Exercise for depression in care home residents: a randomised controlled trial with cost-effectiveness analysis (OPERA)

Evidence for the use of the structured exercise programme

Physical activity is ‘any bodily movement produced by skeletal muscles that results in caloric expenditure’. 41 Exercise is a subcategory of physical activity; it is planned, structured, repetitive and results in improvement or maintenance of one or more facets of physical fitness. There are several types of exercise, including aerobic exercise (submaximal, rhythmic, repetitive exercise of large muscle groups, during which the needed energy is supplied by inspired oxygen) and resistance exercise (any form of active exercise in which a dynamic or static muscular contraction is resisted by an outside force). 42

The endorphin hypothesis suggests that physical activity and exercise cause an increase in the release of ß-endorphins, believed to be related to a positive mood and an enhanced sense of well-being. 43 Although it was originally thought that these ß-endorphins could not cross the blood–brain barrier,44 the endogenous release of central opioids with strenuous exercise may directly account for the sensations of euphoria associated with this type of exercise. 45 Some of the endorphin-related effects may be elicited during aerobic physical activity via an increased discharge from the mechanosensitive nerve fibres within contracting skeletal muscles. 46,47 The level of these increases in endorphins also appears to be directly related to the use of more intense resistance exercise. 48

The monoamine hypothesis suggests that exercise leads to an increase in the availability of brain neurotransmitters (such as serotonin, dopamine and noradrenaline) that are often decreased in depression. 43,47 Exercise also appears to cause increases in levels of monoamines in the blood and urine, and may do so centrally. This theory is based on animal work, and the connection to central increases in monoamines following exercise in humans is as yet unproven. 43

Neurotrophins, especially brain-derived neurotrophic factor, can counteract the hippocampal atrophy that appears to be associated with the high plasma cortisol levels seen in stress and depression,49 and may make the brain more resistant to stress. Neurotrophins are a family of closely related proteins that control many aspects of survival, development and function of neurons in both the peripheral and central nervous system. 50 Exercise appears to activate cellular cascades, such as increases in brain-derived neurotrophic factor expression, which may be both time and intensity dependent. 51,52 Moderate intensity aerobic exercise for a period of 12 months may have the capacity to significantly increase hippocampal size in healthy older adults53 and there appear to be links between resistance exercise and mood changes via the brain-derived neurotrophic factor system,54 although the evidence for the direct effects of resistance exercise on brain-derived neurotrophic factor levels have so far been inconsistent. 55

Engaging in exercise may have the capacity to counteract common symptoms of depression, such as negative thought patterns, low self-esteem and anhedonia (the inability to gain pleasure from enjoyable experiences)56 through distraction, mastery experience,43,57 improvements in self-evaluation58,59 and a sense of achievement. 60 A further theoretical underpinning comes from the theories of positive psychology, considering the potential of exercise interventions to build on ‘virtues’ and ‘character strengths’, such as love of learning (mastery of new skills), bravery (not shrinking from challenge or difficulty) and hope (expecting the best and working to achieve it). 61

Based on our literature review we developed a theoretical model for how exercise might reduce depressive symptoms (Figure 1).

FIGURE 1.

Theoretical model of effect of exercise on depressive symptoms.

A systematic review of the evidence for the effects of exercise and physical activity on depression in older people concluded that aerobic and resistance exercise are likely to be the most effective forms of exercise for decreasing depressive symptoms in this group. The review also found that interventions using moderate or high intensities of training appeared to be the most effective, signifying that there may be a dose–response effect. 62 A second review examined the evidence for the effect of physical activity and exercise interventions in older people with dementia and depression. It concluded that although there is no clear indication of effect on depressive symptoms or quality of life, there appears to be good potential for people with dementia to participate in physical activity and group exercise programmes, including the use of strength training and higher exercise intensities. 38

Exercise participation for older people with dementia

Kitwood's principles of ‘positive person work’ emphasises the need to understand people with dementia as individuals with very different experiences of life, and different needs, feelings, likes and dislikes63 (Box 1). The features of a person-centred dementia care approach are most likely to support well-being and enable any intervention to create the best possible effects.

Theoretical concepts supporting the use of a whole-home approach

Older adults are more likely than any other age group to lead a sedentary lifestyle despite an awareness of the benefits of being physically active. They cite deterrent factors such as lack of interest, shortness of breath, joint pain, perceived lack of fitness and lack of energy. 84 Frail older adults with multiple comorbidities and high disability levels residing in care homes are even more likely to have very limited physical activity levels. 85,86 To influence the physical activity behaviour of as many residents as possible, and maximise exposure to the exercise sessions, any activity/exercise intervention should offer a variety of physical activity opportunities without placing unreasonable demands on care home staff.

A large number of factors create barriers to physical activity within care homes: resource limitations, perceived role boundaries, limited staff training opportunities and restricted access to appropriate health-care providers. 86–92 Hence, any intervention to increase exercise participation should include approaches to create the culture change needed to effectively address these barriers. 92,93

A whole-home intervention, using an organisational approach to encourage all residents and staff in efforts to increase the residents' level of physical activity, is more likely to achieve the positive effects sought than simply providing group exercise sessions. Such whole-home interventions need to consider the full range of physical, organisational and attitudinal barriers that may be encountered.

Ethical issues

Cluster randomised trials have become increasingly common in health service research for a range of pragmatic and methodological reasons. However, the nature of cluster randomised trials, where randomisation, and often also the intervention, is at the level of the cluster rather than the individual, means that the standard ethical requirement of informed consent from all research participants prior to participation cannot be achieved.

The diverse range of models for cluster randomised trials means that how researchers need to secure appropriate consent, while respecting the rights of research participants, is complex. 96–98 MacRae et al. 99 draw on the moral foundations of informed consent and international regulatory guidelines to offer a framework for consent in cluster randomised trials. 99 This includes justificatory reasons for waiver of consent, when consent to randomisation may not be necessary, and the level of information required to be provided following randomisation. When the unit of randomisation in a cluster randomised trial is a care home, the ‘cluster guardian’, who provides consent for participation in the study, will usually be the care home manager. The manager also acts as a gate keeper, controlling access to individual residents for recruitment. The importance of care home managers and staff in the recruitment of participants to research in care home settings has been noted previously. 100 Reasons for excluding access include the views of care home staff on the resident's capacity to consent, the resident's or their next of kin's (NOK's) physical frailty, and the value they place on their privacy. There is a difficult ethical balance to be struck by care home managers in protecting their residents' interests while not denying them the opportunity to participate in research if they wish.

A further difficulty in conducting research in care homes is the substantial number of residents who have a degree of cognitive impairment and who therefore may lack capacity to consent. Cognitive impairment, per se, is not determinative of whether or not a person has capacity to make a specific decision such as consenting or refusing to participate in a research project. There is no direct correlation between Mini Mental State Examination scores and capacity to consent,101 and there is considerable variability in judgements of the capacity of people with Alzheimer's disease to consent to research, even among experienced psychiatrists. 102 A key message from these studies would appear to be that researchers must take responsibility for assessing capacity of participants for their specific study, and recording the process and reasoning for their assessment. It is not clear from the literature that this is currently regarded as standard research practice. A systematic review of cluster randomised trials in care homes found that of the 46 papers that reported consent processes, three trials relied on the opinion of care home staff to assess capacity and seven assumed lack of capacity in all participants by virtue of being resident in a dementia specialist home. 94

Some potential research participants in care homes will clearly lack capacity to consent to take part in research. International research ethics guidelines permit research involving participants who lack capacity if specific criteria are met, for example that the risk of harm to the participant should be no more than slightly greater than that of ordinary medical care103 or that consent is given by a legal representative. 104

The Mental Capacity Act 2005105 requires researchers to consult with either a personal or nominated consultee to seek advice on whether or not the person who lacks capacity would have objected to taking part in the research if they were able to consent. Confusion among research ethics committees and researchers regarding this consultation process has been identified. 106,107 There are also concerns about the impact of this requirement on recruitment of participants who lack capacity to consent. Two UK studies that specified recruitment rates for care home residents through proxies or consultees reported levels of 61% and 41%, respectively. 100,108

In developing this study we identified several ethical issues:

1. the ethics of cluster randomisation and the role of care home managers as cluster guardians providing consent for participation in the study

2. recruitment and consent of individuals for assessments and data gathering in an institutional setting, including the role of home staff as gatekeepers

3. obtaining valid consent for individual participation in a population that will include a high number of cognitively impaired participants, including assessment of capacity to consent

4. the involvement of personal and nominated consultees in decisions about recruitment of residents who lacked capacity to consent

5. the challenges of fluctuating mental capacity throughout the study.

Developing the process evaluation

The process evaluation pilot had two main aims: to inform (1) the development and implementation of both the whole-home intervention and the control intervention for the main study, and (2) the design of the process evaluation of the main study. We used a formative approach in the pilot phase to ensure that any problems were identified and addressed in the main study protocols and procedures. As part of the process evaluation of the pilot study, one researcher (DE) spent time in the three pilot study homes observing the research staff activity and talking to staff and residents. Although we aimed to conduct non-participant observation, it proved impossible to maintain non-participant observer status in this setting. The researcher was inevitably drawn into the activities in the home; residents became familiar with him and expected him to interact with them socially. We used both non-participant and participant observation in the main study process evaluation as these approaches are complementary. 114

Other challenges were identified in the pilot process evaluation13 and influenced our development of the process evaluation protocol including:

1. We were working with a vulnerable population with varying degrees of cognitive and physical abilities. We had to take care to be inclusive in our approach to ensure that our findings reflected the experiences of the whole population of the care homes.

2. In many homes space to carry out interviews in private was limited. We needed to establish standard operating procedures that as far as possible respected residents' right to privacy.

3. A normal day in a care home is structured with set times for meals, breaks, activities and drug rounds, which can leave little time for a researcher to engage with residents. During the pilot study, we found that there was often a narrow window of opportunity to engage in individual interviews. Residents tended to be more receptive in the mornings (a common finding in populations with a high prevalence of dementia), with many homes running activities at these times for this very reason. This coupled with prearranged interviews being cancelled when the resident was found to be asleep further narrowed the opportunities for carrying out interviews.

4. Observations and feedback from team members revealed that the process of recruiting participants and, in particular, gaining informed consent, took considerably longer than originally envisaged. During the pilot phase, a number of problems with the process of gaining consent were noted, including the content of the consent forms and the difficulties of ensuring that participants had given informed consent.

In a focus group with the recruiting team, to discuss issues in the process of gaining consent, the team reported finding the process challenging, and thought the consent form was too complex for residents. It took up to 20 minutes to complete the process for each individual resident, and some of the residents found the information difficult to understand and to retain. Those study recruitment staff who had carried out the assent/agreement process (with relatives of residents deemed unable to give informed consent) found it easier, with fewer tick boxes, making it simpler to complete. This led to the consent form being shortened to make it easier for residents to complete.

A number of other changes to the conduct of the main study were implemented as a result of the pilot study process evaluation, including changes to data collection forms. Other recommendations arising from the pilot process evaluation are described in Box 2.

Homes

We recruited homes from two geographical locations: Coventry and Warwickshire (C&W) and north-east London (NEL). The practicalities of delivering the intervention meant that all care homes needed to be within reasonable travelling distance for the physiotherapists delivering the intervention: either Warwick Clinical Trials Unit for C&W or Barking & Dagenham Primary Care Trust's physiotherapy department for NEL. Together, these two locations provided a wide range of populations broadly representative of the social mix within the UK, ranging from prosperous rural South Warwickshire through to deprived multicultural urban communities in Coventry and NEL.

We identified all care homes within the relevant primary care trust areas (Barking & Dagenham, Coventry, Havering, Redbridge, Waltham Forest, Warwickshire and West Essex). Following primary care trust approval we approached all care homes with between 16 and 60 beds, but from 11 February 2009 increased the maximum number of beds to 70. Also, from this date we specified that a minimum of six residents needed to have been assessed to be included in the study; our lower limit for home size was based on being able to recruit a minimum of six participants from each home. We excluded smaller homes to ensure that we could collect, from each home, follow-up data from at least one resident with five or more depressive symptoms at baseline. We excluded larger homes because the reduced statistical power from larger clusters meant the additional time taken in recruitment and delivering the intervention was an inefficient use of available resources. With the agreement of the Trial Steering Committee we decided to exclude homes in which more than half of the beds were registered as dementia specialist beds. We took this decision because recruiting in the first dementia specialist home for the main study confirmed our observation from the pilot study, that too few residents in dementia specialist homes were able to provide data on the primary outcome measure. With the agreement of the Trial Steering Committee we changed our protocol to exclude these homes after the first of them had been randomised.

We approached, by post, all apparently eligible homes, apart from three in C&W included in our pilot study. Interested homes were then visited by a member of the recruitment team to assess their suitability. At this time homes were excluded if fewer than six residents were likely to be able to take part in the study; more than half of the residents had severe cognitive impairment; the majority of residents were non-English speaking; or, after discussion with the study team, the home felt they were too busy to participate. Towards the end of the study some interested homes were unable to participate because we had reached our target for home recruitment.

Inclusion and exclusion criteria

Ethical considerations

Obtaining appropriate valid consent from participants is essential to any research study and there are specific ethical concerns regarding consent in both cluster randomised trials and research conducted in care homes (see Chapter 1, Ethical issues). Many residents were likely to have substantial cognitive impairment and therefore may have lacked capacity to consent to taking part in a research study, which involves completing a series of assessments and allowing access to their records.

In the OPERA study the clusters were the care homes where the home managers not only acted as gatekeepers to individual residents, or their NOK, for the purposes of consent to the assessments, but were also the cluster guardians for the home giving consent for the home to participate, and actively participated in the whole-home intervention. 96

To ensure that participants had given valid consent particular attention needed to be paid to assessment of capacity during the information-giving process. We therefore developed a two-stage approach to gaining consent; first assessing capacity and at a second visit obtaining consent (see Obtaining consent, below). All of the research nurses and physiotherapists who were part of the recruitment team received specific training in assessment of capacity and taking consent by the study ethicist.

Enrolling participants who are unable to consent into a study is ethically problematic, as it can be difficult to argue that participation is necessarily in their best interests. For those residents who were unable to consent to take part in the assessment component of the study we followed the requirements of The Mental Capacity Act 2005,105 which provides the legal framework for involving people who lack capacity in medical research. Initially, the NOK was approached as the personal consultee, as defined in the Act, and if no personal consultee could be identified then a nominated consultee, as defined in the Act, was sought.

If the consultee agreed that the resident should take part in the study then the resident was enrolled as a participant. The process of setting up the study and obtaining ethics approval took place very soon after enactment of The Mental Capacity Act 2005, when researchers were still familiarising themselves with the practical implications of its requirements for research. As a result, our documentation referred to assent/agreement of the consultee to the resident taking part in the research. We are aware that there has been discussion in the literature about the initial failure of researchers and research ethics committees to fully understand the precise intention of the Act in this regard. 106

There has been a subsequent movement towards seeking a consultee's views about the likely wishes of the individual (the resident) about participating in the research, rather than their agreement on their behalf. Our own research practice has mirrored this development in thinking. However, as our study documentation, which had been subjected to ethical review, included the terms assent and agreement, we will continue to use these throughout this report to describe this research process.

Residents who lacked capacity to consent to take part in the research may still have been able to give consent, or refuse, to take part in an individual assessment. Therefore, at each assessment their consent or agreement was sought for that particular assessment. If at any time during the assessment a participant indicated that they did not want to take part their wishes were respected, whether or not they had formally consented or agreement had been provided by NOK.

Ethical review for the study was provided by the Joint University College London/University College London Hospital Committees on the Ethics of Human Research (Committee A), now known as Central London REC 4. The REC reference for the study is 07/Q0505/56. The Committee also approved 10 Substantial Amendments for the study.

Obtaining consent

At the first visit to the care home the care home manager, or delegated staff member, was asked to exclude any residents whom they felt it would be inappropriate to approach to take part in the study, for example those with a terminal illness. A specially trained research nurse or physiotherapist approached the remaining residents to tell them about the study and assess their capacity to give consent (stage 1). Those who, in the opinion of the recruitment team, lacked capacity to give consent to participate were not approached further; we did, however, approach their NOK. Residents not interested in taking part in the study were excluded.

Information sheets were given to residents who were considered competent to consent, and left in a prominent place so that they could discuss it with their relatives or friends. Large print or audio versions were available for those with visual impairment. Mutually agreed appointments were made for study assessments within about 1 week of the initial approach, giving residents time to consider taking part in the study.

Residents had the option to consent, or NOK had the option to give assent, either to the assessments, the use of care home data or to both. Residents (or NOK) could also consent (or assent) to the results of assessments being shared with the care home manager and their NOK, and/or their general practitioners being informed of their participation in the study. The research nurse/physiotherapist conducting the recruitment assessment checked that the resident had read, or otherwise availed themselves of, the information sheet and understood the broad nature of the study, and answered any questions that the resident had. In practice, family members were sometimes present for these discussions. Each statement on the consent form was discussed and initialled by the resident; if they were in agreement then the consent form was signed by the resident and witnessed by the research nurse. For those residents with sight impairment a member of the care home staff witnessed the consent process. The process of going through the information sheet and consent form and checking understanding provided a further check on the resident's capacity to consent (stage 2). If at this stage it was clear that the resident lacked capacity then the process was stopped and a personal consultee was approached.

Care homes wrote to the NOK of residents who lacked capacity to consent, enclosing information about the study and an expression-of-interest slip. Residents who were unable to communicate were identified by the care home manager at the initial meeting, and agreement was requested for access to care home data only. NOK who expressed an interest in their relative/friend taking part in the study were contacted by telephone to answer any queries and arrange for an assent form to be sent to them by post or to be completed with the research nurse at the care home.

If after 2 weeks there had been no response to the request from the initial letter the relatives were contacted by telephone. It had been anticipated that if a resident had no known NOK then an appropriate nominated person could be contacted as permitted by The Mental Capacity Act 2005. 105

We consulted with local social services departments and the Independent Mental Capacity Advocacy (IMCA); neither organisation felt able to provide this service. In view of the relatively small numbers of residents involved we decided not to explore other possibilities for a nominated consultee and therefore residents who were unable to give consent and who had no contactable NOK were not recruited.

Residents recruited after randomisation

At the 3- and 6-month visits the care home manager was asked to identify any new permanent residents (see Follow-up, below). We did not do this at 9 months to ensure that all those contributing to the cross-sectional analyses had been exposed to the environment within the care home long enough to be exposed to at least 4 months of the environment of that home. If the resident was eligible and interested in taking part in the study, the research nurse/physiotherapist explained the study and assessed their capacity to give informed consent, allowing the resident at least 24 hours to consider taking part in the study. NOK of residents unable to give informed consent were contacted in the same manner as those approached prior to randomisation. All processes for data collection were the same as for participants recruited prior to randomisation (see Obtaining consent, above).

Care home data

We collected demographic data (age, sex, ethnicity, social class, age at leaving full-time education), comorbidities (cancer, stroke, dementia, depression, anxiety, osteoporosis, chronic lung disease, urinary incontinence) and data on length of residence from the care home records. Data on current medication use were obtained directly from residents' Medication Administration Record Sheets. The resident's key carer, or carer looking after them on the day of data collection, was asked to complete a proxy EQ-5D, Barthel Index and Social Engagement Scale (Table 2). 113,116,119

Follow-up data collection

At 6 and 12 months we sought to complete follow-up assessments on all residents from whom we had consent or agreement to take part in study assessments. The research nurse/physiotherapist confirmed with the care home manager that it was an appropriate time to approach the resident. Residents who agreed to take part in follow-up assessment completed the same questionnaire instruments as baseline apart from the physical assessment Short Physical Performance Battery, which was not administered at 6 months.

Residents who were known to have died or had moved from the home were reported to the study team by the care home manager at each visit. At 6 and 12 months residents who were in hospital, or too unwell at the time of assessment, were contacted within 1 month to see if they were now well enough to be assessed. Participants who had moved to another participating care home were contacted and assessed, at the time that assessments were being done in their original home. We were unable to follow-up participants who had moved to a non-participating home.

Care home data were collected for those for whom we had consent/agreement and who were present in the care home at 3, 6, 9 and 12 months. Medication use, visits by health-care professionals, outpatient appointments, accident and emergency (A&E) attendances, and hospital admissions were recorded. The resident's key carer, or carer looking after them on the day of data collection, was asked to complete a proxy EQ-5D and the Social Engagement Scale at the 3-, 6-, 9- and 12-month follow-up visits.

Residents recruited after randomisation

Follow-up assessments were completed in the same manner as those recruited prior to randomisation with an assessment 6 months after they had joined the study. All participating new residents had an end of study assessment when their home came to the end of its 12-month study period. The majority of new residents were identified and recruited at 3- and 6-month follow-up visits to the care homes; a small number requiring assent from NOK were recruited at between 6 and 9 months post randomisation.

Evaluation of recruitment processes

As part of the process evaluation, recruitment processes, including home recruitment and recruitment of individual participants, were evaluated (see Process evaluation, below).

Independent observations of the process of consenting participants were carried out by the process evaluation research fellow. This involved shadowing the recruiting staff on a number of occasions and observing the consent and assessment process in different homes. Results of these observations are reported in the quality control section and are mentioned in the consent substudy section (see Ethics substudy, below).

Primary outcome measure

Our primary outcome measure was the Geriatric Depression Scale-15. 112 This brief instrument consists of 15 yes/no questions and has been well validated in care home populations. It avoids using potentially somatic features of depression that may be misleading in this age group, focusing more on mood and functional symptoms of depression. 121 It is one of the most widely used measures in this field. It is simple to complete, with 97% of cognitively intact nursing home residents producing analysable data, and showed high internal consistency in a large sample of US nursing home residents. 122 The Geriatric Depression Scale-15 can be interpreted as an indication of the presence/absence of depressive mood. A score of ≥ 5 appears to give the best sensitivity and specificity for presence/absence of depressive mood. 120,123 Nevertheless, the Geriatric Depression Scale-15 is not a substitute for a clinical diagnosis of depression. However, in this report we will refer henceforth to the presence of five or more depressive symptoms, or its equivalent, on the Geriatric Depression Scale-15 as depression (Table ). Some individuals completed fewer than 15 items on the Geriatric Depression Scale. The recommended score indicating depression is ‘5’ when 13 or more items are completed, ‘4’ when 12 or 11 items are completed and ‘3’ when 10 items are completed (see Table 3). When nine items or fewer were completed the Geriatric Depression Scale-15 score was set to missing. Henceforth the presence of the appropriate number of positive items is referred to as depression. The Geriatric Depression Scale-15 has been used as a continuous measure in other randomised controlled trials based in care homes. 95,124 We also use the Geriatric Depression Scale-15 as a continuous measure in some of our analyses.

Other participant reported outcomes

Physical assessment

Proxy measures

We collected proxy data on all of those for whom we had permission to collect care home data. Thus, more proxy data are available on participants than data from the participant-completed measures.

Care home data

Safety data

Health service activity

Control intervention

Our control intervention mirrored as closely as possible a realistic depression intervention which the National Health Service might introduce in care homes that would constitute a best usual care implementable by the National Health Service. The depression awareness training was designed to be short, inexpensive, deliverable by a non-expert clinician, and to address relevant learning objectives for staff in the care home sector. We used a scripted interview that was filmed using a trained simulator (actor) playing a 67-year-old care home resident with clinical depression being visited by a community psychiatric nurse. This video lasted six minutes and was used in the middle part of the depression awareness training session to illustrate key clues to diagnosing depression and possible ways to treat depression.

Each clinician delivering the depression awareness training intervention session in the care homes attended a 120- to 150-minute training session. The session reviewed the definition and epidemiology of depression, the place of depression awareness training in the OPERA study, the care home sector and its staff in the UK, and exercises including role play in delivering the session. Clinicians were able to deliver the depression awareness training only if, after a training session, in the opinion of an experienced training clinician, they had achieved the expertise to deal with this. The implementation of the depression awareness training intervention was studied as part of the process evaluation.

The depression awareness training was delivered to staff in every care home in OPERA. Each session was delivered by a clinically qualified member of the research team (e.g. nurse/physiotherapist/general practitioner) and lasted approximately 30–40 minutes. Staff members at care homes were asked to attend on one occasion. Sessions were designed for 3 to 10 staff. Requirements for the session were a room which could hold the staff group and was quiet, and a session time suitable for the care home staff; this depended on care home arrangements but was typically during the middle part of the day, around staff handovers. A method of delivering a video to staff was needed; either a DVD player and television, or a laptop computer with loudspeakers. In many care homes, more than one session was required to allow as many staff as possible to attend.

The stated aims of the session for staff attending were to be able to:

• say what depression is

• distinguish depression from normal sadness

• name two things that can be done to help people with depression in a care home.

The session was structured as shown below:

• introduction (3–4 minutes)

• talk/discussion on what is depression, supported by written materials (6–8 minutes)

• video of simulated resident/nurse encounter (6–7 minutes)

• case discussion (15–18 minutes).

At the end of each session care home staff completed feedback forms that recorded attendance, satisfaction with the sessions, and suggestions for development.

Active intervention

The OPERA intervention was designed to test the effects of exercise and increased physical activity on depression and other important outcomes (Figure 3). Our plan was to deliver a whole-home exercise intervention, consisting of training for care home staff backed up with a twice-weekly physiotherapist-led exercise groups.

FIGURE 3.

Design considerations for OPERA intervention.

Whole-home intervention

The aims of the whole-home intervention were to:

• increase the physical activity levels of care home residents over the length of the study in each home (12 months) using simple approaches

• engage residents' participation in twice-weekly exercise groups delivered by the OPERA physiotherapists.

To promote the effectiveness and sustainability of the OPERA intervention it was envisaged as a whole-home, ‘ecological’ intervention153 directed at changing the whole-home environment and the culture within care homes. To this end the intervention adopted the WHO ‘settings approach’,154 aiming to strengthen the resources for health in a ‘setting of everyday life’. Such systemic interventions have been shown to be effective in significantly changing behaviour within cluster randomised trials in other settings. 155

The intervention was limited to specific pathways and treatments thought to affect depression. It was not intended to provide physiotherapy more broadly. The approach to safety and risk management was grounded in available evidence, supplemented by pragmatic opinion of professional bodies and internationally respected opinion leaders. The aim was to include and address the problems of as many residents of care homes as possible, including those with dementia.

The whole-home intervention consisted of training for residential and nursing home staff backed up with a twice-weekly physiotherapist-led exercise class. This approach, with the regular exercise groups at its core, was used to ensure that residents had maximum opportunity to engage in the exercise groups and to maintain physical activity outside the formal exercise session.

The development of the whole-home intervention drew on components of both Greenhalgh's ‘diffusion of innovations’ model156 and Cialdini's ‘model of persuasion and influence’. 157 A dementia/person-centred care approach was robustly promoted within the OPERA intervention along with the use of an ‘unconditional positive regard’. 63

The OPERA intervention

The OPERA study was located in UK care homes in which up to 70% of residents are likely to be living with dementia. 167 A dementia/person-centred care approach was robustly promoted within the OPERA intervention and team members, along with the use of an ‘unconditional positive regard’. The intervention was delivered by a physiotherapist who was assigned to work with an individual home for 12 months.

The physiotherapists were allocated two half-days per week to each care home – more in some larger homes in which we ran more exercise groups. There were seven components to the intervention (Figure 4).

FIGURE 4.

OPERA intervention flow chart.

Initiation and assessment period:

1. Initial meetings with the care home manager and senior care home staff.

2. The identification of a physical activity champion from within the care home staff.

3. Carer training in depression and activity awareness.

4. Physical activity assessments.

Ongoing physical activity and exercise programme:

1. 5. Physical activity (and mobility safety) recommendations for care home staff.

2. 6. A twice-weekly group exercise programme. The exercise groups were the most resource intensive and most tangible component of the intervention.

3. 7. Ongoing informal discussion and co-operative problem-solving between physiotherapists and care home staff.

Initial meetings with care home manager

The physiotherapist set up an initiation meeting with the care home manager and senior carers aimed at establishing rapport with the care home staff, understanding and agreeing roles within the intervention, and to sort out logistical details such as locations for the exercise groups.

Physical activity champion

A ‘champion’ (a key individual in an organisation) may assist in the assimilation of an innovation, so we sought to identify a physical activity champion in collaboration with the care home manager and senior care staff to act as a key link within the home. Their role was to assist with planning the exercise groups, and to attend the groups whenever possible, to act as a link between the physiotherapist and home, and to facilitate the implementation of the physical activity recommendations for each resident.

Care home staff depression awareness and activity training

The aims of the care home staff depression awareness and activity training, in addition to the aims of the depression awareness training programme described above, were to:

• highlight the importance of promoting and increasing physical activity to aid the physical and mental health of residents

• establish a collaborative approach towards increasing physical activity within the home

• describe the specific elements of the OPERA intervention and how the staff would be involved.

The session was based on the best current adult pedagogical practice and previous experience of what works in training in care homes. To this end it involved short sessions, used a variety of teaching methods including the depression awareness DVD, and included discussion of cases.

Staff training sessions took about 60 minutes to complete and were repeated when new staff needed to be trained. The training materials used were the same as in the control arm depression awareness training sessions, with the addition of posters illustrating the physical and mental health benefits of physical activity and cycles of activity avoidance related to falls and pain. The training booklets for care home staff had additional pages for copies of these posters plus a mobility safety advice section outlining important enabling techniques and considerations for the care home staff.

Physical activity assessment and recommendations

In consultation with the care home manager all eligible residents were invited to undergo a mobility and physical activity assessment, and received a tailored physical activity recommendation to promote day-to-day activity alongside an invitation to attend the exercise classes. The assessment took place in the resident's room or other locations according to the resident's preference or abilities. The physiotherapist assessed functional capacity, medical history, pain, fear, anxiety, sensory and communication abilities.

On the basis of the assessment the physiotherapist determined a plan of action for the intervention programme elements. The first was a bespoke physical activity programme tailored to each resident and aimed at increasing the level of habitual physical activity, developed in co-operation with the physical activity champion/senior carers. This included the provision of mobility aids, advice on footwear, and manual handling tips to enable mobility. The second was to determine the appropriate level of exercise activities for the group exercise programme.

Progressive resistance exercise

In order to achieve an increase in muscle strength or aerobic conditioning, the demands placed on the musculoskeletal and cardiopulmonary systems must progressively increase over time and stress them beyond their accustomed loads. 169,170 Informed by American College of Sports Medicine161 and by several existing evidence-based exercise programmes for older people, we aimed to increase general muscle work and strength.

We decided not to use the one repetition maximum to calculate starting weights or progressions, because many of the residents in the pilot study were not able to be accurately tested with free weights or a dynamometer. Instead, our starting weights (Table 5) were set according to functional ability level, which could be adapted according to individual preference for acceptability and promotion of adherence.

TABLE 5 - Starting weights for the progressive resistance exercise section

Level	Soft hand weights	Ankle weights
1	Juggling ball (∼200 g)	0.5 kg
2	0.5 kg	1.0 kg
3	1.0 kg	1.5 kg

For the OPERA study we chose to use an observed form of the Borg Scale of Perceived Exertion for resistance exercise. 171–173 It is unlikely that many residents would be able to subjectively report their rate of perceived exertion owing to the high levels of observed cognitive decline observed in residents in the pilot study, and the high prevalence of dementia in the study population as mentioned earlier. We are not aware of any studies that have examined the use of such observed measures in older adults.

Owing to the high levels of frailty common in UK care home residents, it would be likely that few residents could attain or tolerate the higher level of Progressive Resistance Exercise intensity used in previous studies, which demonstrated positive effects on the symptoms of depression. 174–176

Aerobic exercises

Aerobic exercises are activities that are most likely to increase cardiorespiratory demand, and need to be sustained for 15–20 minutes. Aerobic intensity was monitored by rating perceived exertion using the adapted 10-point Borg rate of perceived exertion scale. 173,177

Given the frailty of this population, and the setting, the aim was to achieve (the higher end of) a moderate level of intensity, built up over time, using the ‘start low, go slow’ maxim. 178,179

Warm-up and cool-down sections

These sections included activities designed to gradually increase or decrease heart/respiratory/circulatory rates as recommended by American College of Sports Medicine. 161

Progression of exercises

In order to ensure and maintain a training effect over time, each resident was progressed within the exercise activities at their starting level or up to next level (Figures 5 and 6). Residents could progress between levels on an individual basis by changing their position, for example by standing up in some parts of the aerobic section (level 2). The group could also progress as a whole through the use of changes in music tempo and routines. The exercise progression models were, again, set according to the American College of Sports Medicine Guidelines,161,162 and informed by several existing evidence-based exercise programmes for older people (see Figures 5 and 6). 174,180–182

FIGURE 5.

Flow chart for the progressive resistance exercise progressions. RPE, rate of perceived exertion.

FIGURE 6.

Flow chart for progression of aerobic exercise. RPE, rate of perceived exertion.

Directly attributable adverse events

We defined a directly attributable adverse event as an event needing external medical attention as a consequence of participation in OPERA. This included events occurring during an intervention exercise group, or when participants in the groups were getting to and from the session. Events to be included (but not restricted to) were injurious falls, myocardial infarction, angina, stroke, respiratory distress, musculoskeletal injuries, severe psychological distress or death. For any such events, which were suspected to be a directly attributable adverse event, the treating physiotherapist completed a notification form and notified the study manager. Clinical members of the team then assessed if this event was a directly attributable adverse event. In the event that the home manager became aware of such an event after the physiotherapist had left the home they completed a suspected serious adverse event form and notified the study manager. We used the same system for assessing any suspected adverse events that occurred during study assessments.

Summary

We designed a structured exercise and physical activation programme, within a whole-home approach, to affect the physiological and psychological processes most likely to be able to decrease or prevent depressive symptoms in older people residing in care homes in the UK. The intervention adhered to the principles of exercise prescription, person-centred care and innovation adoption, and was structured with sufficient flexibility to be delivered within a range of different care home settings, to a variety of resident profiles. The training and support package was practical to allow for roll out within the National Health Service if the intervention proved effective.

Objectives

Our primary objective was to compare depression levels between intervention and control care homes, addressed via three primary analyses that represent different ways of expressing depression levels:

1. the prevalence of depression (defined as proportion of participants with five or more positive responses to the Geriatric Depression Scale-15 or equivalent) in intervention homes with that in control homes in all residents contributing data 12 months after homes were randomised (cross-sectional analyses)

2. the number of depressive symptoms at 6 months between intervention and control homes in residents who were depressed at pre-randomisation baseline assessment (depressed cohort analyses)

3. the number of depressive symptoms at 12 months between intervention and control homes in all residents who were assessed prior to randomisation (cohort analyses).

In our original proposal we planned to include, as part of our primary analyses, comparisons of the rates of remission from depression at 6 months and the number of depressive symptoms at 12 months in those who were depressed at pre-randomisation baseline. Early in the study it became clear that the average cluster size and follow-up rates would be smaller than originally anticipated and that we would be unlikely to obtain sufficient data to produce robust comparisons. We therefore changed our primary analyses with the agreement of the independent Trial Steering Committee, Data Monitoring and Ethics Committee (DMEC) and the funder (see below). Remission of depression became a secondary outcome.

Our secondary objectives were:

1. to compare, between intervention and control arms:

   1. remission of depression

   2. cognitive function

   3. health-related quality of life

   4. mobility and exercise tolerance

   5. pain

   6. fear of falling

   7. social engagement

2. to report descriptive statistics on:

   1. the prescribing of antidepressants, hospital admissions

3. to report the following safety outcomes:

   1. injurious falls as indicated by peripheral fractures, mortality rates.

For the depressed cohort we just report a 6-month analysis. For remaining analyses we report on each time point at which we collected data.

Sample size

Unusually, three linked primary analyses were specified for this study. The funder's brief specified a cluster randomised trial with the outcome of interest to be remission of depression. In designing the study, the research team developed a whole-home intervention that would have a long-term impact on all residents. Thus, the outcome of greatest interest to the research team was the difference in the prevalence of depression (as indicated by five or more depressive symptoms or equivalent) at 12 months. (If 15–13 items on the Geriatric Depression Scale-15 have been completed a score of ≥ 5 is indicative of depression; if 12–11 items on the Geriatric Depression Scale-15 have been completed a score of ≥ 4 is indicative of depression; if 10 items on the Geriatric Depression Scale-15 have been completed a score of ≥ 3 is indicative of depression.) For completeness we also considered the long-term effect the intervention may have on the mean level of depressive symptoms at 12 months for all study participants with baseline assessments prior to randomisation.

The remission of depression question originally specified by the funders required the largest number of participating homes; it was therefore this comparison we used to set our original sample size of 77 homes. To reduce treatment costs we planned an unbalanced randomisation of 1 : 1.5 in favour of control.

Randomisation

This was a cluster randomised trial with care homes as the unit of randomisation. Care homes were first stratified by location (NEL or C&W) and then minimised into intervention and control arms using the MINIM programme (www-users.york.ac.uk/∼mb55/guide/minim.htm). This programme uses minimisation with a random element, allocating care homes to the arm of the study that gives the better balance with a probability of 70%. The minimisation factors were type of home (local authority, voluntary, private and care home, private and nursing home), and size of home (32 beds or fewer, > 32 beds). Initially, we had intended to use a third randomisation factor, dementia home or not, but only one dementia home was recruited (see Homes, above).

Baseline comparisons

We compared baseline characteristics of homes and characteristics of participants in the intervention and control arms. We compared the baseline characteristics between intervention and control groups for the cohort and the depressed cohort (these individuals are included in our cohort comparisons). We also compared these characteristics for those participants who joined the study after randomisation.

Analysis

Our full analysis plan is included in Appendix 3. In this overview we describe the general principles for our primary and secondary analyses. Safety analyses are described later in this chapter (see Safety analyses, below).

We conducted available case analyses using intention-to-treat principles. This meant that, for the cohort analyses, when we were able to obtain data for an individual at a specified outcome point (6 or 12 months), we analysed them in the care home they were recruited to regardless of whether they were still in this care home, had moved to another OPERA care home, or had moved to a home outside OPERA or elsewhere. For the cross-sectional comparisons we analysed data according to which care home they were in at the time data were collected. We did not impute missing values in our clinical analyses.

To account for cluster randomisation, we used mixed-effect models with care home as a random effect, and intervention and minimisation factors as fixed effects. Baseline values of outcome and other covariates were also used as summarised in Table 8. All covariates were selected prior to any analyses, as specified in the analysis plan, and all were retained in the analysis, unless there was evidence of multicollinearity. To assess any potential multicollinearity we examined correlations between covariates. If severe multicollinearity was present we dropped one of the collinear covariates, retaining the covariate providing the best fit.

An appropriate link function was used in the analyses depending on the type of outcome:

• continuous outcomes an identity link function

• binary outcomes a logit link function

• count outcomes a Poisson link function

• ordinal outcomes cumulative logit link function.

We used model diagnostics to check distributions of residuals and homoscedasticity as appropriate.

Data collection

Data for the economic evaluation covered a 12-month time period from when the home was randomised. We collected data needed to calculate costs at the individual-level through the care home records at 3, 6, 9 and 12 months and from routine primary care trust data provided after the study was finished (Table 12).

We used the EQ-5D to measure gains in quality-adjusted life-years. The research team collected proxy EQ-5D data on all participants at baseline, and 3, 6, 9 and 12 months after randomisation. Additionally, we collected self-completed EQ-5D scores at baseline and 6- and 12-month assessments on participants able to complete these. They are not reported here but will be used to look at a secondary analysis of the relationship between EQ-5D and proxy EQ-5D.

For residents joining the study after their care home was randomised, we collected data in the same way but participation in the study is deemed to start from the time they were first assessed, i.e. 3, 6 or 9 months after randomisation (see Table 12). These participants were included only in our secondary analysis.

Costs

Individual-level resource use was combined with unit costs to calculate the total health-care cost for each resident. The data received from the primary care trusts on inpatient services, outpatient services and A&E visits included diagnostic related groups and associated tariffs. These costs included staff costs, consumables and procedure costs. For those A&E attendances that recorded arrival ‘via ambulance or air ambulance’, the primary care trusts confirmed that most, or indeed all, arrivals were via road and the cost for a road ambulance journey188 was added to the tariff for the visit. For resource use coming from care home records and community data provided by primary care trusts, we obtained unit costs for the UK from published sources. 189,190 Unit costs are inclusive of ancillary staff costs, overheads and training costs. We used unit costs for the year 2010, actualising unit costs for inflation using the Healthcare Price Index when necessary. 189 Table 12 shows all unit costs derived from published sources. We were unable to procure data on diagnostic tests.

Unit costs for medications were obtained from the Prescription Cost Analysis database for 2010. 190 We included all medication costs where reliable data could be collected, as there were unlikely to be many health costs that could not plausibly be related to the intervention. We obtained details for medications that were being taken by study participants over the 1-week time period that preceded the data collection visit (see Health service activity, above). We calculated the total medication costs using the average cost per dose for each product obtained from the database, and the mean quantity taken per day of each product during the 7 days prior to data collection. These costs were extrapolated for the 91 days prior to the follow-up visits at 3, 6, 9 and 12 months. If the daily drug cost was missing for a time point then we used the midpoint of the adjacent time points.

Where a dose range is cited as ‘as required’, we coded it as the midpoint of that range. For common analgesics such as paracetamol and co-codamol, the dose was considered to be two tablets unless another dose was clearly specified. If the dose of the drug had not been recorded, the standard dose for that drug was assumed. If the quantity was not recorded, the standard quantity for that drug was assumed. Good data were available for tablets, liquid medicines, patches, suppositories and metered-dose inhalers. Owing to inconsistencies in the way dosing was recorded within care homes, the following items were excluded from our analyses: all wound dressings (including bandages), creams and lotions, sprays, eye drops, ear drops, injections (including insulin), glucose strips and lancets, catheters and catheter solutions and appliances.

Quality-adjusted life years

The proxy EQ-5D is a well-accepted measure that has been used for stroke patients. 138,191 Responses from proxy EQ-5D questionnaires were transformed into quality of life weights (utility) derived from at UK general population sample using an algorithm developed by Dolan et al. 192 We specified a priori that we would use proxy EQ-5D data rather than self-completed EQ-5D. This was because of an anticipated poor completion rate and to avoid problems with participants whose health deteriorated making them unable to complete the EQ-5D at follow-up time points. In practice, the proxy EQ-5D was completed for a larger proportion of participants than the self-reported EQ-5D.

In order to ensure that quality-adjusted life-year calculations were as accurate as possible, we chose to use multiple imputation to guard against any bias that may result from missing proxy EQ-5D scores. As the study was cluster randomised rather than individual randomised, the data were multilevel. To adjust for this characteristic of the data, a multiple imputation model that accounted for clustering (by means of a random cluster effect) by care home was used to generate the missing proxy EQ-5D scores at each of the five time points (baseline and 3, 6, 9 and 12 months). The imputation model was multivariate for the five EQ-5D scores at each time point, and used the auxiliary variables home size, baseline age of resident and proxy EQ-5D scores at all five time points. Five imputed data sets were created in this way.

The EQ-5D scores, including those scores that were generated by the multiple imputation, were then used to calculate five quality-adjusted life-years, one for each imputed data set. We calculated the total utility for each participant from point estimates at baseline, 3, 6, 9 and 12 months, using the ‘area under the curve’. To estimate the total utility, we assumed that utility for each person followed a linear trend line between these point estimates. 193 We then added up these utility estimates. As many study participants died over the course of the study, we assumed a linear relationship from last proxy EQ-5D measurement to death.

Analyses

Our primary analysis was a cost–utility analysis over 12 months examining the cost per quality-adjusted life-year gained for all participants, who were assessed for proxy EQ-5D prior to randomisation and had primary care trust data extracts. Importantly, the health economics criteria for inclusion were slightly more restrictive than those for the statistical analysis. The clinical effectiveness analysis included residents with and without primary care trust data extracts, whereas the health economic analysis included only those residents with primary care trust data extracts. This should not be confused with the notion of complete or missing data, as all participants with a primary care trust data extract were included irrespective of the completeness of their records. This analysis included those who were not depressed as well as those who were depressed. This was a more meaningful analysis than the alternative of only examining those who were depressed at baseline, as the costs and any potential benefits were related to all of those exposed to the intervention. In order to be aligned with the clinical effectiveness analysis, we included only those who provided the primary outcome, in this case proxy EQ-5D, prior to the home being randomised. Unlike the clinical effectiveness analysis, we included those residents with partially incomplete proxy EQ-5D scores, as these scores could be imputed for our analyses. In line with an intention-to-treat principle, any participants who moved between care homes during the study were analysed as if they had remained in their original care home.

For each cost category, we multiplied the number of items, or contacts with each type of health-care service, by their unit cost. We calculated the mean total costs per participant from an National Health Service provider perspective adding the cost of consultations, admissions, equipment, rehabilitation and physiotherapy services, prescriptions and applicable intervention costs.

As the outpatient data had a large proportion of missing costs, and these missing data were more frequent with some primary care trusts than with others, we imputed these data using the same methodology as we did for the missing EQ-5D scores before calculating the total outpatient cost. We log-transformed costs before imputing, to make the assumption of normality more plausible. The imputation model also accounted for clustering using random home effects and used the auxiliary variables home size, baseline age of resident and the specialty code provided by the primary care trust. Five data sets with imputed data were created. From these sets, the total outpatient cost per resident was calculated and added to the total cost from the other cost categories to produce five total cost data sets, which were used in the baseline analysis.

To compare differences in mean resource use and mean costs between participants in the intervention and control arms of the study, a univariate linear-mixed model was used. To aid in the comparison of costs between the two groups for the outpatient data and total costs, the univariate linear-mixed model was applied to one of the data sets generated by multiple imputation.

Aims and objectives

The aims of the process evaluation were to

• assist in the interpretation of the results of the main effectiveness study

• develop a set of transferable principles regarding the whole-home intervention to inform its implementation on a wider scale.

The objectives were to

• collect quantitative data on the delivery of the intervention in both intervention and control care homes

• observe the exercise programme delivery and the depression awareness training (see Interventions, above)

• document the care home environment in case study care homes and document any changes observed

• carry out in-depth interviews with care home managers, care home staff, patients and carers in a purposive sample of the care homes

• collate and interpret the resulting data.

Methods

All of the OPERA care homes contributed quantitative data to the process evaluation; eight purposively sampled care homes (six intervention and two control) were selected as case study care homes where additional quantitative and qualitative data collection took place.

Data were collected from all care homes using specially designed case report forms and from other sources, including the Commission for Social Care Inspection reports for the care homes (i.e. data from the last published report prior to randomisation), and the records of exercise classes kept by the physiotherapists. Box 4 shows the data that were collected from each participating care home and from the residents and staff within each care home.

All staff in the care homes (control and intervention) were invited to participate in a brief training session on depression awareness. Staff in the intervention homes received additional information about promoting and increasing physical activity amongst residents. Key elements included staff response to these sessions, how confident the presenter was in engaging/adapting to residents' needs, the environment used for the training, timing, and the level of interest of participants.

In addition to observing some sessions, feedback on the value and quality of the training package was sought from attendees using an evaluation form distributed immediately after the session. A follow-up questionnaire with a return envelope was mailed to care home staff 4 weeks later exploring how useful the training had been and if it had been put into practice.

Data collection

Data collection methods for items 1–3 are described in the above sections Case study homes and Data collected from OPERA staff, respectively.

We conducted three focus groups with elderly people who were not currently living in a care home. Initial attempts at recruitment using advertising through local contact with two national, elderly care-related voluntary organisations were unsuccessful. Final recruitment was through local support groups providing day care or activities for the elderly. The managers of centres within C&W were contacted by the research team via telephone and asked if they would be interested in participating. The first two centres contacted expressed an interest and were sent information and posters about the study. Centre managers spoke with different groups within their centres, using posters and participant information sheets provided by the study team. Three groups from two centres agreed to participate and members of these groups were given copies of the participant information sheet and a consent form. Focus groups were organised on days and times specified by the centre managers to coincide with regular timetabled group meetings and were carried out by two researchers: a lead and a scribe (note taker). All participants provided written informed consent on the day of the meeting. As an introduction to the focus group topic the researcher opened the group discussion with a question about the process of consent that the participants had just experienced. A schedule of open questions followed by prompts where necessary guided the main discussion. Questions focused on experiences, perceptions or feelings about the involvement of elderly people in research, particularly about how decisions to take part in research are made (a copy of the schedule can be found in Appendix 6). All focus groups were audio-recorded and field notes were taken. Key points were identified on a flip chart and fed back to participants to check that we had accurately captured their views.

We conducted interviews with key informants, who were purposively selected to provide a range of points of view from different stakeholder and professional groups. We initially contacted 40 potential participants by e-mail with an invitation to participate in the study. Following an expression of interest we sent an information sheet and consent form to complete and return to us. Interviews took place either face to face or by telephone, according to the participant's preference. All interviews were audio-recorded and transcribed.

Interviews were structured around a topic guide that focused on their views about health research in elderly or care home populations. Specific prompts were given, where necessary, regarding potential difficulties in conducting research in these populations and on the question of consent. Finally, participants were asked whether or not they would be prepared to be involved in a consent process for research involving a relative who lacked capacity to consent, and who if anyone would they wish to be involved in such a process for them. Personal involvement in this kind of research, for example as a researcher, was explored. (A copy of the interview schedule can be found in Appendix 7.)

Analysis

All interviews were recorded, transcribed and analysed using NVivo 7 software (QSR International, Southport, UK). Focus group field notes and flip chart records were used as the basis of analysis with reference to the recordings for clarification of points and illustrative quotes. We used a thematic content analysis based on the framework approach and matching that we used in the process evaluation (see Process and evaluation, above). We analysed the process evaluation interview data separately looking for any factors or themes relating to the process of recruitment and consent. We then compared themes emerging from all sources of data (key informant interviews, focus groups, process evaluation interviews and observations) to look for congruence and areas of divergence. In identifying and mapping themes across the data we framed our analysis within the theoretical foundation of informed consent to research and current policy and regulatory frameworks on research ethics. Using a process of reflective equilibrium between empirical data and ethical theory as described by Ives and Draper,208 we were able to achieve a contextual understanding of the ethical requirement of informed consent to research in older people in care homes.

Recruitment (consent and assessments) and data security (in homes)

As part of the process evaluation a researcher (DE) included quality control assessments as part of the work within the eight case study homes (five in NEL and three in C&W). The researcher observed the consenting of residents; the initial Short Physical Performance Battery; and the completion of baseline outcome measures (e.g. Mini Mental State Examination and Geriatric Depression Scale-15). A formal check was also made that OPERA data/files were being maintained and stored correctly within the care home.

Depression awareness training (control intervention)

Observations were carried out during training sessions. Checklists were completed to ensure training was delivered as per protocol. Interactions with the care home staff during the training were also noted.

Intervention delivery

Although the lead physiotherapist assessed the fidelity of the delivery of the exercise intervention in terms of the individuals (physiotherapists) delivering it [see Intervention fidelity (physiotherapist), below], independent observations were also carried out by the process evaluation research fellow. Observation at these times included how the session fitted into the day, how the residents were reacting, how the physiotherapist was interacting with the residents, staff/home involvement and what happened when it ended.

Data security and data collection accuracy at base

All of the electronic OPERA data are stored on the OPERA web application. The OPERA web application is a secure, password-protected, ASP.net web application developed by the Warwick Clinical Trials Unit programming team. The OPERA web application is hosted on the Warwick University Clinical Trials Unit SQL Server 2005 Enterprise edition database. Full back-ups of the OPERA database are completed once every 24 hours.

Baseline and follow-up data were collected on laptops or on paper case report forms. Data entered on to laptops were directly uploaded on to the OPERA web application. Paper case report forms were returned to the study office. Case report forms received, which had not been inputted on to the online database by the recruitment team, were then inputted directly by a data entry clerk, after having been checked by the study manager. The OPERA web application was equipped with computerised validation criteria to minimise data entry errors. These validations placed limitations and checks on data fields to ensure that only the expected responses could be inputted. The database would flag up unexpected responses, prompting the data entry clerk to investigate the query and to correct it. The programmer produced a regular report to identify any possible anomalies on the database in order that they may be investigated in a timely fashion. After data entry had been completed, a 100% data check was conducted on the primary outcome measure (Geriatric Depression Scale-15). This involved checking all of the data points on all of the Geriatric Depression Scale-15 case report forms. For all other outcomes a random 10% check was carried out. This involved checking all of the data points on a randomly selected 10% of the case report forms that contained secondary outcome data. The data checks revealed an error rate lower than the pre-set maximum error rate of one error in 1000 data points. All data management and quality control checks were in line with the standard operating procedures of the Warwick Clinical Trials Unit.

Intervention fidelity (physiotherapist)

The quality and fidelity of the intervention delivery was checked via a site visit to each physiotherapist at one of their assigned care homes on at least two occasions – at least at 6 weeks and at 6 months after the intervention started in that home. These visits were made by the lead (C&W) physiotherapist, or during annual leave an assigned deputy. The delivery of the exercise group and use of the whole-home approach was assessed by observation, and sample of the intervention case report forms and registers were checked. If any major issues were identified, in which case all were checked, that physiotherapist was then visited again within the next few weeks.

In-depth discussions with individual physiotherapists around strategies to maintain and improve treatment fidelity were held on each visit whenever possible, but particularly whenever an area of concern was observed, especially if that difficulty persisted on later visits. The intervention session content and skills demonstrated by their physiotherapist were assessed using a checklist that included items on administration, group locations, resources, self-assessment, progress of participants, working with the care home and mobility safety.

Research questions

Methods

We planned to recruit all 35 of the OPERA intervention homes and a sample of 18 of the control homes. Within this sample we planned to invite all eight case study homes from the process evaluation (six intervention and two control) to take part in both the interview and the observational parts of this follow-up.

This follow-up study followed homes as they reached approximately 6 months after the end of their participation in OPERA. However, as a few of the homes would not reach this 6-month milestone until October 2011, some were approached at about 4 months to ensure their inclusion.

Interviews with care home managers or their delegates were carried out either face to face (in case study homes or some local homes in C&W) or via telephone. Questions explored reflections on the care homes' experiences of OPERA (e.g. interaction with OPERA team, depression awareness training, exercise groups, what went well and what we could have done differently/better), the impact of OPERA interventions on the home during the period of the study and since its end (e.g. awareness of depression, continued use of information, continuation of OPERA-type exercise interventions or other activities), and the impact of the withdrawal of the physiotherapist and the exercise intervention (intervention homes only) (see Appendix 10). Analysis was grounded on contemporaneous notes taken during interviews, with transcripts and original recordings available, when needed, for clarification.

We repeated the ethnographic observations that we carried out in process evaluation case study homes during the OPERA intervention period. The method for the activity sweeps carried out during these observations is explained fully in the process evaluation section above (see Methods).

Recruitment of homes

There were 323 care homes across the two localities. Of these, 180 met our inclusion criteria and were sent letters inviting them to take part in the study (Figure 8). Around half of these (98, 54%) expressed an interest in participating; 17 of these were excluded after a visit by the study team because they had too few residents at the time of recruitment, were too busy, had too many residents with severe cognitive impairment, had too few residents aged ≥ 65 years or the residents were too frail/disabled. After having completed service-level agreements and consent processes, we recruited 78 of the 81 eligible homes to the study between January 2009 and March 2010; follow-up was completed 1 year later (see Figure 8).

FIGURE 8.

Recruitment of homes. a, Did not want to take part.

Home characteristics

We recruited equal numbers of care homes from C&W and NEL, and 18 (23%) were care homes with nursing. One residential home was a dementia specialist home, 61 (78%) were privately owned (Table 13), 16 (21%) were voluntary or charity owned and one (1%) was owned by the local authority. Fifty-one of the privately run homes were in independent (small company) ownership. Most homes had a large number of residents with cognitive impairment; 22 were registered for dementia care. 209 Most of the participating care homes were rated as good (54/78, 69%) or excellent (11/78, 14%) in the last reported Commission for Social Care Inspection data (Table 14). 209

Care homes ranged in size from 17 to 65 beds across the study; 2140 of the 2450 (87%) of beds were occupied when recruitment started (Table 15). Three-quarters of participating homes (60/78) were classed as residential, 15 were classified as nursing and a further three homes were mixed nursing/residential homes (see Table 13). Most of the care homes (67/78, 86%) also offered (temporary) respite care.

These results suggest that the care homes in OPERA were broadly representative of care homes in England, with respect to home ownership, and Commission for Social Care Inspection rating. Local authority-run homes were, however, under-represented in our sample, with only one home included (local authority homes represented 10% of care homes in England in 2010),209 and no very large homes were included. The allocation to intervention and control homes was appropriately balanced in terms of size and location (see Tables 13–15).

On average there was one member of staff for every four home residents during the morning and one for every five in the afternoon. At night there was on average of one member of staff for every 10 residents. Few nurses were employed by homes. We are unable to report the different staff types within the homes, as the self-reported data from the homes did not differentiate type of staff and reported numbers may include ancillary staff, thus staff–resident ratios are likely to be overestimates.

Exercise groups at baseline

Nearly all of the care homes reported that they offered established exercise classes, with two-thirds reporting running a class at least once per week (Table 16). About half of the classes were reported as being run by members of the care home staff and around half by external fitness instructors but only two care homes reported external physiotherapist-led exercise classes. Reported attendance at these groups ranged from 3 to 25, with a median of 10, but many care home managers (22) failed to give an estimated attendance. However, observations carried out in the case study care homes (see below) and anecdotal reports by the field staff suggest that exercise groups outside the OPERA intervention were, in fact, uncommon. Observations showed that care homes commonly advertised activities on weekly calendars but that these did not always take place.

Individual recruitment

When we started individual recruitment in the care homes we identified 2133 residents. This is slightly different to the number of residents who were present at the time homes were recruited (2140) because of resident turnover. Fifty-five residents (3%) were not eligible as there were aged < 65 years or were not permanent residents. We screened the remaining 2078 for study entry. Some residents were excluded because the care home manager felt it was inappropriate to approach them or they were too ill (245 and 279, respectively) (Figure 9). We obtained consent from 607 residents and assent from the NOK of 300 (Table 17). Ninety-four per cent (853) agreed to both assessment and use of care home data, 50 to use of care home data only and four to assessments only. Of the 907 residents screened and providing consent or assent before randomisation, six died prior to randomisation (five of whom had provided some assessment or proxy/care home data), baseline data collection was completed on a further six within 6 weeks of randomisation, one did not complete assessments until 3 months and one completed assessments around 6 months. A further two residents did not provide outcome assessment or proxy/care home data at any point. In total, 897/2078 (43%) of residents screened provided some baseline outcome data prior to randomisation. After excluding the six who died before randomisation, it is the remaining 891 who form our population of interest for the cohort analyses. For a few care homes the number of residents providing a baseline Geriatric Depression Scale-15 prior to randomisation was five rather than six. This was because the lower cluster size criterion for study entry applied to the number of assessments completed rather than the number of baseline Geriatric Depression Scale-15 scores obtained (see Table 17). Proportionally more participants in C&W had given assent than in NEL. This was because care homes in NEL had proportionally fewer cognitively impaired residents and some time constraints for the NEL research team limiting time to obtain assent.

FIGURE 9.

Reasons for being unable to gain consent/assent prior to randomisation. a, Eligible.

Individual characteristics

We had consent/agreement to directly assess 857/907 of the residents who consented or assented to join the study prior to randomisation (see Table 17). We assessed 783 (91%) of these. Two died between assessment and the day the home was randomised. (We do not have data on time of death. Some deaths may have occurred on the day of randomisation and prior to the time randomisation was performed.) It is the remaining 781 residents who constitute our population of interest for participant-reported outcomes. We obtained a Geriatric Depression Scale-15 score from 765 (98%) of this population at baseline. Overall we recruited an elderly (age range 65–107 years), predominantly (76%) female sample. Of the 781 individuals included in the analyses for patient-reported outcomes, 94–98% responded to each item of measurement at baseline except for self-assessed EQ-5D for which the response rate was only 81%. We anticipated this low response rate prior to the study and consequently also collected proxy EQ-5D. The majority of residents were not in pain, but almost half were afraid of falling and almost one-third were on antidepressants. Baseline characteristics of those who provided data prior to randomisation were similar in the intervention and control groups (Tables 18–21).

Prior to randomisation, we had consent/agreement to obtain care home data on 903/907 residents. Six of these died between consent/agreement being obtained and the day the care home was randomised, and a further 16 had no care home data collected prior to randomisation. It is the remaining 887 participants who constitute our population of interest for care home data. We obtained proxy EQ-5D and Social Engagement Scale data on 88% of these participants.

The prevalence of comorbidities was high and similar in the intervention and control groups (see Table 19).

Depressed cohort baseline characteristics

Forty-nine per cent (374/765) of those who provided GDS-15 data at baseline were classified as depressed (GDS-15 score of ≥ 5 or equivalent) (Table 22). It is these participants who are our population of interest for all the depressed cohort analyses. For participants in the depressed cohort, GDS-15 scores were inevitably higher (mean = 7.5), but other baseline characteristics, including the proportion on antidepressants, were broadly similar to those in the overall cohort (Tables 23–25). Baseline characteristics of the depressed cohort were also similar in the intervention and control groups. Of those in the depressed cohort, 25% (92/365) had been diagnosed with depression according to care home records and 30% (112/371) were taking antidepressants (see Table 22).

The prevalence of comorbidities was high and similar in the intervention and control groups. Prevalence of comorbidities was similar to that observed in the overall cohort (Table 26).

Participants joining study after randomisation

A further 163 residents provided baseline data after randomisation; of these 150 provided assessment data and care home data, and 13 provided just care home data. Ten were screened for eligibility prior to randomisation but the recruitment and/or baseline data collection processes were completed only after randomisation (six by 6 weeks, one around 3 months and one around 6 months). Fifty-seven were residents who were already living in the homes (26 recruited around 3 months, 19 around 6 months and 4 around 9 months), 45 were new residents recruited at the 3-month data collection visits, 59 were recruited at the 6-month visits; the remaining two were recruited at the 9-month visits (Table 27). In total, 127 of the 163 were present at the end of the study. Of these, 107 contributed direct assessment data and 126 contributed proxy data to the cross-sectional analyses.

The baseline characteristics of participants included in the cross-sectional analyses who were first assessed after randomisation were similar to those recruited before randomisation. Characteristics of intervention and control groups were similar, although larger numbers were recruited in the intervention than in the control arm (Tables 28–30).

Secondary outcomes: cross-sectional analyses

As for the cohort analyses (see Follow-up and Primary outcomes: Geriatric Depression Scale-15, above), cross-sectional analyses including all those resident in the home at 12 months do not provide any evidence of an effect of the intervention on any secondary outcomes (Tables 44 and 45).

Medication use

We obtained baseline drug use data on 869/887 (98%) of participants in the care home data cohort. We obtained data on progressively fewer residents at each time point falling to 615 at 12 months. We obtained data on 91% of those who were alive at 12 months. A large proportion of these were taking psychoactive drugs; with one in six taking hypnotics or anxiolytics, one in three taking antidepressants, and one in nine taking antipsychotic drugs. There were no differences in the number of participants using these drugs, over time or between the intervention and control (Tables 46–48).

There was no evidence of a change in participants' use of psychoactive drugs during the lifetime of the study. Few residents either started or stopped these medications, and there was no suggestion of a difference in the pattern of change between intervention and control homes (see Table 47).

Only a minority of residents used any medication from each of the three selected groups of psychoactive drugs. The mean number of defined daily doses taken was small for antipsychotics, larger for hypnotics and anxiolytics and largest for antidepressants. There was no evidence of a difference in amount of these drugs used between intervention and control homes (see Table 48).

Indirectly attributable adverse events

There was a total of 90 fractures identified from home records over the study period; these occurred in both study participants and in non-participants. There was no evidence of a differential fracture rate between intervention and control homes (Table 49).

Among those who had provided consent or assent to have their data examined, there were 56 definite fractures, eight probable fractures and nine possible fractures. Again, there was no evidence of a difference in fracture rate between the intervention and control homes (Table 50). Of the definite fractures, i.e. those requiring hospital admission, 39/56 (70%) were fractures of the femur.

There were 241 deaths among the 1054 study participants: 119 in the intervention group and 122 in the control group. There was no evidence of differential death rates between intervention and control groups (Table 51).

Directly attributable adverse events

Five residents in intervention homes, including one resident who was not a study participant, experienced adverse events during the exercise groups or when preparing for the groups; none was serious (Table 52). Although two residents fell in the lounge of their care home this was before the group started and no injuries were sustained. Two residents had other health problems that became apparent during the groups and one suffered a minor injury, not requiring treatment. Thus incidence of directly attributable serious adverse events was < 1 : 30,000 attendances.

Costs

Resource use was broadly comparable between the control and intervention homes (Table 53). Participants were frequent users of community services and general practitioner home visits. Use of general practitioner surgery visits, mental health team visits and practice nurse visits was infrequent. Using the univariate linear-mixed model, no significant differences were seen between the two arms of the study. All event types were more frequent in the intervention arm.

The inpatient and A&E cost data collected from the primary care trusts had some missing data. For the outpatient attendance cost data, 36% of attendances had missing cost values, and 96% of these missing values came from two of the primary care trust data sets. Consequently, the outpatient costs were calculated using data for missing values that had been generated by multiple imputation. Outpatient attendances, however, constitute only a small proportion of the total costs.

The distribution of costs was skewed, as is typical for health-care costs (Figure 13).

FIGURE 13.

Distribution of participants' costs at 12 months, across all homes. (The outpatient costs used in this figure were derived by taking the mean cost of the five data sets generated by multiple imputation.)

As with the resource use, no significant differences were found (Table 54). For the costs associated with medications, the intervention arm had lower costs. For all other categories, the mean cost of the control arm was lower. The mean National Health Service cost per individual in the intervention arm over 12 months was £4639, whereas the cost per participant in the active control arm was £4251 (mean difference = £388, 95% CI −£604 to £1380).

Active control and intervention costs

The time and materials needed to deliver both the depression awareness training and the intervention programme were collected alongside the study. Unless otherwise noted, all costs associated with delivering the active control and intervention programmes were taken from the study records (Tables 55 and 56). In the control arm the cost of depression awareness training sessions was the highest cost, followed by the materials. For the intervention arm, the cost of the exercise sessions was the highest cost, followed by the mobility assessments.

The weighted cost per participant in the intervention homes was £322. For participants in the active control homes, it was £5. The societal cost per participant in the intervention home was £314 and for the participants in the control home it was £7. The lower societal cost in the intervention care homes was due to home staff time saved by the assistance and occupation of the physiotherapists while in the home, which outweighed the care home staff time spent training. The higher societal cost in the control care homes was due to the staff time spent in depression awareness training.

Quality-adjusted life-years

The level of missing data ranged from 4% to 16% per time point for the proxy EQ-5D scores; consequently, the quality-adjusted life-years for the primary analyses were calculated from the five data sets (one for each time point) with missing values replaced by values generated by multiple imputation.

Analyses

All reported incremental costs and quality-adjusted life-years are from multilevel models to recognise clustering.

Base-case analysis

In the control arm the multiple imputation averaged correlation between total cost and quality-adjusted life-year is −0.13, whereas in the intervention this correlation is −0.05.

Each of the five sets of total costs and quality-adjusted life-years were analysed using multilevel models to account for clustering. These produced an incremental cost of £374 (SE = £525, 95% CI −£655 to £1404). The incremental quality-adjusted life-year gain was −0.0014 (SE = 0.036, 95%CI −0.0727 to 0.0699). The intracluster correlation coefficient for the total costs was 0.07, and the intracluster correlation coefficient for the quality-adjusted life-years was 0.14.

To obtain a graphical representation of the uncertainty of the cost-effectiveness estimates, we used bootstrapping. We sampled at random with replacement each of our five cost-effectiveness sets including the imputed data, producing 2000 incremental cost and incremental quality-adjusted life-years estimates for each data set. These estimates were plotted in a cost-effectiveness plane. Figure 14 shows one of these graphs. The graph of bootstrapped estimates provides a graphical description of the distribution of the incremental costs and quality-adjusted life-years and the uncertainty surrounding these values. From Figure 14, it is apparent that more points lie above the x-axis, indicating that the intervention is more costly, and that more points are to the left of the y-axis, indicating that the intervention produces fewer quality-adjusted life-years. This also displays how other combinations of costs and quality-adjusted life-years that were consistent with the data produced sample means in all four quadrants.

FIGURE 14.

Cost-effectiveness plane generated from bootstrapped mean cost and quality-adjusted life-year differences for residents over 12 months (positive values indicate that the intervention was more costly or more effective).

The incremental cost-effectiveness ratio was not calculated as the intervention was more expensive with a net reduction in quality-adjusted life-years. Under these circumstances, disaggregated data on costs and effects are considered to be more informative. The probability of the intervention being cost-effective at a willingness-to-pay threshold of £20,000 per quality-adjusted life-year was 33% and at a threshold of £30,000 was 37%.

The case study care homes

We recruited four homes each from NEL and C&W. In each area one home was a control home and three intervention homes. The three intervention homes were selected so that one was an independent home (where the owner owned fewer than 10 homes), one was part of a larger chain, and one was designated as a nursing home (Table 58). All of the case study care homes were located in residential areas – several with local shops and amenities within a short walking distance, and several had access to nearby parkland. They all had gardens, usually well maintained, although one was poorly tended with very limited access. Smokers (staff and residents) were the main users of the outside space.

Of the four homes in NEL, one was a purpose-built home (which in the past was in local authority control) and had two floors. Two were converted houses, while the remaining home was a converted block of flats. With one exception, public areas (e.g. lounges and dining rooms) were all on the ground floor. The purpose-built home had an additional little-used small lounge on the first floor (used as a store room). All care homes had ground floor bedrooms and had lifts to the upper floors. Two of the homes had separate dining rooms next to the kitchen, while in the other two food was served either in the main lounge or an area used during the day as a ‘quiet’ room for reading or craft activities. Three of the homes had multiple areas for residents to sit and/or become involved in activities, while the other had a large multiuse room.

In C&W, there was one purpose-built home, one converted hospital building and two converted houses. Two of these homes had public areas on each of the floors. There were residents' rooms on all floors and all of the homes had lifts to all floors. In the purpose-built home there were also small kitchen/dining rooms in each of the three wings. One of the converted houses had one large central space divided into three areas: a dining area, a lounge for activities and a TV lounge, with no physical dividers between the areas.

Interviews undertaken

The process evaluation fellow (DE) carried out 48 interviews with 35 people; 21 were from NEL and 14 were from C&W (eight care home managers, six carers, three senior carers, four activities co-ordinators, eleven residents and three NOK). Where direct quotations are used to illustrate themes, an assigned code identifies the respondent and time point within the study. A full table of the interviewees and their corresponding codes is given in Appendix 12.

We could not complete as many interviews as we had planned. Care home managers and staff were very busy and often had to cancel interviews. We could interview residents only if they had, themselves, consented to the OPERA study, which restricted the number of potential interviewees. These numbers reduced further as people left the home, were in hospital or died. We also found that participants were reluctant to be formally interviewed. It was particularly difficult to arrange interviews with NOK as we had no access to their contact details.

Evaluation of recruitment processes

Impact of recruitment on the care homes

In addition to being asked about their interactions with the physiotherapists, care home managers and carers were also asked about interactions with the OPERA team (including the recruitment/assessment staff, research fellows and contact with the study office). All managers and a number of the care staff commented that the visiting OPERA team, although sometimes viewed warily at the start, became welcome in the home. Care home managers noted that appointments were made for visits and the team were no burden on the homes' residents or staff. Several reported good outcomes from problems that were raised.

I think it was excellent. There wasn't any problem at all. I mean the residents, they [the OPERA team] did not impose. They knew which residents were not able to make the decision and so they contacted relatives. I think it was done very well – no problem.

19 manager

And it was a pleasure having [Names] here. I mean all of you ... I've never had any problems. You've actually fitted in and people have accepted you.

20 manager

Even when you did the interviews here, we did not have any problems. Everybody was just so respectful and kind and just nice people – just a nice bunch of people.

32 senior carer

Delivery of care home staff training

We delivered at least 142 depression awareness training sessions to at least 902 care home staff; eight of these training sessions were observed (five intervention and three control) (Table 59). Registers are missing from one home where at least one training session was delivered. One home had six sessions delivered covering most of the staff. Overall less than half of the care homes' staff were exposed to the training. OPERA research staff in both arms of the study reported difficulties at times in organising the training sessions. On at least 15 occasions no care home staff attended a booked training session. It proved to be particularly difficult to deliver training in some care homes, and after several failed attempts training was abandoned in one home in the control arm.

We received 916 completed feedback forms at the end of the training sessions (this included 14 forms from the care home with a missing register); 174 (19%) participants had worked in a care home setting for < 1 year, while 421 (46%) had worked in such a setting for > 5 years. Ethnographic observations in the case study care homes and interviews suggested that a majority of care staff have been working in the sector for many years. We observed cases of a changing staff population but there was a large core of staff with long-term experience.

Specific observations of the delivery of the training revealed some of the challenges faced by the research team delivering the session. Quiet space within homes for staff to attend training was not always available and the attendance varied. Some of the care home managements did not pay staff for time spent in training with the result that there was little incentive to attend, especially if staff had a day off or if training took place outside their shift hours. A number of staff attended the training during their normal shift but in some cases this resulted in them being called away from the training to carry out a caring task. Some home managers actively encouraged all staff to attend (including ancillary staff and night staff) but this was not universal.

Observations during the training sessions generally supported the results of the feedback questionnaire described below (Table 60). Care home staff appeared to enjoy the DVD and it generated discussion. The levels of engagement with the training were variable – some got involved and asked questions or made comments, while others said very little.

The immediate participant feedback questionnaires suggested that the training was felt to be relevant and provided new information. However, the response rate to the mailed, 3-month questionnaire was very poor (15%). Investigation suggests that there was no monitoring of returns and no reminders were sent. Those who did respond provided interesting feedback, more detail of which is provided in Appendices 14 and 15; in summary it suggested that:

• the training was well liked

• the materials were useful

• the training sensitised people to be aware of low mood and depression

• the training changed people's perceptions of what to look for in a resident.

However, there was not much evidence in the feedback that the training changed the way people did their jobs. Most respondents said they would have liked more training.

Interviews with care home staff and home managers support the findings of the training feedback. Most of the managers had attended the training themselves and several had received positive feedback from their staff. Although the training was seen as a good thing there were comments about the level of the training.

... to be honest, most of it I felt like I already knew really

28 activity co-ordinator

I thought it was very basic and I would have liked something more sophisticated because I think basically what she said, we already knew and we really needed the next level up.

23 manager

Equally, there were positive comments such as:

I think it was an eye-opener. There was the DVD she brought with her as well. I think it did get people questioning and there was a lot of interacting going on in that session.

20 manager

Yes, I mean the whole thing that [Name] and [Name] did was good. It wasn't too technical and it wasn't talking down to people. It wasn't talking under people. It wasn't too complex, too simple, too fussy, too thin or whatever.

25 manager

Many of the research staff who delivered the care home staff training reported that managers and staff would have liked more than just a one-off depression awareness session, and some care homes would have liked the training to be in more depth, supporting the findings from the interviews. One care home requested additional depression training for staff after follow-up was completed (post OPERA). Several physiotherapists were aware that the training helped the home fulfil its requirements for the Care Quality Commission supported by the certificates provided to attendees.

Set-up of the exercise groups

Thirty-five homes were randomised to the ‘whole-home’ intervention: 18 in NEL and 17 in C&W. After randomisation to the active arm of the study, one care home in NEL was found to be unsuitable for the intervention during the physiotherapist assessments of residents and no exercise groups were delivered in this home. Thus the following data, for the delivery of exercise classes, are from 17 homes in each area. Data presented in this section include all residents within the care homes and are split by those who are study participants and those who are not. There was considerable variation in the time taken getting the intervention started (between 3 and 14 weeks) (Table 61). There was a shorter average set-up time in C&W and hence a longer period of actual intervention delivery.

Before the exercise intervention started, there were some anxieties among staff of the homes. Both care home managers and carers anticipated a number of potential barriers, including difficulties in fitting into the care homes‘ routines and sustaining the participation of residents.

Well, one thing that could stop you from doing it is if none of the residents joined in, obviously ... but other than that, I do not really think there is anything that would stop you from doing it.

1 manager

It's a bit of hard question because they might say ‘yes’ and when it comes to it, they'll probably change their minds – the residents, because they like doing that to us. I mean one or two might do it.

12 carer

Well, I think some barriers have probably already come across and that is ... and I said this from the start, that some residents ... You know, they're having it twice a week – the exercises and some residents, they may want to be totally committed one week but then the next week they cannot be bothered, and this happens.

19 manager

There were also some reported problems arising from differences in expectations, although these were all resolved relatively quickly.

With regards to myself and the staff, I think she's just a little bit overpowering and I do not think they understood that she would have such an active part that it would be twice a week and it would be so intense, and she would want so much from the other staff. We did not expect that, so I think that's been a bit of a learning curve really.

44 manager

Well, I think it's a bit over the top with [Name] occasionally. You know ... all I can do as a manager is just put it to the company and it's not my decision. It's the company's decision whether we change things.

19 manager

The residents interviewed at baseline could see no barriers to the research team delivering OPERA into the home and in general thought it was a good idea although two residents doubted that people would participate:

But people will not join in, will they?

11 resident

Yes, that would be good [the exercise groups]. They do not always join in anything here.

34 resident

Dose delivered

In one of the 35 intervention homes the dose delivered was zero because residents were too disabled to participate. We report here on activity in the remaining 34 homes. Delivery of two exercise groups per week over 12 months (allowing for set-up time and key public holidays) gives a potential 100% dose delivered of 92 groups in a home (Table 62). The study achieved a 90% dose delivered: 90% in C&W and 89% in NEL. Four homes were so big, or laid out in such a way, that they required more than one pair of exercise classes per week. Some planned exercise classes failed to happen. The physiotherapist team reported a number of reasons for this including infection control issues (e.g. outbreaks of diarrhoea and vomiting, norovirus and the influenza H1N1 epidemic), and adverse weather conditions (e.g. snow and ice, or days that were too hot for residents to exercise).

The 34 intervention homes had a total resident population of 1439 (based on study registers during the period of the intervention delivery); of these, 494 (34%) were study participants. Not all residents were judged by the physiotherapists to be fit enough to participate in the exercise groups. There were around 1256 (87%) residents who were eligible to attend groups of whom 478 (38%) were study participants.

During the study physiotherapists delivered 3191 exercise groups in 34 homes. In total 65,196 person sessions were made available, calculated by summing the number of groups each of the 1256 eligible residents would be able to attend (i.e. excluding sessions for which they were ineligible or when they were not at that point a resident in the home). On average residents attended about half of the available sessions (31,705). An average group size of 10 (5.3 study participants and 4.6 non-study participants) was calculated by dividing the total number of person attendances (31,705) by the total number of groups delivered (3191) (Table 63). Figure 15 groups study participants into the number of groups they attended. The reasons for not attending the available sessions are discussed in more detail below.

FIGURE 15.

Study participants: number of groups attended.

The most cited reason for non-attendance was the unwillingness of participants. Unwillingness to attend appears to be consistently higher in the NEL homes at 31% compared with 21% in C&W. We observed 21 exercise groups in case study care homes; seven in the early stages of their introduction (five homes once and one home twice), six at about the midpoint (once in each of six homes) and eight at the end of the 12-months (four homes once and two homes twice).

During the interviews little was said about why residents attended or did not attend. However, a care home manager and an activities co-ordinator noted that the time of day could have an impact on attendance.

The only thing we actually found was that someone could not come in the morning, but wanted to come in the afternoon and they were not having none of it (the residents) ... They know their own mind ... And no, they're all tired and they've had lunch and no, and it's really funny but that day none of them wanted to do it.

31 manager

... what I've found is the best time for me to do it, like you've seen I come in the mornings and I do it straight off in the morning, because the best time is to get them in the morning because they do it before their lunch, they're sitting there, they're tired, they do not ... if you allow them to, they'll sit and they'll smoke away in the chairs and be relaxed and just listen to the television or watch the television. Get them up and moving, and then it's easier to get them more involved in other activities later on in the day, where if I do other activities and then just put the exercise at the end of the day, they're normally you've got less of a chance to get them to participate in the afternoon, as you would in the morning.

50 activity co-ordinator

Home observation suggested that some residents preferred groups in the morning and some in the afternoon.

A priori we set the threshold for the minimum effective dose as attending 51 sessions, an average of one per week; 36% of our participants achieved this. Nine per cent of our study participants attended no groups, 20% attended 10 groups or fewer (Table 64).

Overall, there was little attrition in the attendance at groups over the year (Figures 16–18 and Table 65).

FIGURE 16.

Average monthly attendance to exercise groups: all homes (± SD).

FIGURE 17.

Average monthly attendance to exercise groups: NEL homes (± SD).

FIGURE 18.

Average monthly attendance to exercise groups: C&W homes (± SD).

Perceptions of exercise groups

Interviews at the end of the study illustrate the impact that the groups had on the residents and care home staff. Most were positive about the planning and execution of the groups, noting that the regular and consistent time slots were appreciated and that the physiotherapist liaised well with the care home managers. Care home staff also appreciated having a professional in charge of the exercise class.

I mean it's very good because it's always Monday and Wednesday ...

29 manager

... and [Name] was very good. You know they would come in and tell us what they were doing and how they were going to do it and what was going to happen so the communication even from them to us was very very good.

51 manager

You know, at first even when I used to do my little five minutes and workouts or we'd do a daily dance, it could be quite worrying; ‘Oh my legs hurt so I'm not going to do anything’, whereas because they've got a professional guiding them, using the weights and even telling her what weights to use ... You know, there's a sudden interest in their own well-being.

27 activities co-ordinator

The physiotherapists reported that residents, care home staff and visiting relatives enjoyed the exercise groups. Physiotherapists felt that they gained increased confidence in running a group, job satisfaction, and liked being involved in research. The practicalities of running a group raised issues relating to variation in the amount of care home staff input. Examples were given where there was no-one to help get people to the group or help with running it. This was therefore more time-consuming especially if there was little flexibility in time allocated to OPERA. For the physiotherapists, having an activity co-ordinator or member of staff designated to assist was a great help, and they felt that continuity in the staffing was important, and that the bond and understanding of the residents' idiosyncrasies made it easier to gauge the group. However, some of the care home staff felt that the group was a time during which they could ‘do other things’ (sometimes a highly appreciated aspect of the intervention).

Interviews with physiotherapists, care home managers, care home staff and residents revealed that the delivery of the group worked well and was easy to follow. Equipment was considered appropriate and the music an important part. Some explanations for the positive impact of music were that it helped to establish a routine, increased enjoyment, promoted activity, encouraged singing, which helped lung function, and promoted discussion about the music and artist.

The positive impact of the exercise groups was mentioned by many interviewees, talking about changes in residents' mood and physical ability and about the changes they have seen in residents' attitude towards exercise:

They have become happier in their mood and I think it has achieved what it set out to do. Certainly towards the end of the programme once everyone got into the swing of it, it really picked up.

48 manager

I'd often walk by and see them and they were always smiling. There was always ... a good rapport in there. There was always laughter in there. The music would be going and you would see them all doing this [gesture] and I never saw anybody sitting there sort of looking glum.

32 senior carer

You know it has been wonderful, it has been good for our residents, it has been excellent in their development; you know they are happier

51 manager

I just feel a lot better. I mean I can ... (shows me he can stand) I can (So is that better than it was?) Oh yes (note: was very reliant on wheelchair at start of study).

39 resident

During observation of the exercise groups the challenges faced by the team and the general enjoyment of the residents were noted. As a relationship developed between the physiotherapists and the residents the groups appeared easier but there were still some residents who were reluctant to be involved and others with challenging behaviour. During the groups there was laughter and smiles, and at times the groups became competitive. The music was well liked and residents often sang along, and indeed knew which exercise they should be doing with a particular tune. It was clear in the homes that there were a number of residents who looked forward to the groups.

The physiotherapists running the groups noted that the exercise involving a velvet rope promoted inclusion and involvement, and gave a tactile experience. They also noted that the use of equipment (e.g. balls and weights) promoted activity, especially in the cognitively impaired, but could be problematic. Some of the physiotherapists found the weights hard work and some residents were initially resistant and said that the weights, for example, ‘looked like manacles’. In some cases, weights and the weighted balls created some competition between residents, with banter about how much they had done that day in the session.

There were some issues around safety, for example, when throwing and kicking balls, or the potential to wrap the rope around one's neck. Space, temperature, arguments over seats or routines, and group disruptions were also challenges that had to be faced.

The whole-home intervention and role of the physiotherapist

The aim of the intervention was to achieve a change in the whole culture of the home, so that the importance of helping residents to be active was embedded permanently. Home champions were intended to be part of this intervention and it was planned that these would be identified early on. However, this did not happen in all homes. Some of the activity co-ordinators in homes became involved but not all felt it was in their remit. Again, some carers were identified as ‘champions’ but problems with shifts, time or enthusiasm limited their effectiveness. It seems that the role of home champion was unrealistically demanding for most staff in most homes.

The physiotherapists felt that it was sometimes difficult to explain their remit to care home staff and that some staff were very wary of them. Nevertheless, the physiotherapists reported that clinical assessments of the residents gave opportunities to meet residents and staff, to build a rapport and increase the understanding of the importance of physical activity and lay good foundations within the home. Several care home staff commented on how staff were encouraging residents to walk more.

She said ‘Wheelchair’ and I said ‘No, you can walk.’ She said ‘Oh I don't know whether I could’ and I said ‘But you can because you used to walk all the way up there.’ She said ‘Oh yes, I did’ and she walked down there fine. She was slow but she walked there fine because everybody keeps sticking her in the wheelchair to take her to bed, but she can actually walk and I think if she can walk then for G**'s sake, let her walk.

1 Carer

But generally, sort of one of the ladies downstairs, she was having the wheelchair quite a lot and then they said ‘Well, maybe if we encourage her to walk little bits at a time and then a bit more and a bit more’ and it's worked really well and she's walking everywhere now.

2 Senior carer

Physiotherapists reported seeing their suggestions being acted upon, as for example, where residents were seen to be walking instead of using a wheelchair. The physiotherapists also felt that they had reinforced good principles of manual handling and promoting activity, and had broken down barriers to mobility. Physiotherapists encouraged exercise progression by setting individual goals and helping care home staff to recognise and reinforce these.

Care home managers were generally happy to have a physiotherapist working in the home, and communication and understanding increased as the intervention progressed, although not all staff were receptive to the physiotherapists' advice.

The staff definitely do not feel that (the physiotherapist is an imposition) and as time has gone on with [Name], they've really got to know [Name] and she sits and has her lunch with them – not for long, but she does and they look forward to her coming in, to be honest

19 manager

Yes, because I think that gave us a sort of opening as well to actually think, ‘Well, I'll do ... this person can actually do more than what they're actually saying that they can actually do’, so they brought a sort of ... they brought a lot of that out

31 manager

The hardest part I think for me was trying to get the staff on board to realise what this was about, who it was for and how people would benefit.

32 senior carer

One manager noted the challenge she faced in trying to increase mobility.

It is a collective protective culture and that's why they come into care. They think people should be protected, but I think in this situation because we deal with people who have very high dependency needs that it sometimes is that we de-skill them, rather than enhance them because it's a protective mechanism.

44 manager

And another talked about the challenge in bringing about changes to thinking:

Yes, that's what I say to the staff because I mean when people talk about activities I think it's still there – the old belief that it is about doing the art work, doing bingo and you need to get away from that because that's not what it's about.

20 manager

Two interviewees reported that OPERA has had a positive impact on external views of the home. One home found the study mentioned approvingly in a Care Quality Commission report and in another home a social workers' report highlighted the positive effect of OPERA.

Yes, she'd actually heard of it (Care Quality Commission inspector). She must have gone to another home before she came in here and we actually saw the file and she said it was a very good study and she actually had feedback from another home.

32 senior carer

Social Services, they have to have regular reviews and one of the residents who's involved in this study, she had her review and they ... The reviewing officer always sends me a copy of the review ... and I read it yesterday and I was so proud of, you know ... Her and her family could not thank the home enough but also, OPERA was mentioned in that review ... and about how she's enjoying it and that's recorded.

19 manager

By 6 months and at the end of the study interviewees were positive about the intervention. Most commented on how the relationship developed between the care home (the manager, residents, staff and sometimes relatives) and the physiotherapist and in all cases the physiotherapist became seen as ‘members of staff’, ‘members of the team’ or ‘one of us’.

Brilliant and that's never petered off. In fact, it's just got more positive. They're brilliant with her. They trust her with everything and that in itself is a feat.

40 manager

I think the staff ... well, certainly we will miss her and they really will. The residents ... my G** they'll miss her

43 manager

One senior carer felt that the number of falls had reduced and attributed some of this to OPERA physiotherapist input.

There was one resident here with a walking stick, ‘Where's his stick?’ and he kept falling. We had a terrible history of falling with him. With a lot of perseverance, [Name] got the frame and now he actually walks with a frame very well. He's reduced his risk of falls. We've had hardly any falls with this man and he doesn't mention his stick any more ... Yes, so we have seen some good results ... (falls). They have been reduced, I'd say definitely ... And I think if you've got the right aids – and a lot of them didn't have, it has minimised the falls. They now use the proper things they should be using.

32 senior carer

The professional conduct of the physiotherapist was seen as very important, as was having the same physiotherapist over time.

Well that's very important, yes because the residents get used to that person coming in and they build up ... They kind of bond with them don't they?

28 activities co-ordinator

... yes and I think it definitely makes a difference what physio you use because I think it depends on the person they send down as well. I mean [Name] is very ... She's sort of a bubbly character. She's not afraid to motivate them to join in.

29 manager

It doesn't matter what sector you come from, whether it's general practitioner, physio or district nurse, (continuity) is really important because each time a new person comes in you're starting from the beginning again.

40 manager

However, despite the intervention lasting for a 12-month period physiotherapists felt there was little time to facilitate a change in staff behaviour around encouraging residents' mobility. The barriers to facilitating change included staff turnover, staff attitudes towards manual handling, staff morale and time constraints.

The key facilitators were the physiotherapists and several care home managers saw this as an important benefit for the residents. Physiotherapists were seen as supplying both practical advice on mobility aids and more general advice on navigating the services available.

That time when we'd got a problem with ... oh that lady who passed away eventually. We were really struggling with the doctors and the mental health team. It was evident that she needed something and [Name] just came and said, ‘You can refer her online to the mental health team’ and we didn't even know that did we, and because she's a professional, we got somebody here the same day.

43 manager

Yes, that was another asset I think that [Name] and the team brought in, which was yes, we have people walking with the walking aids and the frames and the walking sticks, but the good thing is that for me to get everybody checked here ... It would be crazy of us. [Name] came in and [Name] and they assessed each person ...

20 manager

[Name] was excellent there. We were using the wrong aids and we still had walking sticks on the premises we shouldn't have had. [Name] was brilliant that way and that saved us a lot of time I think, referring them from the general practitioner to a physio, for them to come in and then now, the (National Health Service) physios don't actually come on the premises.

32 senior carer

The importance that care home managers placed on having access to a physiotherapist was no doubt influenced by the difficulties they reported in accessing routinely available physiotherapy services and mobility aids.

There are the initial referrals – especially via the general practitioner. Sometimes you get some of the district nurses that um and aah and all the rest of it and then eventually do and I said to [S] last week, I'm finding it quite hard to understand why she's put in her referrals and the equipment had come within a couple of days. If you go via the other channels we've waited an age.

16 manager

Activity in case study care homes

A total of 109 activity sweeps were carried out: 53 at baseline and 56 at follow-up, with 12 to 14 observations in each home at each time point. Across the case study care homes a substantial amount of residents' time was spent in activities associated with eating and drinking with active social interaction only amounting to between 5% and 15% of daily activity. The proportion of time spent in recreational activity, passive social interaction and being socially inactive appeared to vary across the homes. With only one exception (home 4), no exercise groups or ‘exercise’ activities were observed (Figure 19).

FIGURE 19.

Summary of day-time activity in homes, baseline vs follow-up. BL, baseline; FU, follow-up.

In contrast with the different patterns of activity and inactivity seen between the homes, within homes the patterns of activity were remarkably consistent between baseline and follow-up. None of these data was analysed until all data were collected, so it is unlikely that the researcher was influenced at follow-up by the findings at baseline (see Figure 19).

The observation sweeps do not show any major changes in activity within the homes; two homes had some increase in recreational activity and one a decrease, and in one social interaction increased. These may be chance differences because of the large number of comparisons made. General observation in the homes supported the conclusion that there was little change in overall activity levels within the homes.

Ending the intervention

When the intervention was coming to an end, staff in the homes expressed some concern about what would happen when the study was finished. Without exception all interviewees expressed a feeling of sadness, and in some cases concern, about the withdrawal of the exercise groups and the physiotherapist.

Oh yes they will miss it as well. They used to come to the exercise class as well so. Yes they will miss it. And I think it is a shame that it has come to an end but everything has to come to an end, and hopefully when all the powers that be have done their research and they have got all of their figures and everything together maybe ...

48 manager

No, I'm just hoping that something can be done for them when this course is over and done with. I just don't like to see them be let down. I've seen it too often and it can actually damage them – and I have seen it, so ... and if you can do anything for the staff that would be fantastic. We need to be de-stressed.

17 carer

Yes, they love it. They really love it and it's actually a shame if it actually stops, because it's something that they've got there that they have every week. You know, it is a shame ...

31 manager

Many of the care homes brought gifts for the physiotherapists and had celebrations at the end of OPERA and there were tears and sadness on both sides. Several of the physiotherapists have stayed in contact with some of the homes and residents. Some physiotherapists felt unprepared, as the closing process was not in original paperwork. Some homes have kept exercises classes going in a modified form, some have purchased some physiotherapist input, and other homes have no funds to continue.

Experience within the OPERA study

Data external to OPERA

The ethical challenges of recruitment and consent to participation in research studies involving adults who may lack capacity were further explored through analysis of data from the focus groups and key informant interviews. There was a consensus that conducting research with elderly participants was important and focus group participants were prepared to consider participation in a wide range of research studies. Key informants considered that excluding people because of frailty, comorbidities or cognitive impairment was discriminatory. They identified a number of practical and ethical difficulties in organising and conducting research in this population, including protectiveness of research ethics committees or research governance processes stopping research proceeding, the need to minimise the burden on frail and/or ill people in the research design, the demands placed on care home staff in facilitating research, and the difficulty of assessing capacity and obtaining valid consent. Two key themes emerged from the analysis, identified by both focus group participants and key informants from their different perspectives. These were (a) difficulties with the standard consent process, and (b) the importance of trust.