for capacity in this way in accordance with the Mental Capacity Act 2005,133 and they or their consultee had agreed that they would take part, were residents eligible to participate in the trial. We treated consent as a continuing process rather than a single decision, and discontinued interviewing whenever residents did not want to engage or became distressed by the assessments.

We also asked staff to consent to participate in the trial, both by answering questions about themselves and by providing information about consented residents. We emphasised to all participants, both residents and care staff, that choosing not to take part would not disadvantage them in any way. The resident’s GP was also given information about the study and notified that their patient was participating.

Ethical arrangements

We know of no documented harmful effects of using functional analysis-orientated interventions to manage CB. Nevertheless, we devised a trial-specific procedure for reporting serious adverse events (SAEs) to the chief investigator. In addition to agreeing with care home managers that they would update us on any changes to the residents’ circumstances, the researchers contacted each care home every month to check on adverse events that might not have been reported to us. Criteria for judgement of whether or not the event was deemed serious, or could be defined as an ‘untoward occurrence’ for the participant, were as follows:

• resulted in death

• was life-threatening

• required hospitalisation or prolongation of existing hospitalisation

• resulted in persistent or significant disability or incapacity

• was considered medically significant by the investigators

• included alleged or suspected abuse or neglect, according to the protection of vulnerable adults procedure agreed for the trial.

If so, the researcher completed a SAE reporting form for the chief investigator to assess whether that SAE could have been related to the intervention or deemed as ‘unexpected’. We then reported any adjudged intervention related or unexpected (or both a fortiori) to the REC, the trial Data Monitoring and Ethics Committee (DMEC) and the sponsoring NHS trust R&D department, within 15 days.

Randomisation

Given the CRT design, in which the care home was the unit of randomisation, consent was obtained from care home managers prior to computerised randomisation of care homes between the experimental and control groups, undertaken remotely by NWORTH. Our plan was for randomisation to take place gradually, home by home, once the consent process had been completed in each. However, this became problematic at an early stage, as the lengthy process of setting up computers and internet facilities in experimental homes (see Chapter 2) had not been anticipated. Thus, we needed to know which homes would contribute to the experimental arm of the CRT in advance, to ensure that facilities were in place in order to commence intervention as close to the baseline data collection point as possible. Therefore, we gained consent from managers at all homes and began the process of randomisation early on. In order to not affect baseline responses, as far as possible we communicated with managers, but not care staff, about technology required for the intervention, and maintained blinding to the treatment condition for researchers who collected the data.

Homes were randomised using two stratifying criteria:

1. ‘high or low CB’, namely whether more or less than 40% of residents had a CBS incidence score of > 10 at screening

2. size of home in terms of registered beds, that is, small (25–39 beds), medium (40–49 beds) or large (≥ 50 beds).

Blinding

As the intervention was pervasive, it was neither possible nor desirable to blind participating staff or residents to homes allocated to the experimental or control group. Instead, the researchers responsible for data collection remained blinded during collection of all baseline and follow-up assessments of care home staff and residents. An unblinded researcher informed the care home manager of their allocation to the experimental or control group, and ensured that experimental homes had the IT equipment and adequate internet (including broadband) facilities that were required for the intervention. An unblinded specialist dementia care therapist then acted as facilitator, supporting staff champions in training and using the decision support e-tool to develop action plans for target behaviours in CB residents identified by care staff.

After follow-up data collection was complete in each home, and in order to take into account the quality of blinding, the blinded researchers completed a perception sheet that asked them to predict the treatment group to which that home had been allocated.

Data analysts applied an agreed analysis plan to the data set, which was blinded for treatment group with anonymous codes.

Data collection

Our aim was to assess all participating residents at baseline and as close to 4 months as possible after that. However, because of problems in delivering the intervention and associated change of plan (see Chapter 2, Changes to protocol: the ResCare trial), follow-up in control homes ranged from 4 to 7 months after baseline [mean 5.15 months (SD 1.31 months)]. In intervention homes, the range was from 5 to 11 months [mean 7.86 months (SD 1.48 months)]. To adjust for this significant difference (p < 0.001) and the potential resulting bias, the agreed analysis plan was updated to use participants’ time to follow-up as a major covariate.

We conducted all interviews with residents and care staff in the care home and this usually took 2 or 3 days per home. In most homes, two researchers interviewed residents and care staff concurrently in separate rooms. In larger homes more than two researchers sometimes worked at the same time; in smaller homes mostly only one researcher conducted all the interviews. Second visits to complete interviews were arranged, for example when participants became tired.

Measures

(See Appendix 6 for details of the instruments and their scoring methods.)

Data management

Data were collected in questionnaire packs by researchers, during interviews with participating care home staff and residents, and from residents’ care home records, including the resident’s Medication Administration Record for their prescribed medication. Questionnaires were scanned into Verity Cardiff Teleform (Teleform Desktop V10.4.1; Cardiff Software, Inc., Vista, CA, USA), and were verified and validated. Data were then exported into IBM SPSS Statistics version 19.0 (IBM Corporation, Armonk, NY, USA) and SPSS was used to check data for inconsistencies, values out of range and missing data. All corrections to the SPSS file were logged, along with the reason for each change. A 5% sample of each type of questionnaire was checked against the final SPSS file to ensure consistency and to identify variables needing further in-depth checking. For example, the medication data needed further checking against the questionnaires, because numerous text fields in the recording form had increased the likelihood of human and scanning error.

Data analyses

The main statistical analyses applied SPSS version 20 (similar to version 19 in all relevant respects) to three participant populations: all residents, those with CB and staff. Participants were compared by treatment allocated (previously known as analysis by intention to treat) to draw pragmatic inferences about the real world in which participants vary in their compliance with planned interventions. Health economic analyses used SPSS version 19.

Economic analyses

Randomised allocation

Figure 4 shows the flow of care homes, staff and resident participants through the trial. In total, 111 care homes were considered for inclusion in the trial, with 63 ultimately being randomised and 58 completing baseline assessment. Table 7 shows the distribution of the 832 residents across the homes, with 555 in the CB group and 277 in the ‘other residents’ group. The main reasons for the lack of baseline data for those residents who were initially assessed for inclusion were withdrawal of care home, death, and non-selection or unwillingness of residents. Baseline assessments were completed by 609 staff.

FIGURE 4.

The ResCare trial: participant flow through the trial by treatment allocated. HN, home numbers; RN, resident numbers; SN, staff numbers. Note that the loss of 43% (SN = 135) from the staff intervention group, but only 21% (SN = 61) from the staff control group and 28% (RN = 117) and 14% (RN = 57), respectively, from the resident group are both largely attributable to the difference in the period of follow-up.

Of those who completed baseline data collection, in the intervention group there were 28 care homes, 420 residents (286 residents in the CB group and 134 in the other residents group) and 315 staff. In the control group, there were 30 homes, 412 residents (269 residents in the CB group and 143 in the other residents group) and 294 staff. A further 23 staff joined the study after baseline, mainly because some original staff were not available at follow-up and these staff were asked by the manager to answer questions about the residents in their place. Within the intervention homes, 92 staff trained as champions, 88 of whom completed study measures.

Dropouts

Seven homes withdrew from the study, five before baseline data collection and two after. Two homes were lost to follow-up (22 residents at baseline), both from the intervention group: one withdrew after the staff champions had completed the e-learning but before they had developed action plans, and the other withdrew after action plans were completed. Another intervention home failed to identify any staff to act as champions and, therefore, received no intervention, but provided data at follow-up, so these were analysed by treatment allocated (17 residents at baseline and 12 at follow-up). No control homes dropped out of the study between baseline and follow-up. Fifty-six homes were therefore followed up.

The dropout rate for the CB sample was 22.9%, with 428 of the initial 555 in the CB group completing follow-up. This was higher than the original estimate of 15%, which was based on previous local studies. 63,131

The ICCs from the NPI and CMAI data were between 0.03 and 0.06 (except for the NPI distress subscale; see Table 25), and so were higher than the value of 0.03 used in the original sample size estimation. The dropout rate in the other residents group, consisting of residents with dementia or with CB but not both, was 17% and, therefore, closer to our original estimate of 15%.

Table 8 outlines reasons why residents dropped out, broken down by their group allocation. Death was the most common reason, with 10.2% (n = 113) of the overall resident sample being lost to follow-up for that reason.

The mean age of the 113 residents who died before follow-up was 87.85 years (SD 7.01 years) and the mean age of survivors was 84.68 years (SD 7.58 years); 73% of those who died were female, as were 76% of survivors. There appeared to be little or no difference between those who died and those who survived, except for the expected difference in age.

Logistic regression was carried out to identify any factors and covariates affecting the dropout rate, that is, not just those who died. The factors included were gender, smoking status, size of care home, proportion of CB, type of home, treatment group, whether or not there had been organisational change and CB/other resident group. Covariates were age and length of time from baseline to follow-up. Based on the 174 residents who dropped out, the likelihood of dropout from the intervention group was increased by living in a home with a larger number of beds; male gender; older age; and longer time between baseline and follow-up (see Appendix 8, Table 71). The number of residents who dropped out of the intervention group (n = 117) was double that of the control group (n = 57). However, this may be attributed to the longer follow-up times in the intervention homes than those in the control homes because of the delays encountered in training the staff champions following baseline data collection (see Chapter 2).

The dropout analysis, conducted as part of the primary effectiveness analysis, included 31 cases. This figure was arrived at by excluding those residents who had died or whose care home had withdrawn, as these were not considered to be random. Whether or not the resident was in the CB sample group was the only factor found to affect dropout in this case, with these residents having a 2.8 times higher risk of dropout than a resident from the other residents group.

Those who dropped out in the CB group alone (n = 127) were therefore examined further. The proportions dropping out due to death in the intervention (50/64) and control (31/43) groups were compared, excluding those in the care homes that withdrew from the study. No significant difference between the proportions [χ²(1) = 0.51; p = 0.498] was found. Furthermore, no significant differences were found when the CB dropout group was compared with those who remained in the study at follow-up on baseline CBS, NPI and CMAI scores (Table 9).

Dropout was the reason why intervention (i.e. an action plan for a resident) was not provided for 87 residents (30% of the original 286 residents with dementia and CB) from the intervention group. This was mainly because the resident had died or the care home had withdrawn or declined the intervention (see Chapter 2, Figure 3).

Analysis of the staff who completed the baseline questionnaire (n = 609) showed that those who dropped out were significantly younger [t(606) = 2.4; p = 0.015], with a mean age of 37.25 years (SD 13.45 years), than those who remained at follow-up [mean age 40.02 years (SD 12.86 years)]. They had also worked with older people for less time, though this difference was not statistically significant.

The most common reason for staff not being followed up was their lack of availability because of either annual leave or sickness (Table 10). As with the figures for resident dropout, more staff dropped out in the intervention homes than in the control homes, but this may again be explained by the longer duration to follow-up in the intervention homes. After excluding the two homes that withdrew from the study, based on the chi-squared/Fisher’s exact test, two factors appear to significantly influence staff dropout: working in a large home and experiencing organisational change (see Appendix 8, Table 72).

Maintenance of ‘blind’ follow-up assessments

One or more researchers completed a blinding perception questionnaire for each of the 26 experimental and 30 control homes at follow-up. Table 11 indicates that, when researchers were able to make a judgement as to which group the care home had been allocated to, they were more likely to be correct than incorrect in their prediction. However, they were certain of their judgement in only 27.4% of instances, most of which were correct. In 41.2% of cases the researchers stated that the allocation could have been ‘either group’ and in another 31.4% researchers only ‘thought they knew’, as opposed to ‘definitely knew’. Some unblinding at the follow-up assessment point may have occurred, but the proportion of correct definite judgements remained low, at just under one-quarter, reflecting the considerable remaining uncertainty.

Descriptive analyses

The baseline descriptive data have been split into a number of tables describing the care homes (see Table 12), the care home staff (see Tables 13–16) and the residents (see Tables 17–20).

Analysis of serious adverse events, reactions and unexpected reactions

Our rigorous procedure for identifying, validating and reporting SAEs uncovered no SARs and, therefore, no suspected unexpected SARs.

Our analysis of SAEs focuses on the 58 homes and their 832 residents who completed baseline assessments. The 28 intervention homes reported 80 SAEs, namely 59 in the 286 residents with CB (comprising 50 deaths, seven hospital admissions and two instances of ‘persistent or significant disability’) and 21 deaths among the 134 other residents. The 30 control homes reported 55 SAEs, namely 40 in the 269 residents with CB, (comprising 31 deaths and nine hospital admissions) and 15 among the 143 other residents, comprising 11 deaths and four hospital admissions. As the 420 intervention residents were at risk for an average of 7.86 months, while the 412 control residents were so for an average of 5.15 months, the SAE rates per resident per year were 0.291 in the intervention group and 0.311 in the control group. None of these comparisons is at all close to reaching a statistically significant difference between the intervention and control groups.

Analyses of primary outcome measures

A residual analysis showed that the assumption of normality was acceptable for both NPI frequency and severity. The relationship between the baseline and follow-up measures was similar across all the care homes in both intervention and control groups, so there were no difficulties in using the models, outlined in Methods, for the chosen analysis.

The sample size was 555 residents from the CB sample, and 81 deaths were excluded from the imputed data sets. Fifty-seven care homes were included in the analysis because one home had only one resident in the CB sample at baseline and did not subsequently complete follow-up measures.

Table 21 shows the NPI frequency scores at follow-up for those residents with measures at both baseline and follow-up. There was only a small difference in the scores at follow-up. To be significant at the 5% level, the 95% CI for the mean should not contain zero. Subsequently, because the CIs contain zero for both models for the complete-case analysis, we found no significant difference between the intervention and control groups. This non-significant finding was also found for the imputed data set.

Table 22 shows the results of the analysis of the NPI severity scores. Again, there was only a small numerical difference between the intervention and control groups at follow-up. The results of fitting models I and II, for both the complete case and multiple imputed data, again failed to find any of the differences to be significant at the 5% level. In model II, we adjusted for the differential duration of follow-up between the two groups, as this was included as a covariate.

In conclusion, the primary analysis did not demonstrate any evidence of differences between the two treatment groups on the primary outcome measures of NPI frequency and severity. We therefore found no evidence of an effect of the intervention.

Analyses of secondary outcome measures

To determine whether or not the experimental intervention improved staff experience and quality of life for the residents living in the care homes, the following variables for the CB sample were used: NPI distress, NPI total score and CMAI frequency. Variables for the whole resident sample were CBS frequency, CBS total score, EQ-5D and QoL-AD. For the staff the following measures were used: MBI, ADQ, EQ-5D and SES.

The same combination of two models (I and II) and two data sets (complete and pooled results of five imputed data sets) were used for the secondary outcome measures.

A residual analysis showed that the assumption of normality was acceptable for all the variables, and as the relationship between the baseline and follow-up measures was similar across all the care homes in both treatment groups, there were no difficulties in using models I and II for the chosen analyses.

The CBS frequency showed that for model I there was almost a significant effect of the intervention for the complete-case data, whereas the imputed data did find a significant effect. However, model II did not confirm this effect (see Table 23 for the results of the secondary analyses). There was a significant effect of the intervention in relation to the EQ-5D index for model II but not for model I. However, when the correction for multiple comparisons was applied, the possible effect of treatment on EQ-5D index disappeared. No significant effects of the intervention were found in the staff measures for either of the models.

Data for the residents’ responses were not imputed, as there were so many missing data (only 376, 174 and 214 out of the total 832 residents answered the EQ-5D index, VAS and QoL-AD, respectively). There were no significant effects.

Other significant covariates were CDR-SB affects CMAI verbal non-aggressive; the time between baseline and follow-up affects EQ-5D index; and the group (CB or other residents group) affects CBS frequency and total score, both EQ-5D measures and QoL-AD.

In conclusion, we did not show that the experimental intervention consistently affected any of the secondary outcomes.

In order to address whether or not the intervention had an impact on the prescriptions of psychotropic and pain/discomfort relief medication, a multilevel analysis as above for the binary outcome of whether or not the resident was prescribed the drug was attempted. However, there were computational difficulties because of the distribution not being normal, and so the preferred analysis could not be performed. We then decided that the clustering effect was less important than the effect of the resident and so the results of comparing those residents who changed medication between baseline and follow-up in the two treatment groups are presented. This potentially results in an analysis that would find more significant effects than would be found if the appropriate multilevel model was fitted. The medication information was collected only for the CB sample and eight medication categories were of particular interest: antipsychotics, antidepressants, hypnotics, anxiolytics, anticonvulsants, dementia drugs, pain relief medication and laxatives.

Table 24 shows details of the residents who were present at both baseline and follow-up (n = 428) and whether or not their medication was changed. The majority of residents were prescribed the medication either at both baseline and follow-up, or at neither time point. Those residents whose medication changed received it either at baseline and not at follow-up, or received it at follow-up but not at baseline. Using only the change figures and comparing them between the two treatment groups, a chi-squared test to compare the proportion found no significant differences between the intervention and control groups.

In conclusion, we found no evidence of changes in the prescription of medication between the intervention and control groups.

Effect sizes

The effect sizes found were not as large as had been expected when the study was designed, and the largest effect sizes in this study did not occur for the measures that had been expected. Table 25 presents the effect size in two ways: (1) the naive effect size is calculated as the difference between the change from baseline to follow-up (follow-up – baseline) between the two groups (intervention group – control group) divided by the pooled SD of the change; and (2) the effect size with clustering taking account of the clustering in model I.

In summary, the effect sizes do not approach the 0.3 value that was used for designing the study. The majority of values are < 0.2, which is usually classified as a small effect size. For the staff-reported residents’ measures the maximum naive effect size is 0.277 for the CDR-SB and with clustering the value for the CDR-SB shows a similar value of 0.215. For the resident-reported residents measures the effect size with clustering value for the QoL-AD is 0.236, but this is not mirrored by the naive effect size value of 0.114. For the staff measures, the effect size with clustering value for the SES is 0.151, but again this is not mirrored by the naive effect size value of 0.098.

Intraclass correlation coefficients

The intraclass correlation coefficients values were calculated from model I output and were found to be of the same magnitude as had been expected when the study was designed assuming a value of 0.03.

Dose effect of the intervention

As we found no evidence of effect of the intervention, we carried out an investigation of the ‘dose effect’. The dose was measured by two home-level outcomes: the number of action plans expressed as a proportion of the number of CB residents per home and the number of champions per home. These two measures were included in models I and II, replacing the treatment groups, to investigate the intervention effect. The number of inpatient days was an additional response besides the primary outcome variables of NPI frequency and severity. For each of the responses, two models were fitted and both considered the clustering. Model 1 included the dose measures and baseline measure of the response. This was to test significant changes of the response caused by the dose measures (the intervention) before adjusting for other factors and covariates. Model 2 added other factors and covariates. At the individual level these were age, gender and baseline measure, and at the home level they were size of home, proportion of residents with a CBS incidence score of > 10, care home type, organisation change, and length of time between baseline and follow-up.

In summary, the proportion of action plans did not affect the NPI frequency significantly according to either model. (See Appendix 9, Tables 73 and 74, for the results of models 1 and 2, respectively, for NPI frequency, Tables 75 and 76 for NPI severity and Tables 77 and 78 for the number of inpatient days.) We found borderline evidence in model 2 that the NPI severity was affected by the number of champions, but this evidence was not strong. The dose measures showed significant effects on the NPI severity using the data after the 25% missing rule, but this was not then confirmed using the pooled results from the imputed data sets. Likewise, the dose measures did not affect the number of inpatient days significantly according to either of the models.

One might have expected that when the number of champions was higher, the NPI severity would have been found to be lower. However, this was not the case: when the number of champions was higher, the NPI severity was also higher. One explanation for this might be that where the NPI severity was higher, the home manager sent more staff to be trained as champions.

Further analysis of residents with inpatient stays

As a point of interest, a further analysis of residents with inpatient stays (medical, surgical, other and mixed) compared with those without any inpatient stays was conducted. The NPI frequency and severity, and the CBS frequency × difficulty scores of the residents were visually compared. Box plots were used to summarise those with no inpatient stays, whereas dot plots summarised those with inpatient stays (see Appendix 10, Figures 17–21). There was no visually apparent difference in the three measures between baseline and follow-up or between those with or without hospital inpatient stays. In relation to the cost of medication in the 4 months up to follow-up, the median costs were similar across all the groups. There was nothing to suggest that those residents with inpatient stays, irrespective of treatment group, were in any way different from those who had no inpatient stays during the study.

Economic analyses

Cost-effectiveness and cost–utility analysis

Follow-up data collection occurred approximately 4 months post intervention; however, due to problems implementing the intervention in some care homes, the length of time between baseline and follow-up was longer than anticipated. To account for the differential timing, it was assumed that the NPI and EQ-5D changed linearly over time and individuals’ outcome changes were adjusted to 4 months. This adjustment meant that our analysis of the data kept closely to our original analysis plan. As shown in Table 32, over 4 months the mean cost per resident in the intervention group was £187 less than the mean cost per resident in the control group (bootstrapped 95% CI of –£744 to £300). For the cost-effectiveness analysis, we reverse scored the total NPI so that an increased score indicated an improvement. NPI total scores range from 0 to 144 points. Residents in the control group had improved their mean NPI score by 2.90 points more than residents in the intervention group (bootstrapped 95% CI of –0.60 to 5.09). No significant difference between groups was found in NPI severity or NPI frequency score in the primary effectiveness analysis; however, the economic analysis used raw change scores and did not include variables accounting for clustering or other characteristics. This was because we adopted a bootstrapping approach to the analysis, and, as there was a very low intraclass correlation between homes, clustering was not seen as necessary. The main base-case economic analysis produced an incremental cost-effectiveness ratio (ICER) of £64 per NPI point improvement (95% bootstrapped CI of £14 to £303). There is no NICE-recommended threshold for a 1-point improvement on the NPI; however, we extrapolated from the £20,000–30,000 per QALY gain threshold that an equivalent threshold for the NPI would be approximately £140–210 per 1-point improvement on the NPI (which has a range of 0–144 points). Figure 5 shows the cost-effectiveness plane with 1000 bootstrapped ICER estimates. The majority of plots (75%) fell in the south-west quadrant of the effectiveness plane, where the intervention is less costly and less effective than the control.

TABLE 32 - The ResCare trial: cost-effectiveness analysis

N/A, not applicable.

Measure	Incremental		ICER point estimate (£) (bootstrapped 95% CI)	Probability intervention is cost-effective at
	Cost (£) (bootstrapped 95% CI)	Effect (bootstrapped 95% CI)		£20,000 per full unit of effect (%)	£30,000 per full unit of effect (%)
Main analysis
NPI proxy (n = 428)	–187 (–744 to 300)	–2.90 (–5.09 to 0.60)	64 (14 to 303)	0	0
QALY proxy (n = 428)	–187 (–744 to 303)	0.01 (–0.01 to 0.02)	N/A: intervention less costly and more effective	76	77
QALY self-report (n = 206)	–72 (–1085 to 714)	0.02 (0.00 to 0.05)	N/A: intervention less costly and more effective	85	90
Subgroup analysis
25–39 beds (small) (n = 188)	–595 (–1963 to 477)	–2.28 (–4.97 to 0.47)	261 (44 to 1953)	6	6
40–49 beds (medium) (n = 148)	49 (–318 to 419)	–3.24 (–6.77 to 0.60)	N/A: control less costly and more effective	5	5
≥ 50 beds (large) (n = 92)	–86 (–885 to 617)	–3.81 (–8.80 to 0.94)	23 (14 to 344)	7	7
Sensitivity analysis
Staff cost: action-planning done by senior care assistant (n = 428)	–242 (–776 to 269)	–2.90 (–4.95 to 0.68)	83 (17 to 327)	1	1
Removing secondary care costs (n = 428)	115 (–7 to 230)	–2.90 (–5.21 to –0.81)	N/A: control less costly and more effective	1	1
Removing outlier in control group (n = 427)	14 (–79 to 674)	–2.91 (–5.13 to –0.71)	N/A: control less costly and more effective	1	1

FIGURE 5.

The ResCare trial: cost-effectiveness plane for the NPI with 1000 bootstrapped ICER estimates.

The cost–utility analysis showed a different picture. When repeating the analysis using QALY area under the curve values, calculated using UK tariffs for the EQ-5D completed by staff as proxies for the residents, the intervention group had a QALY gain of 0.01 (equivalent to 3.65 days) larger than the control group (95% bootstrapped CI of –0.01 to 0.02). Figure 6 shows that the majority of the plots in the cost-effectiveness plane fell in the south-east quadrant, where the intervention is less costly and more effective than the control. Using resident self-reported QALY values (Figure 7), the majority of plots in the cost-effectiveness plane fell between the north-east and south-east quadrants, indicating that the intervention was slightly more effective than the control.

FIGURE 6.

The ResCare trial: cost-effectiveness plane for the proxy-rated EQ-5D with 1000 bootstrapped ICER estimates.

FIGURE 7.

The ResCare trial: cost-effectiveness plane for self-report EQ-5D with 1000 bootstrapped ICER estimates.

As can be seen in Figure 8, the probability that the intervention is cost-effective at a threshold of £30,000 per full NPI improvement is only 11% when the NPI is used as the measure of effect. The probability that the intervention is cost-effective at a threshold of £30,000 is much higher when the unit of effect is the QALY. In Figure 9 it can be seen that there is a 77% probability of cost-effectiveness at a threshold of £30,000 per QALY using proxy values, and in Figure 10 there is a 90% probability that the intervention is cost-effective at a threshold of £30,000 per QALY when using residents’ self-reported values. The issue is to what extent these two health measures (NPI as a dementia-specific measure and EQ-5D as a generic health-related quality-of-life measure) are in any way comparable. We have concerns for both the NPI and EQ-5D with regard to relating differences in costs to extremely small differences in outcomes, which may not be of clinical significance.

FIGURE 8.

The ResCare trial: cost-effectiveness acceptability curve for NPI.

FIGURE 9.

The ResCare trial: cost-effectiveness acceptability curve for proxy-scored EQ-5D.

FIGURE 10.

The ResCare trial: cost-effectiveness acceptability curve for self-rated EQ-5D.

Summary of findings

This is the largest study in the UK to date (December 2016) of an intervention for the management of dementia and clinically significant CB in care homes. Two thousand one hundred and eighty-five residents aged > 65 years living in 63 care homes were screened for CB. Homes were then stratified across treatment conditions for high versus low proportions of clinically significant CB. Within the cluster randomised study design, 555 residents with dementia and CB (the CB sample) and 609 staff were recruited to this study. In the intervention homes, 92 staff champions, nominated by their manager, attended training and co-ordinated delivery of action plans for the management of residents with dementia and CB in their home.

There were no significant demographic differences between the residents and staff in the two treatment groups. Comparison of the two groups on the primary measure of NPI (incidence and severity scores) failed to find significant differences that could be attributed to the intervention. This was true for both ‘complete cases’ and those derived by imputing scores for residents who dropped out but did not die, as well as for all of the resident and staff secondary measures, including comparisons of medications that were prescribed for residents. Post hoc analyses found that the proportion of action plans delivered within this intervention in a given home did not affect the NPI scores. There was borderline evidence that the number of champions who received training and contributed to delivery of action plans was associated with an increased NPI severity score. This finding was not examined qualitatively post hoc, but training may have contributed in part. For example, improved observation of communications through behavioural expressions by the resident (see content of e-learning in Chapter 2 and Appendix 1) could have enhanced staff observations of resident behaviour or, conversely, post-intervention completion of the NPI interview may have triggered staff awareness of resident communication of their severe distress through behaviour. Comparisons of health- and social-care costs, over and above care home costs, did not differ between groups over 4 months. However, cost–utility analysis indicated that residents in the intervention group had higher QALYs at the 4-month follow-up than those in the control group.

Limitations and implications for research

There are a number of limitations to design of the study and delivery of the intervention, some of which were unavoidable. We did not conduct extensive feasibility checks, as recommended by the MRC in its framework for complex interventions. 155 For example, we did not survey stakeholder homes prior to inclusion into the study on the availability of computer equipment and opportunities for e-learning (including space for staff to do this) prior to planning the CRT, as we were keen to offer the intervention as widely as possible in the real world of the range of care home settings; nor did we test and validate the e-tool prior to its inclusion in the CRT, as difficulties with the engineers delayed progress of the study. Therefore, the CRT was perhaps premature in relation to the e-tool, which required a more thorough developmental process. Furthermore, we did not evaluate whether or not action plans were being implemented or limit our experimental group data collection to staff (champions) who had attended training and were charged with co-ordinating delivery of the action plans in the home. Finally, although we did not collect data on ethnicity of staff or their migrant status in respect of the national picture, the dementia care workforce includes many migrant workers, and it is likely that the care homes in the ResCare trial were not atypical. Hussein and Manthorpe160 reported that the dementia workforce contains significantly more non-British workers (19.1% vs. 15.3%) than other parts of the care workforce. Non-UK workers among the dementia workforce are most commonly from the Philippines (17%) followed, equally, by Poland and India (10.8%). This may have an influence on training and staff practices, but this study was not designed to investigate these.

Potential contributing factors to the non-significant results of this intervention on CB outcomes in care homes will be considered next.

First, we examine the design of this intervention. Functional analysis-based approaches to intervention for CB and dementia were developed from modelling167 studies of face-to-face staff training workshops for dementia and CB,63 single case studies of functional analysis-based interventions in care homes,13,68,69 and our description of this intervention outlined in our Cochrane review82 (see also Chapter 1). The content of the range of supportive investigations and interventions for the management of CB and dementia was derived from Dementia: A NICE–SCIE Guideline on Supporting People with Dementia and their Carers in Health and Social Care (p. 210). 12 A functional analysis approach offered a systematic means of choosing the most appropriate interventions and testing these out in a given case. In order to widen its impact for future delivery in the present intervention, we additionally conceived an ICT solution for both staff training and intervention planning (see Chapter 2). The ICT solution comprised an e-learning course and a decision support IMS (the e-tool), to develop case-specific functional analysis-based action plans for residents with dementia and clinically significant CB. To this we added support from a specialist dementia care therapist, as this is seen as an important component of delivery in an individualised case-specific intervention such as this. 82 After we had adapted the delivery of staff training to care workers’ need for facilitated learning away from their work environment, this was well received (see Chapter 2). However, the action-planning process required a significant time commitment from the specialist dementia care therapist to interact with the e-tool, rather than supporting care home staff to use the iterative approach that is common in functional analysis-based interventions in dementia (see Chapter 2). This study was not designed to compare therapist-supported functional analysis-based interventions with the present ‘ICT-plus-therapist’ intervention described in Chapter 2. Our aspirations to widen the scope for the delivery of functional analysis-based interventions using ICT may have undermined any potential robust conclusions about use of this promising168 approach to intervention in care homes. A review of computerised clinical decision support systems suggests that these may improve practitioner performance but, to date, the effects on patient outcomes are understudied or inconsistent. 169 The present study found that the decision support software was not enough to impact on resident or staff outcomes. In contrast, an earlier study, FITS (Focused Intervention Training and Support), demonstrated good impact on prescribing practice with the use of bespoke interventions delivered by trained therapists working intensively with a small number of care homes. 27 A trained therapist is usually required for such an intervention,82 as a key skill in working in this area is the ability to be flexible and have the capability to alter care plans as new needs arise, meaning that the process remains iterative, but the functional analytical skills remain at the core of the plan. Future research with decision support software could explore the feasibility of training specialist dementia care therapists in working with care homes to deliver interventions for the management of dementia and clinically significant CB.

Second, this study adds to knowledge about staff training in care home settings. Our e-learning course, with its training on how to deal with common CB using simulated ‘real-world’ scenarios in the care home setting, is what was advocated in a recent focus group study of e-learning with 21 stakeholders in south London. 127 In the present study almost one-third of care home staff had never had any dementia training, let alone specialist training on the management of CB in dementia, which was the focus of our e-learning course. Staff champions were positive about the case-specific individualisation of the approach to dealing with CB and the qualities of the specialist dementia care therapist who facilitated the e-learning course (see Chapter 2). However, our findings indicate that, whatever delivery method is used, training alone is insufficient in its impact on the lives of residents with dementia and CB; future research needs to incorporate a sustained period of joint working and supervision with a trained dementia professional. 60,170 This conclusion stands true even for a course such as ours which, in contrast to many which are not empirically based,170 was informed by the growing evidence on case-specific understandings and intervention for CB13 and also on the NICE–SCIE guidance12 (see Chapters 1 and 2).

Third, organisational factors associated with engagement of care homes impacted on delivery of the intervention. For example, even in the context of providing extra arrangements to deliver the training away from the care home at no financial cost to the home, additional courses had to be arranged, as some homes struggled to send their staff to pre-arranged training (see Chapter 2), whereas homes with high NPI scores nominated more staff champions to support the intervention, suggesting that engagement with our e-intervention would be better focused on homes where there are more residents with CB, as staff may be more aware of their learning needs171 in the management of dementia with CB. Moreover, at the start of the intervention two homes dropped out and a third large facility failed to engage with the intervention at all. In contrast, no control homes were lost to the study. Organisational commitment to an intervention such as this is an important consideration for future research. Barriers to delivery of the intervention to reduce antipsychotics in the FITS implementation study included unclear communication across management levels and values and ethos that contradicted the intervention approach. 84 Other studies have noted that staff perceptions of empowerment are important in the provision of individualised care;172 values and beliefs that guide behaviour in an organisation can have an impact on depressive symptoms in nursing home residents;173 and the impact of training initiatives on staff confidence and standards of care is often dependent on strong leadership and the home culture. 174 We will now consider our findings in the context of fidelity and adherence to the intervention.

Fidelity and adherence to the intervention

Our intervention was designed to empower staff to access clinical expertise for the management of dementia and CB, using a broad view of training that was complemented by an IMS to develop action plans (the decision support e-tool) and access to a specialist dementia care therapist. A study of training in decision-making using the internet in primary care practices also noted disappointing use of the resource. 175 In the present study we can conclude that the effects of training nominated staff champions did not diffuse to non-exposed staff in the homes. This may have been attributable to our unavoidable need to depart from our original plan (see Chapter 2) and deploy research therapists to actively work with staff to gather data for the e-tool, and a specialist dementia care therapist to assist with developing the action plans. It is possible that active intercession of the therapists may have increased feelings of dependency among the staff champions and reduced their sense of ownership of the action plans, which in turn may have reduced the likelihood of them providing leadership in driving forward the steps associated with the intervention. However, a more likely explanation is associated with our proposed mechanism of change, which we will consider next.

The proposed mechanism of change envisaged was that once staff champions had completed the e-learning course, and produced an idiographic action plan for a resident using the decision support tool, they would use the action plan for a given resident. Furthermore, we envisaged that the bespoke training would be used to improve their own approaches and interactive skills and the action plans would further guide them to try solutions to meet the residents’ needs and also to facilitate informed approaches of their colleagues. To achieve the latter, the staff champions needed to be able to influence and motivate their colleagues to work differently and consistently in accordance with the action plan. It was also thought that in those homes in which staff champions operated in this facilitative manner, there would be improvements in residents’ quality of life, staff attitudes and other secondary outcomes. Limiting factors to behaviour change in staff champions include lack of practice in delivering the intervention, as 30% of the residents in the CB intervention group did not receive an action plan (see Chapter 2); no specific component of support or training given to assist them to implement the action plans and/or disseminate their new knowledge to their colleagues; different grades and experiential backgrounds, where some may have been unable within the hierarchy of the home culture to speak up and share insights with their colleagues; the low ‘dosage’ of the intervention in terms of the small percentage of staff champions in each home, the lack of relevant and ongoing supervision, particularly over the implementation phase of the action plan; and the experiences of the specialist dementia care therapist, who provided the perhaps crucial time required to mentor and support staff to implement the action plans, with interaction with the e-tool. This may have also undermined the expert clinical role of the specialist therapist who was cast into a different, more technical role with time taken to troubleshoot computer problems. Therefore, just one action plan was developed for most of the 199 residents, even though residents often displayed more than one CB, which may or may not have been related to each other. In understanding the barriers to integration of internet-based instruction by nurse trainers or other experts working in care homes, Irvine et al. 73 note that the technological role can undermine comfort levels of some clinical experts involved in training, even if they are frequent internet users.

The lack of effect on prescribing practices was particularly disappointing. Our baseline data on prescribing practices indicated that these are contrary to emerging clinical recommendations for antipsychotic use in dementia105,112,176,177 and clinical guidelines advocating benzodiazepines prescription for no longer than 28 days. 178 The action plans for such residents would have signposted staff to seek a review of these from the resident’s GP or a psychiatrist, at a time when high-level NHS targets for the reduction of antipsychotics in dementia care were emerging. 112 Such action plans would have also recommended clinical risk–benefit evaluation for the high use of benzodiazepines, hypnotics and anxiolytics (B/Z/A products) given the cautionary guidelines associated with adverse effects of these in older people. 178

It may be that staff working in care homes were unable to persuade other professionals, for example the GP, to assist with action plans such as review of drugs, as attitudes of professionals, including physicians and psychiatrists, to the management of dementia and CB can vary. 179,180 Indeed, there is some evidence from process notes (see Chapter 2) that the ‘unofficial hierarchy’ that was noted in a study of care home staff and physician interactions104 may have undermined implementation of these action plans. Variability in the culture of some homes, with respect to antipsychotic use,181 could have also contributed. The unwillingness of nurses and GPs to discontinue use of antipsychotics has been noted as a significant barrier to changed practice182 where the authors suggest that these barriers may be overcome with more complex multidisciplinary interventions. The FITS implementation study used the equivalent of specialist dementia care therapists to demonstrate impact on prescribing practices, but only in organisations with commitment to this type of intervention. 84

The fidelity of the intervention could have additionally been compromised by the needs of the specialist dementia care therapist who supported delivery. Although a percentage of action plans were checked by the clinical expert team, this may not have been enough, particularly for the residents who had presentations of severe CB. Routine access by the therapist to the clinical expert team was not incorporated into the planned intervention here. Where care homes are committed to provide care to residents with dementia and clinically significant CB, future cost–benefit evaluation is warranted. However, this will need to include in-reach by a trained dementia care therapist. However, care will be needed to ensure that trained dementia care therapists have access to clinical experts, as one of the lessons learned from integrating IMS into clinical practice is that experts operate at a higher level of complexity,183 and residents with dementia and clinically significant CB often have significant comorbid health conditions184 and live in a complex environment reflecting a system of varying cultures and practices in a care home. This type of cost–benefit study for a training intervention such as this would need to take into account the level of ‘scaffolding’,185,186 such as supervision, mentorship and access to clinical experts needed for the specialist dementia therapist, as well as that provided by the therapist to staff in care homes, as confidence in ongoing management of CB can vary. For example, the relevance of scaffolding new learning was demonstrated empirically in a study of experienced cognitive–behavioural therapy therapists who, following training, demonstrated an early dip in therapy performance, because of an initial undermining of confidence and requirements of the participants to change existing ways of working before further training and support raised their levels of competence. 187 An alternative investigation of ‘dose’ could be to explore substituting many product champions for an in-home specialist dementia care therapist, as our study appears to suggest that this low dose is not effective but that a higher dose, with access to clinical expertise, may be more effective in the management of dementia and clinically significant CB in care homes. For statistical examination of dose effect, our indices were necessarily limited. We were unable to capture the extent to which action plans were implemented and any resulting changes in CB. The likelihood is that many action plans were implemented only partially, if at all, and the good practice of reviewing and adapting plans in the light of the response to the initial set of actions could not be followed through. Future studies need to document each step on the putative pathway to change and allow for longer follow-up, in order to allow the intervention to become embedded in the care home. This was the case in all of the three care home intervention studies in our Cochrane review,82 where the intervention phase was 4, 6 or 10 months, with specialist support following training to support staff in implementing care plans described as occurring weekly or twice weekly.

Health economics findings

The health economic analyses did not identify any significant difference in service use by residents between the intervention and control arms over the follow-up period (health, social care and medicines, excluding residential costs). Although it was possible to calculate an ICER of £64 (95% CI £14 to £303) per 1-point improvement on the NPI, this is difficult to interpret as there is no agreed threshold of societal willingness to pay for such an improvement. In general, the lack of evidence for effectiveness in the trial is not compensated for by any difference in costs. The cost–utility analysis appeared to support the intervention, in that, using EQ-5D as a source of utility weights, the intervention dominated the control group, that is, it was more effective and less costly than usual care. There was a 77% probability that the intervention could be cost-effective at a payer threshold of £30,000. However, the mean QALY gain for the intervention group, as compared with the control group, was only 0.01 and so this finding must be treated with caution.

Aim

To understand key implementation processes for delivery of e-learning and computer-assisted decision support (e-tools) for the management of CB and dementia, including perceived and actual barriers to delivery.

Methods

We used the normalisation process theory (NPT) and framework analysis to re-analyse data from the ResCare trial and examine how innovations may become embedded in everyday work. Barriers to, and facilitators of, change in care homes were studied by considering ‘process problems’ in social care settings; and ‘structural problems’ affecting the integration of new systems into those settings. Following analysis of contextual data collected during the trial for the intervention homes, a typology of ‘organisational cultures’ for computer-assisted intervention was developed. From this, four ‘case study’ homes were extracted and seven (managers, senior care staff and care assistants), from a sample of 14 participants, were interviewed individually; and a specialist dementia care intervention therapist and a research nurse who collected data during the study were interviewed together. Nine additional qualitative interviews with care home staff included those from the control group, and three focus groups using nominal group techniques with a maximum variation sample (n = 22) of wider stakeholders provided opportunities to consider how far the findings of the ResCare trial resonated with their experiences and how they interpreted the trial and its findings.

Results

Three explanatory themes for the findings of the ResCare trial emerged: variation in care home managers’ trust of their staff; variation in the extent to which managers commissioned training; and variation in cultures of training and practices within care homes. The findings also suggest that care homes are not ideal environments for implementing new approaches, but implementation can be feasible, in smaller care homes and in those with less hierarchical structures.

Conclusions

The implementation of interventions for the management of CB and dementia, using specialist NHS support and resources, depends on the readiness of care homes to invest in innovation. Capable leadership and collective willingness are also important. The toolkit developed for implementing online interventions in care homes has scope for informing future practice innovations and research.

Studying implementation processes from findings of the ResCare trial

Normalisation process theory is a relatively recently developed method to assist understanding(s) of how new effective ways of working or service improvements may become a ‘normal’ part of daily practice. 191 It has particular strengths in the systematic use of data from complex interventions192 to gain a deeper understanding of the contextual processes involved in delivery of an intervention. By systematically capturing the many human processes at work, when individuals and groups encounter a new set of practices or aspirations to improve service innovations, it allows clear specification of the processes of implementation that can then inform the transfer of the innovation into routine practice.

Normalisation process theory is used in the present investigation as a way of understanding and describing the processes that inhibited or promoted a functional analysis of the causes and consequences of CB in care home residents. Using the theory-based approach provided by NPT and its sociological ‘toolkit’ to understand implementation, we focused on the dynamic, rather than the linear, processes that affected how an innovative approach to training and delivery of functional analysis-based interventions for CB in dementia may become embedded in everyday work in care homes.

There are several ways of using NPT to investigate implementation. Some of these have been developed in care homes193 and have also been used to inform computer-assisted decision support systems in the NHS. 194 In this present study we adopted a retrospective approach, following completion of the ResCare trial (see Chapters 2 and 3), as some implementation variables may have changed during the course of the trial. For example, in a journal dedicated to practitioners, including care home staff and managers in the UK, Flint and Cream127 conclude that there is a pressing need for e-learning programs to manage CB in dementia. Other efforts made in the social care sector to improve care homes’ access to IT have also been recommended for consideration by employers, who have to balance the time staff spend on training against the demands of their work and personal commitments;195 and websites, such as that of the SCIE, have begun investment in dementia management video-assisted demonstration and training resources. 196

In this chapter we report the evaluation that explored the processes at work when people and groups encounter, and seek to embed, a new set of practices or ways of doing things. 191,197 We used process data collected during the trial and also asked a sample of those involved in the study to reflect, with the benefit of hindsight, on the processes of the trial. In addition, during stakeholder consultations, we asked those who had not been engaged in the intervention, or in the trial, to think about the findings of the study and how these might relate to their own current practice or service.

Normalisation process theory was used to address two important problems commonly encountered when attempting to translate an intervention or service improvement into practice:

1. process problems – about the ‘implementation’ of new ways of thinking, acting and organising in health and social care

2. structural problems – about the ‘integration’ of new systems of practice into existing organisational and professional settings.

Normalisation process theory is built around a set of four core constructs and specific components that represent ‘generative’ mechanisms of social action, namely the different kinds of work that people do as they seek to implement; for example, the findings of a trial of a complex intervention. 192 These are the ways that people make sense of the work entailed in implementing and integrating a complex intervention (coherence); how they engage with this work (cognitive participation); how they enact it (collective action); and how they appraise its effects (reflexive monitoring). 198

The present study builds on other recent research conducted in health care, and, to a lesser extent, in social care settings,193,199–202 in which NPT as a conceptual framework has been used to understand the implementation of complex interventions,203 such as the ResCare intervention trial outlined in Chapters 2 and 3. It additionally outlines stakeholder contributions to the knowledge base, in which they discussed and refined the combined findings from the ResCare trial, including those described in Chapters 2 and 3 and from NPT analysis.

Participants: intervention case study homes

Four candidate intervention case study homes were identified from within the emerging typology of 27 intervention homes by researchers (JK and FP) who were independent of the original ResCare trial. This sample represents approximately 14% of intervention homes and was chosen to best represent the range of contextual factors and implementation processes identified from our analysis of existing data (see Table 33). These included home size, ownership and reported ease of implementation across a range of enabling and disabling implementation processes. All four candidate homes agreed to support the implementation study data collection. We aimed to speak to managers and any staff who attended staff champion training from these case study homes. These additional qualitative data were collected between May and June 2013 by a researcher (JK), unknown to, and independent of, the original research project team. The ResCare trial programme manager (CH) contacted homes to find out who was still employed at the home and available for interview; their contact details were then passed on to the researchers. Interviews were conducted over the telephone, electronically recorded and transcribed verbatim. Topic guides were informed by NPT205 and revised iteratively to explore further matters that emerged as important in the initial analysis. Telephone interviews took between 20 and 50 minutes each.

Reflecting the difficulties experienced by the original research team in contacting care homes, potential participants were often hard to contact, and scheduling interviews proved to be similarly problematic. Care home staff would often schedule dates and times with the researchers, but would later be unavailable. A research nurse, who was involved in data collection during the trial, and the specialist dementia care (intervention) therapist were also interviewed face to face, together, by the researcher (JK), to discuss specific homes. Transcripts were not returned to participants for comment and/or correction, as this was deemed unnecessary for a study of a non-personal nature, and would incur significant delays. Julia Keenan led the analysis to identify salient themes and issues; both Julia Keenan and Fiona Poland independently coded and discussed a subsample of four interviews to develop the initial coding frame, which was largely deductive, as informed by NPT, but tailored to the context of the ResCare trial.

Participants: stakeholders

We aimed to recruit a qualitative maximum variation sample of participants and so, in addition to the case study homes that were sourced from intervention homes (see Participants: intervention case study homes), we expanded our enquiry to participants from the other ResCare trial intervention homes (that were not used as case studies) and control homes; stakeholders from localities where the trial had been carried out; and wider stakeholders from other parts of England. These diverse participant groups are here referred to as ‘stakeholders’. Use of qualitative methodology meant that no sample size was determined in advance for the wider stakeholder group discussion. Instead, individuals with a broad variety of relevant experiences working in different agencies, and others with personal interest in dementia care, were approached and recruited. They included:

1. care home managers and care home workers from different parts of England

2. professionals with experience of supporting care home residents and working with care home staff

3. people with experience of training and professional education

4. people with experience of service improvement (including regulators, inspectors, complaints work), described here as ‘service improvers’

5. lay people who had reflected on their personal experiences of visiting and scrutinising care homes as relatives, or friends of residents; and

6. professionals with experience of purchasing places in care homes on behalf of local authorities (commissioners and senior managers).

Recruitment and data collection ceased when thematic saturation was achieved, that is, when no new themes appeared to be emerging from the data.

Two further researchers who were independent of the original research project (JM and KS) invited 22 stakeholders to a discussion event, which comprised a facilitated meeting involving three discussion groups that used modified nominal group techniques. This was held in London, in April 2013. We (JM and KS) further interviewed nine others in person or over the telephone between June and July 2013. Topic guides were informed by findings from the ResCare trial to include development of the intervention (see Chapter 2), the CRT (see Chapter 3) and the case studies that were used during NPT process investigation. These were tailored to stakeholder participants’ knowledge and role, where stakeholders considered whether or not our findings ‘rang true’ (see section Interview topic guides). Stakeholder interviews and discussions at the facilitated meeting were recorded in note form. These data were anonymised and then analysed thematically. Two researchers (JM and KS) undertook the analysis to identify salient themes and issues. Again, transcripts were not returned to participants for comment and/or correction, as this was deemed unnecessary for a study of a non-personal nature, and would incur significant delays.

Interview topic guides

Semistructured tailored topic guides were developed to explore the research questions posed in the case study interviews with managers and care staff (Boxes 4 and 5), facilitated meeting and in the subsequent interviews. The topic guides were discussed with members of the programme steering group, which included a person with dementia, and the wider research team. The case study topic guides were closely informed by NPT constructs and components, but tailored to the individual characteristics of the care home and interviewee role in relation to the ResCare trial.

For practitioners and other experts, the interview schedules were tailored to the activities of the specific agency or professional group being interviewed, whereas for family carers, their expressed experiences of managing CB themselves and their observations of care home practice and staffing were both potential sources of information and of their views. The topics covered included:

1. Do the findings of the ResCare trial (summarised), and about implementing an e-learning/e-tool for care home staff (described), and the introduction of functional analysis approached for planning (explained), ‘ring true’ in your experience?

2. Thinking of homes you have worked with or visited, does our typology for computer-assisted interventions (explained) help anticipate and explain these challenges for implementing e-learning/e-tool and functional analysis?

3. What other barriers might be relevant to consider?

4. What potential resources/enablers might be relevant to consider?

Ethical considerations

In line with other parts of the trial, data reported in this chapter were collected with informed consent and participants were assured of anonymity. To help reduce the risk of bias, the interviews were conducted by researchers who had not been part of the trial as the participants may have wished to give socially acceptable answers to members of the trial research team. Each case study home was assigned a pseudonym, chosen to avoid the name of any actual UK care home at the time.

Data analyses

The first phase of analysis involved close reading and re-reading of various overarching and home-specific data sources, related to the process and outcomes of implementing the intervention. These sources were then manually coded (by JK and FP) so as to identify barriers to, and facilitators of, successful implementation, and then to identify the main (sequential) mechanisms of implementation (see the example outlined in Figure 11).

FIGURE 11.

The ResCare trial: mechanisms of implementation taken from the analysis of process and outcome data from the ResCare trial.

These mechanisms, in conjunction with context data (see Box 2), were used to identify potential factors affecting barriers to, or facilitators of, implementation; and to develop a typology for intervention homes according to whether they provided an enabling or disabling environment for each of these mechanisms. Of the 28 intervention homes in the main study, 27 received the intervention; one home failed to engage with the intervention and one dropped out following intervention (see Chapter 3), so no trial follow-up data were available for this home. Where supporting data were unclear, making initial assignment of a home to an ‘enabling’ or ‘disabling’ environmental category difficult, a note was made, and clarification was achieved in April 2013 during consensus discussions with the research team. NVivo 8 (QSR International, Cambridge, MA, USA) was used to manage the systematic coding of existing and supplementary qualitative data. Participants in any interviews completed for this chapter did not provide feedback on the findings, but the findings from the case study intervention homes’ interviews were shared with the stakeholder group and their reflections recorded as part of the modified nominal group.

Implementation of the intervention in homes

Collective action: enacting the work required for implementing research and the intervention

Home managers were largely responsible for the practical work of co-ordinating resources to ensure that:

• the research team could collect data from residents and care staff

• the appropriate staff were identified for training as champions

• the staff were scheduled to attend training and given time to develop and implement care plans without this undermining the routine running of the home.

Champions also had a central role in enacting the intervention through co-producing and implementing the CB action plans, but there were five barriers identified to enacting the implementation of functional analysis in care homes. These will be outlined next.

First, longitudinal data collected by the research team document various contacts made with individual care homes over the process of the research to initiate and support implementation, and evidence difficulties for researchers in getting in touch with care homes. Telephone calls were the only fruitful way of contacting homes and different homes had different cultures/policies about staff availability to take calls, if they were at work. Absent owners were the most difficult people to reach.

A second barrier, evident in the majority of homes, was that senior staff or managers struggled to co-ordinate staffing to support and enact the research or intervention. The research team tried hard to be flexible and unobtrusive when collecting data in homes; however, this flexibility inevitably incurred difficulties and delays. Staff absences, shift patterns and turnover aside, by far the most significant barrier to managers committing to releasing staff to take part were competing service pressures – when carers would be scheduled to meet with the research team and then called back into service to meet residents’ needs:

. . . It was demands on time; it was like ‘Oh how much longer are you going to be?’ ‘We need you to come and help with bathing,’ or you know, lunch, or somebody’s come to see someone, can you come and deal with it, so there were constant interruptions, nobody was, and they were feeling almost as if ‘I’ve got to rush, I’ve got to rush’.

PA2 focus group

The changes made to (1) deliver e-learning off-site; (2) have a therapist lead the training; (3) deliver the action-planning; and (4) offer to compensate care homes to release staff were appreciated by managers interviewed, although some still complained about sending staff away from the home. Even supportive managers, such as the one at Happy Haven, who managed to schedule all the time and staff required, felt that in hindsight the home should have claimed more money back:

. . . I probably short-changed ourselves to what [staff] we actually did provide to be honest. But if we was to do anything like that again, I’d just have a separate diary for the days the people come in because they’d cancel and then we’d reschedule and all like this but obviously I’d already got them staff members in.

Happy Haven, manager

A more general concern was that managers were seen as unhappy with the amount of staff time implementation would take, irrespective of whether or not it was being paid for by the research team. Many of the managers did not actually claim back the money to cover staff time. From the manager’s point of view, it is burdensome to arrange cover from existing staff to co-ordinate this leave on top of existing training demands:

. . . we don’t have the amount of free staff I suppose, for a better word, to be able to . . . you know, with all the staff training, I mean believe me we have a huge staff training matrix what we have to deliver in a year anyway, [. . .] And if staff aren’t always easy to pin down to say ‘can you do extra shifts to cover the shifts’, that makes it quite difficult really.

Lifelong Lodge, manager

Another barrier was that managers frequently sent inappropriate staff to be trained as champions. Interviews suggested that managers had a strategy for selecting staff for training, but that this was perhaps wrong for the intervention. In Lifelong Lodge, for example, priority was sensibly given to selecting full-time over part-time staff to champion the intervention. However, the therapist’s notes reveal that it had taken several months to get to the stage where the manager at Lifelong Lodge released these staff for training, of whom one was quite junior and another had a booked hospital admission for a knee operation the week after training, and would be off work for a significant period of time. Only one champion actually took part in the action-planning side of the intervention in this home and she was too junior to be able to enact the medications part of the care action plans for residents. More junior staff were usually keen, but they were more likely to complete entries in a resident’s day book than be fully involved in writing individualised care plans within the home, nor did they have a sufficient understanding of residents’ medications to fully engage with care action-planning. Conversely, there were also issues, depending on homes’ staffing structures, when senior staff were distanced from residents and their day-to-day care needs, personalities and life histories:

. . . if you think about it, you’ve got this gap between people writing care plans and people doing the work, so [. . .] it’s not really ideally going to work with just like you know, one person being a bit more senior and they do care planning but don’t do hands-on stuff because you’re not getting that depth of information there. And certainly you know, it’s difficult to then sort of have a prescription of an action plan almost.

Therapist

Thus, in homes with a gap between those doing care planning and those with in-depth knowledge of the person with dementia it can make it difficult to implement this intervention on a one-carer-to-one-resident basis. Across the case study homes there were also differences in how care plans were routinely written and used by those sent to train as champions. In Lifelong Lodge the day seniors and manager write the care plans:

. . . I’m just doing that myself now and updating the care plans and risk assessments and making sure that staff are aware that changes have happened [. . .] she [carer 1] does not take over any care plans or risk assessments . . . they write a daily life record on what’s happened on her shift but . . .

Lifelong Lodge manager

Right and that feeds into the care plan does it, that daily life record or . . .?

Researcher

No, no . . . they’re writing care plans every day but they don’t probably review the changes and . . . they get told back in handover you know, at the beginning of each shift really but it wouldn’t be that they would go look at a care plan and read it thoroughly every shift you know, that doesn’t happen really at all. [. . .] No, I mean major changes or things what are affecting people day by day are changes where they would get given verbally at handover and then like I say it’s left to the senior team to write the care plans.

Lifelong Lodge manager

This is confirmed by the champion from this home, who also noted that neither champion at this home was routinely involved in writing care plans.

Finally, in this section, our NPT analysis in respect of collective action and enacting the research/intervention revealed that a major barrier to implementation was that staff were not given time to develop action plans with the therapist in the homes. The research nurse subsequently felt that this part of the intervention was initially thought to be much more straightforward, that the e-tool would simply produce individualised care action plans; and homes had bought in and signed up to participate on this understanding:

To be fair, right at the very beginning, we told them that this tool . . . there’d be like three trainings [modules] and then the fourth one would write you a care plan. And then it ends up that [Therapist]’s going in with them. So they were hoping to just do that [e-learning] and out comes this piece of paper with like [. . .] and they don’t have to check it because that’s dead right because computer’s done it. [. . .] So they weren’t expecting to spend a lot of time with you [. . .] Press a few buttons and there’s your answer.

Research nurse

However, from the perspective of the research team, if care action plans were to be fine-grained enough to impact on levels of CB, quite detailed information about the person with dementia was required. Once the therapist actually got into the home, champions often were not allowed enough extra time to develop these action plans, and:

. . . a lot of people have had to do a lot of stuff in their own time, and you know, whether they’re prepared to do that or not has made a difference.

Therapist

This was particularly the case at Home Court, where the champion took the main role in providing information about the CB residents for action-planning. This senior care assistant did not appear to see this as a problem and had previously attended training on her non-working days.

To summarise, the five barriers identified above are:

1. difficulties for researchers in getting in touch with care homes

2. managers struggling to co-ordinate staffing to support the research or intervention

3. managers selecting inappropriate staff for training as champions

4. different staffing models within homes sometimes meaning that champions did not have sufficiently detailed knowledge of, or a working relationship with, those residents with dementia to effectively implement care action plans

5. managers not providing adequate time or support for care staff to co-produce action plans.

These barriers were addressed by the research team and steps were taken to work around these problems with care homes, with the consequence of significant delays and additional costs. The real and ideal conditions for implementation and strategies to promote the enactment of interventions in care homes are summarised in Table 38.

Table 39 provides an overview of findings across the four case study homes with respect to the collective action processes of implementing the intervention.

Reflexive monitoring: appraising the effects of implementing the research process and intervention

To successfully embed interventions for CB in residents with dementia, care home staff needed to review their experiences of implementing it and, if necessary, adapt their practice to suit local circumstances and individual residents. Key examples of reflexive monitoring are seen through the changes made by the research team to adapt implementation, such as responding to difficulties about installing computers in homes, providing financial resources to release staff for training and arranging more training events with smaller groups to help some managers who were unable to release staff as planned. However, there were barriers to reflexive monitoring within the homes themselves. These were:

• lack of any systematic feedback on the impacts of the intervention on residents with dementia from the research team, because the trial was ongoing

• positive feedback from staff who attended training not being shared with managers

• lack of momentum and support for staff to share and embed learning and change their practices.

Case study care home managers were aware that the nature of the research design meant that they would not get to know the outcomes of the study (in terms of impacts of the intervention on measured levels of CB among people with dementia) until data had been analysed and the study had ended, but managers at Careful Place and Lifelong Lodge reflected that they felt short-changed in that they had not yet had what they understood as feedback from the research team since data collection had been completed, as described below:

. . . it would have been nice to have had some general feedback you know, to say well the trial’s now finished you know and a general summary at a location.

Careful Place, manager

The manager at Happy Haven, however, recognised that she had received informal feedback along the way and that more formal findings would be coming. Intervention care homes in the study were written to after data collection and thanked and offered three free additional licences (active for 3 months), a certificate and a free place at a dissemination conference.

In terms of appraisals of therapist-led e-learning training, all staff interviewed who had attended training felt that it had exceeded their expectations. This supports the findings of the evaluative questionnaires completed by those who attended training. An aspect of the intervention judged as particularly good was that the e-learning:

. . . wasn’t as complicated on the computers as I’d thought [it would be].

Happy Haven, care staff 1

Most noted was the package providing a much more active and engaging learning experience, which related well to their own experiences of working with people with dementia:

I’m more of a hands-on, ‘let me read it, let me do it’, than having someone just sit there and go on and on and on and on for hours. And then say to you ‘Take a test’ or something, [. . .] it kept your mind going [. . .] some of them scenarios that we did I just thought oh that is so like so and so where I work.

Home Court, care staff 1

Also valued was that the e-learning process gave the learner instant feedback on their understanding:

. . . it made you think ‘Oh, oh yeah, I get it’; do you know what I mean? We put in what we would have done but then the computer told you the correct way.

Happy Haven, care staff 2

Some staff were more effusive in their praise than others, whereas on a basic level the relatively junior champion sent from Lifelong Lodge reflected that:

. . . it does open your eyes a bit.

Others were clear about the specific value the learning added in terms of improving individualised care:

. . . where before you just thought like everybody was the same, dementia was . . . but it was explained better and everybody’s is not the same.

Happy Haven, care staff 2

These individual responses corroborate the therapist’s notes on which champions in each home performed and engaged best with the process – in that these staff could offer a more detailed account of what they learned. The manager from this care home who attended training remarked that:

. . . if it had widely been available, I certainly would have sent more staff on it.

Champions who attended training and completed care action plans appraised its perceived impact on practice in lukewarm ways in interview. Although the impacts could be interpreted as being fairly profound, the intervention prompted some subtle changes, which were perceived to have stayed with staff, as in these comments:

No, I don’t think I’ve changed anything on the way I deal with things. I think it just makes you more aware [. . .] you know, if a situation arises, I think it just gives you a better insight into it, on how to deal with it.

Happy Haven, care staff 1

I don’t know how to explain it really. Something pops up in the back of your mind what you learnt and you do like tend to deal with it differently.

Lifelong Lodge, care staff

. . . it did work sometimes and it still works now sometimes but you just have to find what kind of mood they’re in that day and maybe that one won’t work today but you go and you try it tomorrow and it works fantastically.

Home Court, care staff

When probed, these champions could provide illustrative examples of the intervention in practice as follows:

Just the way you calm them down and things like that, like your tone of voice.

Lifelong Lodge, care staff

Just things like trying to go along with a resident when the resident’s sort of maybe going off on a tangent about their mum or dad or something . . . just to listen and agree [. . .] just things like that instead of constantly battling with them and trying to explain that they actually live here.

Home Court, care staff

Similarly, subtle examples were noted by the therapist when she had returned to homes. One example was the reduction in the number of falls experienced by one resident. Following training, staff had noticed that the resident had difficulty differentiating between a counterpane and the carpet. They changed to counterpane and there were no further falls at this location again.

Given the overwhelmingly positive feedback about the e-learning experience from care staff, it is surprising to find that the two managers interviewed in case study homes who did not attend training saw little value in what staff had learned:

. . . when I spoke to the staff, they said well they didn’t really get a lot out of it and it was kind of dismissed.

Careful Place, manager

And how does that compare to their reaction to other sorts of training or would you expect . . .?

Researcher

No they’d . . . you know, like they’ve just done an end-of-life course and dignity in the home and things like that and their reactions to those were quite positive.

Careful Place, manager

Only the manager at Happy Haven, who had attended training herself and thought that it had been useful for staff, gave any feedback on the perceived impact of the action plans at her home on levels of CB among people with dementia. Furthermore, the therapist’s notes mention that it had been fed back to her that the project action plans had received praise and been identified as being extremely good, in terms of both clarity and person-centred individuality, by the CQC team during their recent audit; the manager in this more recent interview evaluates them differently:

. . . I think it was more like the problem areas, if things was more sort of like . . . we learnt more about the person and sort of like what their conditions was [. . .] [therapist] gave a lot of insight into that which helped staff, rather than more of the individualised . . . we already had it really in the care plans.

Happy Haven, manager

. . . Because we’re not a home that have got a lot of challenging behaviour to other care homes but I saw it more as that we could help like the project develop [. . .] The staff did get something out of it but I think the project got more out of us than we did.

Happy Haven, manager

Again, this reflection that the intervention might have a greater impact on homes or individual residents with higher pre-existing levels of CB should be considered when looking at recommendations for wider implementation.

Interviews with care staff also revealed few existing mechanisms through which champions could share learning with colleagues within the home who had not attended training. In smaller, family-run care homes (Happy Haven), some wider dissemination could be achieved informally and through general face-to-face interactions and a staff comments folder. Staff in larger homes also mentioned being able to discuss and so to reflect with other staff who had attended training and were still employed at the home. However, the lack of formal support either within the home or from the research team to help champions to embed the intervention was seen as a weakness by the therapist:

I think again, what normally happens when something’s like an incremental sort of development thing, if somebody else comes along and shows you what they’re looking at, what type of things that they’re judging it on and then bit by bit you sort of can do that yourself and you learn to judge yourself don’t you and evaluate for yourself.

Therapist

The research team ran out of time to embed the intervention, as it was hard to address and overcome the initial barriers to delivery within the time scale of the CRT.

For care staff, the experience of becoming champions led to fresh insights, and better strategies, to improve the individualised care of residents exhibiting CB. Champions in smaller, less hierarchical homes could share these insights informally with other care staff.

However, overall, there was little feedback to managers who had not attended training about the use of systematically formulated case-specific approaches for managing dementia and CB and ways in which support could be delivered in the context of a given care home. Therefore, there was inadequate support from managers or the research team to further develop and embed a functional analysis approach to the management of residents with dementia and CB. Understanding of real compared with ideal conditions is required to enable staff to evaluate and adapt case-specific interventions for dementia and CB. These should also suit local strategies to facilitate reflexive monitoring and are summarised in Table 40.

Table 41 provides an overview of the outcomes of implementation in the four case study homes. From this and the other data, Figure 12 summarises what needs to be in place to support implementation in individual homes, as identified by NPT analysis.

Training and functional analysis: ‘ringing true’?

Overall, there was general agreement among stakeholders that much could be done to improve the quality of life of care home residents, especially people with dementia and CB. One experienced trainer and professional educator, whose work covers different parts of the UK, observed:

I think many care homes could improve their support for people who are distressed, confused, angry and expressing how they feel in a huge number of ways. I think the way they prepare staff. The way they help staff to even think about what’s happening for that person and what is their emotional experience like and how is that linked to the way that they would behave?

MD

Many of those interviewed had experience of training care home staff or were professionals who visited care home residents for assessment, provision of clinical care or treatment, or to offer staff advice about care practices. However, concern was expressed by some stakeholders that not all training was necessarily helpful. An experienced trainer and professional educator reflected on their experience:

I’ve seen lots of very poor examples, particularly from external professionals coming in who maybe have a general sense of, well, I know what the tasks that care assistant does in a day. But [they] seem to have no insight into where those people are coming from as non-professionals. They [care home workers] haven’t taken three years out to do training [like nurses or occupational therapists].

MJD

Mixed views emerged around the optimum venue for training. Some stakeholders were not surprised to hear of the ResCare trial findings that staff often preferred to undertake training outside their work environment. For example, one stakeholder drew on their broad experience of training (over 25 years) to argue that training needs to relate to the individual’s experience and yet should be challenging:

I would say they definitely prefer to be trained outside their place of work, usually with others doing the same work but in different locations/situations – so they can learn from others’ experiences. Without a shadow of a doubt the training that is most effective falls into two categories, the first is using actual real live case stories of those in need and illustrating how they were supported/helped by the [care home worker]. The second is using more challenging training that in effect places the learner in the same position as those they seek to support. I use exercises that have the effect of making the learner vulnerable and in need of support so they can really ‘feel’ what it is like to be in this position. These can only be used when the Trainer has achieved a strong rapport and very high level of trust within the group of trainees.

PK

This last point reflects the importance expressed by some other participants that a trainer has a key interpersonal role in addressing dynamics and different responses to the emotions that may surface during training. This was something that many commented on, and reflected trial participants’ positive views of the skills of the trainer who facilitated our e-learning course (see Chapter 2).

However, the view of one care home manager from the experimental arm of the ResCare trial was very different. They thought that training at an ‘out-of-home’ venue was often seen as a ‘jolly away from work’ for many of the staff and found that friends tended to go along together, and that their learning was therefore limited. This manager preferred training to be within the home itself. Similarly, an experienced trainer thought that staff might like to leave the care home for training, but that the benefits of this were more to convey to staff that they were valued:

It doesn’t surprise me at all. It’s a completely different change of context and yes, I suppose my experience would be much more about saying, we value our people as a member of staff and people feel that value. In terms of changing people’s skills, confidence, I’m not sure that it addresses that at all.

MJD

A service improver argued that the effects of training were hard to evidence, but that improving staff morale was valuable for care workers and did have a positive effect on practice. They added that in thinking about the wider dementia care services and training the workforce, it could be easier to arrange training for care home staff rather than home care staff ‘where it’s much harder to get people in same place at same time’ (SF).

In contrast, a care home worker (experimental home) who had undertaken the ResCare trial training away from the work location felt that it had been beneficial, for reasons that related to learning with colleagues:

. . . away day . . . helps you get away from everyday work stress . . . [had gone with a couple of friends/colleagues] . . . you have that support when you’re back in the care home to ask questions and like, you know.

CD

In their view, the training had been very good:

I liked the training; I learnt things I hadn’t known before. Maybe I would have learnt it on the job anyway, but it was good to have a day to do it.

CD

Similarly, a care home manager (experimental home) thought that the ‘out-of-home’ venue worked quite well; if staff were able to go away together, they would learn together, and could then ‘bounce ideas off one another’ on their return. Another care home worker said that they had enjoyed the training and the e-learning program as part of the ResCare trial – in their view the trainer was very enthusiastic and ‘taught very well’. They appreciated having a full day for this, but realised that it must have been difficult for the rest of the staff in the care home to cover their shift.

There was an alternative view that a full day was too long and did not reflect what we know about adult learning. One trainer interviewed thought that:

What works best in training is probably a basic outline plus simple case examples – short ongoing sessions comprising revision plus a new aspect each time.

RJ

This trainer suspected that concentration among care staff was often limited and that 15-minute-long at-work sessions were more suitable. They also noted that there seemed to be a much broader range of ability among ‘hands-on’ care staff than among managers, and concluded that opportunities for refresher and repeat sessions should be available for those who needed or wanted them.

Moreover, the value of formal training on its own was not universally accepted. One care worker, working in a control home in the ResCare trial, did not think that there was a need for training because, in their experience:

Just talking about patients with challenging behaviour at case meetings or in the staff room is good, as it helps us get a better understanding of that person . . . it was good to meet the team . . . I was given time to think about the residents when they asked questions and filled in their paperwork . . . just talking about the residents when they [research team] came each time gave me ideas . . .

EF

Although they were not able to comment on the training offered, they thought that discussion with someone with expertise would be more effective:

I’m sure that the training is good, but I can see why that [talking with a specialist dementia CB practitioner] would work.

EF

How this might work was outlined by a manager of a care home that had been part of the experimental arm of the ResCare trial. They described the study as ‘fitting in’ the home and had found that the trained champion was able to answer questions raised by other members of staff. Training without extra support afterwards was not generally seen to be a good investment, and one part of training that appeared to be strongly supported by some interview participants was the desirability of more senior staff (senior care workers or shift leaders) to model good practice ‘on the floor’:

It’s not only about training. For me, the best services are where you have, whoever the shift leader is, a good knowledge of the people and individuals concerned and the way they feel and how their illness or their degeneration of whatever it is affect the way they behave. They are the role model and they will guide that. They will suggest. They will encourage. They will question.

MD

However, a care home manager who expressed their belief in a ‘train the trainers’ model did not think a single champion who could be referred to for all questions was particularly helpful owing to high staff turnover, but that training someone who could regularly cascade this to all other staff in the home was more useful. In their view, sending all staff on training was not always possible, and they expected all staff to be competent in addressing some things like management of CB.

What seemed to emerge in these discussions was a nuanced view of training and learning. Training may make a difference in the short term, if it encourages more material discussion across the team about real incidents and events. It is here that it seemed to be seen as useful in terms of encouraging openness and discussion about the practice that actually happens in a care home. ‘Ringing true’ or a sense of authenticity about service improvement seemed to be a barometer by which the interventions were gauged.

Enablers and facilitators

One service improver commented that training should not have to be argued to be cost-effective, being confident that ‘the big facilitator is that training leads to improved staff morale and job satisfaction’. Similarly, one of the care home managers suggested that training providers would be well received by care homes if they were able to promise refresher training every year, which would incentivise care homes to invest in training. They added:

Most managers have now come to accept that one-off sessions are futile, and without the offer of second phase, when things are likely to have changed, managers see little point in training events.

MJA

Turning to the ResCare trial development of a championing role in each care home, one manager expressed surprise that having a specialist CB practitioner did not work, as they felt that it had been useful in practice. Unfortunately, they reported that most of the staff whom the practitioner had engaged with had subsequently left, being replaced by younger and more junior staff who found training of this sort harder to understand. This manager described how some of these new staff did not understand some of the basics of care home training (such as how to manage restraint and aggression, for example, through breakaway training), and that they themselves had to explain this to the staff.

In contrast, several reflected that some care home staff were not ‘empty vessels’, but with experience they had accumulated substantial skills that were not always acknowledged or respected. For example, a care assistant, working in a control home in the ResCare trial, felt that speaking to others about the topic was helpful as it made them realise how much they actually knew but had not been utilising or had not considered previously. In their view, more intensive training would have been better, as it was about an area that was of particular interest. Similarly, another care home worker had overall enjoyed the ResCare trial training as it had given them the confidence to do what they had known all along, but did not know if it was correct or not – it ‘gave a grounding’.

This contrasted with the view of a former regulator who noted that one enabler of promoting training was when a regulator can insist on staff being trained to do their job. In their experience:

Care homes get away with doing what is mandatory. Issues like dignity, nutrition, and so on, are seen as ‘add ons’ and there is reluctance to release staff for this. I have been commissioned to provide training for care home workers on completing care plans as they had been pulled up by the CQC. [Locally] I have tried engaging with care homes to attend free training and the response was zero. So . . . unless there is evidence of poor practice the regulator can insist only so much. The essential standards only talk about the ‘suitability of staff’ so if the care homes can prove they are meeting this standard by some means CQC will probably have to be satisfied.

ER

Similarly, a local authority manager spoke of the limits of regulation and inspection in prompting training. They pointed to the differences between their role and that of NHS commissioners:

As a purchaser – I use that word because we buy – we don’t actively commission residential care; by and large we are really just a major funder [of people’s fees] we do not currently require a suite of training. We obviously require that providers are registered with CQC and they in turn will check the training records. At the moment, that is as far as we go.

NP

This local authority manager envisaged some changes, as the local authority was working more closely with the local Clinical Commissioning Group to possibly develop a dementia staff competency framework, with validated training attached. In addition, there were aspirations that staff termed dementia care ‘mappers’ (having attended a course on dementia care mapping) might be able to identify specific training needs among a staff group. As a result, if there were severe problems, the contracts monitoring team in the local authority would be able to check if the care home’s staff had undertaken training and thus might be able to block new placements (those paid for by the local authority) until specific training had been undertaken. All this was at the planning stage and would represent considerable management investment, and possibly could be open to questions about infringements of choice.

The stakeholder interview data show that it is important to think about the interactions of content, methods, timing, location and venue of training in order to draw conclusions about its effects. There were several views that the sustainability of current training was limited, partly because of high staff turnover and reorganisation in care homes, but also because of changes in ownership and management. A trainer and professional educator commented:

I’ve worked with many teams who have lots of training – 3 day[s’] training or even more. I’m not sure that that’s reflected in what I see in their practice 6 months later, 12 months later and so, if it is a training that you go away to and you are taught almost in a classroom setting, then I’m not sure a year later that I see the results of that training any more.

MD

Care homes being commercial organisations in the main, even if not for profit, their managers often adopted an economic perspective in discussions of training investment. One care home manager, for example, noted the presence of other preoccupations for managers: ‘There are only so many hours in the day’. They were uncertain as to how care homes could be incentivised to participate in training and e-learning, thinking that this could be over-reliant on one person – ‘sometimes need a champion in the home, but this has to be the manager as they make decisions’. This manager noted that having a champion should help, but that high staff turnover made it hard for managers to decide on investing in a champion who might then decide to leave in a few months. Like many care home managers and other stakeholders, they described high staff turnover as a major problem: hence specific topic training took a lower priority as recruiting new staff, inducting them, and training them for basic care became more important.

In the context of these pressures, a service improver and researcher commented that smaller care homes seemed to benefit from the establishment of a care home or social care-focused organisation that could offer staff opportunities that they could not purchase or arrange by themselves. In the region where they had recently facilitated a learning set for care home managers, such an independent organisation had been financially supported by the local authority, which was able to recommend that some managers might find certain training ‘useful’ when concerns had been raised about their home’s performance.

Problems and barriers

This dissonance between professional expectations and the reality of care staff’s experience and prior skills was expressed as a barrier to receiving professional advice and adopting decision support tools for planning care. In this section we explore stakeholders’ views on care plans. Several commented that detailed care plans were not commonly found in care homes. One manager said:

They are not used ‘on the floor’ and the subtle nuances of how you respond to [a situation of CB]. You end up with things like, ‘please respond sensitively to X’s [resident’s] aggression’. What on earth does that mean? It’s about getting skilled people to role model that to show people; actually this [care planning] works.

MD

Furthermore, although there might be several professionals going in and out of a care home, it seemed to require taking a deliberate decision to respond to a situation where the level of distress was getting beyond the ability of a care home to cope. There seemed no agreement over who should take the lead to challenge, and to say ‘What’s going wrong?’. Part of the explanation offered by one trainer was the lack of a culture of teamworking in some homes:

I still work with teams who don’t meet. They never have a staff meeting. There is no time at all for more skilled perhaps senior staff to be asking those questions. What happened this week? How did it go? I still work with teams that never do that.

MJD

One of the care home managers interviewed ascribed a lack of interest in taking on further activity as a result of some care homes’ economic, and other, uncertainties. This feeling of being under siege arose from the general impression of several changes in the care home sector in recent years, with the recession and the Southern Cross ‘fiasco’ (closure/break-up of a chain of 750 care homes in England 2011, including some that were part of the present study). As a result of general austerity and reductions in local authority expenditure, managers were managing budgets more carefully, and the first expense line to go could be training: ‘Some feel there is no use for it, or not much use, shall I say’. Another care home manager in the ResCare trial control group, although expressing disappointment that they did not have more training, or had not been offered the e-learning provided to the experimental group, considered that ‘we wouldn’t have had time for that anyway [laughs]!’ At an extreme, although voiced by more than one stakeholder, was the view that some care homes were not engaged with policies and strategies about training.

Further barriers raised by a minority of stakeholders were connected with suspicions that e-learning might be adopted as a cost-saving measure. One trainer thought that employers might start to say:

Can you do this in your own time? You all have access to the internet now at home and it will be good for you and your career if you can say you have completed these modules and so some staff will and some staff won’t.

MD

There were other views, that, although it might be theoretically possible to do e-learning at work, shifts were often short staffed and for some staff the ability to switch from care work to computer study would be ‘mentally difficult’, even if they were not interrupted. Others had observed almost a feeling of ‘distrust across some organisations or an anxiety about staff not being on the shop floor’, which would mean that sitting by a computer could be seen as avoiding work or the ‘difficult’ residents.

For one former regulator, its impression was that, in the current state of the public sector economic climate, care homes would do only what was mandatory and had been identified as a requirement by the CQC:

Unfortunately any non-mandatory training is seen as a luxury and in the current financial climate is not something they would do willingly.

ER

Others too were pessimistic about regulation’s potential to improve training uptake:

By chance I’ve studied a few compliance [CQC] reports in the last month. Very, very little mention of training, if the word training in there at all. Even where I would perhaps have identified a training need in relation to an issue that was described. That’s not my experience; I haven’t seen a lot emphasis on that.

DM

However, this did not reflect the stakeholders’ realisation of care homes’ changing customer base, with more residents having high levels of disability. One care home manager felt that ‘selling’ training to junior staff was hard because many initially thought that all people with dementia were ‘nice little grannies just watching TV’. Initially, new staff rarely expected to encounter CB, but they quickly realised it was common. In this manager’s experience this ‘frightens a lot of the younger ones’. Just to get staff ‘to read the care plan’ gave them some confidence that the problem was not their fault. They felt that more discussions and demonstrations, as well as training around management of CB, could do nothing but good. One care home worker from the control group of the ResCare trial said that she had liked having someone:

. . . professional and experienced in challenging behaviour to talk to – it made me appreciate how difficult some of this is.

IJ

The emotional impact of being ‘on the floor’ and being expected to manage CB was communicated by several stakeholders. One added that having someone to talk to made them confident to approach other senior members of staff with any later queries.

Resources and implementation

There were several reasons why the offer of free e-learning licences to the care homes following the ResCare trial was not seized on (see Chapter 2). One care home manager thought that the free licences were not taken up because many younger staff members were being protective of their work–life balance – ‘few are willing to stay late, even if overtime is paid’. Another, in the control group of the ResCare trial, found that taking part in the study had been valuable – ‘everyone is talking about challenging behaviour right now’ – but had mainly found it ‘helpful to have someone [the data collection research team] to talk to about the residents’. This manager thought that free licences, which could enable staff to do the training in their own time, should work, and confessed surprise that they had not been taken up. It was the personal expert approach that seemed to have left a lasting legacy.

One of the care home workers interviewed thought that refresher training would be necessary, ‘as things change so often’. Their expectations for this were not high. By refresher training they meant having someone to remind them about what they had learnt, perhaps to talk to on an annual basis, even over the telephone or during a visit to the care home, rather than having to go along to another day of training ‘like the first aid courses you do’. However, this person seemed little engaged with professional development and could not offer a view about the free licences, other than speculating ‘maybe the manager had other priorities, or didn’t realise it was free or something’. This lack of knowledge was expressed by other care home workers, illustrating the point to be made below that training decisions were largely the business of managers and that front-line care staff felt that they had no role in such decisions and plans. Placing this in context, a trainer thought that many care homes and care home staff were used to abiding by statutory minimum training requirements and added that taking up training in the form of an internet package is ‘not the way they are used to working. It is unusual and it would be new and different’. They further noted that managers would have to free staff from their work and would not necessarily be able to map what staff were doing, a point already discussed above when reflecting on the surveillance of staff in care homes and the limits of trust in the home.

Making decisions

In this section, we report one of the underpinning themes around innovation in care home practice. Who makes decisions about investment and staff time proved to be a helpful diagnostic aid when thinking about system change, which is in effect what the ResCare trial was investigating. Training, for example, was seen by almost all stakeholders as a subject that the manager made decisions about. This was because it necessitated authorisation of expenditure, it affected the resource of staff time, and it was an area where there was some discretion in choosing to go beyond minimum requirements or not. Both managers and their staff agreed that the manager ‘managed’ training. They were described as sometimes shopping around, for example for a ‘knowledgeable good trainer, with appropriate credentials’. However, one care home manager prided themself on being a prudent purchaser, noting, for example, their view that the language used in training needs to be in ‘layman’s’ terms, as specialist knowledge tended to go ‘over the heads’ of younger, less experienced staff. In the experience of a former regulator, managers often ‘shopped locally’ for training as it was cheaper.

There were very few examples of care home staff feeling that they could influence training purchasing. One care home worker in the control group commented that they would have liked more training, but ‘X [name of manager] is in charge, I do what s/he says’. For some of the care homes that were owned by large corporate providers, their managers’ discretion about purchasing training was minimal. One of these, a manager of a care home in the experimental arm of the ResCare trial, commented that in their care home’s company, the national training manager decided whether or not specific training was required across all homes, and then ‘every home has to have it’. This manager expressed surprise that free licences from ResCare were not taken up, but seemed to have passed on responsibility for training to head office: ‘I leave that to the training department’.

Those who were familiar with the world of the corporate care home described how there was little discretion over the training that homes provided for staff. The exception to this was when something had to be commissioned quickly as a response to some failing or perceived problem. Such prompts could emerge from various directions.

The economy of care home systems was often used by stakeholders to explain why care homes operated as they did. For some homes, aspirations to develop training could evaporate if there was no money left in the annual training budget. However, as commented by an experienced trainer interviewed, training could also be part of the marketing budget, or viewed as such:

There is certainly more of (that) – we are marketing ourselves as having this specialism and this specific expertise and therefore everybody will be trained in that . . . That is the way we will sell ourselves.

MD

There was a mix of views expressed about the potential for training during times when care homes were preoccupied with other changes. Generally, it was agreed that during substantial changes (such as change of management or ownership), there would be turmoil and worry about its possible impact. However, changed ownership, management or a new manager was seen to be likely to want to introduce new things to establish their authority. One regulator interviewed commented:

New face – new initiatives are more acceptable than old face – new initiatives – if you know what I mean.

ER

A trainer reported being brought on board at the time of a company takeover of a care home when the new managers were keen to ‘brand’ their new acquisition:

I’ve come across services which will say, we bought up that service and I’ve visited those services with corporate managers and they have said, we are now putting them on board with our dementia methodology whatever it is. Therefore, the staff are now beginning that route of training, so we will start with the managers and then . . . it’s bringing them on board once that company has been subsumed.

MJD

Limitations to this evaluation

This process evaluation had its limitations. First, the retrospective nature of this exercise meant that there may have been some inaccurate recall by participants over time. Second, although the sample of case study homes was diverse, implementation barriers may have been missed, as staff members who did not attend the training as champions and who would have views about whether learning was shared or cascaded through care homes were not interviewed in large numbers. Third, the sample of stakeholders interviewed may have been tempted to overgeneralise, or to report their own opinions about training and care homes.

Aim

To describe the characteristics and resource use over 6 months of a cohort of people with dementia and CB living at home, and their carers, referred to specialist NHS services across England; and to elicit stakeholder views on CB service provision and the findings from the cohort study.

Design and methods

Cohort study of people with dementia referred for CB to six NHS mental health organisations. Participants were people who met the diagnostic criteria for dementia and CB and their carers (dyads). The primary outcome measure was the RMBPC at baseline, 2 and 6 months; and the extent and cost of formal and informal care – using an adapted CSRI and NHS records of contacts with specialist mental health practitioners. Secondary measures included quality of life for the person with dementia and the family carer; and distress, guilt, mood and coping (sense of competence) in the family carer. Stakeholders debated emerging findings.

Results

Over 15 months we recruited 157 dyads (154 included family carers), in which 61% of people with dementia had mild dementia with clinically significant CB; we followed up 126 and 117 dyads at 2 and 6 months, respectively. Dyads received an average of nine contacts from mental health practitioners over 6 months, but there was little overall change in levels of CB. Increased contact with practitioners significantly reduced levels of guilt (p = 0.016) among carers. There was significant variation in trends for CB among dyads, but no stable clusters of those who improved, remained the same or deteriorated over time were identified. Family carers estimated that they devoted a mean of 112 hours a week to providing care at baseline, rising, though not significantly, to 129 hours at 6 months. They contributed over 80% of the total estimated cost of care. Stakeholder consultations revealed concerns about the equity of access to CB services for these carers.

Conclusions

People living at home with mild dementia can present with clinically significant CB. CB fluctuates for some, even over a short 6-month period. Families require trained practitioners, irrespective of where dementia service pathways are located, to systematically assess their varied needs and provide timely patient-specific interventions. Commissioning practice should reconsider the priority given to specialist assistance for families experiencing CB.

Trial registration

The ISRCTN is 58876649.

Rationale for and background to the present study

What is not clear from the literature outlined above is the extent of the variation of carer characteristics and their reports of BPSD. 242 This is an important clinical question given our definition of CB in dementia (see Chapter 1) that takes into account distress in both the person with dementia and/or the carer to provide interventions for the management of dementia and CB (see Chapter 2). What is also unknown is whether or not current health- and social-care support, provided early on in the development of CB, that is, while the person with dementia is living at home and the family has been referred to specialist services for mental health support for the management of dementia and CB, makes any difference over time to family carers.

The present longitudinal study was conducted in specialist NHS mental health organisations across England, where the delivery of specialist services for people with dementia and CB was co-ordinated by CMHNs working within specialist multidisciplinary CMHTsOP. During the course of this study, changes within some of the organisations resulted in extending this remit to other professionals, such as occupational therapists, working in these CMHTsOP (see Appendix 2). In this chapter, we describe professionals who deliver specialist NHS mental health support as specialist mental health ‘practitioners’. Additionally, the terms ‘carer’ and ‘family carer’ are used interchangeably in place of ‘informal carer’, as only three participant dyads did not include family members.

This study explores, in detail, the noted variation in carer responses to behaviour in the person with dementia. We examine whether or not it is possible to determine subgroups of carers who might improve, remain stable or worsen in terms of their stress, coping and quality of life. We also examine the impact of the support delivered by specialist mental health practitioners; and the extent of health- and social-care service support over time.

Finally, we examine stakeholder views about CB service provision and the findings from the cohort study.

Changes to protocol

Mental health practitioners were trained to use functional analysis-orientated interventions for CB in people living at home with their family, with a decision support system (see Chapter 2). Appendix 2 outlines detailed changes to the FamCare trial protocol, from a planned CRT to a controlled feasibility trial and then to the present observational naturalistic cohort study.

This was because recruitment to the study across all NHS organisations was slow and we could not therefore proceed with an intervention study. Prior to the study, all organisations had provided estimates of participant recruitment potential based on their recorded figures for performance of the CMHTOP in previous years. Reasons for slow recruitment were as follows.

1. The mental health NHS organisations were in the final preparatory year for the PbR initiative121 and all practitioners were engaged for the first time in using online electronic systems for the assessment and clustering of ‘need’243 for every new patient. Many community mental health services for older people were undergoing redesign,244 in a context of widespread uncertainty,245,246 which could be bewildering and disruptive for some,247,248 whereas others continued to try and make sense of this in their practice. 249

2. Triggered by the National Dementia Strategy of 2009250 was the growing national move towards dementia-specific services, with early diagnosis of dementia carried out in memory clinics or Memory Assessment Services (MASs). During this study period, some NHS organisations delivered all mental health services to older people within the CMHTsOP, others had memory clinics for diagnosis and management of early dementia, with the CMHTsOP providing support to older people with dementia later in the pathway, and yet others were redesigning services to develop either MAS provision within their CMHTOP or new memory clinics. This may have affected thresholds for acceptance as a case by the CMHTOP. At the start of recruitment in August 2010, continuing for an average of 31 weeks (SD 6.00 weeks; range 14–46 weeks), 33 CMHTsOP across seven NHS organisations received 5360 new referrals (see Appendix 3, Table 66), of which only 452 (8.4%) patients were potentially eligible for recruitment to this study of dementia and CB in family settings (see Appendix 3, Table 66). About one-quarter (25.8%) of new cases, that is, 1385 people referred for specialist mental health care, were not accepted by the CMHTOP (see Appendix 3, Table 67). They could therefore not be approached by the research team, as to be eligible for the study the referral had to be accepted by the CMHTsOP. The criterion of recruitment through CMHTsOP was extended to include redesigned CMHTOP services, in which new MASs and memory clinics were emerging in some NHS trusts. Consequently, 16.6% of the FamCare study cohort was recruited from MASs and memory clinics.

3. In 2011/12 the contract for delivering services in one NHS trust was in the process of transfer to another NHS organisation; the reorganisation of roles and responsibilities in this NHS trust resulted in withdrawal of all five of their CMHTsOP from the FamCare study. Although not directly affecting this study, but in order to provide context for this organisation with its five CMHTsOP, we further observed that recommissioning of services for the locality covering this NHS trust occurred for a second time in 2014 with transfer to yet another NHS organisation in 2015.

Research questions

The main aims of this study were to investigate the following questions in respect of people with dementia and CB, living at home, supported by a carer:

1. Do levels of reported CB, and carer reaction to this, change over time as measured by the frequency and reaction domains of the RMBPC?111

2. Does the level of support offered by usual care, determined by the number of specialist mental health care service contacts and time spent with the family, influence CB, family coping and/or quality of life of people with dementia and their family carer?

3. What are the predictors of change in CB, measured by the frequency and reaction domains of the RMBPC and the NPI total and distress domains?

4. What are the patterns of health- and social-care service use and associated costs?

5. What are the estimated costs of informal family care?

6. What are the patterns and costs of prescribing medications?

7. What are the total costs of services, informal care and medication?

An additional aim was to elicit the views of stakeholders in relation to CB service provision in dementia care and the findings from the cohort study.

Design

A cohort of people with dementia and CB and their carers who were referred to 28 CMHTsOP in six NHS organisations was followed up over a 6-month period. The information collected focused specifically on CB in family care settings, the stress experienced by family carers, the quality of life of people with dementia and their carers and the range, frequency and cost of health- and social-care services (including prescriptions) accessed by participants recruited to the study.

A participatory design development process used consultations with practitioners and managers across the NHS organisations. At the end of data collection, emerging findings were discussed in groups at two further consultation meetings with a wider group of stakeholders. These occurred in July 2013 in two cities in northern England.

Governance and study approvals

Ethics permissions for the protocol and subsequent changes were applied for and granted by the York REC (REC reference number 09/H1311/28), and later Leeds West REC when York ceased to exist, between March 2009 and October 2013 (see Appendix 12).

Confirmation that Humber NHS Foundation Trust would sponsor the trial was provided in April 2009 by the trust’s R&D department. Site-specific approvals for the participating NHS sites were also obtained from all seven of the original participating NHS organisations’ R&D departments, before commencing local recruitment.

For a description of the management arrangements for this study, see Appendix 5.

Study population

Six community mental health NHS organisations in England were involved in the cohort study: Greater Manchester West, Grimsby, Humber, North East London, Oxford and Buckinghamshire, and Sheffield. Of these, Grimsby was a specialist mental health social enterprise and the rest were specialist community mental health NHS trusts. One NHS trust, consisting of five CMHTsOP, withdrew prior to data collection for this study because of the transfer of its contract for services to another NHS trust. However, this NHS trust provided data prior to its withdrawal to inform our review of all new referrals to CMHTsOP (see Appendix 3) and for our iterative consultations of practitioners and managers (see Design). The CMHTsOP were considered to be broadly representative of the national picture; a mixture of rural and city locations, spread geographically across the country and from a mix of affluent and deprived areas. Researchers undertaking data collection were based in NHS mental health services. Within all study sites, the researchers visited each of the CMHTsOP to introduce the study, explain the eligibility criteria, and to collaborate with practitioners to apply these criteria for study recruitment. Recruitment commenced in August 2010 and ended in October 2011, with follow-up data collection continuing until July 2012.

Eligibility criteria

The criteria applied at the point of recruitment are detailed below.

Sample size

Each of the six NHS organisations expected to recruit at least 30 dyads and, therefore, this convenience sample had a target of 180 dyads (person with dementia and their carer). Evidence from a previous relevant study65 suggested that approximately 40% of participants with dementia living at home may be lost to follow-up over a period of 6 months because of death, movement to other accommodation or care settings, or illness being too severe to allow for completion of follow-up assessments. If this loss was evenly spread, then 13% of recruited participants, that is, 24 out of the 180 dyads, would be lost by 2 months and 72 by 6 months, leaving a total sample of 108 by the end of the study.

Recruitment procedures

Community mental health teams for older people were asked to approach every new patient referred to them who potentially met the inclusion criteria. Each new patient notified to the research team was allocated a study identification number, whether or not they consented to take part in the FamCare study, or subsequently met the inclusion criteria. This enabled documentation of the flow of participants through each of the stages, as specified in the CONSORT statement. 251

Where agreed with team leaders, a member of the CMHTOP contacted the patient or carer who had been referred to them to explain the study and to ask permission to pass their contact details to the research team. When permission was given, a member of the research team then arranged to visit the person with dementia and their carer to explain the study further, provide information leaflets if not received already, and obtain written consent.

Informed consent

The MRC’s recommended ‘cluster representation mechanism’ (CRM) was adopted for this study. A CRM is an individual, body or mechanism that represents the interests of a cluster, a gatekeeper. In this study, the clusters were the mental health teams and so the team leaders for each were deemed to be the gatekeeper. This accords with the MRC’s guidance that the CRM must be independent of the research team so that they can act in good faith in the interests of the cluster. They had the right to exclude or withdraw the cluster from the trial at any time and to exclude or withdraw individual patients. Patients and family carers who met the inclusion criteria, and were referred to those teams, were approached for consent to be included in the study.

Participants had at least 24 hours to consider taking part in the study. In line with the British Psychological Society’s guidance on the Mental Capacity Act of 2015,134 we assumed that each person had the capacity to consent, and explained the research in an accessible manner, increasing their likelihood of being able to consent for themselves. For each dyad, separate informed consent was sought from the person with dementia and their carer. If the person with dementia was judged to be unable to provide informed consent, then the carer was asked to act as a personal consultee and to indicate whether or not they believed the person would wish to take part. It was made clear to participants that there would be no disadvantage to them if they chose not to participate. Only after people had been assessed for capacity in accordance with the Mental Capacity Act 2005,133 and they or their consultee had agreed that they would take part, were they eligible to participate in the study.

We treated consent as a continuing process rather than a single decision, and discontinued interviewing whenever participants indicated either verbally or in any other way that they did not want to engage, or when they became distressed by the assessments.

Ethical arrangements

Our original design of FamCare, as a trial of functional analysis, included a procedure for reporting SAEs to the chief investigator. On converting the trial to a cohort study, this procedure was no longer necessary, given that interviewing participants with the selected questionnaires carried very little risk of a SAE. Nonetheless, we continued to abide by other ethical obligations, such as being aware of potential safeguarding concerns.

Data collection procedures

After initial baseline data collection, there were two follow-up periods of data collection for this cohort study, at 2 and 6 months. For the first follow-up this ranged from 1.80 to 4.40 months (mean 2.44 months, SD 0.54 months) and for the second follow-up it was 5.77 to 9.50 months (mean 6.55 months, SD 0.72 months).

All interviews were conducted in the home of the person with dementia or their carer, unless they specifically requested an alternative location. The questionnaire measures were arranged into booklets, which facilitated their ease of delivery during the interviews. If a participant became tired, or if it was requested by participants or deemed appropriate by the researcher, an interview was occasionally broken off part-way through and then continued on another day.

As well as the data collected from participants themselves via interview, the number, type and duration of mental health practitioner contacts were also collected from internal patient administration systems, for the full period that participants were in the study. This was from baseline to the first follow-up and then from first follow-up to the second final time point.

Measures

(See Appendix 6 for details of the instruments and their scoring methods.)

Data management

Data for the study were collected in questionnaire packs completed by researchers with the person with dementia and their carer. For further details in relation to scanning and data verification see Chapter 3, Data management.

Data analyses

The main statistical analyses for the study applied SPSS version 20 to two participant populations: people with dementia and their carers. Health economic analyses used SPSS version 19.

Missing data

The researchers endeavoured to collect as many data as they could. However, some missing data were inevitable (see Chapter 3, Missing data, for an explanation of the methods used to maximise the available data). We did not impute data for the six people with dementia who died during the study. The first follow-up data sets were imputed simultaneously based on the demographics and corresponding baseline measure. The second follow-up data sets were imputed using demographics and the corresponding measure at baseline and first follow-up.

Outliers

We re-checked the data of all outliers identified during statistical analyses of the cohort, but found no reason to drop any from the data set.

Research question 1

Do levels of reported CB, and carer reaction to this, change over time as measured by the frequency and reaction domains of the RMBPC?

Initially, a linear mixed model (LMM) approach to the analysis of repeated measures was used to assess whether or not there was a significant change in the measures over the three time points. The covariates were baseline measures, NHS organisation, gender, age, carer type and CDR-SB.

As no significant changes were observed, we used a cluster analysis to examine if there were participants who were improving, remaining stable or declining, as measured by the RMBPC frequency and reaction over time. The RMBPC measures were summarised using orthogonal polynomial transformations, and linear and quadratic means. 259 We planned to use different methods of cluster analysis260 on the orthogonal polynomial transformations using those participants with complete data for the RMBPC. If relevant clusters were found, then, for those without complete RMBPC data, we planned to assign them to the cluster they were most similar to, using all available baseline data. The stability of the clustering was assessed by comparing the results from the method described above with those obtained from the fully imputed RMBPC values. We also planned a logistic regression analysis to identify baseline measures that predicted the cluster to which a participant belonged.

Research question 2

Does the level of support offered by usual care, determined by the number of specialist mental health care service contacts and time spent with the family, influence CB, family coping and/or quality of life of people with dementia and their family carer?

The levels of support were measured by the number and length of total contacts with the specialist mental health practitioners during the full period that participants were in the study, that is, between their baseline and first follow-up interviews, and between their first and second follow-up interviews (this measurement of ‘contact’ was taken from the NHS organisation patient administration electronic record and should not be confused with the CSRI data collected during the participant interviews and used in the health economic analyses). The contact data were used to determine their effect on the response measures in a LMM; the response at the two follow-ups were the response measure, the number and duration of contacts were the predictor, and baseline and first follow-up measures acted as the covariate. The response measures here were the primary measures of RMBPC frequency and reaction. The secondary measures were NPI total; for the person with dementia the EQ-5D, using both index and VAS methods, and the QoL-AD; and for the carers they were the GS and the SSCQ.

Research question 3

What are the predictors of change in CB, measured by the frequency and reaction domains of the RMBPC; and the NPI total and distress domains?

The four ‘behaviour’ response measures of RMBPC frequency and reaction, NPI total and distress were further analysed to determine whether or not predictors of change could be identified. We assessed the baseline measures and demographic variables (e.g. NHS organisation, gender, age, carer type, CDR-SB) for inclusion in the LMM used for question 1. The effect of any identified clusters from research question 1 was also included in the analysis. The assumptions of the model were assessed and, where they did not hold, we used transformations of the data to find the most appropriate model. A sensitivity analysis of the imputation results was also conducted.

The missing status of participants was categorised into three groups: complete, monotone missing (dropouts) and non-monotone missing (those who missed first follow-up, but returned for second follow-up). We used logistic regression to assess whether or not the missing status could be predicted from any of the factors and covariates measured at baseline. The variables that were identified to affect the missing status were included in the above models. 261

Recruitment to the study

In total, 289 dyads were considered for inclusion in the study, with 157 completing baseline assessment. Figure 13 shows the flow of participants through the study. Our consultations with practitioners and managers resulted in an awareness of their misunderstandings about the nature of CB in dementia in family care settings; and in-depth discussion to consider where potential participants might be located in their services.

FIGURE 13.

The FamCare study: participant flow through the study.

Consequently, 16.6% were recruited from MASs or memory clinics, as from March 2011 two trusts decided to extend recruitment to within their memory clinics as follows.

1. From May 2011 onwards, one trust, which was redesigning services to develop new memory clinics, recruited 60% of their total contribution to the FamCare study cohort from memory clinics.

2. From October 2011 a second trust, which had two longstanding memory clinics, recruited 28% of its contribution to the FamCare study cohort from one memory clinic.

Of the 132 participants who agreed for their details to be passed to the research team, but then dropped out prior to baseline data collection, 99 (34.3% of the total considered) declined to participate when contacted by researchers. Reasons for this are summarised in Appendix 13. This information suggests that some of the most distressed participant dyads had excluded themselves from this research programme that was specifically designed to help people such as them who may have been distressed because of dementia symptoms.

Follow-up dropout rates

In total, 126 dyads completed the 2-month (first) follow-up, 117 completed the 6-month (second) follow-up and, as three dyads had missed the first follow-up, there were 114 for whom data were collected at all three time points. Therefore, 20% (n = 31) of the initial 157 were lost to follow-up at 2 months; 10% out of 126 (n = 12) were further lost at the second follow-up, and with the three participants who were not observed at all the time points, by the end of the study there were 27% (n = 43) incomplete cases. Three dyads who were lost to follow-up at 2 months returned to the study at second follow-up. Therefore, the overall dropout rate was 25.5%, much less than the 40% predicted at the outset.

Descriptive data

Tables 43 and 44 outline the demographic characteristics of participants. Of the carers recruited to the study, 70.7% were living together with the person with dementia in the same household, 74.5% owned their own homes and most (64.3%) were retired (see Table 43). The majority of the people with dementia were female (59.2%), as were the majority of carers (70.7%). The mean age of people with dementia was 80.34 years (SD 7.66 years) and of carers it was 66.13 years (SD 13.06 years). The most common relationship between dyads (see Table 44) was spousal (52.9%).

As shown in Table 45, although all participants met the criteria for dementia using the DSM-IV, a very small number (n = 5) at baseline assessment were rated by the researchers as falling within the ‘no cognitive impairment’ range on the CDR. At the opposite end of the spectrum, there were relatively few (n = 14) who were scored as having severe dementia.

In the autumn of 2014 the Alzheimer’s Society published an update of its 2007 study on the social and economic impact of dementia in the UK. 235 It did not identify any further UK evidence on the distribution of dementia by severity compared with its estimates in the 2007 Dementia UK: The Full Report265 and, therefore, concluded that estimates for dementia severity remained relevant as follows: 55.4% mild dementia, 32.1% moderate dementia and 12.5% severe dementia. Based on the CDR, the proportions in each group are very similar to those we found in the present FamCare study (see Table 45).

Examination of change over the 6 months of the study (Table 46) indicates that most people either stayed the same or moved up to the next level of severity on the CDR.

Reported problems at baseline

The RMBPC ratings of the daily concerns of carers show that the most commonly reported problems cited by carers (Table 47) were associated with forgetfulness, including asking the same question and misplacing things. These can be triggers for escalation of upset and other CB in families who have not learned how to respond effectively; the least common ones were destroying property and threatening to hurt themselves or others.

Main outcomes over time

Table 48 outlines the outcome measures at each time point. From the sample of 157 participants with dementia, 122 completed some of the self-report measures at baseline.

The average NPI composite score (see Table 48) of 25.75 (SD 19.17) at baseline is strikingly high and is comparable to that found in smaller studies conducted in Spain240 and Australia. 266 There was no discernible change over the 6 months on this measure, or on the RMBPC, which was our primary outcome measure.

The self-reported baseline QoL-AD score averaged at 37.2 (SD 5.3), comparable to values from the REMCARE (REMiniscence groups for people with dementia and their family CAREgivers) study of reminiscence therapy for community-dwelling people with dementia and their carers,267 in which the intervention group average score was 37.5 (SD 5.32) and the control group average score was 37.0 (SD 5.35). As with the REMCARE study,267 carer (proxy) ratings of quality of life of the person with dementia were lower than self-reported ratings, providing further support for the view that patient and proxy ratings of quality of life in people with dementia do not concur with each other;268 perhaps adding weight to findings from another study of BPSD that families may be more aware. FamCare study participants with dementia had a mean EQ-5D index self-reported score of 0.71 (SD 0.28) and carers had a mean self-reported index score of 0.81 (SD 0.21). These values were similar to those found in the REMCARE study. 267 Self-reported scores for participants with dementia were also similar to the age-matched UK general population norms. 161 Although not statistically significant, carers aged ≥ 75 years had a slightly higher average EQ-5D index score (mean 0.79, SD 0.23) than the general population aged ≥ 75 years (mean 0.73, SD 0.27), but younger carers had a slightly lower than normal average EQ-5D index; in the age group of 45–54 years, the average carer EQ-5D index was 0.81 (SD 0.24), compared with the population norm of 0.85 (SD 0.25).

Prescribed medication 3 months prior to baseline

Table 49 outlines the various groups of medications that people with dementia and their carers were prescribed in the 3 months before baseline, whether or not they were taking them for the full 3 months and whether they were on one or more medication in a category. From this, we can see that 8.9% (n = 14) of people with dementia were taking at least one antipsychotic at some point in the 3 months before baseline, 21.6% (n = 34) antidepressants and 6.3% (n = 10) hypnotics and anxiolytics (B/Z/A drugs). Furthermore, 29.9% (n = 47) were prescribed dementia drugs. Baseline data in Table 49 cover an important period between August 2010 and October 2011. This period follows the publication in October 2009 of the landmark report commissioned by the Department of Health, The Use of Antipsychotic Medication for People with Dementia: Time for Action,112 which recommended that reduction of the use of antipsychotics in dementia should be a clinical governance priority for primary and secondary care. Its author estimated that this could be done safely over a 36-month period. In March 2011, the Prescribing Observatory for Mental Health (POMH-UK) programme conducted an audit-based quality improvement programme of prescribing in dementia care within 54 specialist NHS organisations. 269 This noted that 64% of people with dementia lived at home and, of this group, 9.6% were prescribed an antipsychotic. Our data, which were taken from patient prescriptions rather than from audits carried out by staff within their organisation – and where some but not all NHS organisations had participated in the POMH-UK audit – show that 91% of our sample were not prescribed an antipsychotic and thus concur with the POMH-UK data. Disappointingly, the quality of prescribing was suboptimal, as very few people had been taking an antipsychotic for less than 3 months. This may be because participants had been managed in primary care prior to referral to specialist services for CB, and primary care physicians may be less confident about withdrawing patients from antipsychotics than their colleagues in secondary care. 269 Similar to other studies, a large number of participants with dementia (i.e. around 20%) were prescribed at least one antidepressant, but our data note a low likelihood of discontinuing in 3 months, despite evidence indicating limited or no benefit in the management of depression in dementia. 270,271 This pattern of suboptimal prescribing for the hypnotics and anxiolytic (B/Z/A) drugs is also seen, as 6.4% of participants continued with prescription for 3 months or more, contrary to guidelines advocating benzodiazepine prescription for no longer than 28 days. 178