Home-based health promotion for older people with mild frailty: the HomeHealth intervention development and feasibility RCT

Methods

Synthesis of data and data analysis

We inductively developed coding schemes to group types of interventions targeted to pre-frailty (e.g. exercise and nutrition) and summarised the evidence available within each type.

For RCTs, we used meta-analysis to combine the results for similar interventions assessing the same outcome. When studies assessed an outcome using two or more measures, we reviewed the literature and selected the most comprehensive, valid and reliable measure; when studies compared multiple interventions with the same control group, we pooled the mean and standard deviation (SD). 27 For crossover designs, we included first period data only (obtained from the authors). 28 Post-intervention end-point data were used for consistency across trials because intervention and follow-up durations differed. We combined continuous end-point scores using standardised mean difference (SMD) in a random-effects model and quantified heterogeneity using the I²-statistic. For outcome data that could not be synthesised (e.g. change scores rather than end-point scores27), interventions with only a single study and long-term follow-up data we used narrative synthesis. Authors were contacted when possible for further data.

For observational studies, we narratively summarised the data available relating to each modifiable risk factor for frailty progression/adverse outcomes to explore which factors have potential for improving outcomes to include in a new intervention. We intended to synthesise qualitative studies using meta-ethnography,29 but we did not find any qualitative studies meeting our inclusion criteria.

Results

The flow diagram for this review is reported in Figure 1. We identified 855 unique references through database searches and citation tracking of eligible papers. We excluded 727 citations on title and abstract, screened 128 full texts and excluded 114, largely attributable to population (n = 71; e.g. combined outcome data for frailty and pre-frailty, frailty in a specific condition). Three were reported in Japanese and five were unobtainable. Therefore, the review included 10 papers reporting on seven RCTs and four observational studies. Of these, 13 were identified through database searches and one from citation tracking.

FIGURE 1.

The PRISMA flow diagram of studies included in the review.

Synthesis

Key findings

Currently, both observational and experimental research targeted at pre-frail or mildly frail older people is sparse. RCTs are mostly small and/or of poor quality and focused on group exercise to improve mobility and physical functioning, with mixed evidence that group exercise may have some effects on functioning. Observational studies were large, of good quality and suggested that BMI, cognition and vitamin D levels may influence future transitioning to frailty, but that the number and burden of medications is unlikely to have an effect. There were no qualitative studies exploring perspectives on health promotion interventions in populations with mild or pre-frailty.

Implications for intervention development

• No interventions can currently be recommended for widespread use in pre-frail or mildly frail older people, but exercise may be an effective component for a new intervention. Including a nutritional or cognitive component may have potential in frailty prevention.

• Broader, multidimensional interventions targeted to mildly frail populations have not previously been assessed.

• Further qualitative work is needed to help clarify potential components for an intervention tailored for people with mild or pre-frailty. Current interventions for pre-frailty lack developmental input from service users or other stakeholders. Future interventions may benefit from this.

• From observational studies, nutrition, including ensuring the use of vitamin D supplements (as per standard guidance for older people) and addressing low weight/weight loss, may be an important component of an intervention.

Intervention functions

Michie et al. ’s51,52 behaviour change wheel and behaviour change technique (BCT) taxonomy identify and classify the content of interventions aimed at changing behaviour. Their systematic review of behaviour change frameworks identified nine intervention functions (broad, non-exclusive categories of means by which an intervention can change behaviour), namely:

1. education – increasing understanding or knowledge around the behaviour

2. persuasion – inducing positive or negative feelings or action through communication

3. incentivisation – creating the expectation of a reward

4. coercion – creating the expectation of cost or punishment

5. training – teaching skills

6. restriction – reducing the opportunity to engage in a target behaviour or reducing competing behaviours to increase engaging in a target behaviour, using rules

7. environmental restructuring – changing the social or physical context for a behaviour

8. modelling – providing an example for people to imitate or aspire to

9. enablement – reducing barriers, or increasing means, to improve capability or opportunity for a behaviour (beyond education, training or environmental restructuring).

Behaviour change techniques

Intervention functions describe broad mechanisms rather than specific ones. As such, the actual active components of effecting the changes needed are described as BCTs. BCTs are the active components of behaviour change interventions that are observable, replicable and irreducible. 52 Michie et al. ’s53 taxonomy of 93 BCTs was developed through a Delphi study in order to facilitate consistent description and replication of behaviour change interventions, as well as define mechanisms of action to use in intervention development and refinement.

Aims

In order to develop a new intervention from a rigorous theoretical basis, we undertook a systematic review to (1) identify behaviour change components used in home-based health promotion interventions in frail and pre-frail older adults and (2) explore how these components may be associated with intervention effectiveness.

This review is summarised in brief here – see Jovicic et al. 54 for the review protocol and Gardner et al. 55 for a comprehensive report of the findings.

Methods

Results

Out of 1213 unique references identified from database searches and citation tracking, we screened 267 full texts and included 19 full texts describing 22 interventions (see Gardner et al. 55 for flow diagram). We excluded 248 full texts, largely attributable to an irrelevant population (n = 201).

We included 19 studies of people who were frail or at risk of frailty: 16 RCTs,12,13,56–69 two clusters RCTs48,70 and one pseudocluster RCT71 (see Report Supplementary Material 3 for study characteristics). Most (n = 16) compared one intervention to control, with three three-arm trials. Overall, trials were fairly good quality. All were rated as having a low risk of bias on four out of seven criteria and three were rated as having an overall low risk of bias. Interventions largely took the form of a personalised assessment, focused on care or health needs (n = 15), compared with usual care or no treatment (n = 18) and delivered by nurses (n = 21/22). Some studies included more than one intervention. A small number of studies included other professionals such as physiotherapists or social workers. Physical health and functioning outcomes were most commonly assessed (n = 19), followed by mental health and functioning, health and social service use and generic health and well-being (n = 11). Social functioning and well-being was assessed in seven12,57,64–66,70,71 studies and behavioural outcomes in four. 58,65–67

Behavioural targets

Regarding behaviour, only three reported using behaviour change theories. 48,63,64 Most interventions (n = 11/22) targeted a single behaviour,12,61–64,66–68,70,71 while nine targeted two or three behaviours13,56–59,65,69 and two targeted four or more behaviours. 48,60 The behaviours most commonly targeted were medication adherence or management (n = 16),13,48,56,58–61,63–66,68,69 physical activity (n = 11)12,13,48,56,57,59,60,65,70 and diet (n = 8),13,48,56,57,60,69,71 with one or two studies addressing areas such as alcohol, sleep etc. Targeting a particular behaviour did not appear to be associated with positive results for any outcomes (Table 4).

Intervention functions

Intervention functions included:

• enablement (n = 16/22) – increasing older adults’ means or reducing barriers to increase their capability (above and beyond education or training) or opportunity (above and beyond environmental restructuring) to undertake a behaviour

• education (n = 7/22) – increasing knowledge or understanding

• environmental restructuring (n = 4/22) – changing the context for a behaviour (including physical and social)

• training (n = 2/22) – imparting skills

• persuasion (n = 2/22) – inducing or stimulating positive or negative feelings or actions through communication.

No intervention used any other function. Functions could not be identified in five interventions. 12,66,68,69,71 More studies including education and enablement had positive results for physical functioning outcomes than those that did not including education and enablement (see Table 4); however, in other studies, all functions had less than half of the studies showing positive effects.

Behaviour change techniques

A minority, 21 out of 96 possible BCTs, were identified in at least one intervention (median 4.5, range 1–9). Social support (practical and unspecified) and monitoring outcomes of behaviour without feedback were used most commonly; however, only adding objects to the environment and instruction on how to perform the behaviour had a majority of studies with positive findings (see Table 4).

Key findings

• Studies rarely assessed behavioural outcomes, used explicit theoretical approaches or reported intervention components clearly.

• No single BCT or intervention function showed potential for outcomes other than physical functioning. Studies rarely measured specific behavioural outcomes.

• Targeting a particular behaviour did not appear to lead to improvements in physical functioning.

• No intervention functions or BCTs appear to be uniformly effective, but education and enablement and adding objects to the environment and instruction on how to perform the behaviour were more promising.

Implications for intervention development

In the light of the findings from this review, a new intervention for early frailty should be explicit about its theoretical basis, with clear development and definition of intervention components. Within a trial, behavioural outcomes (e.g. physical activity) should be assessed in addition to clinical outcomes in order to explore whether or not behaviour change led to better outcomes. The most promising components were the intervention functions education and enablement, which showed the most promise for improving physical functioning (a key frailty indicator), and the BCTs ‘adding objects to the environment’ and ‘instruction on how to perform the behaviour’. We found insufficient evidence to support the omission of other intervention functions or BCTs at this point.

Objectives

To identify key systematic reviews of home- or community-based health promotion strategies, delivered by non-specialists, in the following domains and to extract key messages to inform components to include in a new health promotion service for mild frailty:

• exercise/mobility/physical activity

• falls prevention

• nutrition and diet

• social engagement

• mental health

• memory.

Methods

In the light of the paucity of studies focused on older adults with mild frailty, we took a broader approach to including reviews that could provide relevant evidence on which to base intervention components. Our inclusion criteria were:

• Community-dwelling older adults who could be defined as ‘at risk’ or frailer than a general population (e.g. non-healthy, at risk of falls or hospitalisation, with frailty).

• Interventions that could be delivered by non-specialists with or without minimal training, in the light of emerging findings from our qualitative work suggesting that a non-specialist support worker would be the most suitable professional (see Chapter 3, Person to deliver the service).

• Any control treatment.

• Outcomes relevant to each area of intervention (e.g. depression, social isolation, strength, frailty).

• Systematic reviews of RCTs or overviews of systematic reviews. As we aimed to synthesise the most up-to-date evidence, we excluded earlier reviews that had been updated and superseded.

We searched three key databases (MEDLINE, EMBASE and Scopus) from inception to April 2015 (see Report Supplementary Material 2 for search terms) and one reviewer screened titles, abstracts and full texts and extracted key data from reviews meeting our eligibility criteria in each domain. Two reviewers then identified the main findings of relevance to the development of our new service from each review, with reference to the original articles. As this was a state-of-the-art review, quality assessment and formal synthesis were not carried out. 72 We summarise the key findings from the included reviews that informed intervention development below.

Key findings of relevance to design of our new intervention

We extracted evidence from 66 reviews and three overviews of systematic reviews (Figure 2). We found the largest evidence base for physical activity and falls prevention (178 systematic reviews identified, including three overviews of reviews used for evidence extraction) and nutrition interventions (38 reviews), with limited evidence for social and mental health interventions and very limited evidence for cognitive interventions. Some reviews contained a mixture of systematic reviews, RCTs and observational studies.

FIGURE 2.

The PRISMA flow diagram of studies included in the state-of-the-art review.

Implications from the ‘state-of-the-art’ review

• Physical activity is a beneficial intervention to improve functioning and reduce falls and should form part of a tailored intervention, with a focus on individualising activity to frailty status and signposting to group classes when appropriate.

• Although promoting social activities has limited evidence, older people may benefit most from being signposted to local groups with a shared interest they can participate in, depending on their preferences.

• The most beneficial ways to prevent depression in later life may be signposting to psychological services or encouraging socialising.

• As there was very little evidence for effective methods to improve memory, this should not currently be included as part of an evidence-based intervention. Strategies to compensate for mild memory loss could be considered.

• Owing to the supporting evidence for some nutritional interventions, it may be beneficial to include these within a new service.

Background

Policy review and analysis helps explain past successes and failures, identify gaps and plan for future reforms. 142 In conducting this policy review we were informed by the work of Walt et al. 142 Our review was framed by theories of public policy as processes including problem analysis, formulation and implementation in which different interests, institutions and ideas interact. 143 These theories include recognition of the exercise of power by different interest groups, including the influence of ageism. 144

Public policy-makers internationally need to address increasing demand for health care and old-age support in the context of a declining labour force. 144 The Political Declaration and Madrid International Plan of Action on Ageing,145 agreed by the United Nations in 2002, addressed three priority areas: older persons and development, advancing health and well-being into old age, and ensuring enabling and supportive environments. Regional implementation strategies have been agreed at ministerial level, to work towards maintaining quality of life and promoting independence, health and well-being throughout the life course, underpinned by policies supporting health promotion and disease prevention. 146

In England, evidence of changing demography and potential impact on public spending costs has been known to governments for some decades and was recently requantified. 3 A ministerial commitment has translated into a range of policies that specify the promotion of well-being for older people as a strategic objective including those for longer working lives,147 for housing,148 and for transport. 149 We now summarise our analysis specifically of health and social care policies for frailty prevention.

Method

The design drew on the interpretative tradition in undertaking a narrative review using a method of documentary analysis. 141 It encompassed policy as created at three levels in the state:143

1. state laws

2. strategies and plans of government-mandated national bodies for the delivery of health and social care

3. government-mandated bodies at local administrative levels for health and social care.

The public policies for inclusion in our review had to be published and/or current to the period of the study (2014–17) and address one of the following:

• a population of older people (without an age-specific definition)

• public health and well-being for whole populations including older people

• publicly funded health and social care services for whole populations, including older people.

Internet searches of government and a representative sample of local government and NHS commissioning websites were conducted in 2015 and updated in 2017. A snowball technique was used to follow linked policies. A total of 79 national level and 78 local level documents were included. Each document was scrutinised for key words of interest, such as ‘older people’, ‘elderly’, ‘frail’, ‘frailty’, ‘health promotion’ and ‘ageing well’. Relevant surrounding text on the problem analysis, the formulation of actions and stated intent as well as absence of attention to the question of interest, was noted. Iterative analysis was discussed within the research team meetings and a final narrative analysis written.

Findings

The findings are reported within the following themes: problem analysis, formulation, solutions and delivery mechanisms.

The policy problem analysis of the ageing population with a changing epidemiological profile and the consequences for society (national and local) was restated at the beginning of every policy document.

State policy solutions included directions for ‘preventative actions’ as one theme for all adults for local authorities,150 the NHS151 and for social care provision. 152 It is worth noting that the term health promotion, which featured in the English National Service Framework for Older People, published in 2001 and re-endorsed by the government in 2014,153 was not evident in relation to this age group. In recent years prevention of ill health has become a priority strategy for the health and social care system in addressing the changing population, as evidenced by policy commitments and subsequent legislation for the public health function within government, local authorities,150 the NHS153 and the care system. 152 For the first time in England, social care policy mandated the promotion of well-being and prevention or delaying the development of needs for care:152 domains previously associated with the responsibilities of the ‘health’ system. The associated guidance provides definitions of primary, secondary and tertiary prevention. 154 This policy shift reflected the wider policy aspiration of integration between the health and social care systems. 155

A shift in the target populations for prevention had also occurred. In 2001, three groups were identified within the older population: the well and healthy, the frail, and then a transition group (i.e. the pre-frail or mildly frail). 153 However, policies reviewed here give little explicit consideration of those with pre-frailty.

Discussion and conclusion

This review has analysed contemporary health and social care policy for health promotion for older people with mild or pre-frailty in England. The review is time limited, our searches may not have been complete and the volume of material identified at the local level, and subject to amendments by changing governments, may have been incomplete. We have tried to mitigate this through our iterative processes.

We found that the older adult population was not always identified separately as a policy priority: the extent to which this represented a positive lack of age discrimination or a lack of attention to specific problems of some older people cannot be judged from the documentary evidence alone. Over time, the discourse changed to be more specific regarding prevention of ill health rather than promotion of health and was targeted either at the most frail and at risk of adverse outcomes or to those in mid-life (i.e. an ‘upstream’ public health solution169 and earlier in the life course). We found an absence of policy consideration for preventative actions for those on a pathway to frailty – a population that is predicted to grow in all countries. Publicly funded or supported services seeking to develop health promotion for older people with mild frailty may find it difficult to provide a policy ‘rationale’ amid other priorities. Addressing the individual and societal consequences of adverse experiences of those with the greatest frailty may require that some attention is paid further ‘upstream’ in the life course.

Methods and sample

The WISH study took place in two study sites: an urban and a semi-rural area comparable to the HomeHealth study sites. The methods for the quantitative feasibility study have been reported elsewhere. 175 Community-dwelling adults aged ≥ 65 years were recruited from five general practices. A total of 454 older people completed baseline Multidimensional Risk Appraisal for Older people (MRAO) assessments and received tailored feedback on their health and well-being. At 6 months, 348 (77%) completed follow-up questionnaires. As part of the evaluation, we interviewed a purposive sample of 30 older people. We sampled for diversity in age, gender, education, ethnicity, functional ability and location. Just over half of participants (17/30) were aged between 65 and 74 years, although a spread of older age bands were included (eight people aged 75–84 years and five people aged ≥ 85 years). Around half were male (17/30) and from the urban study site (18/30), two-thirds (20/30) had received only a basic level of education up to the age of 16 years, and the large majority (29/30) identified ethnically as white (white British or white other). In this community-dwelling population, one-third (10/30) needed assistance with one or more IADL (e.g. shopping and cleaning) and 1 in 10 needed assistance with one or more basic ADL (e.g. dressing and washing).

Within the semistructured interviews, we explored views on healthy ageing and identified barriers to and facilitators of health promotion in later life. Interviews were conducted in the participants’ homes and lasted 45–60 minutes. They were audio-recorded, transcribed and analysed for the WISH study using framework analysis.

The WISH study participants described perspectives on healthy ageing and health and well-being promotion in a comparable population to the current study. For pragmatic reasons, we conducted a secondary analysis of these data. 179 The charted data were reread by three members of the research team to identify themes relevant to the intervention development for the new HomeHealth service and to identify gaps in knowledge which could be supplemented by primary data collection in the current study (see Tailoring a new home-based health promotion service for older people with mild frailty: qualitative study of older people, carers and health and social care professionals).

Key findings of importance to service development for mild frailty

Three overarching themes described older people’s views on healthy ageing and health promotion in later life: ‘maintenance as well as change’, ‘recovery as well as decline’ and ‘social connectedness’.

There was a clear message that preventative activities should not focus on change alone, but that maintenance of current health and well-being status was more, or equally, important. Some people had no desire to become ‘healthier’ and prioritised staying as they currently were for as long as possible. To some extent, this reflected the awareness and practise of healthy ageing activities that many older people were already engaged in (i.e. they already felt that they were doing what they could to stay healthy). For others, their current status matched their age-related expectations. A further theme reflected the view that age-related changes were not necessarily about progressive decline; older people described the importance of addressing ‘recovery’ in response to periods of poor health that could be ameliorated, and that health promotion in later life should be able to be tailored to such circumstances.

Psychosocial aspects of healthy ageing dominated the discussions. These included acceptance of and adjustment to changes and maintenance of previous behaviours and roles, including within networks of friends and family, and the importance of staying connected and actively contributing to the world around them:

I think relationships, you know, friends, family, that certainly keeps me going . . . I’m in contact with my former wife and I’ve helped out when the kids needed help, like looking after her; she’s got Alzheimer’s I’m afraid. But that keeps me active and sort of within the family unit, and I think that helps a lot.

Male, 78 years

Facilitators and barriers to the implementation of a new service broadly fell into four key themes: knowledge, attitudes, health status and access, as described and illustrated with quotations in the next sections.

Implications for service development

Key messages from the WISH study to inform the development of a new health and well-being promotion service are:

• Maintenance of current health and well-being status may be as important as change, fitting an assets-based model of ageing. The language used for promoting the service should address this, as a service perceived as being aimed to promote being ‘healthy’ may deter some people from participating.

• A tailored approach is appropriate to meet the diverse individual needs and expectations for health and well-being in later life. Time should be allowed to enable an understanding of individual circumstances (including their knowledge, attitudes, health status and accessibility concerns) and preferences.

• A focus on maintaining independence and staying socially connected is important rather than ‘health and well-being promotion’ per se. This includes addressing mobility needs, concerns about memory loss and cognitive impairment, mood, social networks and identifying gaps in informal support, access including financial concerns and transport, and suitability of the home environment. Pain management and clarification of information on specific topics, such as nutrition and exercise, were raised.

Aims

1. To explore experiences of current health promotion behaviours and support needs in older adults with mild frailty.

2. To identify potential components for a new home-based health promotion service for mildly frail older people.

Methods

We collected qualitative data from a range of stakeholders, including older people with mild frailty as potential new service recipients, carers as people who may support those involved in a new service, and community health and social care professionals (including home care workers) for their expertise in identifying potential content, training required and to ensure ‘fit’ with existing services. We undertook interviews and focus groups (Table 5), audio-recorded with consent and transcribed verbatim. Table 5 shows the sampling method, sample characteristics and data collection methods for each group. Recruitment of older adults and carers continued until the sample was sufficiently diverse and no new major themes emerged. All participants provided informed consent to participate. Topics explored in the interviews and focus groups are summarised in Table 5 (see Report Supplementary Material 4 for topic guides). We collated the data in NVivo 11 (QSR International, Warrington, UK) and used thematic analysis and constant comparison to analyse the data. 184 Further details of the data collection and analysis process are reported elsewhere. 176

Results

Interviews included discussions of current health promotion behaviours, what helped and hindered them in achieving these and, more specifically, around what the content of a new service to promote well-being and independence in older people with mild frailty might include.

Summary of key findings

Within this qualitative study of older people with mild frailty, carers and health and social care professionals, we found that the service should aim to address a broad range of domains (in particular socialising and mobility) at a low level, delivered over a sustained period through mechanisms such as providing information and psychological support. Stakeholders preferred a single non-specialist worker with good communication skills to deliver the service, rather than a particular health-care professional. A focus on independence was advised to increase uptake of services.

Key points from evidence reviews

• Broader, multidimensional interventions targeted to mildly frail populations have not previously been assessed.

• An explicit theoretical basis for behaviour change should be included within a new intervention.

• A new service should include the intervention functions education and enablement and the BCTs adding objects to the environment and instruction on how to perform the behaviour.

• No intervention functions or BCTs should be explicitly omitted at this point.

• A new service should include the following domains:

  • Physical activity and exercise, to improve mobility and reduce falls. Exercise should be individualised, with signposting to group classes where necessary.

  • Support to encourage socialising.

  • Exploring mood and signposting to psychological services or encouraging socialising to improve mental well-being.

  • Nutritional advice/education, particularly relating to vitamin D and protein and energy intake. Other dietary supplements should not be advised.

  • Memory concerns should be addressed (e.g. signposting to GP memory assessments/memory services and strategies to compensate for milder cognitive impairments, for example use of memory aids); however, cognitive training should not be included in a new service as there is currently limited evidence for effective ways of improving this that have an impact on functioning.

Key points from qualitative studies

• The service should cover a broad range of domains and be tailored to the individual. Socialising and mobility are the key areas to address, but aspects such as mood, memory, finances, form-filling, computing, nutrition and environmental adjustments may be important for some older people. Medication advice should not be within the remit of a new service.

• The service should support the older person through providing information, practical and psychological support, building motivation and signposting to existing services.

• The service should have continuity of a single support worker over a sustained period of time. A health professional qualification appears to be unnecessary to deliver this service and could be detrimental, but communication skills are vital.

• To increase the uptake of a new service, it should be promoted in a way that emphasises independence, targets domains valued by older people and focuses on enabling individuals to carry out behaviours that improve their well-being and preserving assets.

Summary of core intervention components integrating theory with evidence from literature reviews and qualitative studies

• Home-based ‘brief’ intervention, up to six sessions over up to 6 months.

• Includes comprehensive assessment followed by individually tailored support/enablement, with goal-setting, monitoring, education, signposting, advocacy roles, with or without provision practical aids.

• Tailored to individuals, ‘asset’-based maintenance of the things they like to do and overcoming barriers to achieving this.

• Likely to encompass mobility, muscle strength/resistance training, falls prevention, home environment, nutrition, social engagement, mood, managing money, other (e.g. vision, hearing). Delivered by ‘mature’ non-specialist, non-health professional.

Consultation with key stakeholders

To ensure that the new preventative service for older people with mild frailty would align with existing local services and structures, we arranged a series of one to one meetings with key stakeholders in the two study areas, including commissioners for older people’s services and managers and practitioners working across health, local authority and voluntary sectors. Although the new service was conceptualised as a NHS service, commissioners highlighted that the potential breadth of its remit meant that it could be funded from an integrated health and care budget, and delivered through third-sector providers. Commissioners shared useful information on contemporaneous local projects and services for their community-dwelling older population, both to help identify eligible older people, for example, implementation of the electronic Frailty Index (eFI) in GP practices, and to prevent any duplication or conflict with existing services for frail older people in each area.

Managers and practitioners were local frailty leads, from the Clinical Commissioning Groups and local authorities and third-sector providers of the local Care Navigation service (London, UK) and local Age UK providers of services, including dementia support workers. They clarified the eligibility criteria and remit of their services, shared information on the job specification, recruitment and caseload of their workers in similar fields (e.g. care navigation, complex case managers and reablement workers), discussed opportunities for sharing information on local resources and staff training, potential colocation and supervision of HomeHealth support worker(s) and referrals both during and after the end of the new service being developed. The discussions were important in terms of the practicalities of service delivery and helped to raise the profile of the new service and establish good local working relationships.

Service development panels

The purpose of the service development panels was to invite comment and debate within groups of key stakeholders and to flesh out and refine the new service under development. Facilitated group discussions share similarities with focus groups as a useful method to encourage participants to share experiences and elicit views that may be harder to express on a one-to-one basis, for example more negative views. 188 We asked for constructive criticism from panel members. Both separate and mixed panels were held for public representative groups to allow the research team to present information already gathered in a tailored manner, for example, using appropriate language, and to create an environment that allowed participants to speak freely. This was particularly important for the panel of older people with mild frailty, who had no prior experience of this type of group working.

Participants and settings

Three intervention development panels were held sequentially with the following groups. Panels 1 and 3 were co-hosted with Age UK London.

Outcomes

Summary

The key learning points from the consultation and intervention development panels are:

• Confidence in our findings from phase 1 on approach and content.

• Who should deliver the service.

• Modified service delivery (e.g. length and number of sessions).

• Boundary setting and clarity of service remit.

• Fit to other services/resources: how it can integrate and be implemented.

• Changes to study materials/leaflets (separating the research from the new service) and the explanations for participants.

Service content

HomeHealth aimed to be personalised to each individual, but cover four main domains: mobility, socialising, nutrition, and mood/psychological well-being. It addressed other domains raised by the individual as needed. Domains considered outside the service remit included detailed information regarding finances, medication management and nutritional or exercise advice for those with complex health needs for whom specialist support might be needed (e.g. diet advice in someone with renal failure).

Behaviour change content

HomeHealth project workers provided information (and signposting when necessary), skills (e.g. exercises), practical and emotional support and encouraged changes in behaviour. In terms of behaviour change content, the project workers delivered education, enablement, training and environmental restructuring as needed as the main intervention functions. These reflected the components derived from the qualitative work and our behaviour change review. In the light of the sparse evidence that we found in our systematic review of BCTs, we did not actively exclude any functions or BCTs from the intervention. As there was little consistent evidence about which BCTs to use, project workers were encouraged to tailor BCTs used to specific barriers that were highlighted by clients. Behaviour change content was primarily delivered through a core set of techniques: goal-setting, action-planning and monitoring progress, including maintenance of behaviours and developing habits. Goals were divided into three types in a hierarchy. From least to most specific, these were:

1. Outcome goal: the overarching goal that the person would like to achieve from the service. This may relate to an outcome of behaviour (e.g. ‘have more energy’). This is developed through discussion at the first appointment. This would then be reviewed according to progress and modified as needed.

2. Behavioural goal: the specific action(s) the older person wants to do to achieve their outcome goal (rather than what they would like to achieve). This goal relates to a specific behaviour, but phrased in general terms (e.g. for having more energy: ‘eat enough calories’).

3. A SMART goal (specific, measurable, achievable, relevant and timely): a detailed action plan that specifies key elements of the context in which the behavioural goal will be achieved (when, where, what and with whom), is measurable (recordable whether or not the goal was met), is achievable (to build motivation), relevant (important to the person) and timely (achievable within a target time).

If relevant, self-monitoring (i.e. recording behavioural progress towards goals), habit formation (repeated performance of the planned action in a specific context) and social and practical support (both from the project worker and through involvement of others, e.g. family members) were incorporated to assist in achieving goals or coping with setbacks.

Delivery

HomeHealth was delivered through approximately six (minimum of three and maximum of 12) appointments at clients’ homes over 6 months. Participants were told that there would be 3–6 appointments initially, with further appointments if needed (up to a maximum of 12 sessions) within 6 months, but an emphasis that this was a time-limited service. This would include:

• First appointment:

  • learn about the person, focusing on mobility, nutrition, psychological well-being and socialising as key domains to address

  • build rapport and trust

  • identify their overall outcome goal from the service

  • if possible, identify a behavioural goal and a SMART goal and form an action plan, taking into account any barriers using the COM-B model or agree an action before the next meeting.

• Subsequent appointments:

  • review goals and progress

  • address problem-solving, coping with setbacks and motivation

  • modify or create new goals as needed and form an action plan

  • maintain changes in behaviour (if applicable).

• Final appointment

  • reinforce self-efficacy

  • reinforce new things they have learned (e.g. exercises)

  • advise on strategies to maintain motivation and continue behaviours after the HomeHealth service stops

  • provide information on further help or support (e.g. Care Navigators, GP, local Age UK).

HomeHealth project workers

The HomeHealth project worker role was created for this project. In line with the recommendations from the developmental phase of the study outlined above, our main criteria were excellent communication skills rather than a specific professional background, previous experience of working with older people in community settings and experience of engagement and person-centred planning (ideally with older people).

Training

The workers were briefed on the development of the HomeHealth service and the rationale and evidence base for it. They received an initial pre-intervention training programme delivered over four half-day training sessions by members of the research team and external experts as appropriate:

• communication skills, low mood and depression, safeguarding

• physical activity, strength and balance exercises

• nutrition in older adults

• delivering and personalising BCTs.

All sessions were led by an expert in that field. This training supplemented the intervention manual and resource packs developed for the HomeHealth service and was designed to provide basic knowledge in the key intervention areas, with signposting and referral recommended in cases that required specialist input. Training was interactive, with role-play, vignettes and demonstration and participation in exercise training, including the use of weights and resistance bands.

After commencing intervention delivery, one further session 2 months later was given by the health psychologist to reinforce the behaviour change content of the intervention, with a focus on overcoming setbacks and maintenance of behaviour change. It further allowed case-based discussion of applying this in a mildly frail population.

Supervision

In line with recommendations from the professionals within the qualitative study and intervention development panels, the project workers were supervised weekly/once every 2 weeks by an expert in older people and communication skills, with interim contact as needed. Clinical input from GPs and experts in nutrition, exercise and behaviour change were available as necessary to answer queries. Project workers were trained in identifying ‘red flags’ (health or social concerns or potential risks to the individual or others) and given protocols for associated action and follow-up in the intervention manual. This normally involved liaison with their accountable GP in the first instance.

Inclusion criteria

• Older people aged ≥ 65 years registered with a participating general practice

• Scoring as ‘mildly frail’ on the Rockwood Clinical Frailty Scale7

• Community dwelling (including extra care housing)

• A life expectancy of > 6 months

• Capacity to consent to participate (including those with dementia or communication difficulties who retained capacity).

We included people unable to speak English, with the provision of translated materials and translators, if required.

Exclusion criteria

Those people who:

• were living in care homes

• had moderate to severe frailty or who are not frail (according to the Rockwood Clinical Frailty Scale)7

• were on the GP register for palliative care or dementia

• were housebound

• were already case managed

• were lacking capacity to consent

• it would be inappropriate to have an invitation to participate for at this time (e.g. because of recent bereavement), as judged by their GP.

Treatment group

The HomeHealth intervention was a manualised home-based behaviour change intervention, targeted at increasing independence and well-being through addressing key areas of mobility, nutrition, psychological well-being and social isolation (see Chapter 3, HomeHealth: intervention description). The service was tailored to each participant, who developed goals and strategies to achieve them in conjunction with the HomeHealth project worker. Each participant received up to six sessions over 6 months, with a further six if considered necessary following a joint discussion with the participant and project worker. HomeHealth project workers were non-specialists recruited specifically for the project, who underwent training in communication skills, behaviour change, safeguarding, nutrition and exercises for mobility. Each participant also received TAU. We have summarised the intervention according to the template for intervention description and replication (TIDieR)200 checklist in Table 7.

Comparator group

The control participants received TAU, which consisted of primary and secondary care treatment that the person would usually receive and any social care, private or family support as needed. No services or therapeutic interventions were prohibited in either group and participants were not asked to maintain particular behaviours. A supplementary objective of the study was to document ‘treatment as usual’ to define this for a full-scale trial. At the end of the study period (6-month outcome assessment) all study participants received four Independent Age201 guides (Your health and the NHS; Advice for later life; Extra help at home; Healthy, happy, connected), an Age UK and NHS England booklet202 (A Practical Guide To Healthy Ageing) and a list of local services collated by the project worker in each area.

Feasibility and acceptability success criteria

The primary objective for this study was to determine feasibility and acceptability, according to the following success criteria:

1. A minimum recruitment rate of 70% of our target of 50 people within 6 months.

2. A retention rate of 80% at 6 months, excluding people with outcomes measured as part of the study (i.e. deaths).

3. A positive evaluation of acceptability to older people (uptake of the service, satisfaction with service).

4. At least no difference between groups on analyses of the candidate primary outcomes (i.e. no negative effects).

Clinical outcomes

We assessed a range of outcome measures for feasibility of data collection at baseline at 6 months. The outcomes collected at each stage are summarised in Table 8.

Self-report questionnaires were administered verbally in a minority of cases for which participants had impairments preventing self-completion (e.g. visual impairment) or for which participants refused to self-complete but were happy for the questionnaire to be interviewer administered. Participants were given £10 vouchers for each assessment.

Our candidate primary outcomes for a full trial included generic measures of well-being and functioning (ADL), health-related quality of life, capability, well-being and psychological distress. In addition, we collected data to calculate the Fried frailty phenotype score (including gait speed, grip strength, physical activity, weight loss and exhaustion). 10 This is a key outcome of interest, but as it has not been validated as an outcome measure it would therefore not be our primary outcome in a full trial.

Frailty measurements

There are a range of potential approaches to defining and measuring frailty, with some consensus around the importance of ‘physical frailty’. 215 The most widely used measurement of physical frailty is the Fried phenotype, which is composed of five frailty characteristics: weakness, slowness, exhaustion, shrinking and low physical activity. 10

Weight loss

Frailty is associated with weight loss as a result of sarcopenia (decline of skeletal muscle tissue with age) and forms part of the Fried frailty phenotype. 10 We assessed weight (seca 875 Class III digital floor scales; seca United Kingdom, Birmingham, UK) and height (portable Leicester height measure; Chasmors Ltd, London, UK) at baseline and 6 months to calculate:

• weight loss as a continuous measure in participants not obese at baseline (BMI of < 30 kg/m²)

• number of participants losing ≥ 3 kg over the study

• number of participants with a BMI of ≤ 18.5 kg/m² at baseline and 6 months.

Participants were asked to remove shoes for both measures and wore light clothing. Data on whether or not weight loss was intentional was not collected, so weight loss could not be used as in the original Fried phenotype. 10 We used low body weight to define this criterion, as did many other previous studies. 221

Other outcomes

We assessed a range of other potential secondary or intermediate outcomes for a full RCT, including other health behaviours (smoking, alcohol), cognition and falls.

Health economic outcomes

As part of the feasibility study, we aimed to determine intervention costs and the feasibility of our planned health economic analyses in a full RCT.

Demographics

At baseline, we collected date of birth, gender, living arrangements, marital status, current housing, educational level, employment, volunteering, receipt of state/employer/private pension and receipt of state benefits. We determined the local area deprivation based on the participant’s postcode, using the Index of Multiple Deprivation (IMD) in deciles. 236

Feasibility outcomes

We descriptively summarised recruitment, participant characteristics and loss to follow-up to ascertain whether or not our success criteria were met and assessed the acceptability of the intervention through questionnaires and interviews as part of the process evaluation (see Chapter 6, Participant questionnaires). We also assessed the feasibility of trial procedures (see Chapter 4, Feasibility and acceptability of trial procedures).

Outcome measures

Demographic characteristics, baseline variables and outcome variables were descriptively summarised, overall and by group. For continuous measures, we assessed outcomes using the following regression models:

• model 1 – baseline value of the outcome measure, area (therapist effect) and treatment group

• model 2 – baseline value of the outcome measure, area (therapist effect), treatment group, age and gender

• model 3 – baseline value of the outcome measure, area (therapist effect), treatment group, age, gender, deprivation (assessed by proxy by receipt of extra pension above state pension) and number of long-term conditions (as a frailty indicator).

For the majority of outcomes, model 2 provided the best fit, particularly as gender was unbalanced between groups at baseline. Results from model 2 are therefore presented in this report. Model assumptions (normality, homoscedasticity and linearity) were checked using histograms and Q–Q plots. We calculated 95% two-sided confidence intervals (CIs) and reported p-values using a 5% (two-sided) significance level. Number of frailty characteristics was compared using Poisson regression, adjusted for the same variables. Categorical variables were analysed descriptively.

All analyses were conducted using intention to treat with all available data. We intended to use established conventions for imputing missing data; however, given the low number of missing data, this was unnecessary. Sensitivity analyses were planned to define unintentional weight loss by restricting weight loss assessment to non-obese participants and to identify participants losing ≥ 3 kg. Other subgroup analyses were not possible because of the small sample size.

Intervention costs

The mean costs of the intervention were calculated per patient, based on a range of assumptions on the case-load, staff turnover and training requirements for scaling up to a full trial or implementation. For the calculation of costs, we assumed that the person delivering the service would be employed under NHS Agenda for Change band 6 (equivalent to the band project workers were employed as within this study), with a caseload of 50 people per year and an average employment duration of 12 months (as high staff turnover is possible). We also calculated costs if the project worker was employed at NHS Agenda for Change band 5 or 7 to inform service commissioners. All costs are in 2015/16 Great British Pounds and costed based on Personal Social Services Research Unit calculations of mean salary, oncosts and overheads by NHS band. 240

We calculated the hourly costs of specialist trainers and supervision at the equivalent level of the providers in this study. Supervision was costed as an hourly rate for a senior staff member (manager) and training was costed based on consultancy rates of £350 per session charged by our external training providers. As the training was delivered to only two staff members in this small feasibility study and this would be scaled up to larger numbers in practice (in both a full RCT and if implemented), we assumed that four service providers would be participating in a single training session for the cost calculations. We combined these data with the average number of minutes per client, average travel time per client and cost of consumables to calculate the total intervention cost per client.

Feasibility and acceptability of collecting relevant health economic data

We report the proportion of missing data at baseline, 3-month (for CSRI only) and 6-month follow-up for EQ-5D-5L, ICECAP-O and CSRI and the completeness of data extracted from medical notes.

Service use

We report the percentage of patients and mean number of contacts for patients who used the service for each type of health, social care and out-of-pocket health-care costs for baseline, 3 and 6 months. These were costed for each patient using costs from the most recent Unit Costs of Health and Social Care 2016 published by the Personal Social Services Research Unit240 and Reference Costs 2015–2016. 241 All costs are reported in 2015/16 Great British Pounds.

We calculated mean cost per patient plus 95% CIs for the HomeHealth intervention and TAU groups by type of service use at 3 and 6 months, adjusted by baseline service use. As the aim of the study is to assess the feasibility of collecting data for a full trial of HomeHealth and given the small sample size, we have not reported incremental cost-effectiveness ratios or cost-effectiveness curves as this would imply hypothesis testing of the cost-effectiveness of HomeHealth compared with TAU, which is not recommended for feasibility trials. 242

Quality- and capability-adjusted life-years

Mean and SDs for utility scores for the HomeHealth and TAU arms are reported based on response to the EQ-5D-5L and the formula developed by EuroQoL. 243 QALYs are calculated as the area under the curve, adjusting for baseline differences using regression analysis. 232

We also calculated mean capability indexes and CALYs using the ICECAP-O and the index values reported by Coast et al. 190

Societal perspective

Societal costs were also included to capture out-of-pocket costs and the impact on unpaid carers. Unpaid carers (family and close others) often provide essential support and care to frail older patients. Their contribution to care needs to be recognised and valued. If not, this can represent an undervaluing of the total cost of care if an unpaid carer provides a significant amount of care for a patient. This has been costed as equivalent to the cost of an hour of face-to-face time of a home care worker of £24 per hour. 240 There was some missing data for how many times people received help per week and how long a typical visit was but, so that a total cost could be calculated if a value was missing, it was replaced with the mean for completed variables greater than zero at that time point.

Budget impact analysis

Budget impact analysis can provide information to commissioners of public services regarding where costs for new services are likely to be incurred and who will see any of the cost savings. We report costs by different commissioning sectors to demonstrate the potential impact on health and social care budgets of commissioning the HomeHealth service.

Recruitment

The trial was feasible in terms of recruitment. We successfully met our recruitment criterion of a minimum recruitment rate of 70% of our target of 50 people within 6 months. We recruited 51 people within a 5-month period, between 16 December 2015 and 13 May 2016, and most were recruited within 3 months (Figure 6). During our peak recruitment period, we were limited by researcher unavailability to undertake baseline assessments, and higher recruitment rates (in a full RCT) could be achieved with more researcher time. We met our target of recruiting 11–14 people per practice, with 25 from our urban site and 26 from our semi-rural site. The first participant was randomised on 5 January 2016 and the final outcome assessment was on 14 November 2016.

FIGURE 6.

Monthly recruitment for the HomeHealth trial.

We sent out a total of 586 recruitment packs from four general practices. Two of these practices sent a single large mailing to 145 people and two practices sent these out in stages. Approximately one-third (33%) of those approached responded, of which two-thirds were positive response (22% of total mailed) and one-third declined to participate (11%). Many of the group that declined reported on reply slips that this was because they had no health concerns/felt too well. After telephone screening by the research assistant, just over half were eligible for a baseline assessment (55%). As seen in Figure 5, the majority of these did not fit the criteria as they were too fit to participate. A small number declined to participate, owing to caring responsibilities, not feeling well enough for the study, felt other services were sufficient or were simply not interested. A small number (n = 4) lived in the same house as another study participant and so, because of the potential for contamination across trial groups, could not be enrolled into the study. Out of the total approached by post, 8.4% were recruited to the study. Two people (both eligible and randomised) were recruited through clinician referral and one person self-referred, responding to a poster within the practice but was too frail to take part.

After baseline assessment, 20 were excluded, mostly as they did not fit the criteria (n = 17 were too fit). Only two people declined to participate: one disliked the baseline questionnaires and one was unwilling to be randomised.

Retention

The trial was feasible in terms of retention. Our retention success criteria (80% at 6 months) were exceeded, as 48 (94%) participants completed the 3-month and 6-month outcome assessments. Three people withdrew from the study, one in the intervention arm and two in the TAU arm. Withdrawals in the TAU arm were reportedly because of caring responsibilities and a dislike of the assessment questionnaires (the person who disliked the assessment questionnaires consented to medical notes data to be extracted for health economic analyses). One participant withdrew from the intervention arm as they were too unwell (having chemotherapy and anxiety disorder) and found the research and intervention paperwork too burdensome in this context.

Participant characteristics

Tables 10 and 11 show the demographic characteristics of participants at baseline.

The sample was predominantly of white ethnic origin, with a mean age of 80 years and with more women (59%) than men. It appeared more affluent than the general population, with fewer people living in more deprived areas, which reflected the deprivation levels of the study practice areas. The majority owned their own homes (70%) and had an additional pension to the state pension (76.5%). Around half (51%) lived alone and just under half (47%) had an education past 16 years of age.

Apart from a considerably higher proportion of women in the treatment arm, there were no other evident imbalances between the two groups. Between the urban and the semi-rural areas, we did not observe any differences apart from a greater number of participants who had been born outside the UK in the urban area.

Baseline functional, physical, psychological and lifestyle measures

Table 12 shows the baseline clinical data for participants. Baseline measurements demonstrate that the sample was largely independent in ADL on the Modified Barthel Index, but were demonstrating some signs of frailty with a lower grip strength and slower gait speed than population averages244,245 and had an average of three or four long-term health conditions. Mental well-being scores were slightly lower than age-adjusted population norms of 51.6 points246 and levels of psychological distress were high, with the mean GHQ-12 score at baseline (13.4) being higher than the accepted threshold for a ‘case’ of psychological distress of 11/12. 246,247 Quality of life (EQ-5D-5L) was slightly lower than population norms248 and capability (ICECAP-O) was similar to population norms. 190,249 In keeping with other data on older populations in high-income countries,250–252 they were relatively sedentary, with low physical activity levels, and were, on average, overweight (mean BMI of 28 kg/m²). There was only one smoker in the sample, but 29% had a high alcohol consumption, with an AUDIT-C score of ≥ 5 points. 253 The baseline MoCA scores suggest that many of those in the sample were mildly cognitively impaired, with a mean score of 23.6 points (a score of ≥ 26 points is considered normal cognition and < 19 points considered indicative of dementia). 189,254 Most variables were similar in the treatment and TAU arms.

Baseline Fried frailty phenotype

Table 13 shows the Fried frailty phenotype10 of participants at baseline. Although all participants met the criteria for ‘mild frailty’ at baseline, using the Fried phenotype21 and applying cut-off points for the bottom quintile of grip strength and gait speed from the English Longitudinal Study of Ageing (own data),225,255 we found that just over of half the sample were frail and only 43% were ‘pre-frail’. This was largely explained by poor grip strength, low physical activity and low energy levels in the sample. No participants were underweight.

Caring responsibilities

In the total sample, just over one-fifth (n = 11) of participants were carers themselves, mostly for their spouses, with an average of nearly 8 hours a day spent caring. This was causing a degree of carer strain, but most (> 80%) felt able to continue longer term (Table 14).

Functional, physical, psychological and lifestyle outcomes

Table 15 shows the participant outcome data at 6 months. Those receiving the HomeHealth service had significantly better functioning (Modified Barthel Index score), better grip strength and improved psychological distress (GHQ-12 score) than the TAU arm. There were no significant differences in any other outcome.

Frailty status

Frailty status improved slightly in both groups compared with baseline scores, largely explained by differences in grip strength and energy levels (Table 16).

There were no differences in the number of frailty characteristics between groups at 6 months (relative risk 1.008, 95% CI 0.67 to 1.54; p = 0.968).

Caring responsibilities

During the 6 months’ follow-up, the number of carers reduced in the treatment arm (as a result of one spouse dying and one with dementia moving to a care home) and increased in the TAU arm. Overall, at 6 months’ follow-up, the majority felt able to continue caring longer term. The number involved (n = 11) is too small to report on changes from baseline or differences between the treatment and TAU arms (Table 17).

Supplementary analyses

We undertook a planned sensitivity analysis to explore the effects of excluding obese people from weight loss calculations. No participant had a BMI of < 18.5 kg/m² at baseline, therefore no analyses were possible for this subgroup. When excluding obese participants (BMI ≥ 30 kg/m², n = 16) from the analysis (as weight loss in this group may be intentional), there was no significant difference in weight loss between baseline and follow-up in the treatment arm compared with the TAU arm.

Intervention costs

The costs of the intervention are summarised in Table 18, using the assumptions outlined in the methods (see Chapter 4, Intervention costs) and based on the scaling up of the service as delivered in our feasibility service. The total average cost of the intervention per patient is £307, assuming the service is delivered by a NHS band 6 employee, when four staff members are trained together and using conservative assumptions that they stay in post for an average of 12 months, with a caseload of 50 people per year (assuming each person has 6 months of contact with the service, so caseload at any given point would be approximately 25 people).

Quality- and capability-adjusted life-years

The QALYs and CALYs are shown in Table 19. At a p-value of < 0.05, adjusting for baseline differences, CALYs were significantly higher in the intervention arm than in the TAU arm (mean difference 0.017, 95% CI 0.001 to 0.031). There were no significant differences in QALYs between the intervention and TAU arms.

Service use

The number of participants using each service is summarised in Report Supplementary Material 7. The total cost of service use per participant in each group is outlined in Table 20. Total 6-month costs were lower in the intervention arm than in the TAU arm (£1650 vs. £2575), largely explained by differences in secondary care costs incurred by participants; a small number of participants had some expensive treatments during follow-up (e.g. pacemaker fitting, coronary bypass surgery), which were included in the analysis. Therefore, costs incurred were highly skewed with a large range, which coupled with the small sample size resulted in wide CIs in both groups. Given the small sample size, rare expensive events that were unrelated to the intervention have the ability to significantly distort the analysis and, hence, limited conclusions can be drawn from this analysis.

Societal perspective

From baseline to 6-month follow-up there was little change to benefits (Table 21). In the intervention arm, one person began receiving carers allowance at 3 months and one person started receiving attendance allowance at 6 months. One person in the intervention arm moved into sheltered accommodation 1 week before the end of the study, therefore the total additional cost of accommodation was £297 in the intervention arm.

There were some missing data for impact on carers with 66 out of the 252 items (26%) across the three time points having at least one value missing. The majority of the missing data were for estimation of unpaid carer time with 52% of items missing data (44/84 variables were missing at least one response). After this was identified at baseline, the wording was changed for four items from ‘how much help do you receive each week’ to ‘how much help have you received over the past 3 months’ at the 3- and 6-month follow-ups with the aim of making the questionnaire easier to complete. Owing to the high number of missing data, mean values have been imputed for missing responses.

Use of care and support services is summarised in Table 22. Unpaid (e.g. family or friends) and private (e.g. cleaners, ironing services) services were more frequently used than state services (including voluntary sector services, such as British Red Cross or Age UK). Care and support were largely accessed for activities such as shopping and cleaning, rather than basic ADLs (e.g. eating, moving around the house and personal care), in line with a mildly frail population. Total care and support costs were similar between the two groups at baseline. The total cost of help from carers at 6 months was £1563 in the intervention arm and £3632 in the TAU arm, with unpaid help accounting for the majority of the costs (£768 in the intervention arm and £2849 in the TAU arm) (Table 23 shows costs). At 6 months, the cost and use of care and support services increased for all private, voluntary and unpaid services across both groups except for unpaid help, which reduced in the intervention group.

Budget impact analysis

We have developed a tool that estimates the cost per patient dependent on a range of assumptions. This could be used in a full RCT to estimate costs and savings for the NHS and local authorities, respectively. To illustrate this tool, we have used data from the feasibility RCT; however, please note that this is presented as an example and not definitive findings because of our small sample size:

If the NHS was to commission the HomeHealth service, and assuming that the intervention is delivered by a band 6 staff member, based on our feasibility study findings, the total cost per patient of HomeHealth would be £307 with savings of £907 per patient in secondary care and £16 in community care. This would equal a net saving to a NHS commissioner of £616 per patient. This is dependent on the number of patients the service is delivered to, with greater numbers reducing the per-patient cost.

If the local government was to commission a HomeHealth service, and assuming the same cost of £307 per patient of the HomeHealth service, the total cost saving of home-based social care is £170 and a total additional cost of £297 for accommodation. This results in a total additional cost to local government of £434 per patient.

As with all figures in this analysis it is important to note that these are based on small patient numbers and hence are very sensitive to costly outliers. For example, removing the one patient in the intervention arm who moved into sheltered accommodation towards the end of the study would make the total cost to the local government of commissioning a HomeHealth service to be £137 per patient. Similarly, there were a small number of patients in the TAU arm who received costly inpatient care, which influenced these findings.

Data collection

Participants’ views on trial processes

Questionnaires with treatment-as-usual arm

During the 6 months of the trial, study participants in the TAU arm reported the following lifestyle changes:

• none (12/20, 60%)

• dietary changes because of medical reasons (1/20)

• starting balance exercises (1/20)

• starting walking daily (1/20)

• joining an exercise group (2/20)

• joining a day centre following GP advice (1/20).

One participant said that seeing the researcher was ‘uplifting’.

Data from Client Services Receipt Inventory for treatment-as-usual arm

We used self-reported CSRI data obtained at 3 and 6 months and GP medical records at 6 months, to assess TAU in the TAU arm with regard to recent use of NHS services, private health services, local authority and third-sector provided services and other privately funded services. Data for both research arms are summarised in Report Supplementary Material 7 and those regarding the TAU arm are briefly discussed below.

Over the study period, TAU participants mainly used primary care GP services and attended outpatient appointments. Within primary care, TAU participants attended an average of four clinic appointments with a GP over 6 months’ follow-up and received an average of 0.7 GP telephone appointments and 0.3 GP home visits. Participants attended an average of 1.5 practice nurse appointments, 0.5 appointments with other practice staff (e.g. a health-care assistant) and 0.26 appointments with a specialist nurse over 6 months. A minority (13%) received a home visit from an older person community team member. Primary and community care mental health service use was rare (maximum of one person for various services).

Within secondary care, participants attended an average of 2 or 3 outpatient appointments over the intervention period. A minority (17%) had an unplanned admission or a day-patient admission. Planned admissions were rare (n = 1). The A&E/out-of-hours urgent care centres were attended by almost one-third of people, usually not by ambulance, usually one or two times over 6 months (a maximum of four times).

Self-report data for other community health services indicated that NHS optician (43%), private podiatry (39%) and NHS dental services (22%) were most commonly used. On average, these were used one or two times in 6 months for those that used it. A few attended a single hearing clinic (17%) or a NHS podiatry appointment (17%). A minority of participants (n = 1 or 2) used each of the NHS psychological therapy services, private and NHS physiotherapy services, private hearing clinic services, private nail cutting services and NHS osteopathy/chiropractic services; when used, physiotherapy and psychological services were used most intensively (an average of four or five times). No participants used NHS counselling or nail cutting services, or private counselling, psychological therapy, opticians or osteopath/chiropractic services. One-quarter took over-the-counter medicines, only 17% took vitamin D and 43% took other vitamins or supplements.

Regarding care and support services, private and unpaid care and support was most commonly accessed over the 3 months preceding outcome assessment. Privately funded care or support services (e.g. cleaners and ironing services) were used by 65% of TAU arm participants for an average of 1.66 hours per week at 6 months. This might reflect the relative affluence of the study practice areas and participants. Unpaid help from friends or family was used by 65% of participants, for an average of 7.4 hours per week. Only 13% had used state care and support services (e.g. Age UK services), for an average of 0.47 hours per week. It should be noted that high SDs were common in these data and may reflect the difficulty that participants had in accurately recording care or support time received, particularly for unpaid care.

Serious adverse events

A total of 10 SAEs in five participants (one SAE, n = 1; two SAEs, n = 3; three SAEs, n = 1) were reported or collected from medical notes throughout the intervention period. In six cases, it was resolved and four were under ongoing management. In the intervention arm, one participant experienced two SAEs, a chest infection leading to hospitalisation and fall within the home arising from dizziness on the stairs during normal activity. These events were judged to be unrelated to the intervention by the principal investigator (a practising GP). No serious unexpected AEs occurred in the intervention arm. Eight SAEs were reported in the TAU arm, including three falls leading to hospital admittance (n = 1 participant), fainting episodes leading to hospital admittance (n = 1), myocardial infarction (n = 1) and chest pain (n = 3 events, n = 2 participants). No deaths occurred in either arm.

Adverse events

Six AEs were reported. AEs in the intervention arm included two falls leading to injury but not hospital admission; these were not considered to be related to the intervention. In one case, the intervention was temporarily interrupted while the participant recovered. In one further participant, the intervention was temporarily interrupted while they were investigated by their cardiologist for an episode of dizziness and chest pain. In one participant, there was exacerbation of knee arthritis considered possibly related to exercises recommended by the project worker (intervention temporarily interrupted).

In the TAU arm, there was one fall leading to injury but not hospitalisation and one episode of chest pain causing a visit to A&E, subsequently diagnosed as gastro-oesophageal reflux.

Recruitment and retention data

In order to determine the ‘reach’ of the intervention, we documented the characteristics of participants for the RCT and, when possible, the characteristics of those who were excluded as ineligible to take part (see Chapter 5, Participant flow and Baseline data). We do not have data on those who did not respond to the letter of invitation from their GP.

HomeHealth service documentation

Our service providers collected a range of data on intervention appointments in order to assess the ‘dose’ of the intervention received, the content of the intervention delivered and their perceptions prospectively on progress of participants during the intervention delivery.

The following data were collected by the project workers.

• Number and duration of appointments (both face to face and telephone), rescheduled or cancelled appointments, interim contact with the service.

• Topics covered and goals (outcome goal, behavioural goal, SMART objective).

• Project workers’ assessment of progress towards achievement of behavioural goals at each appointment. The project worker assessed the individual’s progress towards achieving their behavioural goals at each visit according to the scale: 2 = goal achieved; 1 = some progress made towards achieving the goal; 0 = no progress made.

Using the above data, the process evaluation team (led by CA) calculated the following:

1. Number of appointments per participant and per area and the overall appointment attendance rate, accounting for cancelled and missed appointments (number of completed appointments/total number of scheduled appointments).

2. Number of participants who received fewer than three appointments (minimum intended to be delivered) and explored reasons for this by reviewing the project worker’s notes.

3. The mean rating by the project worker of progress towards participants’ behavioural goals. This was calculated as follows:

Mean progress towards behavioural goals =(1)[(n of behavioural goals score=2)×2]+[(n of behavioural goals score=1)×1]+[(n of behavioural goals score=0)×0]n of behavioural goals set across all appointments for the participant.

The overall mean behavioural goal rating score had values of between 0 and 2. We created three categories to classify participants by level of progress towards goals:

1. mean behavioural goal rating of 1.33–2.00 – good progress

2. mean behavioural goal rating of 0.66–1.32 – moderate progress

3. mean behavioural goal rating of ≤ 0.65 – poor progress.

Participant questionnaires

Questionnaires (see Chapter 4, Feasibility and acceptability of trial procedures, and Report Supplementary Material 6) were posted to participants with a return envelope for self-completion immediately after their 6-month outcome assessment in the feasibility RCT, with one telephone reminder.

We conducted a descriptive analysis of quantitative questionnaire data on the acceptability/satisfaction with the intervention components and a thematic analysis of open questions.

Semistructured interviews with participants

We conducted interviews with intervention recipients and service providers to explore the acceptability and feasibility of the service, contextual factors, content and how the intervention might work, modifications needed and potential barriers to implementation.

Intervention fidelity: audio-recording of intervention appointments

We requested that all intervention appointments were audio-recorded by the project worker for the purposes of checking fidelity of delivery of the intervention (i.e. that it was delivered as intended in our service manual), with consent of the participant.

Recruitment and retention data

The representativeness of the sample included in the RCT was reported earlier (see Chapter 5, Participant flow and Discussion). We have no specific information on those who were approached by letter from their GP but did not respond. However, we can assume that, as the prevalence of mild frailty on the Clinical Frailty Scale is only 13% in older people7 and we were unable to accurately determine frailty status before approaching participants, a large proportion of those not responding would not be eligible, largely as they would not be frail. Therefore, although only 8.4% of those approached by letter were recruited for the study, it may be that only a small proportion of those not responding would have been eligible. Only one person declined to take part as they did not want to be randomised, but it is possible that more people did not respond because they realised only half would receive the service. Of those who agreed to take part, most of those randomised to receive the intervention went on to engage with the service. Reasons for non-engagement are outlined in Number of appointments and appointment attendance rate.

HomeHealth service documentation

Topics covered by the HomeHealth service

Nearly all (25/26, 96%) participants identified at least one outcome goal in their first or subsequent appointments. One participant did not identify any goals and withdrew from the study after the first appointment. More than half of participants (16/26, 62%) identified more than one outcome goal/topic to work on during the service.

Mobility and physical activity was the most popular goal, chosen by 19 out of 26 (73%) participants. Other domains covered were environment (e.g. home adaptions, decluttering their home), psychological well-being, socialising, finances, diet, transport, memory, supporting carers, and dealing with incontinence or sensory impairment. The number of goals set for each domain are summarised in Figure 7.

FIGURE 7.

Distribution of topics covered by the HomeHealth service (over total number of outcome goals/topics covered).

Fidelity of delivery: analysis of audio-recorded appointments

Analysis of a random 10% sample (13/126) of appointments with the HomeHealth service showed that, in total, 119 out of 165 (72.1%) of items on the checklists that could be assessed by review of appointment transcripts were completed during appointments. By chance, 3 out of 13 randomly selected audio-recorded appointments belonged to a single participant with cognitive impairment, for whom using a behaviour change goal-setting approach was very difficult and the project worker chose instead to use a different, more open and case-management approach. There was one other appointment with low fidelity from a different participant, which was a brief telephone final appointment.

Stratified analysis of this small random sample of transcribed appointments showed that the fidelity of delivery was moderate to high (77%) (range 30–100%) excluding the individual with cognitive impairment, and it was modest (53%) (range 38.9–70%) in the person with cognitive impairment across the three recorded appointments selected.

Many of the items that we have conservatively recorded as ‘not done’ in our fidelity analysis above were originally scored as ‘not appropriate to be done’, but these two categories were merged in analysis as differentiation between them for several items was difficult and subjective. A limitation of the fidelity assessment method that we used is that some items that were not delivered in one randomly selected appointment may have already been covered in other appointments. As we randomly sampled appointments and not participants, we were unable to track if items were covered over a series of appointments; therefore, our estimated fidelity score above may underestimate the true fidelity.

These findings indicate that further refinement of the appointment and fidelity checklists should be considered in order to reflect the flexibility in approach in which a goal-setting behaviour change approach is less applicable. For fidelity analysis in a full trial it may be more suitable to randomly sample participants not appointments, and assess fidelity across all appointments for each randomly selected individual.

Questionnaire data

Questionnaires sent to all intervention arm participants completing the study assessed overall acceptability of the HomeHealth service. Questionnaires also assessed acceptability of the study procedures (see Chapter 5).

Semistructured interview findings

In this section, we report the findings of the semistructured interviews that aimed to explore participants’ motivations for participating in the study, their experiences of the service and acceptability, potential processes of behaviour change and the context in which the service was delivered, as well as providers’ experiences of delivering the service and how it could be modified.

Evidence reviews

Our systematic reviews identified only limited evidence for interventions targeted at mildly frail older people and these appeared to mostly be single domain (exercise/physical activity). Past interventions did not appear to be based on qualitative developmental work incorporating ‘user’ needs or include explicit behaviour change theory and content. There was a limited evidence base for which BCTs have most potential for effectiveness in an older, frailer population. The most promising components we identified were the intervention functions of education and enablement and the BCTs ‘adding objects to the environment’ (e.g. providing aids) and ‘instruction on how to perform the behaviour’. Our state-of-the-art review suggested that physical activity and nutrition had the largest evidence base, whereas home-based social and mental health interventions for frailer older people had limited evidence. Evidence of effective ways of improving cognition that made a difference to clinical outcomes was lacking. Our policy review found that although there was widespread recognition of the problem of the needs of an ageing population, there was little explicit policy on how to address the needs of older people with mild frailty, or prevent worsening frailty in those who were starting to become frail.

Qualitative studies

Qualitative interviews and focus groups suggested that a range of domains needed to be addressed in a new service, including mobility and social activities as core areas but also addressing others tailored to the individual. It further suggested that it should be delivered through low-level support over a sustained period by a trained non-specialist worker with good communication skills. The importance of maintenance of current assets was noted and not necessarily becoming ‘healthier’. The priority for older people was remaining independent and actively contributing. A range of barriers and facilitators was noted, including accumulating health problems (physical, psychological, memory), resources and social capital, perceptions of ageing and willingness/ability to adapt to health declines. Our theoretical framework incorporates these issues, with an assets-based (vs. deficits) model aiming to promote maintenance of independence,186 and the selection of goals to optimise while addressing barriers to compensate for,187 considering the person’s capability, opportunity and motivation to make changes. 51 Our qualitative work further identified potential difficulties in engagement of older people with mild frailty, some of whom felt that accepting help of any kind may indicate the start of dependency and precipitate a decline.

Intervention development

Using evidence from our reviews, qualitative work and theoretical framework we developed our intervention, through a series of intervention development panels that refined the content, and specified the model of service delivery. We also set out to address approaches to engagement of older people with mild frailty, with a focus on language that should be used. From this, we developed a home-based health promotion service, targeted to mildly frail older people and delivered through 3–12 appointments over a 6-month period by a trained non-specialist support worker with good communication skills and previous experience of working with older adults. The service encompassed addressing mobility, socialising, mood and nutrition, and other domains as identified by participants. It would use an explicit behaviour change approach, with goals identified by the individual, focusing on assets that they would like to preserve as well as maintaining independence.

Feasibility randomised controlled trial

We completed a feasibility RCT, testing the feasibility and acceptability of this new service and the feasibility of undertaking a full-scale RCT. This demonstrated that an individually randomised controlled trial was feasible, with recruitment exceeding our targets and very high retention (94%) over 6 months despite this being an older, frailer population. Our battery of outcome measures took some time to complete for frailer participants though was feasible and data were of good quality with very little missing data. It was also feasible to collect health economic data from both NHS and societal perspectives. There was no evidence of contamination of the TAU arm, suggesting that individual randomisation is feasible; however, maintaining blinding for the outcome assessor was difficult in this population, as many had some degree of cognitive impairment and accidentally disclosed their status.

Our study population all had ‘mild frailty’, defined using the Clinical Frailty Scale7 (i.e. they may increasingly need some support with IADL, but are otherwise largely independent in personal care); however, they were classified as a mixture of ‘pre-frail’ and ‘frail’, measured using the Fried frailty phenotype at baseline. 10 They had low gait speed and grip strength for their age, high baseline levels of psychological distress and many had some degree of cognitive impairment. These findings suggest that although only 8.4% of those approached by letter from their GP to take part were eligible and randomised, it had successfully identified a group potentially in need of some support and not the ‘worried well’.

Clinical outcomes for our RCT at 6 months showed no harmful effects and there were no intervention-related SAEs during the course of the study. The intervention shows some promise, in that there were significant differences demonstrated favouring the intervention in our main functioning measure, the Modified Barthel index, in grip strength (a measure of muscle strength and frailty) and in psychological distress, measured by the GHQ-12. All other outcomes showed no significant differences. These findings should be interpreted with caution as the sample size was small and not powered to show differences, but they do provide support for the need to now evaluate this intervention in a large definitive RCT.

Our health economic analysis showed a significant difference in CALYs favouring the intervention, but not in QALYs. The costs of the intervention were modest, at around £307 per patient, using assumptions on the scaling-up of the service, which is slightly lower than a course of CBT currently in the NHS. There were lower NHS costs and carer support costs in the intervention arm, but a large variability and this may be a chance finding. We did not conduct a full health economic analysis, as this is a feasibility study with a small sample size, but we demonstrated this would be feasible alongside a full trial. Further consideration may also need to be given to the time horizon of the analysis given that delaying or reducing a decline in functioning in frail adults potentially has lifelong quality of life, capability and cost implications.

Process evaluation

The process evaluation demonstrated that the intervention was acceptable to participants, who, in general, engaged well with the service, found it helpful, made good or moderate progress towards their goals, and would recommend it to a friend or family. A minority did not engage with the service, for a range of reasons including being ‘too fit’, concurrent complex medical/psychological problems or not liking the approach of goal-setting. The intervention was delivered as intended for most people, incorporating a behaviour change approach, which was well received by most and was endorsed for being motivating and enabling progress and a sense of achievement. However, for some this was less suitable, for example, the small minority who did not like a goal-setting approach and those with significant cognitive impairment. A counselling, reflective listening and emotionally supportive role was also considered an important feature of the service, as was an enablement function, which included practical support in overcoming barriers and promoting capability and opportunity for change. Potential improvements to the service included greater clarity on what it includes at recruitment/initial engagement, more explicit tailoring/alternative approaches to goal-setting when this is less suitable, further carer involvement when there is cognitive impairment, and suggestions for additional training for project workers. An important finding was that the HomeHealth service filled a gap in services for many older people participating who were beginning to encounter increasing difficulties and who did not know where to seek help from.

Published

Jovicic A, Gardner B, Belk, Kharicha K, Iliffe S, Manthorpe J, et al. Identifying the content of home-based health behaviour change interventions for frail older people: a systematic review protocol. Systematic Reviews 2015;4:151.

Gardner B, Jovicic A, Belk C, Kharicha K, Iliffe S, Manthorpe J, et al. Specifying the content of home-based health behaviour change interventions for older people with frailty or at risk of frailty: an exploratory systematic review. BMJ Open 2017;7:e014127.

Frost R, Belk C, Jovicic A, Ricciardi F, Kharicha K, Gardner B, et al. Health promotion interventions for community-dwelling older people with mild or pre-frailty: a systematic review and meta-analysis. BMC Geriatrics 2017;17:157.

Frost R, Kharicha K, Jovicic A, Liljas A, Iliffe S, Manthorpe J, et al. Identifying key components of a new home-based health promotion service for older people with mild frailty: a qualitative study. Health Soc Care Community 2017; in press. 176

Submitted (under review)

Drennan V, Walters K, Avgerinou C, Gardner B, Goodman C, Frost R, et al. Moving upstream in health promoting policies for older people with mild frailty in England? A policy analysis. Health Serv Res Policy.