Support at Home: Interventions to Enhance Life in Dementia (SHIELD) evidence, development and evaluation of complex interventions

Protocol and registration

The protocol was registered with The Cochrane Library and can be found online at http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD005562.pub2/pdf.

Criteria for considering studies for this review

We selected RCTs examining the effect of cognitive stimulation for dementia if they had been published and written in English, peer reviewed and presented in a journal article. Authors were contacted for missing data, such as details of randomisation, means and standard deviations (SDs).

Search methods for identification of studies

The search methods included a combination of the search terms cognitive stimulation, reality orientation, memory therapy, memory groups, memory support, memory stimulation, global stimulation and cognitive psychostimulation, which were used to search ALOIS on 6 December 2011. The studies were identified from the following databases:

• health-care databases: MEDLINE, EMBASE, Cumulative Index to Nursing and Allied Health Literature (CINAHL), PsycINFO and Latin American and Caribbean Health Sciences Literature (LILACS)

• trial registers: metaRegister of Controlled Trials, Umin (University Hospital Medical Information Network) Japan Trial Register and World Health Organization portal [which covers ClinicalTrials.gov, International Standard Randomised Controlled Trial Number (ISRCTN), Chinese Clinical Trials Register, German Clinical Trials Register, Iranian Registry of Clinical Trials and the Netherlands National Trials Register, plus others]

• The Cochrane Library’s Central Register of Controlled Trials

• a number of grey literature sources: ISI Web of Knowledge Conference Proceedings, Index to Theses, Australasian Digital Theses.

Additional searches in each of the sources listed above to cover the time frame from the last searches performed for the Specialised Register to December 2011 were run to ensure that the search for the review was as up to date as possible. A total of 670 references were retrieved from the December 2011 update search. After deduplication and a first assessment, authors were left with 94 references to further assess for inclusion, exclusion or discarding.

Data collection and analysis

Searches were conducted as detailed above to identify all relevant published studies. The date and time of each search, together with details of the version of the database used, were recorded. Additional information was sought, as outlined above, and hard copies of articles were obtained.

Quality assessment

Two reviewers independently screened the identified RCTs for inclusion and the final list of included studies was reached by consensus. Trials not meeting the criteria were excluded. The studies were assessed against a checklist of quality requirements using the Cochrane approach:

• grade A – ‘low risk’: adequate concealment (randomisation; concealed allocation)

• grade B – ‘unclear risk’: ‘randomised’, but methods uncertain

• grade C – ‘high risk’: inadequate concealment of allocation or no randomisation, or both.

Only trials with a grade A or B ranking were included in the review. Again, the reviewers worked independently to ascertain which studies met the quality criteria, and consensus was reached through discussion. Attempts were made to obtain additional information from the study authors when needed.

Data extraction

Descriptive characteristics (such as quality of randomisation and blinding) and study results were extracted, recorded and entered into RevMan 5.1 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark). Additionally, letters and e-mails were sent to some authors of controlled trials asking for essential and additional information (statistics, sources of bias and details of randomisation). The summary statistics required for each trial and each outcome for continuous data were the mean change from baseline, the standard error (SE) of the mean change and the number of patients for each treatment group at each assessment. When changes from baseline were not reported, we extracted the mean, SD and number of patients for each treatment group at each time point, if available. We calculated the required summary statistics from the baseline and assessment time treatment group means and SDs, assuming in this case a zero correlation between the measurements at baseline and assessment time. This conservative approach was chosen, as it is preferable in a meta-analysis. For binary data, the numbers in each treatment group and the numbers experiencing the outcome of interest were sought. The baseline assessment was defined as the latest available assessment prior to randomisation, but no longer than 2 months prior. For each outcome measure, data were sought on every patient randomised. To allow an intention-to-treat analysis, the data were sought irrespective of compliance, or whether or not the patient was subsequently deemed ineligible or otherwise excluded from treatment or follow-up. Discussion between the two reviewers and the other authors was used to resolve any queries.

Data analyses

RevMan 5.1 was used for the meta-analysis. Analyses were adjusted to the random-effects model, owing to the heterogeneity of trials. Because trials used different tests to measure the same outcomes, the measure of the treatment difference for any outcome that we used was the weighted mean difference when the pooled trials used the same rating scale or test, and the standardised mean difference (SMD) (the absolute mean difference divided by the SD) when different rating scales or tests were used. For binary outcomes, such as clinical improvement or no clinical improvement, the odds ratio was used to measure treatment effect. A weighted estimate of the typical treatment effect across trials was calculated. Overall estimates of the treatment difference were employed, presenting the overall estimate from a fixed-effects model and performing a test for heterogeneity using a standard chi-squared statistic. The reviewers achieved consensus on the interpretation of the statistical analyses, seeking specialist statistical advice from the CDCIG as required. Non-randomised studies were described in tabular form and the reviewers discussed and reached consensus on the presentation of the findings in the background to the review.

Studies included in the review

Ninety-four studies were identified since the last review through the literature search. 47 Two reviewers independently assessed eligibility. Of the 94 references, nine studies met the inclusion criteria48–55 and were included in the analysis. Three recent studies were left awaiting classification. 56–58 Our previous review47 included eight studies in the meta-analysis and six of these met the criteria for inclusion in this updated review. 59–64 Two studies from our previous review were excluded this time, as the data needed for the meta analysis were not available. 65,66 Therefore, a total of 15 studies was included in the analysis (Figure 1).

FIGURE 1.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart of the review and selection process of studies.

Quality of included studies

The quality of each study was assessed according to the four criteria outlined in the Cochrane Collaboration Handbook:67 selection bias, performance bias, attrition bias and detection bias. The details of biases and descriptions of studies can be seen in Table 1. Performance bias was difficult to evaluate. With psychological interventions, unlike drug trials, it is impossible to totally blind patients and staff to treatment. Patients are often aware that they are being treated preferentially, staff involved may have different expectations of treatment groups and independent assessors may be given clues from patients during the assessments. There may also be ‘contamination’ between groups, in terms of groups not being held in separate rooms and staff bringing ideas from one group to another.

The latter effect would be reduced with clear therapeutic protocols, the existence of which was not mentioned in any of the studies, although during a face-to-face meeting with Professor Woods he stated that ‘Checks were made to ensure compliance with the therapeutic protocol’ (Bangor University, 2008, personal communication).

In relation to contamination, Wallis et al. 63 and Baines et al. 59 both stated that staff were unaware of the allocation of patients to groups, as they were removed from the setting for treatment, and several other studies described the groups being run in a separate or specific room. 6,55,62,64 Most studies took steps to ensure that at least part of the assessment of outcomes was carried out by assessors blinded to treatment allocation. Only three studies48,60,62 did not report the blinding of assessors. Of course, even independent assessors may be given clues from participants during the assessments, but this was not reported as an issue in the studies reviewed here. Using independent assessors works well for evaluating change in cognition or self-reported mood, well-being and quality of life. Ratings of day-to-day behaviour and function are typically carried out by care staff, who may be more difficult to keep blind to group allocation, unless the group sessions are carried out in a separate location to which all participants are taken. Whether or not the patients were blind to treatment is a controversial issue, depending on how much information was given to them and their level of comprehension.

Several studies reported48,62–64 that the reality orientation groups were held in separate areas, reducing the chance of contamination. Spector et al. 6,55 said that the groups were run in a separate and specific room for the programme. The other studies did not provide information regarding where groups were held. 49,51,53

All of the studies reported on attrition to the programme, although the intention-to-treat analysis plan was described in only two. 6,51 Given the nature of the condition and the age of the participants, attrition in several studies was remarkably small, with zero attrition recorded in six studies48–50,52,59,60 out of 180 participants. The largest attrition rate was reported in the study by Wallis et al. ,63 in which there was 39% attrition in the group of participants with dementia. In this study, patients who attended < 20% of the group sessions were eliminated from the study. Requena et al. 54 reported 32% attrition, but this was over a 2-year period. The two largest studies had rates of 19%59 and 17%6 over periods of 6 and 2 months, respectively.

Detailed treatment protocols were hard to find from the authors of the included studies, so the extent to which the cognitive stimulation was delivered as intended could be questioned. Some recent studies described that staff received training and/or supervision in running the groups. Chapman et al. 51 described weekly meetings held to ensure that their treatment programme was implemented as designed:

Sub-groups were led by a licensed speech-language pathologist and three masters level speech-language pathology students; all underwent two hour training before the groups started; weekly meetings to ensure the programme was implemented as designed.

Onder et al. 53 described how family caregivers were trained by a multidisciplinary team and given a manual and specific schedule for each session. No records were made, however, of how often caregivers delivered the sessions or how closely the manual was followed. The only available data from an early study came from Professor Woods, who stated during a face-to-face meeting that ‘A sample of sessions were tape-recorded and rated to ensure compliance with the therapeutic protocol’ (Bangor University, 2008, personal communication).

Meta-analysis

Data from the included studies were entered into ‘Metaview’ (the Cochrane term for meta-analysis). Data were identified, included and pooled from the 15 included RCTs,6,48–55,59–64 with a total of 718 participants (407 in experimental groups and 311 in control groups). Analyses were adjusted to the random-effects model, owing to the heterogeneity of trials, and SMDs, because trials used different measures to assess the same outcomes.

In order to evaluate the effect of cognitive stimulation on cognitive function, 14 RCTs that included useable data were included in the analysis (n = 658; 377 received treatment and 281 received no treatment or placebo). In comparison with the control groups at the post-treatment assessment, cognitive stimulation was associated with significant improvements on all measures of cognition. The overall results in the cognitive section were significantly in favour of treatment (Figure 2). The overall effect size (SMD) was 0.41 [95% confidence interval (CI) 0.25 to 0.57]. The results were strongly weighted by Breuil et al. 61 (n = 61), Onder et al. 53 (n = 137) and Spector et al. 6 (n = 201), the largest studies. The largest effect sizes can be seen at the 12-month point in Requena’s et al. 54 study [SMD 0.70 on the Alzheimer’s Disease Assessment Scale – Cognitive Subscale (ADAS-Cog)] and the Baldelli et al. 60 study (SMD 0.99 on the MMSE), both of which offered above-average duration of exposure to cognitive stimulation. Other studies with longer exposure62 had below-average effect size and other studies which offer only 10 hours of exposure had above-average effect size (0.63), showing that these effects require replication, as the CIs are broad and cross zero.

FIGURE 2.

Forest plot of comparison for CS vs. no CS: outcome – cognition. ADAS-Cog, Alzheimer’s Disease Co-Operative Study–Cognitive Subscale; CAPE I/O, Clifton Assessment Procedures for the Elderly – Information/Orientation scale; CERAD, Consortium to Establish a Registry for Alzheimer’s Disease; CI, confidence interval; CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance; RCP, Royal College of Physicians.

To evaluate the effectiveness of cognitive stimulation on behaviour, three separate meta-analyses were conducted. One focused on outcome measures seen as a problem, the second focused on ADLs and the third focused on general behaviour outcomes. Three studies used behavioural outcome measures seen as a problem (Figure 3) including 166 participants, showing that no difference was apparent related to cognitive stimulation (SMD –0.14, 95% CI –0.44 to 0.17; z = 0.86; p = 0.39). The ADL measure results also did not achieve significance (Figure 4), including four studies, involving 260 participants. There was no benefit identified associated with cognitive stimulation (SMD 0.21, 95% CI –0.05 to 0.47; z = 1.56; p = 0.12). The third behaviour analysis (Figure 5) (general behaviour) showed a similar picture, with no difference emerging. Eight studies reported data on relevant scales, including 416 participants (SMD 0.13, 95% CI –0.07 to 0.32; z = 1.30; p = 0.20).

FIGURE 3.

Forest plot of comparison for CS vs. no CS: outcome – ADAS-Cog. CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance.

FIGURE 4.

Forest plot of comparison for CS vs. no CS: outcome – MMSE. CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance.

FIGURE 5.

Forest plot of comparison for CS vs. no CS: outcome – other cognitive measure (information/orientation). CAPE-I/O, Clifton Assessment Procedures for the Elderly – Information/Orientation scale; CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance; RCP, Royal College of Physicians.

Five studies, involving 201 participants, used a self-report measure of mood (the Geriatric Depression Scale or the Montgomery–Åsberg Depression Rating Scale) (Figure 6). Cognitive stimulation is not associated with a clear improvement in mood (SMD 0.22, 95% CI –0.09 to 0.53; z = 1.42; p = 0.16). This is of similar magnitude to the finding of proxy report of mood and anxiety, where the SMD is close to zero (SMD 0.05, 95% CI –0.21 to 0.31) (Figure 7).

FIGURE 6.

Forest plot of comparison for CS vs. no CS: post treatment, outcome – communication and social interaction. CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance; MOSES, Multidimensional Observation Scale for Elderly Subjects.

FIGURE 7.

Forest plot of comparison for CS vs. no CS: outcome – GDS. CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance.

Four studies included self report well-being and quality-of-life measures (n = 219) (Figure 8). The analysis showed a significant improvement on this outcome measures following treatment, compared with control groups. The SMD was 0.38 (95% CI 0.11 to 0.65; z = 2.76; p = 0.006).

FIGURE 8.

Forest plot of comparison for CS vs. no CS: outcome – quality of life. CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance; QOL-AD, Quality of Life – Alzheimer’s Disease.

Follow-up

Owing to the great variety in follow-up data, the analyses were divided in two sections: short- and long-term follow-up. Short-term follow-up analysis included Baines et al. 59 and Wallis et al. ,63 who had a 1-month follow-up, and Baldelli et al. ,60 who reported data from a 3-month follow-up. Long-term follow-up data included Chapman et al. ,51 who reported useable data only from a 10-month follow-up, a much longer period in the context of the progression of dementia. For cognitive measures (Figure 9), the three studies with short-term follow-up reported data for 52 participants. The significant advantage for cognitive stimulation on cognitive measures seen immediately post treatment remained at this point (SMD 0.57, 95% CI 0.01 to 1.14; z = 2.00; p = 0.05). For the 54 participants included by Chapman et al. 51 (Figure 10), there was no significant effect on either the MMSE (SMD 0.18) or the ADAS-Cog (SMD 0.12) at the 10-month follow-up. No other significant results were found in the other outcome measures at either the short- or the long-term follow-up analysis.

FIGURE 9.

Forest plot of comparison for CS vs. no CS: post treatment, outcome – mood, staff-reported. CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance; RAID, Rating of Anxiety in Dementia.

FIGURE 10.

Forest plot of comparison for CS vs. no CS: outcome – ADLs. CS, cognitive stimulation; df, degrees of freedom; IV, inverse variance.

Development of maintenance cognitive stimulation therapy sessions and workbook

We selected all relevant and effective interventions from the CST Cochrane review. We contacted study authors with the aim of obtaining additional information, hard copies of the intervention programme manuals and the manuals translated into English. The research team developed a database identifying key themes from the initial CST manual71 guiding principles and sessions, and included the 16 sessions developed for the MCST pilot project. A draft manual (version I) was produced based on the results.

The results from the analysis of the cognitive stimulation manuals, including the original papers, manuals and table database, were presented in a consensus workshop (comprising key academics, research staff and clinical staff involved in CST practice) and used to validate and review a subsequent draft of the MCST manual (version II), which included 24 maintenance sessions and was produced using the results of the consensus workshop. The MCST manual (version II) was discussed in focus groups with care staff (three groups), carers (three groups) and people with dementia (three groups) to review key themes, feasibility and potential modifications. The revised MCST manual was further modified in consultation with the attendees of the consensus workshop and key contributors to other aspects of process (e.g. a sample of care staff, carers, clinical staff and leading experts on CST).

The next draft of the MCST manual (version III) was produced to publication quality and was used in the development of a revised version of the CST training package compromising the revised manual, a CST/DVD including extra maintenance sessions, a PowerPoint^® 2003 (Microsoft Corporation, Redmond, WA, USA) presentation, methods of evaluation and adherence. We developed the training package in consultation with trainers from Dementia UK.

Focus groups: a comparison of the views of people with dementia, staff and family carers

We included qualitative testing of the intervention through focus groups, developing a consultation process with people with dementia, family carers and staff. The aim was to identify improvements for the draft version of the MCST manual. Focus groups were the natural choice, as the aim was to gain comprehensive views of the key stakeholders with regard to the MCST programme. A key strength of focus groups is their ability to ‘tune in’ to the attitudes and perceptions of users regarding services.

Sample

Focus groups were undertaken separately with three user groups that consisted of key stakeholders in the project. Three focus groups were carried out with people with dementia, three groups were carried out with staff and three groups were carried out with family carers of people with dementia. In total, 17 people with dementia, 13 staff and 18 family carers took part in separate focus groups. In the people with dementia groups, there were eight men (47%) and nine women (53%). Their mean age was 78 years, they all scored mild to moderate on the CDR scale46 and were able and willing to consent to participate in the focus groups. The staff groups consisted of three men (23%) and 10 women (77%), with a mean age of 36 years. All were permanent, paid staff, with at least 1 month’s employment, from residential homes, day centres or day hospitals. Their main duties involved caring for people with dementia. The family carers consisted of six men (33%) and 12 women (67%), with a mean age of 53 years. All were former or current carers of a person with dementia and had at least monthly contact during their time as a family carer.

Purposive sampling, for example sex, length of caring experience and dementia type, was used to ensure a wide range of participants. The centres selected were typical in size, organisational structure and management of others in the area. When people were recruited via local carers’ organisations (Uniting Carers for Dementia and Alzheimer’s Society), permission was also obtained from the management committees of these groups.

Procedure

We used the Noticeable Problems Checklist72 to screen potential participants who were then approached for their consent and further screening using the CDR scale. Participants with a CDR score of mild to moderate were included, provided that they did not meet the exclusion criteria (diagnosis of severe learning disability and/or diagnosis of depression, anxiety or other mental or physical illness adversely affecting their ability to participate in focus groups) and were willing to take part in the discussion groups. Seventeen participants met the inclusion criteria and agreed to take part. Members of staff who were working directly with people with dementia in the different centres approached were invited to take part in the groups. Eighteen family caregivers were recruited through Uniting Carers, Dementia UK and the Alzheimer’s Society. Participants were reminded that they could leave the group at any time.

Two researchers conducted self-contained, hourly focus groups. The sessions included a brief presentation about the overall project. One facilitator led the interview, with the second person actively listening and seeking clarification, to ensure the adequacy and accuracy of content as appropriate. 73 The second facilitator also took substantive and methodological field notes during and immediately after the focus groups, as recommended by Burgess. 74 A semistructured interview schedule based on the CST empirical literature was used as cues for open questions. The groups focused on the 24 themes developed for the MCST programme and the cognitive stimulation definition by Clare et al. 11

Design

Focus groups were chosen in preference to individual interviews, as they are useful for stimulating discussion, generating ideas to explore topics in depth, gaining insight and obtaining rich data. 75 A focus group interview schedule was constructed and developed to provide a framework for the discussions, which was piloted and adapted. The participants were encouraged to express their opinions and discuss issues through a series of open questions, which covered various aspects of mental stimulation activities, beginning with a presentation of the programme on a DVD and followed by open questions designed to elicit the activities and themes that participants particularly enjoyed and factors that contributed to the programme activities being stimulating and meaningful. Questions included ‘what do you think about use it or lose it/mental stimulation?’ and ‘could you tell us a bit about what sort of things you do that you find as being mental stimulating for you and you enjoy doing them?’. Pictures and materials used in the different themes and activities were used to help stimulate discussion in the participant groups. All groups used the same schedule, and participants were asked their opinions of what activities/themes they thought would be successful among people with dementia.

Analyses

The focus group interviews were tape-recorded and transcribed. The facilitators then reviewed the transcripts. The notes taken by the second facilitator were consulted to clarify the context of the discussion (e.g. benefits of being part of a group that was expressed by people with dementia focus groups). We used an inductive (data driven) thematic analysis approach76 to code and analyse the data. Inductive analysis allows themes to emerge from the data and is useful when the intent is exploratory and descriptive, as in our case. Consistent with various qualitative research methods, the focus group inquiry allows participants the freedom to provide information that does not necessarily fit with any expectation/hypotheses going into the research. This openness to new and unexpected information allows measurement designers to more fully ‘ground’ the content of new patient-reported outcomes in the concerns and issues that participants think are relevant. 77

To develop a thematic codebook, researchers immersed themselves in the transcripts and rereading to gain deeper understanding of and familiarity with the content. One transcript was revisited and analysed into exclusive in vivo categories; that is, one theme was applied to one unit of meaning and the words used by the participant were adopted as the label of the theme. These themes were then applied to the remaining transcripts using a method of constant comparison to refine themes within the coding manual. During this process themes became more conceptual, with some original themes split into subthemes and others linked under one category depending on the emphasis placed on each theme within the transcripts. The resulting coding manual provided a meaningful way of understanding the views of the participants and assessing similarities and differences across the different groups (people with dementia, staff and family caregivers). Using the final codebook, all transcripts were coded independently by both researchers and then compared to reach 100% consensus.

Mental stimulation: ‘use it or lose it’

This included discussing their opinions regarding mentally stimulating activities, ‘use it or lose it’, in terms of their views and beliefs about the effects of keeping the brain active. All participants showed general agreement about the usefulness of keeping the brain active.

People with dementia expressed the view that keeping the brain active was very important and a way of relieving frustration. They thought that it was essential for a healthy life and preserving their mental abilities, and that engaging in activities would keep the brain going. Some family carers expressed the view that the need for mental stimulation was universal and could bring neurological (building connections in the brain) and mental (helping with mood, anxiety, depression) benefits to everyone, and was important for promoting a healthy lifestyle and well-being. Staff and family carers also thought that there were added benefits to people with dementia in terms of increasing confidence, giving a sense of achievement, satisfaction, retaining skills and enjoyment.

No participants with dementia expressed negative views about the value of ‘use it or lose it’; however, there were concerns about the importance of keeping the brain active from members of staff and family carers groups who gave examples of individuals for whom the idea of ‘use it or lose it’ did not apply. Famous writers and politicians were given as examples of people who maintained a mentally stimulating, active lifestyle but nevertheless developed dementia. Some family caregivers expressed concerns that mental stimulation programmes may result in people with dementia losing confidence, experiencing anxiety or a sense of inferiority if confronted with their own cognitive deficits and difficulties as a result of undertaking a challenging mental activity.

Perceived stimulation in everyday life

Listening to music, singing and dancing, reading, painting, drawing, cooking and knitting were highlighted as being important for people with dementia. Factors that made an activity important included being interesting, being enjoyable, having a relaxing effect and helping to pass the time. Reading was a popular activity for most people with dementia as it raised their confidence and helped to increase interaction and participation when part of a group. In particular, talking and listening to others appeared enjoyable and were highly valued among people with dementia. They thought that talking to another person, to a pet or to the TV maintained their links with important current and past relationships and helped to reduce feelings of isolation.

Being part of a group helps considerably. I think being left on your own is not as effective as being part of a group. I belong to an African Caribbean group and I find that very stimulating. We talk on a number of things what we have done and why we did it. I think talking is very important.

PWD: FG1; 175–179

Family caregivers perceived that activities involving music (e.g. listening to music, singing, tapping and clapping) were those that people with dementia enjoyed the most.

Sounds and music are very important to stimulate something that’s there already so they recognise . . .

FC: FG2; 183–184

Those in the family caregivers group also spoke of having the opportunity to enjoy dinner together and to reminisce together by looking at photographs or sharing memories.

They learned a lot about others, everyone was telling their old past stories of how they went to school in shared shoes and things like this as a family. And we got so much information and it really stimulated them.

FC: FG2; 232–238

Staff caregivers thought that their planned activities were the most valuable stimulating activities for people with dementia. Staff participants identified reminiscence as an activity people with dementia enjoyed and often engaged in. However, there appeared to be little understanding of its value, and concern that it did not relate to the present.

I think we have to be careful with reminiscence, for me one of the best bits of this therapy is the variety of activities that you are going to do, otherwise it can get very repetitive as you tend to do things that you know work well, like reminiscence. You have to set goals within every session . . .

SC: FG1; 60–66

Factors influencing success of a cognitive stimulation programme

The perceptions of what made a CST programme successful were based on the person’s values and beliefs, with their interests and routines reinforcing a sense of identity and sense of belonging. Staff and family caregivers appreciated that the philosophy of the programme should be person centred and that enjoyment was a measure of what made programme activities successful.

People with dementia stated that basic human courtesies were important: being kind, trying to make others happy and not underestimating participants’ abilities.

Nothing that involves cruelty. As long as there’s kindness you can’t fault it.

PWD: FG1; 366

Being able to discuss, learn and make contributions was also mentioned.

I think it is very important to learn new things, you don’t stop learning ‘till you die.

PWD; FG1; 669–672

Other factors highly valued in the groups were activities that included reminiscence as an aid to orientation and activities that were provided in a multisensory way. Activities involving discussion and sharing opinions among the group were highly valued, as were challenging activities and quizzes requiring right or wrong answers.

Some people will remember more things than others you know. But it’s good for the brain . . .

PWD: FG1; 576–579

In contrast, family caregivers and staff participant groups stated that activities in the programme should not be based on right or wrong answers, so that people with dementia did not feel under pressure. ‘Playing it safe’ was perceived as being very important and this helped people feel more secure.

You have to adapt to the person and never ask them to do anything they can’t do because they have a sense of self and it will give them a sense of inferiority or inadequacy.

FC: FG2; 90–94

Most staff acknowledged that it was important to identify participants’ individual preferences, skills and abilities, as this affected their level of engagement in activities, whereas some recognised the need to adapt activities to a participant’s capabilities as a way of providing choice and contributing to their well-being. Some family carers stressed the importance of providing this programme to people only in the mild-to-moderate stages of dementia, as they thought that it would not be appropriate for people in more advanced stages.

The group participants should be of [a] similar level of dementia.

SC; FG2; 176

They also thought that group participants should be chosen based on having similar abilities and interests so that the group could run successfully and be stimulating and enjoyable. Both staff and family carer groups noted that attention needed to be paid to each participant’s level of hearing and vision, as they thought that high levels of impairment in either would limit a person’s participation.

You have to think about the personality dynamics within each group.

SC: FG2; 109

Staff and family caregivers also indicated that the group facilitator’s skills, knowledge, understanding of dementia and attitudes towards participants was also key to running the group effectively. The need for several facilitators was mentioned, as was limiting the number of group participants. Appropriate equipment was identified as another key factor to the success of this programme.

I think the size of the group has to be pretty small as an important factor.

FC: FG2; 388–391

Presented themes

A total of 19 themes was presented to the focus groups of people with dementia, family caregivers and staff. Fourteen themes were from the existing CST programme. Five new themes were developed from the literature review and the pilot MCST study (Table 2).

The five new themes

People with dementia rated useful tips, thinking cards and using objects as very positive themes. They stated that these themes were good for learning, hearing other people’s opinions and giving their own opinions in the group. They thought that the themes could help the group cohesiveness and would trigger conversation. Family caregivers and staff groups also thought that useful tips and using objects were good themes for the session and highly valued the involvement of reminiscence in the activities as an aid to orientation. Some staff expressed concern that the thinking cards theme would not work but others challenged that it would. Some family caregivers thought that the proposed activities for this theme would not be appropriate, as some people did not like ‘closing your eyes and imagining’ and might feel uncomfortable with this. Other carers liked this theme and thought that the questions were a good way of stimulating conversation and possibly helping group cohesion.

People with dementia rated visual clips discussion and art discussion as neutral. Although they liked the idea of group discussion, they did not feel enthusiastic about the topics presented. Staff and family caregivers groups rated these themes positively, as visual prompts were highly valued and they liked the idea of promoting discussion. Some staff had successfully run these types of activities previously with people with dementia and advised that materials be chosen appropriately.

Existing cognitive stimulation therapy themes

Participants were asked to rate and organise the presented themes as very positive, neutral or negative, and to rank them in order of those themes perceived as most and least successful. My life (childhood and occupations), food and orientation were perceived as positive themes; they were applicable to everyone, helped people to keep in touch with themselves and were multisensorial. Quizzes and word games were rated highly among people with dementia, who thought that they were very stimulating and helped the brain to keep working and ‘ticking together’. Family caregivers and staff also rated these two themes very highly. Physical games, sounds, faces and scenes, categorising objects, associated words and being creative were also rated positively by all groups, as these were good for stimulating recognition, reminiscence and discussion, offered multisensorial stimulation and promoted keeping healthy and active.

Number games was the only theme rated very low by people with dementia, who said that, as numbers were not something they related to very well and were a bit meaningless to them, this was not their preferred choice of activity. Family caregivers and staff shared similar views, as their previous experience indicated that number games required more one-to-one work with people with dementia, and were frustrating and pointless unless the numbers were related to pricing.

Using money and current affairs were rated very low by the family caregivers and staff groups but rated highly by people with dementia. Some family caregiver and staff participants stated that people with dementia did not often relate to current affairs (owing to the disease) and it would be meaningless to present topics about news unless reminiscence was used as a context for the information. However, people with dementia expressed a great interest in current affairs and stated that they loved reading newspapers. People with dementia also stated that they would enjoy talking about the value of money, and there were spontaneous comments about the value of money and the cost of bus journeys in the past and the present. In contrast, family caregivers and staff thought that money was too complicated and not a good theme, as it could be a very sensitive topic for some people with dementia.

Opinions about mental stimulation programme and key factors for success

People with dementia thought that keeping the brain active was essential and they acknowledged that this helped with their memory losses and difficulties. This finding supports Barnett’s79 statement that bereavement with memory losses is a major theme for people with dementia, who value the opportunity to be listened to. They also value being in a group. As indicated in other studies, socialising is an important activity for older people in care. 80–82 People with dementia valued quality social interactions, especially when being part of a group, and feeling a sense of belonging, reflecting Kitwood’s83 theory, which claims that positive interactions reinforce the personhood of those with dementia.

Conversely, staff and family carers expressed mixed opinions about the effectiveness of keeping the brain active; they cited examples of public figures who had developed dementia and said that the focus should be on the factors that would make a programme successful or unsuccessful for people with dementia. The main factors to consider when planning a CST group were grouped into participant characteristics (level of dementia, sensory impairments, personality, interests, life history), facilitator characteristics (knowledge about dementia, group skill and personality), group size and materials (multisensorial prompts, age appropriate).

Limitations

Staff focus groups included managers or senior carers, the presence of whom may have affected the opinions expressed by other staff members. Participants sometimes appeared reluctant to give personal examples, perhaps inhibited by the stigma of sharing information in a group, and this may be a disadvantage of using focus groups. We selected a thematic analysis, as it is a method for identifying, analysing and reporting patterns (themes) within data. Although thematic analysis is used widely, there is a lack of consensus regarding its precise methodology. 84

Design

The design was a single-blind, multicentre RCT of CST groups for dementia versus MCST groups (Figure 11). After the completion of the initial CST programme (twice-weekly 45-minute sessions for 7 weeks), participants were randomly allocated to the treatment group (maintenance sessions weekly for 24 weeks) or the control group (usual care). Data were collected at baseline (baseline 0), after completion of the 7-week CST programme (baseline 1), and at 3- (follow-up 1) and 6-month follow-ups (follow-up 2). It was calculated that a sample size of 230 participants was needed at baseline 1 to detect an effect size of 0.39 on the ADAS-Cog,86 with power of 80% using a 5% significance level. Ethics approval was obtained through the Multicentre Research Ethics Committee (reference number 08/H0702/68). The clinical trial was registered as ISRCTN26286067.

FIGURE 11.

Flow diagram of the trial: trial and randomisation stages.

Participants

The participants were people meeting the Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM-IV)87 criteria for dementia graded as mild to moderate on the CDR scale,46 with the ability to communicate, hear and see well enough to participate in the group, with no major physical illness or disability, and without a learning disability. Half of the sample was recruited from care homes and half was recruited from community settings, including CMHTs, day centres and voluntary organisations in London, Essex and Bedfordshire. Of the 21 centres initially contacted, one refused to participate and two were excluded owing to a lack of participants.

Randomisation

The randomisation process in this trial was undertaken in two stages: randomisation 1 and randomisation 2. Figure 11 sets out the two-stage randomisation process. The allocation ratio at randomisation 1 stage was 1 : 1; into either group A or group B, with both groups receiving 7 weeks of CST.

The allocation ratio at randomisation 2 stage was 1 : 1; into either the control group or the treatment group.

The sample was stratified to ensure that equal numbers of participants taking cholinesterase inhibitors were randomised into either the MCST or the control group. The North Wales Organisation for Randomised Trials in Health (NWORTH) was responsible for undertaking the remote randomisation. NWORTH is accredited as a Clinical Trials Unit by the UK Clinical Research Collaboration and funded as part of the Clinical Research Collaboration Cymru, notably for Health Technology Assessment trials.

Blinding

The participants could not be blinded to group allocation owing to the nature of the intervention. The researchers involved in collecting participant consent and in the interviews were not involved in the randomisation process and were blinded to treatment group allocation at follow-up. Our experience in the previous CST trial, as shared by those conducting similar projects, is that participants may occasionally and inadvertently inform researchers of the treatment they are receiving. We aimed to reduce this effect by giving explicit reminders to participants before the assessment visit and by the use of self-report measures whenever feasible. On completion of the two follow-up assessments, the assessors recorded their impression of which arm of the trial each participant belonged to and their confidence in that prediction. This enabled us to test whether or not the inadvertent loss of blinding leads to bias, and to adjust for any bias that was detected.

Intervention

Cognitive stimulation therapy is an intervention for people with mild to moderate dementia, designed following extensive evaluation of the available research, and is an evidence-based treatment. 55 The programme consists of 14 45-minute sessions that are run twice-weekly. Each session incorporates the use of a ‘reality orientation board’, displaying both personal and orientation information, including the group name (as chosen by participants). The guiding principles of CST involve using new ideas, thoughts and associations; using orientation (but sensitively and implicitly); a focus on opinions rather than facts; using reminiscence as an aid to the here-and-now; providing triggers to aid recall; the creation of continuity and consistency between sessions; focus on implicit (rather than explicit) learning; stimulating language; stimulating executive functioning; and being person centred (treating people as unique individuals with their own personalities and preferences). The CST programme aims to create an environment in which people have fun, learn, strengthen their abilities and improve their relationships to other group members, thus maintaining their social and cognitive skills at their optimum ability.

The MCST programme is an evidence-based maintenance group therapy programme for people with dementia. It comprises 24 sessions of MCST, based on the theoretical concepts of reality orientation/cognitive stimulation and grounded in the original CST programme. The aim is to create an evidence-based maintenance group therapy programme for people with dementia, based on the same principles as those of the CST programme. A summary of the CST and MCST programme can be found in Table 3.

Recruitment and training of facilitators

Each CST and MCST group had two facilitators, one from the research team and a cofacilitator who was a member of staff from the recruited centre (e.g. a care home). The facilitators had at least 1 year of experience in dementia care. The main facilitators often had a background in mental health nursing, occupational therapy or clinical psychology, as well as experience in dementia care and group facilitation skills. The use of two facilitators for each group enabled effective debriefing and reflection to occur at the end of each session. All facilitators attended 1-day CST training developed by one of the CST pioneers (AS) as part of the dissemination strategy. The training provided detailed background and description of CST, and used learning methods including group observation, role-playing and small group exercises.

Usual care

The participants allocated to the control group received treatment as usual (TAU). This can vary between and within centres, and may change over time but, in principle, the interventions offered to this group were also available to those in the active treatment groups. Therefore, the trial examined the additional effects of MCST. Our approach to costing the services and interventions received allowed us to monitor whether or not the TAU group had been receiving similar therapeutic interventions. The use of antidementia medication was recorded as part of the costing information collected. It was possible that participants in the TAU group were involved in some form of cognitive stimulation work during the 24 weeks of the study period. However, it is very unlikely that such a structured approach to CST was offered in any of the centres. It is this systematic group-based approach that was the focus of this evaluation.

Risks and anticipated benefits for trial participants

There appears to be no documented harmful side effects of participating in CST groups, and no serious adverse reactions were apparent in the CST study. The participants in the groups consistently reported benefits, including enjoyment and feelings of validation and self-worth. 88 Participants’ inclination to continue meeting following the sessions gave an indication of how valuable they found these benefits. All adverse events were recorded and reports as per the SHIELD adverse events policy. 41

Consent

All recruited participants were in the mild-to-moderate stages of dementia, and were able to give informed consent for participation. Whenever possible, a family member or other supporter was included. It was made clear to participants and family caregivers that no disadvantage would occur if they chose not to participate. In seeking consent, we followed current guidance from the British Psychological Society40 on evaluation of capacity. If a participant’s level of impairment increased, so that they were no longer able to provide informed consent, the provisions of the Mental Capacity Act 200538,39 were followed.

Primary outcome measures

1. Cognition was measured using the ADAS-Cog. 86 The ADAS-Cog consists of 11 tasks measuring the disturbances of memory, language, praxis, attention and other cognitive abilities, which are often referred to as the core symptoms of Alzheimer’s disease.

2. Quality of life was measured using the Quality of Life – Alzheimer’s disease (QOL-AD) scale. 89 The QOL-AD covers 13 domains of quality of life. It has good internal consistency, validity and reliability, and its use is recommended by the European consensus on outcome measures for psychosocial interventions in dementia. 90

Secondary outcomes

1. Cognition was measured using the MMSE,45 a brief, widely used test of cognitive function that has good reliability and validity.

2. Communication was assessed using the Holden Communication Scale. 91 This scale is completed by staff or family caregivers and covers a range of social behaviour and communication variables.

3. Depression was assessed using the Cornell Scale for Depression in Dementia. 92 This scale rates depression in five broad categories using information from interviews with staff and participants. Good reliability and validity have been demonstrated.

4. Anxiety is assessed using the Rating Anxiety in Dementia scale. 93 This rates anxiety in four main categories and uses information from interviews with staff and participants. It has good validity and reliability.

5. Behaviour was assessed using the Neuropsychiatric Inventory (NPI). 94 The NPI assesses 10 behavioural disturbances occurring in dementia patients. It has good validity and reliability.

6. Activities of daily living were assessed using the Alzheimer’s Disease Co-operative Study – Activities of Daily Living Inventory (ADCS-ADL). 95 The ADCS-ADL is a structured questionnaire originally created to assess functional capacity over the range of dementia severity. The sensitivity and reliability of the scale have been established. 95

7. Family caregiver health was assessed using the Short Form-12 Health Survey (SF-12). 96 This scale measures generic health concepts relevant across age, disease and treatment groups. The SF-1296 comprises eight concepts commonly represented in health surveys. It is a self-administered measure that provides a comprehensive, psychometrically sound and efficient way to measure health from the patient’s point of view by scoring standardised responses to standard questions. This measure was used only when a family caregiver was available from the community sample participants.

8. Costs were assessed using the validated Client Services Receipt Inventory (CSRI),97 adapted for this study. Used extensively in studies of mental health and dementia,13 the CSRI gathers comprehensive data on accommodation, medication and services received. In this trial the data covered the previous 3 months (at baseline and after treatment) or the 3 and 6 months’ follow-up points. Unit costs were then attached to services and support received, based on nationally relevant estimates of long-run marginal opportunity costs. Two quality-of-life measures were also included for cost–utility analyses. The EuroQol-5 Dimensions (EQ-5D)98 is a standardised instrument for use as a measure of health-related quality of life. It contains a three-level coding system for five dimensions. The instrument includes a global rating of current health using a visual analogue scale (VAS). The Dementia Quality of Life (DEMQOL)99 uses self-rated reports of quality of life administered by a trained interviewer; there is also a separate scale for family caregiver or members of staff reports, called the DEMQOL-Proxy. It covers five domains of quality of life. The DEMQOL has high internal consistency and acceptable inter-rater reliability, and indicates concurrent validity through moderate associations with the QOL-AD and DEMQOL. 99

Predictors of change in cognition and quality of life between baseline and follow-up

A repeated measure linear model explored the impact of other variables. The model was fitted for age, living situation (community/care home), sex, marital status and taking AChEIs, and using post score as the dependent variable. As a result of fitting the models in this way, the results showed that both age and sex were important factors for the effectiveness of CST. For MMSE, age was a significant predictor of effectiveness of CST, with older participants appearing to benefit more. At the mean age of 82 years there was little difference between the pre and post score, but participants older than this appeared to benefit more, with a possible increase in MMSE score. For ADAS-Cog, age again was a significant predictor of the effectiveness of CST, with older participants benefiting more, in the same way as with the MMSE. Sex proved to be a significant variable in the complete-case study analysis, and showed a strong correlation with cognitive improvement, with female ADAS-Cog scores improving more than male scores.

Living situation was also shown to be an important variable for some of the staff-completed outcome measures (Table 6). For the NPI, a decrease in score was seen for the community sample [18.08 (SE 2.25) to 13.89 (SE 2.24)], while there was a small increase in NPI score for the care-home-based participants [11.35 (SE 2.48) to 13.42 (SE 2.48)]. This indicates a benefit for the community sample.

For the DEMQOL (proxy), both community- and care-home-based participants saw a mean increase. However, the care home increase [94.2 (SE 3.61) to 100.87 (SE 3.39)] was larger than the community increase [99.32 (SE 3.47) to 100.26 (SE 3.26)]. In fact, this can be viewed as the community sample remaining steady while the care home sample were brought into alignment with what was observed in the community.

For the imputed proxy EQ-5D utility values, age was seen as a significant factor, with older participants appearing to benefit more. For the proxy EQ-5D VAS, the community sample showed a mean increase from 60.19 (SE 3.63) to 67.52 (SE 3.36), while, conversely, the care home sample showed a small mean decrease from 65.47 (SE 3.29) to 62.72 (SE 3.04). In summary, we have identified benefits for the community sample on NPI scores and proxy EQ-5D VAS, and there is a benefit for care home sample for the DEMQOL-Proxy.

Change when comparing the results with those of a similar control group

Independent sample t-tests were used to compare the complete-case data set results with the Spector et al. 6 control group, which had a mean age of 84.7 years and a 3 : 1 ratio of female to male. For the ADAS-Cog, the Spector et al. 6 control group showed a mean reduction of 0.3, while the CST group showed a mean reduction of 2.72. This gives a mean difference of 2.42 [t = 2.27, degrees of freedom (df) = 240; p = 0.024], with CIs of 0.33 to 4.51. For the MMSE, the Spector et al. 6 control group reduced by an average of 0.4 points, while the CST group saw a mean increase of 0.9 points. This gave a mean difference of 1.3 points (t = 2.76, df = 293; p = 0.006), with a CI of 0.38 to 2.22 points (Table 7). At follow-up, the CST group had demonstrated significantly better results than the Spector et al. 6 control group on both MMSE and ADAS-Cog. There was no difference between the CST group and the Spector et al. 6 control group on the QOL-AD.

Using an adjusted ANCOVA comparing groups (treatment and controls), Table 8 shows the analysis of the primary end point (follow-up 2) and secondary end point (follow-up 1). At 6-month follow-up the treatment group had significantly higher scores on self-rated quality of life measured by the QOL-AD than the usual care group (p = 0.04).

Cognition on the MMSE also favoured the MCST group but this was not significant. Centre emerged as a significant covariate in relation to the ADAS-Cog, Cornell and Rating Anxiety In Dementia scales. At 3 months’ follow-up, the MCST group had significantly higher scores than the usual care group on proxy-rated quality of life on both the QOL-AD (p = 0.008) and the DEMQOL (p = 0.04), as well as ADLs as measured by the ADCS-ADL (p = 0.05).

The MCST-AChEIs platform adjusted results at 6- and 3-month follow-up showed that the group receiving MCST plus AChEIs scored highest on the MMSE (see Table 3), than those receiving MCST only, those not receiving MCST and those not taking AChEIs. There is a significant interaction term between treatment group and taking AChEIs. Post hoc testing indicated that there was a significant difference between the MCST plus AChEIs group and the usual care and AChEIs group (p = 0.025). No significant interactions were found on any of the other outcome measures.

Numbers needed to treat

The number needed to treat is the calculation of the number of people who need to be treated in a particular intervention for one favourable outcome to be achieved. The number needed to treat calculation uses the proportion of participants who benefit in each treatment group. A number needed to treat analysis using QOL-AD change was used between baseline and follow-up 2, as this was the primary end point. An improvement of one SD on the QOL-AD (large effect size) was taken to indicate major clinical benefit, and 0.5 of a SD change (moderate effect size) was taken to indicate moderate clinical benefit. When calculating a large size effect (1 SD), the proportion benefiting was 0.18 (17/96) for the treatment group compared with 0.07 (6/81) for the control group and, therefore, 9.7 people needed to be treated for 1 to benefit (95% CI 5.0 to 129.9 people). When calculating a moderate size effect (0.5 SD), the proportion benefiting was 42 out of 96 (0.44) for the treatment group compared with 16 out of 81 (0.20) of the control group and, therefore, 4.2 people needed to be treated for 1 to benefit (95% CI 2.7 to 9.2 people).

Methods

The main economic analysis compared MCST with usual care from a health and social care perspective. Secondary analyses also included the costs of unpaid carer time (societal perspective) and the cost of MCST as if it had been provided by health-care assistants rather than by the research team (‘in-practice’ implementation). A subgroup analysis also compared individuals receiving only MCST with individuals using an AChEI in addition. Health and social care services used by participants and inputs from unpaid carers were captured by the CSRI97 and completed with family carers or centre care workers. When possible, unit costs for the services were drawn from the Personal Social Services Research Unit compendium for 2011. 100 We used a 3.5% discount rate (recommended by Her Majesty’s Treasury) for items providing benefit for more than 1 year, such as equipment or adaptations. Medication costs were obtained from the British National Formulary. 101

The economic analysis estimated incremental costs and incremental effects and their CIs with seemingly unrelated regression methods with 1000 bootstrap replications. The information obtained through the economic analysis has led to the calculation of incremental cost-effectiveness ratios (ICERs) and to plotting cost-effectiveness acceptability curves (CEACs), showing the probability that MCST is a cost-effective addition to usual care against a series of hypothetical willingness-to-pay (WTP) values.

Results

Looking at the primary outcomes in the health and social care perspective (Table 9), the mean cost for each 1-point difference on QOL-AD was £266. Looking at the CEAC for this outcome, the probability that MCST would be seen as more cost-effective than usual care alone would be 90% at a WTP of about £1400 (Figure 13). There are no established WTP thresholds for QOL-AD against which to compare this finding, but a cost of only £1400 for a 1-point difference on a 40-point scale is modest. Based on previous studies,44 the effect size of ‘standard’ CST on the QOL-AD scale is around 0.4 SD, which represents a modest increment. In the current study the difference at follow-up for MCST was estimated to be 1.78 points (SD 0.34 points). A 2-point difference in QOL-AD can be considered to be clinically significant and a cost of £2800 to achieve such a result may be attractive to decision-makers. When the outcome was measured in terms of ADAS-Cog, the probability that MCST would be seen as cost-effective was low across all values of WTP (Figure 14).

FIGURE 13.

Cost-effectiveness acceptability curve: MCST vs. usual care – 6 months, health and social care perspective, with effectiveness measured on the QOL-AD scale.

FIGURE 14.

Cost-effectiveness acceptability curve: MCST vs. usual care – 6 months, health and social care perspective, with effectiveness measured on the ADAS-Cog.

Secondary economic analyses indicated that MCST was more cost-effective than usual care over the shorter period of 3 months when outcomes were measured in terms of proxy-rated quality of life (using QOL-AD and DEMQOL) and ADLs (using ADCS-ADL), but less cost-effective than usual care on all other outcomes, including quality-adjusted life-years (QALYs).

In sensitivity analyses we broadened the cost measure to include unpaid care and support, conducting the evaluation from a societal perspective. The cost-effectiveness case for adding MCST to usual care was, again, mixed. Subgroup analyses found that the combination of MCST and AChEI was more cost-effective than AChEI and usual care by reference to a number of outcomes, including cost per QALY.

Predictors of success: who benefits most?

The benefits of CST were found to be independent of the use of AChEIs, which is in line with the Cochrane review findings and other studies. 44,49,51,53 In line with the earlier study,6,104 greater improvements in cognition were associated with female sex. Men might be more reluctant to communicate as they are usually in the minority in most groups, with women outnumbering men in the groups by a ratio of 4 : 1. 6,104 It may be that the sex majority dictates the style in which the groups are run, for example more ‘talking’ in female-dominated groups and more ‘doing’ in male-dominated groups. Further work is needed to explore these sex differences in response to CST and to develop interventions more geared towards the attributes of men.

The greater effect for the older people in this study is an unexpected finding. It may suggest that these older residents are experiencing more excess disability, showing impairment beyond that resulting directly from the dementia. It may be that they receive less stimulation in general than the slightly younger participants, and so benefit more from the new intervention. The findings of differences in outcomes between those living in the community and those living in care homes relate to all measures completed by a proxy: a family member or a member of staff. Their ratings may be affected systematically by different factors (e.g. family carer ratings are typically influenced by their level of strain; staff in care homes may change or have limited contact with residents on which to base their judgements). Further work is needed to explore these differences in perspectives and in the changes reported.

Maintenance cognitive stimulation therapy

This is the first major investigation to compare the short- and long-term impacts of CST for people with dementia. The investigation indicates that after standard CST (7 weeks, twice weekly), 24 weeks of MCST sessions both improves quality of life for people with dementia at 6-month follow-up and benefits quality of life and ADLs at 3-month follow-up, in comparison with the usual care control group. Cognitive stimulation programmes for dementia have a significant positive effect on cognition;44,105 however, there are concerns that the benefits may only persist for a limited period of time. 44 These results indicate that a lower intensity input of once-weekly CST sessions can continue to be effective after the initial twice-weekly CST programme is completed. In comparison with the first baseline before CST, cognitive function in the MCST group had decreased by only 0.35 points on the MMSE, whereas cognitive function in the usual care (CST) group had decreased by 1.65 points at follow-up 2. The ADAS-Cog showed a similar pattern of results. On the basis that MMSE scores in mild-to-moderate dementia are expected to decrease by 2–4 points per year,106 these results suggest that, relative to no treatment, the MCST programme continues to have a protective effect on cognition. Other studies have also found that a longer-term cognitive stimulation intervention can be effective in reducing cognitive decline in dementia. 107,108

However, benefits to cognition alone may not be sufficient to justify an extensive programme of intervention unless they are accompanied by other benefits, and a few studies have explored the long-term impact of psychosocial interventions on quality of life. 109 This study indicates that the MCST programme improves quality of life and ADLs. As CST alone improves both cognition and quality of life,6,105 this provides a strong case for an ongoing programme of CST being supported in health and social care settings.

The MCST-AChEIs subanalysis results confirm the results of similar studies,6 suggesting that the additional effects of cognitive stimulation on cognition are over and above those of medication alone. In addition, the Cochrane review showed that type of control condition (e.g. usual care or social activities) made no difference to outcome, demonstrating that cognitive stimulation, rather than social activities, was responsible for the improvements in outcome.

Previous studies have shown that CST is effective in improving cognition and quality of life,6 and is cost-effective. 13 CST is endorsed in NICE clinical guidelines. 14

Findings from this new trial suggest that, although outcome gains were modest over 6 months, the continuation of CST appeared cost-effective in terms of self-rated quality of life, cognition measured on the MMSE and proxy-rated QALYs. MCST in combination with AChEIs offered cost-effectiveness gains when cognition was measured as the outcome.

The external validity of this pragmatic trial is high, as the sample came from a wide variety of settings. However, participants were almost exclusively white British and, therefore, solid conclusions on the applicability or effectiveness of this intervention to people of other ethnic or cultural groups cannot be made. Having said this, we recently successfully adapted and ran a local CST group in Hindi, and CST programmes have been run successfully in 20 countries.

Limitations of study

The SDs for the cognitive measures were higher than those in the original Spector study,6 suggesting that a larger sample may be needed to achieve significance differences in cognition after MCST relative to CST alone. Proxy measures (e.g. the ADCS-ADL and NPI) were rated by the family carer or staff members who were aware of participant group allocation, and so this might have introduced detection bias into the ratings. The subgroup of people on AChEIs might also have received extra psychiatric services during the trial. However, comparisons of the AChEIs-only group and MCST plus AChEIs group data confirm that the effects of the MCST programme are above and beyond those of medication alone.

Design

The STANDOUT and MONOU studies differed in how staff members were recruited into the research and in their previous exposure to CST. Staff members in the STANDOUT study were new to CST, having had no prior experience of the programme, whereas staff recruited into the MONOU study had either previously attended CST training or purchased the CST ‘Making a difference’ manual. The observational study looked at the effects of CST and MCST on people with dementia in the practice context delivered by staff members. The three studies together provided evidence on both the most effective way of facilitating the implementation of CST, and uptake and effectiveness of the approach in a clinical context (Table 10). Ethics approval was obtained for the project through the Multi-centre Research Ethics Committee (reference number RP-PG-0606-1083). The clinical trial was registered as ISRCTN28793457.

STANDOUT trial

The design was a pragmatic, multicentre, single-blind, two-treatment-arm RCT. The STANDOUT trial sample comprised dementia care staff from specialist and non-specialist dementia care settings. All participants received the training package as TAU, consisting of a CST training day, a training DVD, the CST ‘Making a difference’ manual and the MCST ‘Making a difference 2′ manual, and cluster randomised by centre prior to the training day into the intervention group or to TAU. The intervention group consisted of a local co-ordinator, e-mail support and an online forum. Each staff member was expected to complete three questionnaires, the first before the training day and the others at 6 and 12 months thereafter (Figure 15).

FIGURE 15.

Flow diagram of the STANDOUT trial and assessment schedule.

MONOU trial

The design was a pragmatic, multicentre, single-blind Phase IV trial. The participants were staff members from centres that had previously purchased the CST ‘Making a difference’ manual or had attended a CST training course. Once enrolled in the research, all staff received the free MCST ‘Making a difference 2′ manual and DVD. It was recorded whether participants had the manual only or manual and training before centres were cluster randomised into outreach support or TAU (Figure 16). The time points for completing the questionnaire were the same as for the STANDOUT trial. However, the participants also completed a retrospective questionnaire on their use of CST in practice prior to the research.

FIGURE 16.

Flow diagram of the MONOU trial and assessment schedule.

Observational study of cognitive stimulation therapy in practice

The design was a multicentre, longitudinal observational study with people with dementia. Sites that were running or in the process of setting up CST groups completed minimal outcome measures at three time points with people with dementia participating in the CST and MCST programme (Figure 17). The measures were completed before the group started (baseline), at 7 or 14 weeks depending on how they implemented the CST programme (once or twice weekly) and after the MCST at 31 or 38 weeks. However, this end time point did differ when taking into consideration staff and group member availability and holidays. The intervention of CST was routinely offered in the care setting. The aim of this study was to determine whether or not groups were running in practice and to determine if the positive findings for cognition and quality of life for the person with dementia found in previous CST research5 could be demonstrated in practice.

FIGURE 17.

Flow diagram of observational study and assessment schedule.

Ethics arrangements

Aim

The study explored the experiences and effects of participating in the MCST programme by conducting semistructured interviews with people with dementia and CST facilitators. It also aimed to explore the mechanisms of change and compare the findings with those of a previously conducted study110 on the experiences of CST to determine if the MCST continued the perceived benefits identified from the original programme.

Sample

Ten people with dementia who had completed the CST programme and were part way through the MCST programme were recruited from an East London community day centre for older people. The mean age was 84 years and the sample was made up of four male and six female participants. There was a mean baseline MMSE45 score of 16, indicating a moderate level of impairment. All had completed the CST programme and were up to the sixth MCST session.

Five staff members were recruited that had previously acted as a cofacilitator or main facilitator in running the CST programme and were in the process of running the MCST programme with the people with dementia recruited into the study. The day centre was purposefully chosen as it had previously been enrolled in the MCST research project111 and staff had gone on to run their own groups following on from the research, it was considered that the staff were experienced enough to provide an in-depth experience of the MCST programme. The staff members either had attended the CST training day or had the CST ‘Making a difference’ manual. All staff members were female and between them they had an average of 11 years of dementia care experience.

Inclusion criteria

People with dementia were initially approached if they had completed the CST programme and regularly attended the MCST programme that was currently under way at the centre. However, they also needed to (1) have met the DSM-IV124 criteria for dementia, (2) have scored between 10 and 24 on the MMSE,45 (3) be able to adequately communicate, (4) see and hear well enough to participate in the group, (5) not present behaviour or have a physical illness that could limit their participation, and (6) not have a diagnosis of a learning disability. These inclusion criteria were used to ensure that people could adequately participate and provide useful information in the focus group.

The staff members were considered eligible to participate in the focus group if they had participated in the running of the CST and MCST programme and met with the service users accessing CST at least twice weekly; this was considered sufficient time for staff to be familiar enough with the service users that they could comment on everyday effects that might arise from the MCST programme.

Procedure

Managers were approached 1 month before the focus groups were undertaken to gauge their level of interest and identify potential participants from both staff and people with dementia. A week prior to the focus groups being undertaken, a researcher approached the staff and people with dementia and went through the information and consent form, allowing any questions to be asked. The researcher went through the information sheet and consent form again on the day of the focus group and discussed this with the participant for a second time before informed consent was obtained. Consent was obtained in accordance with the British Psychological Society40 guidelines and ongoing consent was adhered to; participants were reminded that they could withdraw at any time and no adverse consequences would result from this.

Materials

A discussion guide was devised, based on the findings from a previous study,110 to guide the conversation. This allowed a reliable comparison to be made between the person’s MCST experience and the original CST programme. Questions were created with the aim of understanding any mechanisms of change that might be occurring; this was in relation to a person-centred approach (PCA). 125 Swaab’s theory of stimulating underused cognitive ability126 and the biopsychosocial model127 were explored. All focus groups were audio-taped and field notes were taken by a cofacilitator to document any useful non-verbal communication and interaction of group members that might have been lost in the transcription stage of the data analysis.

Data analysis

The data analysis followed the stages of framework analysis128 and followed the principles of thematic content analysis. 129 Each focus group transcript was analysed with the following steps: (1) immersion in the data; (2) categorising the data to identify a thematic framework; (3) coding; and (4) interpretation and association. In the first stage it consisted of the researcher familiarising himself or herself with the general ideas generated from the focus group discussion that were identified from reading the transcripts. In stage 2, ‘open coding’130 was used to categorise the data into general themes, followed by themes more representative of the data. To ensure the validity of the themes, a second researcher also analysed the transcripts so that a final set of themes was agreed by both researchers. The third stage involved the coding of the transcripts in accordance with the thematic framework and this was then followed by the final stage of interpreting the findings in accordance with the existing literature.

Results

After the data were analysed, two main themes emerged across the three focus groups. The first was ‘positive experiences of being in the group’ including the subthemes ‘enjoyable company’, ‘positive feelings’ and ‘benefits of a smaller group’. The second theme identified was ‘cognitive stimulation and cognitive benefits’, with the subthemes ‘importance of and improvement in communication’, ‘importance of variety’ and ‘improvements in memory’.

In the following quotations, to clarify who was speaking, P will be used to identify a person with dementia and S will be used to identify a staff member.

Comparison with Spector et al.110 CST findings

The first theme identified was ‘positive experience of being in the group’, which was identified in the previous study. Within this theme, ‘positive feeling’ was also found in the 2011 study and within this were feelings of enjoyment, relaxation and something to look forward to. The second subtheme, ‘enjoyable company’, corresponds to the identified subtheme in the previous study of ‘supportive and non-threatening environment’, as both recognise the positive effects of relationship building between group members and with the facilitators. A difference in the main theme between this study and that by Spector et al. 110 is that the latter did not find the subtheme of ‘benefits of a smaller group’ and this study did not report the service users gaining and remaining confident for the remainder of the day after CST. However, the increase in confidence for people with dementia had been identified through their family caregivers and, as this demographic was not used in this study, it cannot be assumed that this is different for the MCST programme. However, it is important to note that staff members noticed that for some people with dementia their increased confidence outside the group could be considered evidence of their enhanced confidence (see Importance of and improvement in communication).

Two of the subthemes, ‘cognitive stimulation and cognitive benefits’ and ‘importance of and improvement in communication’, tally with the subtheme ‘positive experience of being in the group’ as identified in the 2010 study. Within this were the subthemes ‘finding it easier to talk’ and ‘sharing a diagnosis’, which can be considered reasons for improvement in communication. An increase of confidence in communication ties in with the first theme of ‘positive experience of being in the group’, as comfort in a smaller group size, valued contributions and friendship can lead to a heightened confidence in expressing oneself. The corresponding subthemes demonstrate that the opportunity that is created in CST to discuss, give and receive opinions is a factor in the perceived success of the CST and MCST programme. It would seem unlikely that conversation would have been an important element without having ‘positive feelings’ and the ‘benefits of a smaller group’. The importance placed on opinion links into the CST key principles of opinion over fact and implicit learning, both of which fall under ‘positive person work’ (PCA). Placing value on what people say rather than on the factual content or accuracy of what they say increases their confidence in expressing their opinion. The subtheme ‘importance of variety’, under the main theme of ‘cognitive stimulation and cognitive benefits’, did not correspond to the Spector et al. 110 study, yet it was a key point of agreement in this study (see Mechanisms of change); however, this might be have been emphasised owing to the extended duration of the MCST programme.

Both a general and a specific improvement in memory were identified in Spector et al. 110 and in this study. It could be argued that ‘improvement in memory’ resulted from the use of reality orientation,131 as a result of the reality orientation board used for every CST and MCST session and the repetitiveness of the implicit information processing (group name, song, location) as well as the session structure as presented in both the CST and MCST manual. This can be related to the subtheme of ‘positive experience of being in the group’, as one might expect that, without an environment conducive to participation and enjoyment, the likelihood of someone becoming actively involved in the discussion and activities might decrease and, in turn, lead to a negative impact on their memory and cognition.

Mechanisms of change

An analysis of the mechanisms of change was necessary to develop an understanding of what is effective in relation to CST. Two of the themes emerged in this study, ‘benefits of a smaller group’ and ‘importance of variety’, that can be considered mechanisms of change. As both the Spector et al. 110 and this study identified ‘positive experience of being in a group’ as a theme, it is apparent that both the CST and MCST programmes created an environment that facilitated ‘positive feelings’ of value and inclusion as well as the building of relationships. This theme is evidence of the benefits of using a PCA. 125 A PCA places emphasis on seeing the person with dementia as a person and an individual, rather than defining them according to their diagnosis.

The ‘benefits of a smaller group’ subtheme appears to fit well as part of a PCA125 as it encourages more in-depth relationships, something one might expect to be harder with a larger group . The possibility of developing friendships in a small group might encourage the active participation of each group member and increase the cognitive benefits of CST. 126 So, although a larger group might adhere to the same principles, it may not demonstrate the same benefits for the group members.

The variety of activities and current affairs to be discussed at the beginning of each session follows another key principle of CST: choice. Allowing choice reiterates that the group belongs to the group members and increases the chance of catering to each person’s needs and interests, further facilitating the use of a PCA. The flexibility of variety is in keeping with the biopsychosocial model,127 which states that treatment should consider the fixed biological factors as well as the changeable psychological and social factors when working with a person with dementia. This model works with CST as it identifies the fixed factors that are reflected in the inclusion criteria of the therapy of scoring 10–24 on the MMSE,45 as well as the tractable factors such as sensory impairment when it is necessary for the person with dementia to see and hear well enough to participate in the programme. 6 The biopsychosocial model does link in with the inclusion criteria used for participation in the programme but it is more strongly related to the psychological factors as recognised in the model. CST as a psychosocial intervention creates a social support network through the group format of the therapy leading to improvements in social relations and uses a multisensory stimulation through the session themes and materials to encourage implicit mental stimulation (psychological tractable factor). An understanding of the person’s history, personality and beliefs before starting CST emphasises a PCA as this can be considered when running discussion or activities. The variety offered by CST also allows for personal psychology and individual differences to be taken into account when creating sessions that can cater for a range of interests.

Limitations

The study focused on the opinions generated from the people with dementia participating in the programme as well as those of the staff members facilitating the groups. As the staff members were familiar with the service users, it was not considered necessary to include family members. However, when conducting the focus group, staff did not feel comfortable commenting on the possible after-effects of MCST. Staff might also be unaware of the person with dementia’s feelings about the group, whereas the service user might be more open with a family member about their experiences. So, focus groups run with family members might be useful when considering the after-effects and a more in-depth understanding of the perceptions of the MCST programme.

Owing to the limited time frame of the study the focus groups were conducted after the sixth session of MCST. The full maintenance programme is run over 24 weeks so the opinions expressed were after one-third of the programme had been delivered. It could be considered that focus groups conducted at the end of the MCST programme would be more beneficial in getting a truer account of the experience of the programme, as in the study the people with dementia could discuss the experience of the first six sessions only. This point might not be as relevant to the staff as they were able to draw on their previous experience of the MCST programme, having delivered it prior to the current group.

As there was a time constraint on collecting the data, only one centre could be recruited into the study. This might have biased the results as the feedback on the MCST programme was specific to the delivery of it in that particular centre. There was also a small number of participants (10 people with dementia and five staff members) and so the amount of feedback was limited to a small sample size. However, the same thoughts about the programme were mentioned across the three groups, so it might be that more participants would not necessarily generate any additional themes but rather would strengthen those already identified in this study.

Aim

To develop a peer-support intervention from the carer perspective in consultation with service users.

Design

A combination of a modified Delphi process132,133 and consensus conference134 was used to explore the details of the peer-support programme from the carer and volunteer perspective. A combination was used to allow participants to respond personally and privately, but also to provide an opportunity to discuss their ideas and concerns.

Participants

Service users and stakeholders were approached through organisations with an interest in dementia, family carers, peer support/voluntary work, NHS organisations and universities. Twenty-five people expressed interest in taking part and completed round 1 of the modified Delphi process. Of these, 21 attended the consensus conference. The delegates comprised eight current and former carers, seven members of voluntary organisations supporting carers and people with dementia or representing volunteers, three clinical health professionals, an academic specialising in the area and two others (did not specify). Round 2 was sent to the original 25 stakeholders as well as to an additional eight who had become involved as the project progressed. In round 2 five questionnaires were returned from three former family carers (one of whom had also worked with carers), one current family carer and one representative of a charity for people with dementia and their carers.

Method

Three weeks before the conference, a round 1 Delphi questionnaire was sent along with copies of the raft participant information sheets and related recruitment materials to provide context to the proposed peer-support intervention. Delegates were required to complete the questionnaire before attending the conference and either post it back to the research team or hand it to the team on the day. The questions concerned the intervention name, the role of the peer supporter, the content and duration of training, desirable peer-supporter characteristics, the criteria to be used to match peer supporters with family carers (matching criteria), support for peer supporters and monitoring the matches for research purposes. These items were developed by the research team with reference to the service user feedback for BECCA, along with ideas and concepts trialled by other research groups in their work with peer supporters. 135–137

During the conference, round 1 data were analysed and the results were presented for discussion. The delegates were split into smaller groups, with careful consideration given to creating heterogeneous groups, as these can be the most productive when aiming to explore uncertainties and develop ideas. 133 Each group was given a question from round 1 to focus on, facilitated by a member of the research team who made notes on a flip-chart. These notes were used to feed back to the larger group and were also a record of the discussions held.

After the conference, the round 1 results were analysed in more detail with particular focus on identifying areas in which consensus had not been reached. Each question was scored on a six-point Likert-type scale measuring strength of consensus (where 1 = unimportant/unsuitable/unfeasible and 6 = important/suitable/feasible). The delegates were also asked to rate their most preferred and least preferred options. The responses to the question items were summarised as medians, with higher medians indicating higher importance/suitability, as these are more robust than means and provide a better indication of distribution of consensus as opposed to central tendency. 133 Percentages were also calculated for most preferred and least preferred options. By using these two response formats, it was possible both to understand the extent of overall consensus (through most and least preferred options) and to assess the strength of consensus by analysing the medians for each item. The scores from all of the delegates were given equal weighting, regardless of perceived ‘expertise’, meaning that the responses of service users were equal to those of other delegates working professionally in the area.

Items with the lowest medians (1 and 2) were excluded from round 2. Items with highest medians (5 and 6) were also excluded because consensus was that these items should clearly be incorporated. Items with median scores of between 3 and 4 were included in the round 2 questionnaire along with clarifying information. This information was based on identified areas of confusion and disagreement from notes taken during the conference. For instance, the difference between ‘giving advice’ and ‘sharing experiences’ was raised as an issue during the conference as it was not clear as to how they differed in reality. An item had been included on ‘giving advice’ to raise debate. ‘Advice giving’ had been explicitly placed outside the remit of the BECCA befriending volunteers, given the potential legal ramifications of providing advice. However, carers would often state that they would ‘value the advice’ of a more experienced peer. Definitions and further information were provided in round 2 to set out some distinction. A summary of the consensus conference discussions held were incorporated into round 2 and more space for open-ended responses allowed delegates to clearly outline their views, reducing the need for a third round. Thematic analysis was carried out for all open-ended responses. The analysis was inductive (data driven) with themes defined as specific patterns of interest within the participants’ responses. 138 A third round was found to be unnecessary as disparate views had been resolved at the end of round 2. The results from rounds 1 and 2 (including the thematic analysis of round 2 comments) are presented together to illustrate the delegates’ contribution to the programme through the consensus process.

Results

Aim

These documents were necessary to enable participants to decide whether or not to take part in the research trial with valid informed consent. The aim of the consultation was to ensure the appropriateness and suitability of the consent documents (recruitment leaflet and information booklets for carers and their relatives with dementia) and improve readability of the documents.

Design

To manage time constraints and reduce the potential burden on readers, postal consultations were used. The materials for consultation had been adapted from the BECCA study,29 so some materials were able to be reused whereas others had to be newly developed. This is a novel approach to developing informed consent documents, which has logistical and practical advantages over existing methods.

Participants

As carers and their relatives with dementia were involved in the SHIELD CSP, both groups took part in the reader consultations. The consultations were anonymous and the packs were sent to readers by post. Family carers were contacted through the Uniting Carers network of the charity Dementia UK. People with dementia were contacted through the East Anglia and North Thames hubs of the Dementias and Neurodegenerative Diseases Network (DeNDRoN), and were members of the Patient and Public Involvement Forum. In both cases, the research team sent consultation packs to identified gatekeepers at each organisation who then forwarded the packs to interested members of their participant involvement panels.

Method

Before beginning consultations, the research team incorporated the outcomes of the modified Delphi process and consensus conference into the draft materials. Although the focus of the reader consultations was on the recruitment leaflet and information booklets, draft consent forms were also included in the consultation to show respondents what participants would be expected to decide on and, therefore, the type of information the booklets should contain. Carers were asked to read the recruitment leaflet, the information booklet for family carers and the consent form. People with dementia were asked to read the information booklet for people with memory problems and the corresponding consent form. The respondents were then asked to complete feedback forms that asked about the clarity of information about the research study, what commitment would be required for taking part, information about right to withdraw from the research study and the layout of the booklet. The respondents rated the clarity of information on a 5-point Likert-type scale (where 1 = no/not clear, 3 = partially clear and 5 = yes/very clear). The respondents were also asked about the overall quality of each document on a similar scale (where 1 = low, 3 = moderate and 5 = high). Space was also provided for respondents to write other suggestions and comments. In total, 12 carer consultation packs and 12 person with dementia consultations packs were distributed, with 6 and 11 packs returned, respectively.

Results

Median scores and interquartile ranges were calculated. Responses with median scores of 1–4 were taken as being unclear or not as clear as possible, and actions were taken to amend these areas of concern. Ease of readability and grade level of the leaflet and information booklets were also assessed using the Flesch–Kincaid tests available in Microsoft Word^® 2003 (Microsoft Corporation, Redmond, WA, USA). 141 The Flesch–Kincaid Reading Ease test is based on the average number of words per sentence and the number of syllables, taking into account the number of words about people in the passage, and the number of sentences addressed to an audience. A score of 90–100 is considered easily understandable by an 11-year-old, a score of 60–70 is considered easily understandable by 13- to 15-year-olds and a score of ≤ 30 is considered easily understandable by graduates. 142 The Flesch–Kincaid Grade Level score rates text in terms of US school years: a score of 8.0 means that an eighth-grade student (aged about 13 years) would be able to understand the information, which is the recommended level of readability for standard documents for the general population. 143

Limitations and implications for service user involvement

The consultation methods proved fruitful for developing the SHIELD CSP. The research team strove to create an environment in which service users would feel empowered, which seemed to be the case for study 1 but less the case for study 2. Anonymised methods of contributing, which may assume a level of expertise or even comfort with the research process, are adequate and suitable ways to involve service users in research when service users will have varying access to different levels of education and other resources. Service users met face to face in study 1 but not in study 2, and it may be that they value methods more that allow them to debate ideas with others. This may also serve to empower them and to ensure that their contribution is experienced as, and seen to be, meaningful.

Conclusions

Often service users are consulted so as to improve a research grant, research methodology or therapeutic intervention, but rarely have the opportunity to help determine specific research topics. Constraints on research processes and resources may explain why much service user involvement is consultative. 143 The current report highlights ways in which service users can still be involved meaningfully while also highlighting the negative impact that some methods can have on power relationships between researchers and service users.

The contribution made by service users makes it more likely that the finally developed intervention will have high levels of acceptability. This is likely to increase adherence and enhance uptake, both of which are important in psychosocial intervention trials, and have been a challenge for many trials evaluating interventions for family carers of people with dementia.

Design

A 2 × 2 factorial, pragmatic, multisite RCT design was used. Participant carers were randomised individually to carer support (CSP) or TAU after consent and baseline data collection, and then rerandomised on a group basis to joint reminiscence therapy (RYCT) or TAU. This gave the four ‘cells’ of CSP, RYCT, combined CSP-RYCT and TAU only. All participants continued to receive usual care. To ensure enough participants to run viable RYCT groups, randomisations between TAU, RYCT, CSP and combined CSP/RYCT were in the proportions 1 : 2 : 1 : 2. Stratification variables were used for each stage of randomisation. At first randomisation, kinship within the dyad (spouses vs. offspring) and locality were used as stratification variables. At second randomisation (2 : 1 reminiscence vs. no reminiscence), the first allocation was added as a stratification variable to keep the four arms in balance. Follow-up data were collected 5 and 12 months after the first randomisation, with the main end point at 12 months.

In accordance with MRC guidance on complex interventions,32 there was a feasibility phase prior to expansion to full trial. The pilot trial ran in two London boroughs with the focus of the first site directed towards appropriateness and acceptability of procedures, the second site focusing on logistics and timing of the interventions in relation to the timetable for recruitment and research interviewing procedures. Monitoring methods were developed in accordance with the draft checklist, and the outcomes of these were systematically recorded alongside the recording of the implications for the main trial. Examples of the monitoring methods used, outcomes of the monitoring process, and decisions made on the potential pooling of pilot and full trial data have been described by Charlesworth et al. 146 As the changes between the feasibility and full trial were minimal, the feasibility phase was considered an internal pilot and data were carried forward to the full trial.

Randomisation procedure

Figure 19 shows the flow diagram for the SHIELD CSP-RYCT trial. Administrators at each site entered participants’ variables into a remote web-based randomisation, which then allocated them in equal proportions between CSP and TAU on an individual basis. Once adequate dyads had been randomised within each site (target 24, range 16–30), they were entered on a group basis into a second randomisation between RYCT and TAU. The (unblinded) administrator at each site then informed carers of their allocation by letter and liaised with the RYCT facilitator and/or CSP co-ordinator, as appropriate.

FIGURE 19.

Flow diagram of the SHIELD CSP-RYCT trial.

Blinding

Owing to the nature of the interventions, it was not possible to blind participants or providers to their allocated intervention. In this trial, for example, both the CSP co-ordinator and the RYCT facilitator needed to know which carers were allocated to the combined interventions. However, the research interviewers who assessed outcomes were blinded, mainly by their being denied access to the web-based management system. The research interviewers recorded their perception of participants’ allocations for use as a covariate in statistical analysis.

Ethics arrangements and research governance

Ethics approval was given by the Outer North East London Research Ethics Committee (09/H0701/54). The clinical trial was registered as ISRCTN37956201. As per the previous studies, a reporting procedure was in place to ensure that SAEs were reported to the chief investigator. The consent procedures were described in the trial protocol. 147 Three formal groups met regularly to manage the trial within the SHIELD programme: the Project Management Group, the Programme Steering Committee and the Data Monitoring and Ethics Committee.

Setting

The trial ran in community settings in England (four boroughs in North East London plus sites in Norfolk, Northamptonshire and Berkshire).

Participants

Measures

Demographic information was collected from family carers for both the carer and the person with dementia, including age, sex, kinship, education and living situation. The nature of the carers’ social network was assessed using the Practitioner Assessment of Network Type. 151 The cognitive status of the person with dementia was measured using the MMSE45 and an interviewer rating of the global functioning of persons with dementia was made using the CDR scale. 46 Information on use of health, social care, voluntary sector and other relevant services for both the carer and person with dementia was collected through interviews with carers, using the CSRI (see Resource use and cost measures). 97

Primary outcomes

The primary outcome was the family carers’ health-related quality of life, measured by the validated and widely used mental component summary (MCS-12) of the UK SF-12. 152 The SF-12 measures general health status from the perspective of the participant and, in addition to the MCS-12, also allows for the generation of a second subscore, the physical component summary (PCS-12). The SF-12 includes eight concepts commonly represented in health surveys: physical functioning, role functioning physical, bodily pain, general health, vitality, social functioning, role functioning emotional and mental health. A higher score is indicative of better mental and physical health. Reliability is 0.74 for the MCS-12 and 0.78 for the PCS-12. 152 Validity is 0.97 for the MCS-12 and 0.67 for the PCS-12. 96

The primary outcome for the person with dementia was quality of life as measured by the QOL-AD. 103 The measure assesses 13 items, namely physical health, energy, mood, living situation, memory, family, marriage, friends, chores, fun, money, self and life as a whole. The 13 items are summed to generate a total score of between 13 and 52. Responses range from poor (1), fair and good to excellent (4), with higher scores indicative of better quality of life. It has good internal consistency, validity and reliability. 89 The QOL-AD proxy was also completed by the carer, which is identical in structure and content to the person with dementia version.

Secondary outcomes for family carers

• Health-related quality of life The EQ-5D98 comprises five items and a VAS. The measure includes five dimensions of quality of life (mobility, self-care, usual activities, pain/discomfort and anxiety/depression), with each dimension scored on a three-level scale ranging from 1, ‘no problem’, to 3, ‘extreme problem’, and is used as the basis for calculating utility scores. It also includes a VAS on which respondents are asked to rate their current health, ranging from 0 points (worst possible health state) to 100 points (best possible health state).

• Anxiety and depression The Hospital Anxiety and Depression Scale (HADS)153 comprises 14 items, with seven evaluating anxiety (e.g. ‘worrying thoughts go through my mind’) and seven assessing depression (e.g. ‘I have lost interest in my appearance’). The HADS is scored on a 4-point scale ranging from 0 to 3, with a higher score indicating higher levels of anxiety and depression. The scale also measures the severity of anxiety and depression; the score for each subscale can range from 0 to 21. Cut-off points have been used to indicate caseness, whereby a score of ≤ 7 indicates a non-case for both subscales, a score of 8–10 indicates a doubtful case and a score of ≥ 11 indicates a definite case. The HADS subscales have been found to have good reliability and validity in a range of contexts and populations. 153,154 Scores for the entire scale (emotional distress) were also calculated, ranging from 0 to 42, with higher scores indicating more distress.

• Emotional loneliness This was measured using the 2-item Loneliness Scale. 155 The questions used are ‘over the past 7 days, how much have you felt distressed by feeling lonely/feeling lonely even when you are with people?’ Both items are rated on a 5-point scale from ‘not at all’ (0) to ‘extremely’ (4). The scores of the items are summed to create a total loneliness score, with a range of between 0 and 8.

• Caregiver distress This was measured using the caregiver distress scale of the NPI. 94 The NPI evaluates the frequency and severity of 10 behavioural disturbances that can occur in dementia (delusions, hallucinations, dysphoria, anxiety, agitation/aggression, euphoria, disinhibition, irritability/lability, apathy and aberrant motor activity) and the associated distress experienced by the carer, rated from 0 (no distress) to 5 (extreme or very severe distress) for each domain. The distress score total is calculated by the sum of all the individual domain scores, with a range of between 0 and 60.

• Positive affect This was measured using the positive affect scale from the Positive and Negative Affect Schedule. 156 Participants rate the extent to which they have felt each of 10 positive mood states (e.g. determined) within a specific time frame along a five-point scale ranging through ‘very slightly or not at all’ (1), ‘a little’, ‘moderately’, ‘quite a bit’ and ‘extremely’ (5). Scores range between 10 and 50, with higher scores indicative of greater positivity.

• Positive aspects of caring These were measured using the four-item positive aspects subscale from the Carers of Older People in Europe (COPE) index. 157 Responses are recorded from the categories never, sometimes, often and always (or not applicable).

• Personal growth The 3-item version Personal Growth Index was used. 158 Respondents report the degree to which they agree or disagree with statements indicating feelings of personal growth on a six-point scale from 1 (strongly disagree) to 6 (strongly agree). The Index score is obtained by summation of the three item scores, ranging between 3 and 18. Higher scores are indicative of higher personal growth.

• Relationship quality The Quality of Caregiver–Patient Relationship (QCPR)159 comprises 14 items measuring expressed emotion along two dimensions: level of absence of criticism and conflict, and level of warmth. Responses are scored on five-point Likert scales ranging from ‘totally disagree’ to ‘totally agree’, with the criticism and conflict subscale items reverse coded so that a higher score reflects a better relationship. Possible scores range between 14 and 70 and scores are dichotomised, with scores of > 42 denoting a good relationship and those of ≤ 42 denoting a poor relationship.

Secondary outcomes for people with dementia

Secondary outcomes for people with dementia included the HADS, QCPR and EQ-5D, as also used with the family carers. In addition, family carers gave a rating of the person with dementia’s functional capacity in ADLs using the ADCS-ADL. 95 The inventory comprises 23 items covering physical and mental functioning and independence in self-care within the last 4 weeks. Each item consists of a series of hierarchical questions designed to determine a person’s ability to perform one of the ADLs, ranging from total independence to total inability. The total score is calculated by summation of item scores, ranging between 0 and 78. Lower values are indicative of greater disability. The inventory has good reliability. 95

Quality of life was also assessed using the DEMQOL,99 both the person with dementia version and the carer-completed proxy version. The DEMQOL covers the domains of health, well-being, cognitive functioning, social relationships and self-concept. Items are rated on a 4-point scale ranging from 1 (a lot) to 4 (not at all) and summed to produce a total score. A higher score is indicative of better quality of life. The DEMQOL comprises 28 items and scores range between 28 and 112, whereas the DEMQOL-Proxy comprises 31 items and scores range between 31 and 124. It has high internal consistency (0.87) and acceptable inter-rater reliability (intraclass correlation coefficient 0.84).

Measuring potential mechanisms action

Modelling of the stress process in family carers of people with dementia has indicated the value of psychological and social resources in influencing carer outcomes. We therefore included measures of coping,160,161 self-efficacy (Revised Scale for Caregiving Self-Efficacy),162 self-efficacy in responding to challenging behaviours163 and social support. 164

Identifying resource use

The CSRI97 adapted for use in this study gathered information on health and social care services used by participants with dementia and by their family carers. For the purposes of the trial the CSRI was completed three times, on each occasion asking about service use retrospectively over the previous 3 months: at randomisation to the trial (baseline), at 5 months after the first randomisation (after 3 months of weekly intervention) and at 12-month follow-up. Carers identified services used by participants with dementia covering the categories of accommodation, hospital services, community services, equipment and adaptations, day services and medication use. Carers also provided records of their own medication use and identified any other services used which fell in one of the other aforementioned categories. Unpaid family carer inputs were also collected.

Data management and statistical analysis

Data were entered into MACRO for clinical trials. Data audits were carried out throughout the trial to check ongoing data integrity. These audits were performed on a randomly selected 10% sample of data for each site at each time point, to ensure that the data entered into the MACRO database were consistent with the hard-copy paper questionnaires. Once data had been entered into the MACRO database and had been screened and cleaned for any inconsistencies or errors, they were then transferred into SPSS. Further accuracy checks were then carried out, such as manually checking for out-of-range values and scoring.

Handling missing data

When individual items were missing within scales or subscales, data were imputed before the calculation of the scale or subscale score. Pro-rating within measures was undertaken at the 20% missing item level (e.g. for a 5-item score if there was one item missing this was completed with the mean of the other items). Multiple imputations at time points were conducted; however, no imputation was completed for a dyad if all measures were missing at a time point.

Effectiveness analysis

As recommended for the analysis of a factorial design, an ‘at the margins’ analysis was carried out for the two main effects of CSP versus not CSP (alone or in combination with RYCT) and RYCT versus not RYCT (alone or in combination with CSP). The interaction term was also included in the analysis. A multilevel ANCOVA was used with 12 months’ follow-up data as the dependent variable and with baseline score as the covariable. Group allocation was treated as fixed effects, together with sex and dyad relationship. The effects of locality (including personnel) and time were treated as random effects and the covariance structure of this was assessed. A sensitivity analysis was performed to test whether or not plausible changes in key variables, such as carers’ ages and relationships to their relatives with dementia, affected findings.

Intervention costs

Costs of the CSP intervention included staff costs and training for the volunteer carer supporters, plus volunteers’ time and expenses. Each site employed a carer supporter co-ordinator for 1 day per week who was part of either a NHS or a voluntary sector organisation. CSP intervention costs were allocated to each family carer/dyad according to the amount of time the carer supporter spent supporting each family carer/dyad, plus the costs of reimbursing volunteers’ travel. The unit cost for providing the carer supporter training included trainer costs, costs of the venue, materials and refreshments, and cost of producing a specially developed DVD. A manual (a 70-page spiral-bound book) was posted to each trainee carer supporter and its development was undertaken during the staff time already identified. Time and expenses data were sorted so that entries for each family carer/dyad were listed together and three ‘time’ totals calculated over the period they received the intervention: (1) carer supporter time spent training, including travel; (2) time spent by the carer supporter supporting the family carer, including travel; and (3) ‘other’ time, mainly when the carer supporter was in contact with the carer supporter co-ordinator (direct support). Expenses and time were also totalled by family carer/dyad over the period they received the intervention. Three values were used for volunteer time: £0 when a public sector cost perspective was taken and two values for when a societal perspective was employed – a replacement cost (health-care assistant) and an opportunity cost (minimum wage). Public sector costs based on the service-level agreements constituted the ‘overarching’ costs associated with activities that allowed the intervention (carer support) to be implemented: recruiting, organising, training and supporting the carer supporter.

For the RYCT intervention, staff costs were the major component and a unit cost (per hour) was estimated for each team member employed by a health or social care agency or voluntary organisation. This included the mean/median salary for that professional group/grade, additional salary on-costs such as employers’ National Insurance and superannuation contributions, and a proportion of the salary to cover direct and indirect overheads. Using this cost per hour, a cost per session for each staff member was estimated and their travel costs were added. Each time a particular member of staff was present at one of the sessions, their unique staff ‘cost per session’ was included, allowing a total cost for any given session to be calculated. Dyad transport costs were calculated separately and included the reimbursed transport costs for each dyad attending at each session. The cost per dyad per session was calculated by dividing the total cost per session by the number of dyads attending that session. The cost per session per dyad is mainly driven by the number of dyads attending each session, but also by the number of team members present. In turn, these figures were totalled for each participating dyad to arrive at a cost per programme per dyad that varied depending on which sessions they attended. Thus, a unique cost for each dyad was calculated which reflected how much of the intervention they received.

Costs of resource use

Unit costs were obtained so as to reflect long-run marginal opportunity costs, drawn where possible from the Personal Social Services Research Unit compendium for 2011. 100 We used a 3.5% discount rate (as recommended by Her Majesty’s Treasury) for items that provide a benefit for > 1 year, such as equipment or adaptations. Medication costs were obtained from the British National Formulary database. 101 Costs for equipment and adaptations to homes were obtained from market sources. Although it was possible to find unit costs in 2011 prices for most items, for a very small number of services we used the Consumer Price Index inflation rate to adjust costs to 2011 price levels.

The costs of unpaid family carer inputs were calculated in three ways, following the approach used for volunteers. For the public sector perspective, a value of £0 per hour was employed. For the societal perspective, the opportunity cost approach assumed that the unpaid carer would be able to find employment with a wage rate equal to the national minimum wage (which was the hourly cost used in this calculation) and the replacement cost was estimated as the hourly cost of a health-care assistant, under the assumption that a care worker would need to be hired to provide care if the unpaid family carer was unable to do so.

Cost-effectiveness analyses

The primary economic evaluation estimated the incremental cost-effectiveness of:

• those in the sample receiving CSP compared with those who did not

• those in the sample receiving RYCT compared with those who did not

• CSP and RYCT combined compared with TAU.

For the primary analysis we adopted a societal perspective, with the main outcome measures being the MCS-12 score of the SF-12 for the carer and QOL-AD for the person with dementia. The secondary analyses adopted both a health and social care perspective and a societal perspective, with effectiveness measured by the carer PCS-12 and QALYs estimated using EQ-5D for carers, and both DEMQOL- and DEMQOL-Proxy-based QALYs for the person with dementia. As these analyses have many possible combinations of outcomes, we followed the standard economic evaluation approach of extended dominance. There were four potential results from each intervention group comparison. For instance:

1. the intervention is less costly and more effective (has better outcomes) than usual care (TAU), in which case the decision-maker would be likely to be attracted to the intervention

2. the intervention is more costly and less effective than usual care (TAU), in which case it would be unlikely that the decision-maker would consider the intervention over TAU

3. the intervention is less costly but less effective than TAU

4. the intervention is more costly and more effective than TAU.

Results (1) and (2) are scenarios that exhibit strong dominance, and the recommendations made to the decision-maker are typically straightforward. For results (3) and (4), however, the decision is not as straightforward and decision-makers will need to weigh the potential outcome differences against the cost differences before deciding whether or not to implement the new intervention. For these instances, we estimated ICERs, defined as ICER = ΔC/ΔE, where ΔC denotes the difference in mean cost between the interventions being compared and ΔE denotes the difference in the outcome measure between the two interventions. These assist in providing a cost per unit of outcome for the outcome measure of interest. Incremental costs and effects, and their CIs, were estimated with seemingly unrelated regression methods with 1000 bootstrap replications. We plotted CEACs when necessary to show the probability that an intervention is cost-effective relative to its comparator against a series of hypothetical WTP values.

One sensitivity analysis recalculated carer costs (used under the societal perspective) by adopting a replacement cost approach. In this case the hourly cost of providing care was set to be equal to the hourly cost of hiring a health-care assistant to provide care for the person with dementia: a replacement cost approach.

Participant flow

A summary of participant flow throughout the trial is provided in the CONSORT diagram (Figure 20). In total, the research team received 640 expressions of interest from carers and 639 were screened for eligibility. Of these, 292 family carers consented into the trial, with the remaining 347 carers excluded from the trial. Of these, 177 carers were excluded for primary eligibility reasons, with the most frequent reasons including that researchers were unable to contact the family carer (n = 66), time constraints (n = 33) and that the person with dementia had recently passed away (n = 12). It is not possible to determine whether or not those 170 who declined involvement would have been eligible for the trial. The reasons for exclusion are shown in Figure 20. No demographic or psychosocial information was collected about potential participants prior to their giving written informed consent. As a result, it is not possible to determine whether or not those who decided to take part in the research differ from those carers who decided not to take part in the research or were excluded from the trial.

FIGURE 20.

The CONSORT summary of participant flow.

Of the 292 family carers who gave informed consent to take part in the research, one withdrew before randomisation. A total of 291 carers completed the baseline assessment and were randomised between January 2010 and March 2012. Of the 291 dyads randomised, only 289 were included in the final analysis. Two dyads were removed as there were no data recorded at any time point throughout the trial.

Follow-up interviews were carried out as soon as could be arranged after the interview due date, this being 5 and 12 months post randomisation. A total of 253 family carers completed the first follow-up (5 months post randomisation), whereas 36 dyads withdrew before the first follow-up, mainly as a result of stress, poor health and time constraints. Two of the dyads were not available to complete the first follow-up assessment, but were available to complete the second follow-up assessment. In total, 241 family carers completed the final follow-up (12 months post randomisation) giving an overall 12-month retention rate of 83%. However, loss to follow-up was slightly greater in the TAU group than in the intervention groups (Table 33).

Baseline characteristics

Baseline psychometric measures

Scores on psychometric measures at baseline are shown in Tables 36 and 37. The groups are mainly equivalent for people with dementia, with the exception of MMSE score where the TAU group was less impaired than other groups. For the family carers, the groups were also equivalent at baseline, with the exception of EQ-5D utility scores (higher in the RYCT/CSP group), positive affectivity score (higher in the TAU group) and personal growth (higher in the TAU group).

Completers versus non-completers

A comparison of baseline characteristics using Fisher’s exact test and Mann–Whitney U-test of those dyads who completed final follow-up with those who withdrew before final follow-up did not indicate any significant differences. The baseline characteristics that were considered included carer sex (p = 0.74), age (p = 0.12), ethnicity (p = 0.67), marital status (p = 0.62), carer/care recipient relationship type (p = 0.37), living situation (p = 0.12), level of education (p = 0.76), type of dementia (p = 0.53), length of time spent caring (p = 0.75) and length of time since the person with dementia was diagnosed (p = 0.56).

Intervention uptake and adherence

In the 49 dyads allocated to the CSP intervention there was a 76% uptake of the intervention, in the 97 dyads allocated to the RYCT intervention there was a 61% uptake of the intervention and in the 97 dyads allocated to the combined intervention there was a 82% uptake of at least one of the interventions.

Effectiveness results

The results of the primary intention-to-treat analysis for family carers (SF-12 MCS-12 at 12 months) and all secondary outcome analyses are shown in Tables 38 and 39. There was no indication of benefit of either peer support or reminiscence intervention over control on any of the outcome measures. There was no indication of a significant interaction between the interventions.

There was some indication of impact on the person with dementia. There was a significant benefit to quality of life as measured by the DEMQOL (p = 0.04) for those allocated to the CSP intervention. The interaction term was significant for daily functioning (p = 0.02), quality of life as measured by the QOL-AD proxy (p = 0.02) and anxiety (p = 0.05), indicating that the assumption of independence between the two interventions was not valid.

Intervention

Costs of health and social care services

For the pre-baseline period, most of the health and social care-related costs were made up by community care services, with an average of £1171 for a 3-month period, whereas (care-related) accommodation costs contributed to a much smaller extent (£88). However, at final follow-up, the average quarterly expenditure for residential care had increased to £1335. Hospital service use typically increased with time, with costs averaging £301 at baseline and £410 at second follow-up. Costs of day-care services and medications barely altered over time, averaging within the £200–300 range at both baseline and final follow-up. For the overall 12-month period, community services remained the highest contributor to total health and social care costs (£4686), followed by accommodation costs (£3225) and hospital services (£1640). The smallest cost component was adaptations and equipment (£269).

Comparing total health and social care expenditures for the person with dementia with those of the family carer at baseline, the cost for the former was about 15 times higher (£2170 vs. £141). This gap was wider at second follow-up (£3475 vs. £154: 23 times higher). For the total 12-month period, the average health and social care costs were £12,497 for the person with dementia and £614 for the family carer.

The only adjusted differences in the distributions of costs between groups found to be significant at the 5% confidence level were residential and accommodation costs at first follow-up and for the whole 12-month period, for family carer at baseline and for some of the cost aggregates, which included intervention-related costs.

Frequency of service use

There was no difference in frequency of service use for either family carers or people with dementia between intervention groups or TAU, although accommodation showed the most variance between groups (p-values ranged between 0.10 and 0.15 depending on the time point). At baseline, a small percentage of people with dementia used care accommodation and this percentage increased over time, so that by final follow-up > 25% of some intervention groups generated accommodation costs. About two-thirds of the people with dementia used hospital services at baseline, declining to around half at final follow-up. At baseline, 95% of people with dementia made use of some sort of community service, this percentage falling to 85% at final follow-up. At baseline, day services were used by half of the sample of people with dementia, with this percentage remaining fairly stable between groups and time points. Every person with dementia was receiving some form of unpaid family care at baseline, given the study design, with this percentage remaining at ≥ 90% at final follow-up.

At baseline, the percentage of carers using health and social care services ranged between 38% and 46%, depending on the treatment group, with this percentage declining slightly at final follow-up. About two-thirds of carers reported using medication, whereas for the people with dementia the mean was about 90%.

Cost-effectiveness analysis

Table 41 displays the incremental costs, incremental effects (both measured over 12 months) and ICERs for the two primary outcomes: MCS-12 for carers and QOL-AD for people with dementia. Results are shown from three perspectives: two societal perspectives using (in turn) opportunity cost and replacement cost approaches to value unpaid care, and a health and social care perspective, which does not include unpaid care. This analytic strategy was repeated for the secondary outcomes: PCS-12 for carers and QALYs for people with dementia derived in turn from the EQ-5D, DEMQOL and DEMQOL-Proxy measures. These results are shown in Table 42.

TABLE 41 - Incremental cost-effectiveness ratios for primary outcomes

Variable	CSP vs. not CSP			RYCT vs. not RYCT			CSP/RYCT combined vs. TAU
	Incremental cost (£, 2010/11), mean (95% bootstrapped CI)	Incremental effect, mean (95% bootstrapped CI)	ICER (£)	Incremental cost (£ 2010/11), mean (95% bootstrapped CI)	Incremental effect, mean (95% bootstrapped CI)	ICER (£)	Incremental cost (£ 2010/11), mean (95% bootstrapped CI)	Incremental effect, mean (95% bootstrapped CI)	ICER (£)
Societal perspective: opportunity cost scenario (0–12 months)
MCS-12	3957 (–1105 to 9019)	0.46 (–1.39 to 2.31)	8601	1492 (–4423 to 7408)	–1.07 (–2.93 to 0.78)	Not-RYCT dominant	7509 (–347 to 15,365)	–1.16 (–3.90 to 1.59)	TAU dominant
QOL-AD	3973 (–1082 to 9028)	–0.03 (–1.66 to 1.60)	Not-CSP dominant	1495 (–4417 to 7408)	0.67 (–1.54 to 2.89)	2220	7520 (–318 to 15,358)	0.50 (–2.45 to 3.45)	15,015
Societal perspective: replacement cost scenario (0–12 months)
MCS-12	3543 (–11,230 to 18,315)	0.46 (–1.39 to 2.31)	7666	–7423 (–23,647 to 8801)	–1.07 (–2.93 to 0.79)	6914	739 (–26,122 to 27,600)	–1.15 (–3.90 to 1.59)	TAU dominant
QOL-AD	3436 (–11,319 to 18,191)	–0.04 (–1.67 to 1.59)	Not-CSP dominant	–7431 (–23,659 to 8797)	0.67 (–1.55 to 2.89)	RYCT dominant	651 (–26,222 to 27,524)	0.49 (–2.46 to 3.45)	1325
Health and social care perspective (12 months)
MCS-12	2887.9 (–850.9 to 6626.7)	0.46 (–1.39 to 2.31)	6312	4356 (230 to 8482)	–1.08 (–2.93 to 0.78)	Not-RYCT dominant	8543 (4152 to 12,994)	–1.16 (–3.90 to 1.58)	TAU dominant
QOL-AD	2888 (–845 to 6620)	–0.04 (–1.67 to 1.59)	Not-CSP dominant	4356 (234 to 8478)	0.67 (–1.56 to 2.89)	6517	8543 (4157 to 12,928)	0.49 (–2.47 to 3.45)	17,428

TABLE 42 - Incremental cost-effectiveness ratios for secondary outcomes

These ICER values arise because RYCT is less costly but also less effective; see text below for discussion.

Variable	CSP vs. not CSP			RYCT vs. not RYCT			CSP/RYCT combined vs. TAU
	Incremental cost (£ 2010/11), mean (95% bootstrapped CI)	Incremental effect, mean (95% bootstrapped CI)	ICER (£)	Incremental cost (£ 2010/11), mean (95% bootstrapped CI)	Incremental effect, mean (95% bootstrapped CI)	ICER (£)	Incremental cost (£ 2010/11), mean (95% bootstrapped CI)	Incremental effect, mean (95% bootstrapped CI)	ICER (£)
Societal perspective: opportunity cost scenario (0–12 months)
PCS-12	3969 (–1088 to 9026)	0.81 (–1.07 to 2.68)	4920	1495 (–4422 to 7411)	–1.26 (–3.21 to 0.68)	Not-RYCT dominant	7517 (–333 to 15,368)	–0.85 (–3.39 to 1.69)	TAU dominant
QALY (EQ-5D)	3966 (–1090 to 9022)	0.02 (–0.01 to 0.05)	208,400	1494 (–4424 to 7412)	–0.03 (–0.06 to 0.00)	Not-RYCT dominant	7515 (–338 to 15,369)	–0.01 (–0.05 to 0.03)	TAU dominant
QALY (DEMQOL)	3956 (–1112 to 9024)	0.01 (–0.01 to 0.03)	292,435	1492 (–4427 to 7412)	–0.01 (–0.04 to 0.02)	Not-RYCT dominant	7508 (–359 to 15,376)	0.00 (–0.04 to 0.04)	183,580,636
QALY (DEMQOL-Proxy)	3994 (–1060 to 9048)	0.01 (–0.01 to 0.03)	636,633	1499 (–4408 to 7406)	0.01 (–0.01 to 0.03)	141,665	7534 (–283 to 15,351)	0.02 (–0.01 to 0.06)	306,490
Societal perspective: replacement cost scenario (0–12 months)
PCS-12	3610 (–11,165 to 18,386)	0.81 (–1.07 to 2.68)	4477	–7418 (–23,647 to 8811)	–1.26 (–3.21 to 0.68)	5865a	794 (–26,059 to 27,647)	–0.85 (–3.39 to 1.69)	TAU dominant
QALY (EQ-5D)	3445 (–11,361 to 18,251)	0.02 (–0.01 to 0.05)	178,979	–7431 (–23,667 to 8806)	–0.03 (–0.06 to 0.00)	247,968a	659 (–26,226 to 27,583)	–0.01 (–0.05 to 0.03)	TAU dominant
QALY (DEMQOL)	3486 (–11,331 to 18,303)	0.01 (–0.01 to 0.03)	258,113	–7427 (–23,658 to 8804)	–0.01 (–0.04 to 0.02)	737,676a	692 (–26,218 to 27,602)	0.00 (–0.04 to 0.04)	35,308,661
QALY (DEMQOL-Proxy)	3669 (–11,077 to 18,416)	0.01 (–0.01 to 0.03)	595,549	–7414 (–23,635 to 8807)	0.0105 (–0.01 to 0.03)	RYCT dominant	843 (–25,946 to 27,631)	0.0243 (–0.01 to 0.05)	34,624
Health and social care perspective (0–12 months)
PCS-12	2886 (–850 to 6623)	0.81 (–1.07 to 2.68)	3578	4356 (233 to 8479)	–1.26 (–3.21 to 0.68)	Not-RYCT dominant	8541 (4153 to 12,929)	–0.85 (–3.39 to 1.69)	TAU dominant
QALY (EQ-5D)	2889 (–844 to 6622)	0.02 (–0.01 to 0.05)	151,027	4356 (229 to 8482)	–0.03 (–0.06 to 0.00)	Not-RYCT dominant	8544 (4151 to 12,938)	–0.01 (–0.05 to 0.03)	TAU dominant
QALY (DEMQOL)	2887 (–849 to 6622)	0.01 (–0.01 to 0.03)	213,369	4357 (230 to 8482)	–0.01 (–0.04 to 0.02)	Not-RYCT dominant	8541 (4155 to 12,928)	0.00 (–0.04 to 0.04)	182,898,801
QALY (DEMQOL-Proxy)	2880 (–857 to 6618)	0.0063 (–0.0126 to 0.0251)	457,823	4358 (238 to 8479)	0.0106 (–0.0076 to 0.0288)	411,373	8533 (4146 to 12,920)	0.0246 (–0.0062 to 0.0555)	346,568

Cost-effectiveness using carer outcomes

For carers, none of CSP, RYCT or CSP/RYCT combined appears to be cost-effective compared with its respective comparator, whatever the study perspective.

Looking at the primary outcome for carers (MCS-12) and taking the primary perspective (societal with opportunity cost values), both RYCT and CSP/RYCT combined are dominated; that is, they are less effective and have higher costs than their respective comparators. CSP alone had marginally better outcomes and higher costs (although neither was significantly different), with the resultant ICER suggesting that the cost for a 1-point improvement on the MCS-12 is £8601. There is no external threshold with which to compare this ICER, but it looks high given that MCS-12 scores can range from 0 to 100.

The findings are the same under a health and social care perspective: RYCT and CSP/RYCT combined are dominated – indeed, the combined intervention is significantly more costly than TAU – and CSP has a high ICER.

Recalculating carer costs using a replacement cost approach, CSP still has a high ICER and CSP/RYCT combined is dominated. RYCT significantly reduces costs but has worse outcomes than its comparator (the ICER is £6914 for each 1-point worsening on the MCS-12). Given that neither the cost nor the outcome difference in this case is significant – and that the ‘cost reduction’ is the imputed value of carer time, and therefore not easily transferable as a ‘saving’ for use elsewhere in the system – it is unlikely that the decision-maker would choose RYCT in these circumstances.

The two secondary outcomes for carers were PCS-12 and QALYs generated by the EQ-5D. Looking first at PCS-12 and taking the primary perspective (societal, opportunity cost assumption), both RYCT and CSP/RYCT combined are dominated by their comparators. CSP has higher costs but better outcomes than not CSP (although neither achieves significance) and the ICER appears high at £4920, although there are no established thresholds for PCS-12 against which to compare this figure.

The findings are similar under a health and social care perspective: RYCT and CSP/RYCT combined are both dominated, and in fact each has significantly higher cost than its comparator, whereas CSP has an ICER value (£3578) likely to be considered high.

Under a societal perspective with replacement cost values for carer time, CSP/RYCT combined is dominated by its comparator (TAU). RYCT has lower costs but worse outcomes (although neither is significant). In a similar vein to the earlier argument, this finding is unlikely to encourage the decision-maker to choose RYCT over non-RYCT because the differences are not significant, and the ‘saving’ is in the ascribed value of carer time that is not easily transferred to another use in the care system. CSP has slightly better effectiveness on PCS-12 than not CSP but also higher costs (although neither is significant) and the ICER appears high.

Using QALY as an outcome measure, whether measured using EQ-5D (for carers) or DEMQOL or DEMQOL-Proxy (for people with dementia), is helpful in cost-effectiveness analyses because it is then possible to compare an estimated ICER value (when adopting a health and social care perspective) with thresholds recommended by NICE, usually quoted as £20,000 per QALY gained.

From both a societal perspective (opportunity costs) and health and social care perspectives the findings are the same: CSP has (non-significantly) higher costs and (non-significantly) lower costs than not CSP (with ICERs considerably greater than the NICE threshold) and both RYCT and CSP/RYCT combined are dominated by their comparators. (The QALY difference for RYCT is small but statistically significant, and under a health and social care perspective RYCT also has significantly higher costs.)

From a societal perspective (replacement costs), CSP is again unlikely to be seen as cost-effective given its high ICER. RYCT generates significantly lower QALYs than not RYCT but also reduces costs, and for the reasons noted earlier is unlikely to be seen as the preferred option. CSP/RYCT combined is dominated by TAU.

Cost-effectiveness using outcomes for people with dementia

The primary outcome measure for people with dementia was QOL-AD. CSP is dominated by not CSP when adopting any of the perspectives and costing approaches. RYCT had marginally but not significantly better effectiveness on this measure, higher cost from the societal perspective with opportunity cost and also the health and social care perspective (with cost difference being significant), and marginally but not significantly lower cost from the societal perspective with replacement cost assumptions. There are no established thresholds against which to compare ICER values for QOL-AD. The CEAC (Figure 21) suggests that RYCT is cost-effective across a wide range of WTP values. For comparison, the CEAC in Figure 22 is for the same comparison but now from a health and social care perspective: RYCT is not cost-effective. CSP/RYCT combined has high ICER values under two perspectives and, under the other (societal with replacement costs), the estimated CEAC (Figure 23) does not suggest that it would be seen as cost-effective.

FIGURE 21.

Cost-effectiveness acceptability curve for RYCT, using QOL-AD as outcome, from a societal perspective (replacement cost assumption).

FIGURE 22.

Cost-effectiveness acceptability curve for RYCT, using QOL-AD as outcome, from a health and social care perspective.

FIGURE 23.

Cost-effectiveness acceptability curve for CSP-RYCT combined, using QOL-AD as outcome, from a societal perspective (opportunity cost).

The secondary outcomes for people with dementia were QALYs measured by DEMQOL and DEMQOL-Proxy. From a societal perspective (opportunity costs), CSP is not cost-effective by reference to either measure, as ICER values are all considerably above the NICE threshold. RYCT is dominated by its comparator using DEMQOL, has a high ICER (£141,665 per QALY) using DEMQOL-Proxy and is not cost-effective. CSP/RYCT combined is not cost-effective, as the ICERs are again considerably above the NICE threshold.

From a societal perspective with replacement costs, CSP is not cost-effective. RYCT is not cost-effective using DEMQOL, but dominates not RYCT using DEMQOL-Proxy and is cost-effective. CSP/RYCT combined has a very ICER using DEMQOL and, although the ICER is much closer to the NICE threshold using DEMQOL-Proxy (although the comparison is not so relevant given the perspective), the CEAC (Figure 24) does not suggest cost-effectiveness as the curve barely gets above the 0.5 probability level, even at high WTP values.

FIGURE 24.

Cost-effectiveness acceptability curve for CSP-RYCT combined, using QALYs from DEMQOL-Proxy as outcome, from a societal perspective (replacement cost).

From a health and social care perspective, none of the interventions is cost-effective.

Bringing these cost-effectiveness results together:

• CSP is not cost-effective whether considering outcomes for carers or outcomes for people with dementia.

• RYCT is not cost-effective when considering outcomes for carers or most outcomes for people with dementia. It is cost-effective when looking at the QOL-AD measure and from a societal perspective with replacement costs, but this was not the primary perspective chosen for the analysis before the trial began.

• CSP-RYCT combined is not cost-effective when considering outcomes for carers or outcomes for people with dementia.

Primary objective

To evaluate the effectiveness of case management approaches to home support for people with dementia from the perspective of the different people involved (patients, carers and staff), compared with other forms of treatment including TAU, standard community treatment and other non-case management interventions, on delaying institutionalisation, improving quality of life and/or reducing the number of hospital admissions.

Secondary objective

To study whether or not other potential mediating variables affect case management outcomes (e.g. key structural and organisational features of case management interventions and also the methodological characteristics of studies).

Protocol and registration

The protocol was registered with The Cochrane Library (http://archie.cochrane.org).

Criteria for considering studies for this review

Randomised controlled trials of case management interventions for people with dementia who lived in the community and their carers were included, if they had been published and written in English, peer reviewed and presented in a journal article. Authors were contacted for missing data, such as details of randomisation, means and SDs.

Search methods for identification of studies

The search methods included a combination of the search terms patient care management, case management, intensive case management, care management, managed care programs, care co-ordination, care pathway and managed care, which were used to search ALOIS, Specialised Register of the CDCIG and EPOC (Effective Practice and Organisation of Care) register between November and December 2008, and was updated in February 2012. The studies were identified from the Specialised Register of the CDCIG. This register contains records from the following major health-care databases: The Cochrane Library, MEDLINE, EMBASE, PsycINFO, CINAHL and LILACS, as well as ongoing clinical trial databases and other grey literature sources for the most recent records. We contacted the first authors of important identified RCTs that were potentially suitable for inclusion to request additional information on related new, unpublished or in-press studies that had not been identified in the main search. We also cross-checked the reference lists and citation reports of trials and relevant systematic reviews as identified by the above methods. Additional searches were made of the following sources: Web of Science [including Science Citation Index Expanded (SCI-EXPANDED) and Social Science Citation Index] and the Campbell Collaboration/SORO database. We also searched the Specialized Register of the Cochrane Effective Practice of Care Group using the search terms dementia OR demented OR Alzheimer in any fields. A total of 10,440 references were retrieved from the November 2008 initial search. After deduplication and a first assessment, the authors were left with 145 references to further assess for inclusion and exclusion or discard.

Participants

Participants were any age and sex with a diagnosis of dementia (Alzheimer’s disease, vascular dementia or mixed Alzheimer’s and vascular dementia, other types of dementia) who lived in the community (excluding people in institutions receiving 24 hours of care), and their carers. Studies which focused on patients only or both patient and carer dyads were included, whereas those which focused exclusively on carers were not included.

Intervention

Participants received any case or care management intervention delivered in the community (not in hospital/residential care settings). Interventions were screened to ensure that they predominantly focused on the planning and co-ordination of services. The intervention needed to be compared with TAU, standard community treatment, other non-case management or waiting list controls.

Quality assessment

Three reviewers independently screened the identified RCTs for inclusion and the final list of included studies was reached by consensus. Trials not meeting the criteria were excluded. The studies were assessed against a checklist of quality requirements using the Cochrane approach (see Figure 22), as follows.

• Grade A: ‘low risk’ – adequate concealment (randomisation; concealed allocation).

• Grade B: ‘unclear risk’ – ‘randomised’, but methods uncertain.

• Grade C: ‘high risk’ – inadequate concealment of allocation or no randomisation, or both.

No studies were excluded because of poor quality in the review. Again, the reviewers worked independently to ascertain which studies met the quality criteria, and consensus was reached through discussion. Attempts were made to obtain additional information from the study authors when needed.

Data extraction

Descriptive characteristics (such as quality of randomisation and blinding) and study results were extracted, recorded and entered into RevMan 5.1. Additionally, letters and e-mails were sent to some authors of controlled trials asking for essential and additional information (e.g. statistics, sources of bias and details of randomisation). The summary statistics required for each trial and each outcome for continuous data were the mean change from baseline (if reported), the SD of the mean change and the number of patients for each treatment group at each assessment. The baseline assessment was defined as the latest available assessment prior to randomisation, but no longer than 2 months to after randomisation. Categorical responses were extracted for categorical outcome data (e.g. admitted to hospital/not admitted). For the meta-analysis of follow-up data, the most frequently assessed time points were combined across trials (3, 6, 12, 18 and 36 months). To allow an intention-to-treat analysis, the data were sought irrespective of compliance, whether or not the patient was subsequently deemed ineligible or otherwise excluded from treatment or follow-up. If intention-to-treat data were not available in the publication, ‘on-treatment’ data were sought (i.e. the data of those who completed the trial). Discussion between the three reviewers and the other authors was used to resolve any queries.

Data analysis

RevMan 5.1 was used for the meta-analysis. The meta-analysis required the combination of data from trials using the same rating scale/test or a different rating scale/test to assess an outcome. The measure of the treatment difference for any outcome was the weighted mean difference when the pooled trials used the same scale and the SMD, the absolute mean difference divided by the SD, when different rating scales were used. Summary statistics (n, mean and SD) were required for each rating scale at all assessment points, for both treatment groups in each trial for change from baseline. For continuous data (or ordinal data approximating a normal distribution), the mean change from baseline, the SD and the number of patients for each treatment group at all assessments were analysed. For binary outcomes, such as clinical improvement or no clinical improvement, the odds ratio was used to measure treatment effect. The treatment differences from both fixed- and random-effects models were examined and a test for heterogeneity was performed using a standard chi-squared statistic and an I²-statistic. When a significant degree of heterogeneity was present, a random-effects model was preferred. The reviewers achieved consensus on the interpretation of the statistical analyses, seeking specialist statistical advice from the CDCIG as required. Non-randomised studies were described in tabular form and the reviewers discussed and reached consensus on the presentation of the findings in the background to the review.

Quality of included studies

The quality of each study was assessed according to the four criteria outlined in the Cochrane Collaboration Handbook:171 selection bias, performance bias, attrition bias and detection bias. Details of biases and description of studies can be seen in Figure 26. Performance bias was difficult to evaluate. With psychological interventions, unlike drug trials, it is impossible to totally blind patients and staff to treatment. Patients are often aware that they are being treated preferentially, staff involved may have different expectations of treatment groups and independent assessors may be given clues from patients during the assessments. There may also be ‘contamination’ between groups, in terms of groups not being held in separate rooms and staff bringing ideas from one group to another.

FIGURE 26.

Risk of bias for all studies.

Introduction

People with dementia are at increased risk of acute admission to hospital and generally have poorer outcomes than those without dementia. There is an increasing body of evidence reporting on physical health problems among people with dementia, including falls, fractures, seizures, infections and pneumonia, which are highly associated with hospital admissions. Psychiatric factors, particularly behavioural problems, are also a key risk factor for hospital admissions and many people admitted to hospital with dementia enter institutional care on discharge, rather than returning to their own homes.

Aims

To conduct a systematic review and meta analysis of literature of the factors leading to hospital admission for people with dementia in comparison with (1) people without dementia acutely admitted and (2) people with dementia remaining in the community.

Methods

Data collection and analysis

Results

Studies comparing hospital admissions for people with or without dementia

The risk factors reported below were all categorised into broader themes by the research team. The key themes and risk factors can be found in Table 44.

TABLE 44 - Primary reason for admission for people with dementia vs. people without dementia

RR, relative risk.

Paper states the number of patients admitted at least once but not the total number of admissions, so the RR is calculated in relation to community population rather than to admissions.

Paper quotes RRs, so approximate numbers of admissions are calculated from these and other data in the paper.

Primary reason system	Subcategory (where specified)	Study (year)	Level of evidence CEBM	Proportion of all admissions (dementia group)	Proportion of all admissions (non-dementia group)	RR (95% CI; p-value) of given factor being primary cause of admission
Orthopaedics	Falls/injury	Nourhashémi et al.178 (2001)	4	22/118	698/6891	1.84 (1.25 to 2.70; 0.002)
	Fractures (all)	Malone et al.183 (2009)	4	955/5396	428/5396	2.23 (2.00 to 2.48; < 0.001)a
	Hip fracture	Natalwala et al.182 (2008)	4	167/2561	835/53,123	4.15 (3.53 to 4.87; < 0.001)
	General (including falls/fractures)	Tuppin et al.184 (2009)b	4	20,855/263,985	4595/63,985	1.10 (1.07 to 1.13; < 0001)
Respiratory	Infection	Carter and Porell181 (2005)	4	110/1782	123/2195	1.10 (0.86 to 1.41; 0.450)
		Sampson et al.166 (2009)	2	62/262	29/355	2.90 (1.92 to 4.37; < 0.001)
		Natalwala et al.182 (2008)	4	331/2561	1614/53,123	4.25 (3.81 to 4.76; < 0.001)
	General (including infection)	Tuppin et al.184 (2009)b	4	16,103/263,985	2168/63,985	1.80 (1.72 to 1.88; < 0.001)
Urology/renal	Infection	Carter and Porell181 (2005)	4	51/1782	47/2195	1.34 (0.90 to 1.98; 0.146)
		Sampson et al.166 (2009)	2	37/262	17/355	2.95 (1.70 to 5.12; < 0.001)
		Natalwala et al.182 (2008)	4	116/2561	440/53,123	5.47 (4.47 to 6.68; < 0.001)
	General (including infection)	Tuppin et al.184 (2009)b	4	33,262/263,985	6718/63,985	1.20 (1.17 to 1.23; < 0.001)
Gastrointestinal	Infection	Carter and Porell181 (2005)	4	80/1782	66/2195	1.49 (1.08 to 2.06; 0.015)
	General (including infection)	Nourhashémi et al.178 (2001)	4	17/118	676/6891	1.47 (0.94 to 2.29; 0.090)
		Tuppin et al.184 (2009)b	4	17,951/263,985	4834/63,985	0.90 (0.87 to 0.93; < 0.001)
Cardiology	General	Tuppin et al.184 (2009)b	4	15,839/263,985	3199/63,985	1.20 (1.16 to 1.25; < 0.001)
	Acute cardiac syndrome	Sampson et al.166 (2009)	2	12/262	43/355	0.38 (0.20 to 0.70; 0.002)
	Cardiovascular	Nourhashémi et al.178 (2001)	4	11/118	1073/6891	0.60 (0.34 to 1.05; 0.076)
Neurology		Tuppin et al.184 (2009)b	4	39,334/263,985	1799/63,985	5.30 (5.06 to 5.55; < 0.001)
		Nourhashémi et al.178 (2001)	4	9/118	1072/6891	0.49 (0.26 to 0.92; 0.027)
Psychiatric	General	Tuppin et al.184 (2009)b	4	16,631/263,985	443/63,985	9.10 (8.28 to 10.00; < 0.001)
	Behavioural disturbance	Nourhashémi et al.178 (2001)	4	31/118	93/6891	19.47 (13.53 to 28.00; < 0.001)
Other	Syncope/collapse	Natalwala et al.182 (2008)	4	102/2561	777/53,123	2.72 (2.22 to 3.33; < 0.001)
	Fever	Nourhashémi et al.178 (2001)	4	13/118	951/6891	0.80 (0.48 to 1.34; 0.394)
	Infection (general)	Albert et al.177 (1999)	4	29/70	52/191	1.52 (1.06 to 2.19; 0.024)
	Dehydration	Natalwala et al.182 (2008)	4	79/2561	147/53,123	11.15 (8.51 to 14.61; < 0.001)

Physical health-related factors

For risk factors within the following four physical health-related themes, relative risks (RRs) were either directly extracted or calculated by the research team from the relevant studies. The data were then pooled in meta analyses and presented in forest plots for each theme.

Orthopaedics

Four prognostic cohort studies178,182–184 reported that people with dementia are at greater risk of an orthopaedic crisis (e.g. falls and all types of fractures) resulting in hospital admission than people without dementia. The RR for dementia inpatients versus non-dementia inpatients was 1.19 (95% CI 1.16 to 1.22; p < 0.001) (Figure 29).

FIGURE 29.

Orthopaedics meta-analysis and summary statistics.

Respiratory

Four prognostic cohort studies166,181,182,184 investigated respiratory crises resulting in hospital admission. All but one of these reported that people with dementia are at greater risk of respiratory crisis resulting in hospitalisation than people without dementia (see Table 44). The RR for dementia inpatients compared with non-dementia inpatients was 2.00 (95% CI 1.92 to 2.08; p < 0.001) (Figure 30).

FIGURE 30.

Respiratory meta-analysis and summary statistics.

Urology/renal

Four prognostic cohort studies166,181,182,184 reported results on urological/renal factors leading to hospital admission. All four reported that people with dementia are at greater risk than people without dementia of experiencing a urological crisis resulting in hospital admission (see Table 44). All except the Carter and Porell181 study reported statistically significant findings and, according to the meta-analysis, the RR for dementia inpatients compared with non-dementia inpatients was 1.23 (95% CI 1.20 to 1.26; p < 0.001) (Figure 31).

FIGURE 31.

Urology/renal meta-analysis and summary tables.

Gastrointestinal

Three prognostic cohort studies178,181,184 informed the meta-analysis for gastrointestinal factors leading to hospitalisation, but there was no consistent pattern. Nourhashémi et al. 178 and Carter and Porell181 reported that people with dementia were more at risk of experiencing a gastrointestinal crisis resulting in hospitalisation than people without dementia. However, the evidence in the Nourhashémi et al. 178 study was not statistically significant. In contrast, Tuppin et al. 184 found that people with dementia had a lower risk of experiencing a gastrointestinal crisis resulting in admission than those without dementia (see Table 44). A meta-analysis gave a lower RR of 0.91 (95% CI 0.88 to 0.94; p < 0.001) for dementia versus non-dementia inpatients (Figure 32).

FIGURE 32.

Gastrointestinal meta-analysis and summary statistics.

For risk factors within the following physical health-related themes, RR were again either directly extracted or calculated by the research team from the relevant studies. However, reported data were not pooled by meta-analysis for these factors owing to potential heterogeneity between risk factors included within these themes.

Cardiological factors

Three prognostic cohort studies166,178,184 reported data on cardiological factors leading to hospital admission (see Table 44). Tuppin et al. 184 reported that people with dementia had a significantly higher risk than those without dementia of acute hospitalisation because of general cardiological factors (RR 1.20, 95% CI 1.16 to 1.25; p < 0.001). In contrast, Sampson et al. 166 reported that people with dementia had a significantly lower risk than those without dementia of experiencing a crisis leading to hospitalisation due to acute cardiac syndrome (RR 0.38, 95% CI 0.20 to 0.70; p = 0.002). Similarly, Nourhashémi et al. 178 reported that people with dementia had a lower risk than those without dementia of experiencing a cardiovascular crisis resulting in hospital admission (RR 0.60, 95% CI 0.34 to 1.05; p = 0.076).

Neurological

Nourhashémi et al. 178 and Tuppin et al. 184 reported contrasting evidence on neurological crises resulting in hospital admissions. Tuppin et al. 184 found that people with dementia had a much higher risk than those without of being hospitalised secondary to a neurological problem (RR 5.30, 95% CI 5.06 to 5.55; p < 0.001), whereas Nourhashémi et al. 178 found that people with dementia had a significantly lower risk of having a neurological crisis leading to hospital admission (RR 0.49, 95% CI 0.26 to 0.92; p = 0.027).

All other physical health-related factors

Several other physical health-related factors were reported across the studies. Natalwala et al. 182 reported a significantly higher risk of people with dementia being admitted to hospital due to syncope/collapse (RR 2.72, 95% 2.22 to 3.33; p < 0.001) and due to dehydration (RR 11.15, 95% CI 8.51 to 14.61; p < 0.001) compared with people without dementia. Nourhashémi et al. 178 found that people with dementia had a lower risk than people without dementia of experiencing fever leading to admission, but this finding was not statistically significant (RR 0.80, 95% CI 0.48 to 1.34; p = 0.394). Finally, Albert et al. 177 reported that people with dementia had a significantly higher risk than those without of being hospitalised due to infection (all types) (RR 1.52, 95% CI 1.06 to 2.19; p = 0.024).

Psychiatric factors

Two prognostic cohort studies reported statistically significant findings relating to hospital admissions caused by a psychiatric crisis. Tuppin et al. 184 found that people with dementia had a much higher risk than of those without psychiatric crisis resulting in hospitalisation (RR 9.10, 95% CI 8.28 to 10.00; p < 0.001). Similarly, Nourhashémi et al. 178 found that people with dementia had a much higher risk of behavioural disturbance resulting in hospital admission (RR 19.47, 95% CI 13.53 to 28.00; p < 0.001).

Discussion

This is the first review to systematically assess the physical and psychiatric risk factors for admission for people with dementia. In particular, people with dementia were more likely to have either orthopaedic (e.g. falls/fractures) or respiratory/urological (e.g. infections) precipitants of admission than those without dementia. Most of the published work to date has focused on general hospital admissions, and the causes of psychiatric admission in people with dementia remain a relatively neglected area of study. Only two studies180,185 specifically investigated the causes of psychiatric admission, of which one focused only on cases of fronto-temporal dementia, whereas the other also included admissions to day hospital facilities. 185 However, one study178 included admissions to a dementia-specific unit among its analysis of all admissions, and three further studies167,179,184 included some analysis of the influence of psychiatric and behavioural disturbances on admission to acute general hospitals. As expected, psychiatric and behavioural disturbance was found to increase the risk of admission for people with dementia in relation to those without, and sleep disturbance emerged as a particular risk factor for admission. Furthermore, the finding that disruption to the social and environmental milieu often precipitated admission for people with dementia185 provides one of several possible explanations for subsequent behavioural and psychological disturbances. The link between acute physical illness and behavioural change in people with dementia has long been appreciated, and yet psychiatric and general medical services are not always as well integrated as they might be to equip them to deal with these problems.

People with dementia had higher risks of acute admission for respiratory and urological indications (likely to include a high proportion of respiratory and urinary tract infections), and for all-cause infections. Potentially preventable hospital admissions, such as those for urinary tract infections, were 78% more common in people with dementia. 186 There are several potential explanations for this finding. First, people with dementia might be more prone to infection, secondary to factors such as reduced mobility, inadequate fluid intake and the impaired performance of daily living tasks concerning personal hygiene. Second, people with dementia might have reduced or delayed help-seeking behaviour, through either reduced recognition of symptoms or impaired communication skills. Consequently, an acute infection might not be diagnosed in a person with dementia until the physical symptoms become severe, or secondary behavioural or psychological symptoms of delirium develop and, even then, carers or health-care staff might misattribute behavioural disturbance to a direct effect of the dementia rather than seek an underlying physical health cause. Therefore, physical health conditions that can be controlled in the general population through effective community management and treatment, also known as ‘ambulatory care sensitive’ conditions, are more likely to result in acute hospital admission for people with dementia. This highlights the importance of a more holistic approach to providing community dementia care, encompassing the cognitive, behavioural, psychological and physical needs of people with dementia and providing responsive care packages and health education to both people with dementia and their carers.

The need for proper community assessment and supportive interventions for people with dementia is further highlighted by the findings that they are at consistently higher risk than the general population of sustaining falls and fractures necessitating acute hospital admission, and that they are more likely to have significant difficulties with ADLs, including bathing, dressing, toileting and eating. In some cases, an acute decline in ADL abilities might be the result of concomitant physical illness. However, physical health can decline as a result of reduced mobility, poor nutritional intake or impaired personal hygiene. Furthermore, patients with dementia may be admitted to acute hospitals as a result of their social circumstances, often when they have been unable to care for themselves and without appropriate support. For example, neglect and poor provision of care at home could lead to problems such as malnutrition, urinary tract infections, chest infections, falls and fractures.

The UK National Audit Office187 reported how people with dementia receive poorer-quality care in acute hospitals and how the lack of skill of acute hospital staff in caring for people with dementia can have adverse effects on the symptoms of dementia. The National Audit of Dementia Care in General Hospitals188 found that two-thirds of hospital staff felt that they had received insufficient training in dementia care. Given what we now know about the poor outcomes of hospital admission for people with dementia, unnecessary acute hospital admissions must be avoided, and this requires the provision of early community assessment, recognition of problems and appropriate interventions by multidisciplinary teams, including medical staff, social workers, occupational therapists, nurses and trained care staff.

One limitation of this review is that several risk factors could not be analysed using a meta-analysis either owing to potential heterogeneity within risk factors (e.g. cardiovascular, psychiatric and neurological disorders) or because there were insufficient studies for a meta-analysis of the identified risk factors (e.g. social/environmental factors and ADLs). However, two studies166,178 concluded that people with dementia were less likely than those without to be admitted for an acute cardiac syndrome. This might reflect a lower recognition of acute angina or myocardial infarction in those with dementia, owing to reduced reporting of specific symptoms or a tendency to interpret any accompanying confusion as part of the established clinical syndrome of dementia.

Conclusion

Many acute hospital admissions for people with dementia are potentially preventable. Poor outcomes of acute hospital admission for people with dementia have been widely documented; therefore, efforts need to be focused on developing responsive and integrated community services. People with dementia are at increased risk of hospitalisation owing to behavioural and psychological disturbances, and many older people’s HTTs tend to focus on prevention and management of psychiatric crises and reducing psychiatric admissions. This review highlights the need for recognition of the physical health risks in these patients, including a high index of suspicion for conditions such as urinary tract and respiratory infections, and a low threshold for early treatment in the community. This approach could potentially prevent acute admission to general hospitals and reduce the associated risk of delirium, so that mental state and behaviour remain more stable. This highlights the importance of integrated working between services for older people’s mental health, primary care, social welfare, intermediate care and hospital liaison.

Aims

Our aim was to conduct a systematic review of evidence to evaluate the effectiveness of crisis resolution/home treatment approaches to support older people with mental health problems at home compared with other forms of treatment and to scope home treatment services for older people with mental health problems, produce a typology of such services and review the typology in the context of policy and research findings.

Methods: literature review

The method used was a literature review; controlled comparison studies, including RCTs, controlled before-and-after studies and interrupted time series were eligible for inclusion in this review. Observational studies, theoretical papers and government frameworks and policies were also reviewed.

Types of interventions

Types of participants

The inclusion criteria for participants in the studies included in the review were those aged ≥ 65 years, diagnosed with a mental health condition and living in the community (excluding 24-hour care institutions).

Types of outcome measures

Primary outcomes included the following: number of admissions, length of hospital stay, maintenance of community residence and patient quality of life. Secondary outcome measures included cognition, use of medication, ADLs, patient/carer satisfaction, service use, health and social care costs and mortality.

Search methods for identification of studies

Electronic searches: we searched the Specialised Register of the CDCIG, which included records from The Cochrane Library, MEDLINE, EMBASE, PsycINFO, CINAHL and LILACS, and many ongoing clinical trial databases and grey literature sources. We also searched Web of Science and the Campbell Collaboration/SORO database.

Data collection and analysis

Selection of papers: titles and abstracts of citations obtained from the search were examined and irrelevant articles were discarded. The full texts of studies were obtained and assessed against our inclusion criteria. Attempts were made to obtain additional information from the study authors when necessary. A second reviewer independently reviewed the selected studies and agreement was reached on papers for inclusion.

Assessment of validity

The studies (primary research) were assigned a level of evidence according to the CEBM guidelines (www.cebm.net/levels_of_evidence.asp#levels). Levels of evidence ranged from 1 to 5, with lower numbers indicating higher quality. Studies rated between levels 1 and 4 were included in this review.

Procedure

Both reviewers assigned levels of evidence to each study independently and, in cases of disagreement, discussed the studies until a conclusion was agreed. The outcomes across the studies were then assigned an overall grade of evidence from A to D according to the CEBM criteria. Grade A represents consistent level 1 studies indicating the best evidence; grade B represents consistent level 2 or 3 studies; grade C represents evidence from level 4 studies or extrapolations from level 2 or 3 studies; and grade D evidence is the lowest level and represents level 5 evidence or troublingly inconsistent or inconclusive studies of any level.

Methods: scoping exercise

From consultations with colleagues, a review of policy and research literature, plus a scoping exercise we sought to describe the diversity of models of HTTs identifying three principal models. Five crisis resolution home treatment teams (CRHTTs) for older people across the UK were visited to identify what kinds of crises take place, key attributes associated with crises, models of service provision, areas of need and key themes to be included in a HTP. Recent case examples were discussed. The services had common aims: to provide a rapid response to older adults presenting in mental health crisis, to offer assessment and treatment at home as an alternative to hospital admission wherever possible and to facilitate timely discharge from psychiatric inpatient units.

Results: literature review

There were 8222 references found, of which 8185 were excluded by reference to title and abstract alone as they either did not cover crisis interventions or did not include people with mental health problems, and 27 were excluded as they did not consider older people in the sample. Ten references were included: three cohort studies,170,189,190 one descriptive study,27 one survey-related research study,191 two theoretical papers192,193 and three government policy documents. 23,194,195

Cohort studies and descriptive study

Three cohort studies and one descriptive study were included in the review (Table 45). All were assigned level 4 evidence according to the CEBM guidelines. Dibben et al. 190 studied the impact of a CRHTT 6 months before and 6 months after the service was implemented. Participants in this study were older people with mental health problems. There were 65 participants in the pre-CRHTT (comparison) group and 102 in the post-CRHTT (intervention) group.

Ratna170 investigated the efficacy of a 24-hour community-orientated old age psychiatric crisis service in North London, UK. The study concerned 142 older patients presenting in mental health crisis who had been referred to the service over a 6-month period in 1976 and followed up for 2 years. The data were compared with retrospective data from a study196 of two other psychiatric services, one a community service in Chichester (n = 216) and the other a hospital-centred service in Salisbury (n = 121), neither of which provided a crisis service. For the purpose of this review, the data from the two services in the Sainsbury et al. 196 study were combined and included as one comparison group.

Doyle and Varian189 described a crisis intervention service operating between 9 a.m. and 5 p.m. on weekdays. Participants in this study were 70 older people with mental health problems, followed up for 3 years from referral (1989–92). The findings were compared with data from the Ratna170 study.

A prospective descriptive study27 described the activity and outcomes of an outreach support team based in a day hospital for older people with mental illness. This service operated 12 hours per day, 7 days per week. Routine clinical, demographic and service use information was described but no attempt was made to draw a causal relationship between the activities of the outreach support team and subsequent maintenance in the community. The participants were 40 older people with mental health problems. There was no comparison group.

Primary outcomes

Secondary outcomes

Results: scoping exercise

We visited all five mental health home treatment services for older people identified across England in 2009 (Table 46). We compared and contrasted the three service models that these services followed in practice.

Discussion

This review found insufficient evidence for the efficacy of CRHTTs in supporting older people with mental health problems to remain at home, as a result of a lack of properly designed research studies. Owing to the lack of consistent and good-quality evidence, it is not possible to say whether or not crisis resolution services for older people with mental health problems reduce length of stay in hospital from the three studies which considered this outcome. In particular, we were not able to calculate the average length of stay for the entire cohort in the Doyle and Varian189 study. The level of recommendation for maintenance of community residence was grade D. Only two studies looked at this outcome. 170,189 It is possible that the 24-hours crisis service can sustain people at home for longer, but this apparent finding might also have arisen as a result of changes in service provision, such as the shift towards care/nursing home placements, with long-term hospital beds becoming increasingly rare. The effects of crisis resolution services on mortality rates, service use and service costs were all given an overall grade D of recommendation. This was due to weakness in the study design, contrasting results, inappropriate comparison groups or lack of data.

From our scoping exercise of HTTs we were able to establish some key themes. Although we visited only five teams across the UK, these services were representative of the three models found in the Cooper et al. 191 UK survey. The home treatment services had varying definitions of crisis. Comparison of efficacy between different models is made difficult by variations in aims and referral criteria between different services. In addition, broad outcomes such as reduction in hospital admissions may be of limited utility in this instance, as baseline admission threshold varies widely between geographical areas on the basis of availability of inpatient and community resources and clinician preference, among other considerations.

Aim

The aims of the focus groups were to identify factors precipitating crises and to identify interventions that may help manage crises for people with dementia living at home and their carers.

Method

Analyses

Two researchers conducted the focus groups. One acted as the moderator by facilitating the group, keeping the discussion going and encouraging all group members to participate fully. The second researcher actively listened, sought clarification and ensured the accuracy of content as needed during the interview,73 as well as recording field notes during and immediately after the focus groups. 74 Participants gave permission for the focus groups to be audio tape-recorded and these were transcribed verbatim. The data were qualitatively analysed to identify broad similarities and differences, using a data-driven inductive thematic analysis76,209 and ‘long table’ approach to code and analyse information. Thematic analysis is useful for summarising key features of a large body of data, as well as highlighting similarities and differences across the data set. 209 This can generate unanticipated insights, which is consistent with focus groups inquiry that allows participants the freedom to provide information that does not necessarily fit with any expectation going into the research. Inductive thematic analysis searches for themes that emerge directly from the data as being important to the description of the phenomenon. 210 This approach complemented our research questions, which were designed to be exploratory to extract descriptive data.

Two researchers independently analysed the focus group transcripts as a method of quality control and validation. In addition, both researchers read the focus group transcripts twice to familiarise themselves with the data. Initial codes were generated by coding interesting features within the transcripts in a systematic and rigorous way across the entire data set, collating data relevant to each code. The codes were then collated into potential themes, gathering all data relevant to each theme. The themes were then applied to all of the transcripts again to assess the applicability, and were reviewed and refined as necessary, generating a ‘thematic’ map of analysis. The themes were continually reviewed through ongoing analysis, and clear definitions and names were finally applied to the agreed themes. A codebook was produced, which highlighted the views of the three stakeholder groups, allowing us to assess the similarities and differences between them.

Tables 47 and 48 present the themes derived from this study according to category and stakeholder group for causes of crisis and interventions that could help to prevent or manage a crisis. The numbers of counts illustrate the number of times the theme was discussed by participants in each stakeholder group. There were five categories relating to themes for causes of crisis: (1) carer related; (2) environmental; (3) vulnerability; (4) behavioural/psychological; and (5) physical health (see Table 47). There were four categories relating to themes for interventions: (1) home environment; (2) carer related; (3) professional health-care support; and (4) social and home care support (see Table 48).

Results

Discussion

We explored the views of a diverse range of stakeholders involved in crises with people with dementia. People with dementia worried about the risks and increased vulnerability associated with their declining cognition but wanted support to allow them to remain at home safely. However, rather than expecting a lot of input from professionals, they valued informal support such as local support from family, friends and neighbours, notes and reminders, mobile phones, and aids and adaptations around the home to help manage risks related to gas, electric, cooking and fire. Family carers also had concerns about vulnerability and safety, and valued home adaptations and specialist assistive technology as being especially useful to them and their relative with dementia during times of crisis. Staff also mentioned that specialist assistive technology should be more widely available for families in crisis.

Family carers had a broad understanding of the range of factors precipitating crises but their main concerns related to increased carer burden, poor carer mental health and lack of support from other family members or other services. Carers valued education and training as being useful to help prevent, and to help them cope during, times of crisis.

Clinical and policy implications

Crises faced by people with dementia and their families are complicated and distressing. 192 Interventions need to be flexible and tailored to both the individual person’s needs and their crisis situation. 211 The findings from this study support the involvement of services users and carers in service planning.

Conclusion

People with dementia and family carers have much to offer in their understanding of the most important causes and the most useful interventions in times of crisis. Although health-care professionals often emphasised more costly and intensive interventions (such as extended-hours services and multidisciplinary interventions), people with dementia often preferred support from family and friends, notes and reminders, and home adaptations to reduce risks.

Methods

Procedure