Cognitive behavioural therapy-based intervention to reduce fear of falling in older people: therapy development and randomised controlled trial the Strategies for Increasing Independence, Confidence and Energy (STRIDE) study

Physical treatments for fear of falling

The optimal management strategy for fear of falling and its adverse physical and psychosocial consequences is poorly understood. Clinical and laboratory observations show that concerns about falling have a clear effect on gait patterns in older people. For example, experimental data in older people who are undergoing gait and balance studies on elevated walkways show disproportionately slow walking speeds and other dysfunctional gait adjustments17 alongside abnormalities in postural balance compared with younger subjects. 18 These experimental data and the observation of higher rates of falls and physical frailty in older individuals with fear of falling suggest that physical interventions may help to ameliorate fear of falling. Much of the research focusing on physical treatments includes home- and community-based exercise interventions, t’ai chi and multifactorial interventions aimed at reducing fall rates, with fear of falling reported as a secondary outcome in the majority of these studies. 8 A meta-analysis of nine studies examining t’ai chi in the management of fall prevention, fear of falling, and balance in older adults concluded that insufficient evidence existed to recommend such an intervention in this context. 19,20 A more recent randomised controlled study in 176 elders, randomised to one of three groups (intensive t’ai chi with cognitive–behavioural strategies, t’ai chi alone and a usual care control group) showed improvements in fear of falling as measured by the Falls Efficacy Scale13 (FES) in the cognitive strategies group compared with the other two groups. 21 So, although t’ai chi can help to prevent falls in older adults,22 its role specifically in the management of fear of falling is less clear.

A previous systematic review found 12 high-quality randomised controlled trials (RCTs) that included fear of falling among the outcomes that they assessed,8 but only one of these trials was primarily aimed at reducing fear of falling. 23 The interventions were conducted across a variety of settings – home-based exercise, community t’ai chi and home-based multifactorial interventions – and all improved fear of falling. 8 In a subsequent trial, a geriatric outpatient-based multifactorial intervention study found no such benefit. 24 More recently, a Cochrane review of exercise (three-dimensional exercise, such as t’ai chi and yoga, balance training or strength and resistance training) for reducing fear of falling in older people living in the community found that exercise ‘probably’ reduces fear of falling to a limited extent immediately after intervention, with inadequate evidence of an effect in the longer term. 25 The authors called for further evidence, with priority being given to the establishment of core outcomes, including fear of falling, in all trials of exercise intervention in community-dwelling elders. 25

Psychological treatments for fear of falling

Although such physical interventions may be of benefit in selected populations, the profile of the disorder and its psychosocial complications suggest that well-designed psychological interventions may help ameliorate the fear of falling more definitively. Several studies have examined an explicitly cognitive–behavioural therapeutic approach in fear of falling in community-dwelling elders, or used cognitive–behavioural therapy (CBT) techniques as part of a wider multifactorial intervention strategy. CBT is an evidence-based form of therapeutic intervention, which seeks to alter difficult emotional and physical states by altering the behaviours and cognitions that maintain them. As used in the CBT interventions for fear of falling studies described below, fear of falling is usually conceptualised as an anxiety disorder maintained by negative reinforcement (avoidance behaviour) and fearful cognitions. Tennstedt et al. ’s26 Matter of Balance study assessed the ability of an eight-session, 4-week group cognitive–behavioural approach with exercise instruction to improve fear of falling and related activity restriction. A total of 434 community-dwelling elders aged ≥ 60 years were randomised to intervention and control groups, with significant differences seen in fear of falling as measured by the FES13 and activity during follow-up. The magnitude of improvement in FES scores attenuated over time, prompting the authors to suggest that a booster session should be used in future studies and in clinical practice. 26 Clemson et al. 27 similarly used what they described as a ‘small-group learning environment’ (although, in practice, some of the methods used included cognitive–behavioural techniques) of 12 individuals per group for 2 hours per session over 7 weeks to improve self-efficacy and reduce falls. 27 The intervention incorporated a variety of learning strategies to facilitate behaviour change, including education regarding exercises to reduce falls risks, medication and home environmental review, and medication management. 27 There was a 31% reduction in the number of falls [relative risk 0.69, 95% confidence interval (CI) 0.50 to 0.96; p = 0.025] in the intervention group, although, interestingly, there was no corresponding change in FES scores. 27

More recently, Zijlstra et al. 28 conducted a RCT of a multicomponent cognitive–behavioural group intervention in older community-dwelling elders. Five hundred and forty participants were drawn from a random sample of 7431 individuals sent questionnaires who reported ‘at least some fear of falling’, although the method of assessment was not specified. Following randomisation, the intervention group underwent a structured 2-hour group CBT intervention (CBTi), based on the investigators’ previous work, once weekly for 8 weeks, with a booster session 6 months following the last session. A primary outcome was not specified and, although a power calculation was supplied based on a difference of 2.5 points, the authors fail to specify on which scale. The beginning of the trial predates widespread use of the FES–International version (FES-I)29 (advocated by the same group as the most appropriate measure for such studies),8 instead using a single-item question on fear of falling as well as an unspecified scale, probably the original FES from the description and reference supplied. 28 Other outcomes included perceived control over falling and daily activity as well as falls. There were no measures of physical function despite the prior evidence base suggesting improvement in fear of falling with the exercise-related interventions as described above. All outcomes showed significant differences between control and intervention groups at 2 and 8 months’ follow-up, with between-group differences persisting at 14 months in fear of falling and perceived control over falling, but not in the other outcome measures. There was a 30% attrition rate in the intervention group and 19.6% attrition rate in the control group. 28 The study intervention was carefully developed and grounded in cognitive–behavioural theory, but interpretation of findings is hampered considerably by the lack of clarity on sample size calculation and outcome measures, and the absence of generic quality-of-life measures and measures of physical functioning. The same research group later reported a before-and-after design study20 implementing a similar protocol into routine health care in 125 community-living older people, with significant improvements in concerns about falling.

The relative effectiveness of group-based psychological interventions thus remains uncertain. A theoretical re-examination of models of fear of falling and a recent cohort study of 500 older adults both suggest that the fear of falling population is a complex and heterogeneous one, in which psychosocial and physiological interventions also need to be individualised. 30,31 This suggests that individually based interventions may be more appropriate but, to date, trials have examined only group-level interventions.

Inclusion and exclusion criteria for patients

Consecutive patients aged ≥ 60 years of both sexes. Those with significant fear of falling, as defined by a FES-I score of > 23,40 were eligible for the study. Patients with cognitive impairment [Mini-Mental State Examination (MMSE) score of < 2441] were excluded from the study, as there is a paucity of data on the use of fear of falling measures in this patient group, and given the difficulties of conducting CBT interventions in those with significant cognitive impairment. Regionally, the population under study had a particularly low proportion of ethnic minorities. Therefore, given the level of communicative nuance needed for the development of a CBT intervention, we excluded those who did not speak English. To avoid participant burden, patients recruited to the CBTi development study were excluded from the Phase I process evaluation work (described later in this chapter) and vice versa.

Sample size

Fear of falling is one of a range of falls-related psychological concerns,42 with emotional and behavioural components including anxiety and activity avoidance, respectively. 43 In the CBTi development study we anticipated interviewing between 10 and 12 patients, although up to 30 were allowed for if data saturation was not reached with fewer numbers. Our rationale for this was that clinically significant anxiety (which fear of falling to be can be a manifestation of) tends to have a fairly narrow range of maintaining factors, which, most usually, are catastrophic beliefs and activity avoidance. The psychological construct of fear of falling as it appeared in the literature suggested that a similar range of maintaining factors was likely to be in operation in our patient group. We therefore expected that the number of patients needed to consistently identify such factors to be relatively small. In fact 15 patients were needed as the construct was more heterogeneous than we anticipated.

Identification, recruitment and consent of participants

The initial approach to patients for the CBTi development study was by a member of staff at the North Tyneside Falls Prevention Service. The staff member provided a brief oral explanation of the research to those patients who met age, FES-I score and cognitive impairment eligibility criteria. Those expressing a preliminary interest were given a Patient Information Sheet (PIS) and an expression of interest (EoI) form (with stamped addressed return envelope addressed to the University) to take home and discuss with family and carers as they felt appropriate. There was no further direct study-related contact with the patients unless they returned the EoI form indicating their willingness to participate in the study. On receipt of an expression of interest form, a telephone call was made by the interviewer (VD) to discuss participation. Questions about the process were actively sought, and if patients required further time for reflection or discussion with family or carers this was encouraged. If patients consented, the interviewer (VD) contacted them by phone and arranged an appointment for the interview. Written consent was obtained prior to the commencement of the interview. These were conducted in the patient’s preferred location, at either at the North Tyneside Falls Prevention Service or, in most cases (14/15 patients), at the patient’s home.

Interview format

The CBTi development interviews were undertaken by VD, a qualified cognitive–behavioural therapist and practitioner health psychologist. The main form of data collection was notes taken by the interviewer as the interview was being conducted, and notes and reflections made in collaboration with the interviewee. All sessions were also audio-recorded in case there should be any need to clarify points from the notes.

The format was a standard CBT assessment, as used by cognitive–behavioural therapists and clinical psychologists. It brings to light which factors are key in having caused the patient’s problem, and which factors are key in keeping the problem going. The structure is often referred to as the ‘three P’s’: what factors predisposed the patient to the problem (such as a history of anxiety); what factors precipitated the current problem (such as a fall or loss of physical capability) and what factors might be perpetuating the problem in terms of cognitions, behaviours, emotional, physical and social factors. Using this model as a template, we investigated the relevant factors in the cause and maintenance of the individuals’ fear of falling.

The interview began with a general open question about the main presenting problem, for example ‘tell me a bit about your concerns about balance and falls’. It then asked more detailed questions about how the problem was affected by different situations, times, people, places and events. Then more detail was sought on what could and could not be done as a result of the problem, and on what made the problem better or worse. Coping strategies and beliefs were then asked about, as well as any current treatments. Next the onset and development of the problem was explored, from what was happening in the interviewee’s life just before they started being concerned about falls, up to the present moment. More general questions were then posed about the interviewee’s current mood and mental state, and about the type of person they perceived themselves to be. This was followed by some questions about any past experiences of emotional or physical problems. Finally, a brief family and social history of family, spouse or partner, carers and social contacts was taken, as well as details of previous and current employment and relationship status as appropriate. The interview ended with the interviewer feeding back to the interviewee some of the information that might help make sense of their fear of falling problem, and with a general discussion of what the interviewee thought of as being the important factors in the cause and management of their problem.

Analysis

The cognitive–behavioural framework (of predisposing, precipitating and perpetuating factors) described above further distinguishes between physical, emotional, cognitive, behavioural and social factors in each domain. Finally, it looks for interactions between the perpetuating factors (such as how anxious thoughts may increase physiological arousal, and vice versa). As a first step of analysis, collected data were sorted into this framework, then their perpetuating factors were identified, and their possible interaction was discussed with participants to attempt an individualised formulation of their fear of falling. This analysis was discussed with them at the end of their interview, and their input was used to develop and refine it. As a last stage of analysis, the interviews were written up as case formulations for each interviewee in terms that attempted to model the onset and maintenance of their condition in terms of the cognitive–behavioural framework. Finally, comparisons were made between individual interviews and a tally of common factors and themes in different domains were made. Factors, and themes and patterns of interaction, were analysed to see if there were recurrent themes between interviewees with significant fear of falling.

Primarily psychological factors

About one-third of the interviewees conformed to the prevailing picture of fear of falling in the medical literature whereby fear, often but not always occasioned by a fall, is maintained by avoidance of activity, leading to loss of confidence, physical weakening and more fear of falling. However even in these cases, the picture was more complicated than this. Interviewee 1, who we will call Amelia, was the closest to the picture that we see in the literature. Her problem had begun 4 years prior to the interview when she had tripped and fallen on her face. Although nothing was broken, she vividly recalled the sense of being out of control and the ‘concrete coming up to meet my face’ (all quotes are directly from participants). Subsequent to this she had become very cautious about walking outside. She altered her walking in that it became much slower and she would almost continually look down. If possible she would ‘link up’, i.e. link arms, with a family member or a friend and increasingly took to either not going out on her own or sticking to areas that she knew well. To refer back to the CBT framework, these behaviours were reinforced and maintained by cognitions. She had come to perceive walking as dangerous, she had the image of the previous fall often in her mind when walking, and she was very aware of the fact that she could no longer ‘stride out’ as she used to (it was Amelia’s story that inspired the name of the intervention). These cognitions and behaviours in turn affected her mood. She was generally anxious when walking, and her mood was low (though she was not depressed) because her life contained much less enjoyment than previously. Her inability to ‘stride out’ and her lack of independence in turn generated a fair amount of frustration about her loss of ability. Physically, she reported being largely OK, but more tired than previously, and specifically mentioned that this type of walking (cautious and slow) was more tiring than how she used to walk, and also much less enjoyable.

As such it was clear from this interview that in Amelia’s case, there was a traumatic precipitating event, although with no obvious predisposing factors. This was the case with most participants, where there was no past history of anxiety disorders. Furthermore, there was a clear interaction of physical, affective, cognitive and behavioural factors that were further reinforced by social isolation (she lived on her own). In CBT literature, the interaction of factors is often presented to the client in the form of a ‘hot cross bun’:44 Amelia’s hot cross bun is presented in Figure 1.

FIGURE 1.

Amelia’s hot cross bun.

Thus, it was clear that Amelia’s fear of falling was maintained by safety and avoidance behaviours, and by anxiogenic and catastrophic cognitions. As such, the obvious intervention would be to target these cognitions and behaviours with the appropriate evidence-based techniques.

However, she was one of the few participants who so neatly fitted this fear/avoidance model. Even when the factors maintaining the inactivity and avoidance were psychological, they were not always primarily to do with anxiety about falling. For instance, one participant was also largely physically capable, although he benefited from walking with a stick. However, he led a largely sedentary life because he was very self-conscious about being perceived as ‘old’. Although self-deprecatingly dismissing his own concerns as ‘vanity’, he nevertheless was clear that the main barrier to him going outside was that he felt very overlooked by his neighbours, and did not want to be seen as the old man with a stick. He was further restricted by social isolation and lack of companionship, and for him the main thing he thought would help would be someone to walk with. Another participant reported that since losing her best friend 2 years prior to the interview, she no longer saw the point in going out. The interview uncovered an interplay of ongoing grief and low mood, which were the main impediments to her engagement with the outside world.

Primarily physical factors

For another proportion of the interviewees, the primary impediments to walking were physical. For instance, one participant, whom we will call Fred, had multiple physical problems. He was overweight, had angina and breathlessness, a frozen right shoulder and a ‘shattered’ left shoulder (from a previous fall). As a result of pain and arthritis in his knees and ankles, he could walk only with sticks, and then with considerable pain. All of these physical ailments had left him severely limited in his physical capacity. He felt very vulnerable when out walking, as he was aware of being only one fall away from being bedbound or hospitalised. Although this latter anxiety was certainly a barrier to activity, he perceived his principal barriers to be his breathlessness, pain and fatigue. He also reported what emerged as a recurring fear among the cohort – that hospitalisation was a ‘one-way ticket to death’ – ‘once you go in you don’t get out’. Although he worked hard to maintain a social life, his activity was haunted by this fear of loss of independence and ultimate death. His ‘hot cross bun’ is presented in Figure 2.

FIGURE 2.

Fred’s hot cross bun.

About one-third of the cohort reported similar issues, with pain and fatigue, as a result of other physical problems, being the most commonly reported barriers to activity, rather than primarily fear.

‘The stoics’

The third group were a surprising finding. These were participants who had considerable psychological and/or physical impediments to activity but who, nevertheless, continued to be active. We perceived an element of this in Fred above. Despite considerable physical problems and very real fears, he continued to pursue his hobby of photography, and he saw this as a ‘passion that keeps me going’. Another more marked example comes from a participant whom we will call Violet. She was the oldest, and perhaps the most physically frail of the people interviewed, and had a very unsteady gait. She also highlighted the heterogeneity of fear-of-falling precipitants. She had become aware of her unsteadiness only when her husband of 60 years had died, and she was no longer able to walk arm in arm with him. Now that he was gone she became aware that she would ‘suddenly list’ and ‘stagger unpredictably’. She had also seen her husband decline very rapidly after hospital admission and die there. Like Fred, she had the fear that admission to hospital was a ‘one-way ticket to loss of independence and ultimate death’. Despite both these physical and psychological ‘inducements’ to inactivity, Violet was the most active participant interviewed (Figure 3). Indeed she was difficult to pin down for interview as she was on (solo) holidays to Spain, Italy and the Netherlands during the study period.

FIGURE 3.

Violet’s hot cross bun.

Violet’s experience was not unique. Several other participants reported ‘keeping going’ despite obvious reasons to stop. This provided a new avenue of enquiry for these interviews: ‘What keeps you going?’. For each person it was different. For Violet it was the belief that ‘you are only here once’. She did not believe in God or an afterlife, and so was committed to get the most out of her life while she still could. Seeing her husband decline, although still a source of grief for her, was a further motivation to make the most of the time she had left. Another participant also struggled with both illness and anxiety, but kept going because she was committed to looking after her sisters, both of whom were housebound. She also had a dog to which she was committed to walking daily. In these cases it was clear that there was something that mattered more, a value that superseded the urge to avoid physical or psychological distress, and allowed the latter to be tolerated. We encouraged the interviewees to reflect on this, with the express purpose of helping us to help others with our intervention.

Duration of health-care assistant training

The training was delivered by VD and NSa over a period of 5 days in a classroom at the University of Northumbria.

Training content

The training gave an overview of CBT and then looked in more detail at the three P’s (predisposing, precipitating and perpetuating) model. Central to the HCAs’ training were the skills of engagement with the client, and assessment and formulation development. Standard CBT interventions for identified problems were described and practised: graded exposure for anxiety; activity monitoring, graded activity and behavioural activation for pain, fatigue and low mood; and sleep management for fatigue. For other recurrent issues HCAs were taught specific cognitive techniques, such as identifying thoughts; making explicit the links between thoughts, feelings and behaviour; thought diaries; thought challenging; and cost–benefit analysis for catastrophic and otherwise unhelpful cognitions. Furthermore, the HCAs were taught how to structure assessment and treatment sessions by working from formulation through to treatment plan in the first one to two sessions, then to specific treatment strategies in the middle three to four sessions, and then focusing on consolidation and discharge planning in the final sessions. An eight-session format was agreed as being most suitable in terms of intervention replicability, with the aim of weekly delivery.

The treatment plan was guided by end of treatment targets, which were worked out in collaboration with the clients, meaning that each client had specific and measurable goals that they were working towards throughout treatment. These individualised and personally meaningful goals aimed to capture that element of ‘valued activity’ that we found had kept the ‘stoic’ participants going, despite physical and/or psychological distress. The weekly implementation of treatment plans was carried out using specified interventions for specified targets in the form of client homework. Subsequent sessions were structured by homework reviews and problem-solving any additional barriers that had been identified before agreeing further target-setting for the next session. Finally, treatment review, skills embedding, and relapse prevention and preparation for discharge shaped the final sessions (Table 1).

Training delivery methods

A mixture of chalk and talk, observation, modelling and practice of the above techniques by the HCAs, as individuals and in pairs, formed the bulk of the training sessions. A strong emphasis was placed on ‘real play’, through which the HCAs were encouraged to use their own anxieties and experiences to work on in assessment, formulation and treatment. This was done with consent, and participants agreed that it helped to bring the techniques to life. Real fear of falling cases, derived from the interviews outlined above, suitably anonymised, were used to present the range of issues involved. The content of the training was summarised in both therapist and participant manuals (available from SWP) and the HCAs were encouraged to develop their own assessment and treatment props, which they did (see Appendix 2).

Post-training supervision

In the months following the above training the therapists engaged in both individual and group supervision. The former was of 1 hour every fortnight with NSa within which the focus was on skills development and consistency of intervention delivery. Individual client session audio tapes were shared and scrutinised with the supervisor. Discussion about internal and external challenges to the process of the study were also explored, as were individual issues that were pertinent. Group supervision was led by VD on a bimonthly basis, and focused on cases brought by the therapists for sharing and exploration of practice and interventions.

Interviews with patients and carers

A sample of 14 patients (aged 60–85 years, nine female) and two family members (both female, age not specified) were interviewed. In contrast with the intervention development interviews, these explored patient and family members’ perceptions about the proposed CBTi, practical aspects of the CBTi (including how to describe the intervention, proposed delivery) and perceived relevance of the CBTi to their situation. Patients were recruited from the North Tyneside Falls Prevention Service to reflect inclusion criteria for the RCT, using maximum variation sampling (for variety in age, gender and FES-I score). When appropriate, family members of these patients were invited to take part. Participants were interviewed in their own homes.

Interviews with professionals

The six professionals (three geriatricians, two physiotherapists and one HCA) who constituted the multidisciplinary team at the North Tyneside Falls Prevention Service were interviewed. VD and NSa (hereafter psychologists) were also formally interviewed using a semistructured interview schedule (three interviews). These interviews explored perceptions of key factors that might impede or facilitate the proposed CBTi being effective and workable in practice.

Observation

Clinical sessions at the North Tyneside Falls Prevention Service (n = 5) were observed to gain an understanding of the clinical context of the study. Professionals sought verbal consent from patients and any companions for the observation. No audio or audiovisual recordings were undertaken of these observations.

The regular meetings of the CBTi development team (VD and NSa) were audio recorded and observed by CB (n = 9 meetings). These data were supplemented with field notes that were taken of informal discussions (n = 8) with VD outside these meetings. Other related stakeholder engagement activities, including a briefing meeting for clinic staff, were also observed.

Data analysis

All qualitative data were analysed thematically using the constant-comparative technique. 46 This process allowed for the meaning of the data, and themes represented within them to emerge freely without the constraints that might be imposed on the data if coding to a prespecified coding frame. Themes were identified individually by CB and TF, who developed initial ideas and undertook coding independently, and then came together to discuss and compare ideas. We systematically worked through different data sources (professional interviews, patient interviews, intervention development meetings, interviews and informal discussions with the intervention development researchers, and field notes) identifying themes separately for each type of data prior to producing an overarching coding frame. The team members responsible for intervention development reviewed and commented on the coding frame thus providing some respondent validation. There were no formal opportunities for respondent validation by patients and carers; however, the identification of similar themes in the intervention development interviews conducted by VD, and those conducted as part of the process evaluation, provided some evidence of trustworthiness.

The process evaluation component of the study was designed not merely to evaluate the development of the CBTi, but also to collect data to inform and optimise the CBTi for delivery within the RCT. This was achieved through iterative data collection and feedback loops (that included formal and informal project-related meetings, and face-to-face and e-mail communications with the psychologists and the wider project team) to enable practical resolution of problems as they arose. The impacts of the process evaluation activities on the development and delivery of the CBTi are highlighted below.

Data reported here are attributed using the unique identifier allocated to the participant. Prefixes are as follows: ‘P’ denotes professionals; ‘C’ denotes patients interviewed for the process evaluation; ‘F’ denotes family members (caregivers); and ‘R’ denotes study team members.

Achieving coherence in randomised controlled trial structures and processes

The process evaluation interviews that were conducted with patients and clinic staff indicated a number of other issues that needed to be addressed to maximise the feasibility and acceptability of the RCT in which the CBTi was to be evaluated. These issues were iteratively fed back by staff responsible for the process evaluation (CB and TF) to key members of the project team to allow the development and implementation of strategies to address them prior to the start of the RCT.

Perceived acceptability of cognitive–behavioural therapy intervention to older people

A key issue raised by all stakeholders was the extent to which CBT would be acceptable to older people, as illustrated by the following comments:

I imagine that some stoical North Easterners might think that it was wishy washy mumbo jumbo that wasn’t necessarily going to help, you know the idea of going and sort of seeing like, effectively having a sort of a counselling type sessions, some people might not take that seriously.

P2, interview, 6 December 2011

So I think although she would agree to any kind of help I think there’s a part of her that possibly feels she doesn’t need it, and I think it’s getting over that hurdle.

F2, interview, 9 March 2012

A key issue concerned the name of the CBTi. Patients and clinic staff indicated that the use of terms such as ‘psychological’ and ‘psychotherapy’ were potentially alienating to the intended client group:

what does psychotherapy treatment mean to you?

Sitting on a couch with a shrink!

And would you then sign up for that do you think?

I don’t know, I don’t know whether that would help me with the problems I’ve got.

C2, interview, 30 January 2012

All stakeholder groups participating in qualitative interviews were asked for advice and suggestions. It was seen as essential to ensure that the title was positive: avoid the word ‘falling’ (as well as any words beginning with ‘psych’) and to consider including terms such as ‘confidence’, to which people would be able to relate. In addition to being non-threatening, we wanted the name to convey a sense of the purpose and focus of the intervention so that it would not be mistaken for an exercise class. Following discussions, we agreed on STRIDE as the brief study title (Strategies for increasing independence, confidence and energy).

Although there were clear reservations about the concept and name of the CBTi, stakeholders were generally positive about the practical aspects of the CBTi, for example the number, frequency and duration of sessions, the timing of the follow-up session, and the intention to deliver the intervention at home. Patients emphasised the need for flexibility and for the sessions to fit around their other commitments. Several patients suggested group sessions, either instead of, or alongside, individual sessions, and although the likelihood of individual preferences was noted it was agreed that this could not be accommodated in the RCT.

Perceived value of cognitive–behavioural therapy intervention for fear of falling

Clinic staff expressed some reservations about the proposed CBTi for fear of falling, in particular the possibility that, for some patients, fear of falling may actually be protective with regard to falling:

I don’t know whether it’s protective or not which would be my other question, is it a good thing people have a fear of falling, if you take that fear of falling away are they going to fall because their sense of caution has gone?

P3, interview, 7 December 2011

Clinic staff also expressed reservations over patients’ willingness to engage with an intervention that focused primarily on fear of falling:

I’m not sure how receptive they will be to addressing fear of falling as a primary factor in the absence of anything else, any modifications to gait or aids or physical performance, so I don’t know how often it’s a stand-alone problem.

P3, interview, 7 December 2011

As the North Tyneside Falls Prevention Service focused on identifying underlying medical causes for falls, it was not surprising that staff were uncertain about the value of psychological interventions such as CBT. In light of their key role in ‘selling’ the CBTi to eligible patients in Phase I of the study, we organised a briefing meeting for clinic staff in which the intervention developers fed back the results of the interviews, described the intervention and illustrated how it would be relevant to different types of clients. This enabled staff to recognise the similarities between the CBTi and their existing practice:

I mean I’ve used it a lot with working with older people and it’s not actually called CBT but it’s what you do with them from the point of view you’ve gradually got to build their confidence up through very simple increasing their activities and things and the way you talk to them and the way you encourage them and things like that as well, so yes I think it would be very, very helpful actually.

P5, briefing meeting for clinic staff, 17 February 2012

During this briefing meeting, we were able to provide reassurance about the venue of the treatment (usually in patients’ own homes) and also to address other concerns, for example mechanisms for informing patients’ general practitioners (GPs) about their participation.

Recruitment of patients to a randomised controlled trial

Observation of the clinical setting in which the RCT was to take place revealed that the service specialised in fast paced, routinised, structured assessments. Potential barriers to patient recruitment were time pressures and the lack of any routine discussion of fear of falling within the service. To facilitate recruitment in these difficult circumstances, we produced a user-friendly leaflet and a ‘script’ that staff could follow. This was designed to have minimal time implications and to allow staff to follow their existing approach to referring patients for strength and balance classes provided by a local voluntary organisation.

Relevance and acceptability of the primary outcome measure

A final issue that emerged from discussions with clinic staff related to the adequacy of the FES-I as a means of identifying potential participants and the primary outcome measure despite clear published evidence of its validity. 29 These concerns over the outcome measure undermined the perceived validity of the proposed RCT. A key concern was the perceived lack of correlation between apparent fear of falling and FES-I scores:

She’s an agoraphobic and was very, very frightened of falling [. . .] she won’t go to the hairdresser and that sort of thing, but because the FES was so low from that point of view then obviously she wouldn’t fit your criteria then would she?

P5, briefing meeting for clinic staff, 17 February 2012

Staff felt that responses on the FES-I often reflected functional limitations (e.g. arthritis or other pain) that affect mobility rather than fear of falling. Given the extent of clinic staff concerns over the use of the FES-I as the primary outcome measure, we added a visual analogue scale, on which patients could rate their fear of falling, to be included alongside other measures used in the RCT.

Primary objective

To determine the effectiveness of a new cognitive–behavioural therapy-based intervention (CBTi) plus usual care compared with usual care alone (the control condition) in reducing fear of falling in community-dwelling older adults attending falls services.

Secondary objectives

To:

1. measure the impact of the intervention compared with the control on fall and injury rates in the trial participants, and its impact on functional abilities

2. measure the effectiveness of the intervention compared with control on anxiety and depression, quality of life, loneliness, social isolation and social participation

3. investigate the acceptability of the intervention for patients, family members and professionals

4. further investigate the professional and organisational factors that promote or inhibit the implementation and integration of the intervention

5. measure the costs and outcomes of the intervention in this setting.

Objectives (c) and (d) relate to the process evaluation, the methods and findings of which are described in Chapter 5; objective (e) relates to the health-economic evaluation, the methods and findings of which are described in Chapter 6.

Inclusion criteria

• FES-I score of > 23.

• Aged ≥ 60 years.

Exclusion criteria

• Cognitive impairment [Mini Mental State Examination (MMSE) score of < 24/30]. 41

• Life expectancy < 1 year or unlikely for any other reason to be unable to complete 1-year follow-up

• Requiring psychosocial interventions that are unrelated to fear of falling.

• Current involvement in other investigational studies or trials, or involvement within 30 days prior to study entry.

• Patients who had taken part in Phase I (i.e. qualitative work to develop the intervention) of the study. A small number of these patients acted as ‘pilot clients’ and received the intervention (as part of the training of the HCAs) but did not complete any of the outcome measures.

During the course of the study, it became apparent that some couples both fulfilled study eligibility criteria. To avoid contamination, it was decided that only one member of each household should be randomised. Therefore, if a spouse or partner of an otherwise eligible individual was already participating in the study, the second individual to be identified was excluded at the second stage of screening by research assistants (RAs) (see below).

Primary outcome measure

There is considerable variation in the literature on what constitutes fear of falling and how it should be measured. 12,25,43 Single-item questions are sometimes used, but have been criticised for their inability to measure variability in degrees of fear of falling. 12 We included a single-item, 11-point rating scale of fear of falling (see next section) in our battery of outcomes, but felt that a multi-item scale would be a more appropriate primary outcome measure. A number of scales have been developed following Tinetti’s initial FES,13 but its modified version, the FES-I29 has gained currency as a measure of fear of falling because of its simplicity, ease of administration and widespread validation across patient groups49–52 and cultures. 40,52,53 Tinetti’s original FES13 measured purely self-efficacy, an individual’s belief in his or her own capabilities in performing specific activities,54 but the FES-I modification (see Appendix 1) more explicitly measures concerns about falling. 29,55 We re-examined this issue in 200 older people using a more detailed principal factor-analytic approach than the original validation study,29 and found that the FES-I more robustly measures concerns about falling than had previously been found. 56 An integrative review of tools to measure fear of falling, considering papers published between 1982 (when the concept of fear of falling was first defined) and 2010, concluded that the FES-I is the best assessment tool for assessing fear of falling in community-dwelling elders. 57

Secondary outcome measures

Identification and screening of participants

Potential participants with significant fear of falling (FES-I score of > 23) were identified prospectively and retrospectively by staff at the participating community falls services, as described in greater detail below; the majority were identified and approached prospectively. When the FES-I was not administered routinely at a particular site, consent was taken prior to administering this measure. At the majority of sites, all other screening assessments were part of routine clinical practice; when a screening measure was not part of routine clinical practice, consent was taken prior to administration of that measure.

A screening log was kept at each site to document details of subjects who were invited to participate in the study. For subjects who declined participation, this documented any reasons that were available for non-participation. The log also ensured that potential participants were approached only once.

Phase I participants who screened eligible and had expressed an interest in the CBTi intervention were offered the opportunity to receive the CBTi, but were not randomised and did not complete any outcome measures.

Recruitment procedures

Consent procedures

Informed consent discussions were undertaken by trained RAs/interviewers, with opportunity for participants to ask any questions. Following receipt of information about the study, participants were given reasonable time (minimum of 24 hours) to decide whether or not they wished to participate. Those wishing to take part provided written, informed consent by signing and dating the study consent form, which was witnessed and dated by the RA/interviewer. Written informed consent to study-specific procedures/investigations was obtained prior to randomisation. The original signed consent form was retained in the Investigator Site file, with a copy placed in the clinical notes and another copy provided to the participant. Participants specifically consented to their GP being informed of their participation in the study.

Interpreters were available for all visits of patients who required them, either for verbal translation or for deaf subjects wishing to take part in the study, via local NHS arrangements. In practice, no patients required an interpreter or signer.

Primary analysis

The difference between FES-I at the end of treatment (12 months) and FES-I immediately before treatment (baseline) in two arms was analysed using analysis of covariance. The dependent variable was the FES-I score at 12 months; baseline FES-I was included as a covariate. Stratification variables (site, gender, pain when walking, and referral for strength and balance training) were included as fixed effects. A two-tailed test was performed, with significance level set at 5% (95% CIs of parameters of interest). The primary analysis was on the basis of intention to treat (ITT) with patients included in the group to which they were randomised.

Secondary analysis of primary outcome measure

The above analyses were extended to include intermediate measures of FES (8 weeks and 6 months). Within a multilevel model both variation between occasions within an individual participant and variation between participants were included as random effects (assumed to have a Gaussian distribution). The hypothesis to be tested was that fear of falling would be lower in the intervention group at 8 weeks (i.e. on completion of therapy) and the difference between groups would be maintained until the end of follow-up at 12 months.

If > 25 patients in the intervention group fail to complete the full course of CBT we planned to investigate whether or not there was an association between the effectiveness of the intervention and the number of CBT sessions attended. We indicated that this would be achieved by adding the number of sessions of therapy as a patient-level covariate to the above model.

Secondary outcomes

Each of the remaining health status and quality-of-life measures took the form of a score that was treated as a continuous variable and analysed using analysis of covariance with an assumed normal error structure. The dependent variable was the response observed at 12 months. Baseline response was included as a covariate; differences between randomisation strata were included as fixed effects (a binary indicator of pain when walking, gender, referral for strength and balance training, and treatment centre). Finally, the difference between groups was fitted as a fixed effect. A 95% CI for the estimated impact of CBTi is presented. Differences between strata were then removed from the model and unadjusted estimates of the impact of CBTi are presented for comparison purposes.

Change over time in these outcomes was investigated using a sequence of multilevel models with both variation between occasions within an individual participant and variation between participants included as random effects (assumed to have a Gaussian distribution). Differences between strata were included as fixed effects. An initial model assumed a common linear trend over time in each of the study groups. In a second model, a separate trend was fitted for each study group; 95% CIs are presented for each estimate of trend and for the difference between them.

Analyses were designed to:

• measure the impact of the intervention compared with control on fall and injury rates in the trial participants and its impact on functional abilities

• assess the effectiveness of the intervention compared with control on anxiety (HADS), quality of life [EQ-5D (5L), SF-6D, WHOQOL-OLD], social isolation and social participation at 6 and 12 months post randomisation, through analysis of covariance

• summarise the acceptability of the intervention for patients, family members and professionals (see Chapter 5)

• investigate further the professional and organisational factors that promote or inhibit the implementation and integration of the intervention (see Chapter 5)

• measure the costs and outcomes of the intervention in this setting (see Chapter 6).

For all analyses, examination of residual plots was used to check model assumptions.

Scoring of questionnaires and handling of missing data

The FES-I scores were calculated in accordance with the authors’ instructions:29 if data were missing one more than four items, the data were declared as missing; if no more than four of the items were missing, the score was calculated by adding the responses to all completed items, dividing by the number of items completed and multiplying the value thus obtained by 16, and then rounding to the nearest whole number.

In respect of secondary outcome data, quality-of-life scores were calculated in accordance with their authors’ instructions. 61,64,67–70,86 For the SF-36, scoring software provided by the developers and distributors of the instrument was used (www.optum.com/optum-outcomes/what-we-do/health-surveys.html, accessed 1 October 2015). In the absence of any authors’ rule for handling missing data, simple imputation was used, provided that at least half of the items in any scale had been completed (imputed missing value = mean value of non-missing items). Secondary outcomes were analysed using mixed models that made use of all data available.

To address differential dropout in the two groups, in both the protocol and statistical analysis plan, we originally proposed to use the method of joint modelling described by Henderson et al,87 based on the assumption that time to dropout would be predictive of final outcome. The rationale behind this approach is that there is some underlying characteristic of the participant, usually referred to as ‘frailty’, which is associated both with dropout and outcome.

In practice, dropout did not occur steadily through this study but rather was triggered by particular activities arising from participation in the study. A large proportion of the dropout occurred prior to week 8; a smaller proportion of dropout accrued over the remaining 44 weeks. The fitting of a Cox proportional hazards model to ‘time to drop out’ would appear to be very questionable, given the pattern of missing data within the two arms of the trial. Instead, to assess the impact of missing visits, we adopted an approach based on multiple imputation. 88

Multiple imputation was used to attribute missing values for the FES-I score at each of the follow-up time points, in those cases when the instrument was not completed at all, or could not be scored. The imputation model included the following variables measured at baseline:

• age

• sex

• gait and balance score

• handgrip strength

• functional reach

• falls efficacy (FES score)

• concern about falling (0–10 scale)

• pain when walking (0–10 scale)

• participation score

• loneliness score

• site.

In addition, the number of sessions of CBT completed was included as a predictor variable.

Four sets of imputation were undertaken. The FES score at 1 year was estimated using a regression model with the above variables included as predictors. Then FES scores at 8 weeks, 6 months and 1 year were estimated simultaneously, assuming a multivariate normal model.

Each of these methods was undertaken:

1. separately for each study group (intervention and control)

2. with both groups together (with group added as a predictor variable).

In each case, 50 data sets were imputed.

Imputation was carried out using the ‘mi impute’ procedure in the Stata statistical package. Two different multivariate data imputation models were used. In both cases, missing values for all of the specified variables were imputed simultaneously. The first method used a regression model approach (‘mi impute regress’ in Stata); the second method assumed a multivariate normal distribution (‘mi impute mvn’ in Stata). These methods were applied using two different approaches. First, the imputation was carried out using all participants; second, the imputation was done separately for each group. Thus there were four sets of imputed data. In each case, 50 imputations were undertaken, giving a total of 4 × 50 = 200 imputed data sets. Inference on each of the four sets of 50 data sets was carried out using the ‘mi estimate’ procedure in Stata. Coefficients and standard errors are adjusted for the variability between imputations according to the combination rules set out by Rubin. 89

Safety (harms)

The aim of the analysis of harms was to estimate the difference in the rate of adverse events experienced by the two treatment groups. For serious adverse events, we determined 95% CI for the hazard ratio (i.e. risk of a serious adverse event) for a patient randomised to CBTi plus usual care compared with a patient randomised to usual care.

Compliance

The number of CBT sessions attended was documented and the date of each session was captured. Descriptive analysis focused on the number of sessions attended and the time interval over which the first eight sessions were completed. We also analysed the distribution of time between completion of the first eight sessions of CBT and the scheduled 8-week and 12-month follow-up assessment of outcomes.

Screening within the sites

Overall, 2355 patients were initially screened (using information recorded in clinic databases) by the clinic staff across the three recruitment sites (Table 8). Of those screened, 1346 were considered to be eligible for the trial (57.1%). Some eligible patients explicitly declined the initial study information that was available at the clinic.

Rates of eligibility were similar between the Falls and Syncope Service and the North Tyneside Falls Prevention Service, but were markedly higher at the Galleries Day Unit. Rates of decline of study information by those who were deemed to be eligible varied significantly across the three sites, with patients at the Galleries Day Unit most likely to decline information.

One reason for ineligibility was recorded for each patient within the screening log. The reasons for non-eligibility at initial screening are shown in Table 9; most commonly these were FES-I score of ≤ 23 (78.4%), MMSE score of < 24 (17%) and patients aged < 60 years (1.6%). There were marked differences between sites in the stated reasons for ineligibility. This in part reflected inter-site differences in patient case mix and variations in practice (see Chapter 3) with respect to routine administration of the FES-I and MMSE, and therefore availability of these data in assessing eligibility.

Screening prior to randomisation

The RAs made a further check on eligibility by reviewing each of the inclusion and exclusion criteria with patients who returned an expression of interest form. This was to ensure that their eligibility had not changed prior to randomisation. Some additional participants were excluded at this stage, the reasons for exclusion are shown in Table 10.

Rates of non-contact did not vary significantly between the three sites. Among the 682 patients contacted, rates of eligibility were 91.4% for the Falls and Syncope Service, 90.7% for the Galleries Day Unit and 92.6% for North Tyneside Falls Prevention Service. A total of 627 of those contacted were deemed eligible, of whom 211 (33.7%) did not consent to randomisation. Rates of non-consent were 40.6% for the Falls and Syncope Service, 43.2% for the Galleries Day Unit, and 26.6% for the North Tyneside Falls Prevention Service; pairwise comparisons indicated that the rate of non-consent at the North Tyneside Falls Prevention Service was significantly different from the rates at the other two sites. There appeared to be some variation across sites in the reasons for declining consent, but this may be due, in part, to inconsistency between RAs/interviewers in how the reasons offered were categorised (Table 11).

The overall participation rate (i.e. number consenting and randomised as a percentage of those returning an expression of interest form) was 58%, with site-specific rates of 51.0%, 50.5% and 64.7% for the Falls and Syncope Service, Galleries Day Unit and North Tyneside Falls Prevention Service, respectively. Pairwise comparisons showed that the participation rate for the North Tyneside Falls Prevention Service was significantly different from the rates at the other two sites. When reasons for non-participation were grouped as ‘no contact’, ‘ineligible’, ‘declined consent’ and ‘died post randomisation’, there were no significant differences between sites.

Recruitment

Monthly recruitment is shown in Figure 5.

FIGURE 5.

Recruitment graph.

Randomisation

The screening, recruitment, randomisation and trial follow-up are summarised in the CONSORT diagram shown in Figure 6.

FIGURE 6.

The STRIDE CONSORT diagram.

Recruitment by strata

A total of 415 participants were randomised; 240 (57.8%) came from the North Tyneside Falls Prevention Service, the first site to open; 125 (30.1%) came from the Falls and Syncope Service; 50 (12.1%) came from the Galleries Day Unit). In addition to recruitment site, the sample was stratified by three other variables: gender, pain when walking (dichotomised as yes/no from ratings on a 11-point scale) and referral to strength and balance training (yes/no) (Table 12). Overall, 124 (29.9%) of those randomised were men (95% CI 25.5% to 34.5%). Two hundred and one participants (48.4%; 95% CI 43.5% to 53.4%) reported that they suffered pain when walking; 288 participants (69.4%; 95% CI 64.7% to 73.8%) had been referred for strength and balance training.

Strata were included as fixed effects in analyses of covariance. To inform this approach, the level of collinearity in these potential explanatory variables was examined.

There were systematic differences between the three sites only with respect to the percentage of participants referred for strength and balance training, reflecting variation in ‘usual care’ between these sites, as described in Chapter 3.

Overall, 50.8% of men and 47.4% of women reported pain on walking [relative risk (RR) 1.04, 95% CI 0.02 to 1.18]. Of those referred for strength and balance training, 50.3% had pain on walking, 49.7% did not [RR for an individual referred for strength and balance training having pain on walking = 1.09 (95% CI 0.87 to 1.36)]. Across all sites, men were more likely to be referred for strength and balance training than women (77.4% vs. 66.0%, RR 1.17, 95% CI 1.03 to 1.32).

Recruitment by arm

Two hundred and five patients (49.4%) were randomised to the control arm; 210 patients (50.6%) were randomised to receive CBTi (Table 13).

Falls Efficacy Scale–International version

Completion of the individual items within the FES-I at baseline was excellent. Four participants did not complete the item ‘Going to the shop’. Greatest concern was expressed about ‘Walking on slippery surface’; least concern was expressed about ‘Preparing simple meals’ (see Appendix 4). The scale was internally consistent; Cronbach’s alpha, based on 402 complete cases, was 0.89. The total FES-I score could be computed for all 415 participants (Figure 7). FES-I score was not associated with age (correlation –0.034 (95% CI –0.130 to 0.070)].

FIGURE 7.

The FES-I scores at baseline.

Fear of falling 11-point scale

In the baseline interview, participants were asked to rate their concerns about falling on an 11-point scale, for which ‘0’ meant ‘no fear of falling’ and ‘10’ meant ‘fear of falling as bad as it could be’. All 415 participants responded to this item (Figure 8). The majority of participants scored between 5 and 10 The correlation between FES-I scores and fear of falling numerical rating was 0.53 (95% CI 0.46 to 0.60) (CI based on bootstrap procedure with 1000 replications). There is clearly a strong association between FES-I scores and fear of falling, as assessed by the rating scale method (see Appendix 4 for a graphical presentation of this association).

FIGURE 8.

Distribution of fear of falling, 11-point scale numerical rating.

Pain when walking, 11-point scale

In the baseline interview, participants were asked to rate their pain on walking on a 11-point scale, for which zero ‘0’ meant ‘no pain’ and ‘10’ meant pain that was ‘as bad as it could be’ (Figure 9). This variable was dichotomised as ‘no pain’ (0) versus ‘pain’ (1–10) for purposes of stratification. ‘Pain when walking’ ratings were weakly but positively correlated with FES-I scores [ρ = 0.35 (95% CI 0.27 to 0.44)] (see also Appendix 4) and with fear of falling numerical rating scales [ρ = 0.32 (95% CI 0.22 to 0.41)].

FIGURE 9.

Distribution of pain when walking, 11-point scale numerical rating.

Social participation and social isolation scale score

Total participation scores ranged from 0 to 37, with a mean of 9.49 and a median of 7 (Figure 10). Internal consistency, measured by Cronbach’s alpha, was poor at 0.27. Total participation score was not correlated with age [ρ = –0.006 (95% CI –0.101 to 0.084)].

FIGURE 10.

Distribution of social participation and social isolation scale score at baseline.

De Jong Gierveld Loneliness Scale

High levels of loneliness were reported by participants. Scores ranged from 0 to 11, with a mean of 8.20, median of 9 and mode of 11 (Figure 11). Internal consistency of this scale was poor, with a Cronbach’s alpha (α) = 0.42. Loneliness score was not correlated with age [ρ = –0.035 (95% CI –0.145 to 0.070)] but was positively, albeit weakly, correlated with total participation score [ρ = 0.148 (95% CI 0.054 to 0.235)].

FIGURE 11.

Distribution of De Jong Gierveld Loneliness Scale scores at baseline.

Short Physical Performance Battery (gait, balance and sit to stand)

Short physical performance battery scores ranged from 0 to 12, with a mean of 5.66 (Figure 12). Lower scores are associated with poorer physical performance, with higher scores reflecting better physical condition in relation to these activities. The mean for our participants represents an average performance for this patient group.

FIGURE 12.

Distribution of SPPB scores at baseline.

Handgrip strength and functional reach

Data on handgrip strength were not available for 13 participants, with reasons for non-completion recorded for eight (one refused; in two cases the test was not attempted because the participant felt unsafe; in one case the participant attempted the test but could not hold the position unaided; and in four cases, the participant attempted the test but failed). Among those completing the test, scores over three attempts ranged from 0.3 kg to 54.4 kg, with a mean of 16.5 kg (Figure 13). The mean is low compared with normative data for older patients, with an overall mean for this age group of 31 kg. 91

FIGURE 13.

Distribution of mean handgrip strength (kg) at baseline.

Data on functional reach were not available for 40 individuals, with reasons recorded for 39 of these (four participants refused; one participant could not understand the instructions; in sixteen cases the participant felt unsafe, and in five cases the RA felt that it would be unsafe; in five cases the participant attempted the test but could not hold the position; in five cases, the participant attempted the test but failed; other reasons were given for three participants). Among those completing the test, scores averaged over three attempts ranged from 1.3 cm to 51.1 cm, with a mean of 22.8 cm (Figure 14), with > 22 cm representing the average normal value for older patients. 92

FIGURE 14.

Distribution of mean functional reach (cm) at baseline.

Self-completion questionnaires

Baseline self-completion questionnaires were returned by 323 participants (77.8%). A minority of respondents (20.8%, 95% CI 16.5% to 25.5%) reported that they had help completing the questionnaire; the percentage requiring assistance was similar across the two arms (Table 15).

Lubben Social Network Scale–6

Lubben Social Network Scale scores ranged from 0 to 30, with a mean of 14.69 (Figure 15). Internal consistency of this scale was adequate, with a Cronbach’s α = 0.83.

FIGURE 15.

Distribution of LSNS-6 scores at baseline.

European Quality of Life–5 Dimensions (5 Level)

The EQ-5D (5L) ratings of ‘Health today’ ranged from 5 to 100, with a mean of 58.6 (median 60, mode 50; Figure 16). Applying UK tariff values, EQ-5D (5L) ‘Index values’ ranged from –0.20 (a state worse than death) to 1.0 (perfect health), with a mean of 0.51 (median 0.56; Figure 17). Internal consistency of the five items in this scale was adequate, with Cronbach’s α = 0.73.

FIGURE 16.

Distribution of EQ-5D (5L) ratings of ‘health today’ at baseline.

FIGURE 17.

Distribution of EQ-5D (5L) index values at baseline.

Hospital Anxiety and Depression Scale

The HADS Anxiety scores ranged from 0 to 21, with a mean of 8.2 (median 8.0, mode 7.0; Figure 18). Scores of ≤ 7 are considered ‘normal’ and 45.5% of participants had scores in this range at baseline; percentages with ‘mild’ (8–10), ‘moderate’ (11–14) and ‘severe’ anxiety were 23.7%, 23.9% and 6.9%, respectively. HADS Depression scores ranged from 0 to 17, with a mean of 6.8 (median and mode both 7.0; Figure 19). Scores of ≤ 7 are considered ‘normal’ and 62.8% of participants had scores in this range at baseline; percentages with ‘mild’ (8–10), ‘moderate’ (11–14) and ‘severe’ depression were 22.1%, 12.6% and 2.5%, respectively. Internal consistency of both scales was adequate, with Cronbach’s α = 0.85 for HADS Anxiety and 0.77 for HADS Depression.

FIGURE 18.

Distribution of HADS Anxiety scores at baseline.

FIGURE 19.

Distribution of HADS Depression scores at baseline.

World Health Organization Quality of Life questionnaire – older adults module

The WHOQOL-OLD total scores ranged from 8.3 to 88.5 with a mean of 57.2 (Figure 20). Internal consistency for the whole scale was adequate, with a Cronbach’s α = 0.86. Cronbach’s alpha for the individual subscales ranged from 0.66 for ‘Autonomy’ to 0.95 for ‘Intimacy’.

FIGURE 20.

Distribution of WHOQOL-OLD scores at baseline.

Short Form questionnaire-36 items

Internal consistency of the seven multi-item domain scores was adequate, with Cronbach’s alphas ranging from 0.72 (General Health Perception) to 0.93 (Role-Emotional). SF-36 Physical Component Scores (PCSs) ranged from 8.4 to 63.8 with a mean of 32.1 (Figure 21). SF-36 Mental Component Scores (MCSs) ranged from 17.2 to 69.7, with a mean of 45.2 (Figure 22). PCSs and MCSs are norm based, with a mean of 50 and SD of 10 in a general population. The observed mean scores for this study population indicate poor physical health relative to a general population, with the mean score almost two SDs below the norm of 50. MCSs were also below the norm, but to a lesser extent.

FIGURE 21.

Distribution of SF-36 PCSs at baseline.

FIGURE 22.

Distribution of SF-36 MCSs at baseline.

Baseline comparability

The study arms were well balanced at baseline (Table 16).

The relationship between baseline scores on the FES-I and stratification variables was examined (see Appendix 4 for detailed graphs and tables). In general, the difference in mean scores between the strata across a range of health status and quality-of-life measures was fairly small.

The FES-I scores at baseline were higher at the Galleries Day Unit (indicating that participants have a higher level of concerns about falling) than at the other two sites (one-way analysis of variance indicated that differences between sites was statistically significant (F_2,412 = 8.69; p < 0.001). Similarly, numerical ratings of fear of falling appeared to differ between sites. The mean fear of falling 11-point numerical rating scores were 6.64 for participants recruited by the Falls and Syncope Service, 7.62 for participants recruited by the Galleries Day Unit, and 6.51 for participants recruited by the North Tyneside Falls Prevention Service; patients from the Galleries Day Unit reported a higher fear of falling per the 11-point scale than patients from the other two sites, a finding consistent with that reported for FES-I scores above. This difference between sites persisted even on allowing for a difference between sites in the proportion of patients reporting pain when walking. For other measures, participants recruited at the Galleries Day Unit tended to have poorer health status and reported poorer quality of life at baseline than patients from the other two sites.

The FES-I scores at baseline were similar for men and women. As expected, men were observed to have greater handgrip strength and greater functional reach than women. Otherwise, differences between men and women were fairly small.

There was a strong positive association between pain on walking and FES-I score at baseline. Scores on the fear of falling numerical rating scale were also positively associated with pain when walking at baseline. The mean fear of falling rating scale scores were 5.99 for those with no pain when walking and 7.42 for those reporting pain when walking, a difference of 1.43 (95% CI 1.03 to 1.83); participants who suffered pain with walking reported significantly greater fear of falling than other participants. Of all of the four randomisation strata ‘pain when walking’ was most highly correlated with the health status of the participant; those who reported pain when walking tended to report poorer health status and quality of life.

As already noted, referral for strength and balance training was confounded with site. There was no evidence of an association between FES-I score at baseline and referral for strength and balance training. Patients who were referred for strength and balance training actually had higher handgrip strength and greater functional reach than those who were not referred. Other outcome measures were not associated with referral for strength and balance training.

Falls Efficacy Scale–International version completion

The figures in the green boxes indicate the number of participants who completed the FES-I sufficiently to allow it to be scored; the figures in the unshaded boxes indicate the number of participants for whom no FES-I items have been scored. We have responses at all four visits for 297 participants (Figure 23).

FIGURE 23.

Number of participants visited by RA and asked to complete the FES-I.

Attrition was slightly greater in the group that was randomised to CBTi (Table 18).

Completion of falls diaries

The majority of diaries (3288/3309) were completed, as anticipated, on paper. The number of diaries that were returned fell over the period of follow-up. In general, participants in the control arm were better at completing and returning falls diaries than participants who were randomised to CBTi (Table 19).

Sixty-eight participants (16.4%) returned no falls diaries at all (Figure 24); 106 participants (25.5%) returned all 13 diaries. The mean number of diaries returned was 7.97 (95% CI 7.44 to 8.49). On average, participants randomised to the CBTi returned 1.18 (95% CI 0.20 to 2.11) fewer diaries than participants in the control arm.

FIGURE 24.

Distribution of the number of diaries returned.

Timing of randomisation and follow-up data collection relative to cognitive–behavioural therapy intervention delivery

Eighty-five per cent (95% CI 81.2% to 88.4%) of participants were randomised on day of the baseline assessment. The week 8 visit took place within 44 and 120 days of randomisation. For those randomised to the CBTi, the interval from the eighth CBTi session to the week 8 visit could be calculated for 137 individuals (in the remaining instances, one or both dates were missing either because the participant had withdrawn from the CBTi or did not have a week 8 visit, or both). In 124 (90.5%) of these cases, the week 8 visit took place on or before the date of the eighth CBTi session; the delay ranged from zero (one participant) to 141 days, with a mean of 31 days, median of 26 days.

Although there was a wide visit window observed at each follow-up point the difference in the mean time between date of randomisation and date of visit between control and intervention participants was not statistically significant (Table 24). At 1 year, the mean difference in timing of ascertainment of primary end point, vis-à-vis randomisation date, was 1.9 (95% CI –3.7 to 6.8) days.

Analysis of Falls Efficacy Scale–International version scores at 1 year (the primary end point)

The difference between FES-I at the end of treatment (12 months) and FES-I prior to treatment (baseline) in two arms was analysed using analysis of covariance. The dependent variable was the FES-I score at 12 months; baseline FES-I was included as a covariate. Stratification variables (site, gender, baseline pain when walking and referral for strength and balance training) were included as fixed effects.

This was achieved by fitting a simple linear regression model. The regression coefficients are given in Table 25. FES-I at 1 year was highly correlated with baseline FES-I.

There were significant differences between sites, with participants at the Galleries Day Unit having higher FES-I scores (reflecting poorer falls-related self-efficacy and fear of falling) than participants from the other two sites. Participants who reported pain when walking had higher FES-I scores, whereas participants who had been referred for strength and balance training had lower FES-I scores. The difference between men and women was not significant. Finally, the difference between study arms was statistically significant; the estimated impact of CBTi on FES-I scores was a reduction in mean score of 4.02 (95% CI 2.10 to 5.95).

The baseline SD of FES-I was 9.37. The reduction in mean score of 4.02 can be expressed as a standardised effect size by dividing by this SD, yielding a point estimate of effect size of 0.43 and an interval estimate of between 0.22 and 0.64.

The change in R² on adding the effect of therapy to the model including baseline FES-I scores and the effects corresponding to each of the strata is shown in Table 26. R² increased from 0.45 to 0.48.

Plots of standardised residuals suggest assumption of a normal error distribution for the residuals is reasonable.

Impact of non-response on primary outcome measure

As previously described (see Chapter 3), multiple imputation was used to impute missing values for the FES-I score at each of the follow-up time points, in those cases where the instrument was not completed at all, or could not be scored. A pooled estimate of the impact of the intervention at 1 year across data sets was based on an analysis of covariance model with adjustments for baseline FES and randomisation strata. The parameter estimates are given in Table 27.

Adjusting for missing data reduces the ITT estimate of the effect of the CBTi; all methods of data imputation result in a smaller estimated impact of the CBTi on fear of falling. This is consistent with the observations that (1) those patients who withdrew from the CBTi arm had lower FES-I scores at baseline and (2) in the analysis of imputed data we have a much greater proportion of patients in the CBTi arm who did not receive the intervention than in the complete case analysis. Point estimates of standardised effect sizes, adjusting for differential loss to follow-up, range from 0.27 to 0.41.

For comparison purposes we can undertake an ‘as treated’ analysis using the imputed data sets. Here we replace the study arm variable with a CBTi variable that ranges from 0 (no sessions of CBTi) to 1 (nine sessions of CBTi). This yields an estimate of the effect of the CBTi for those patients who received nine sessions (Table 28).

The ‘as treated’ analysis suggests a mean improvement in FES-I score of approximately ‘4’ for those patients who complete the CBTi. The results of the ‘as treated’ analysis are very similar to those obtained from the ‘complete case’ analysis, which, perhaps, reflects the poor outcome completion rate among those failing to complete the planned CBTi. Point estimates of standardised effect sizes for this ‘as treated’ analysis range from 0.39 to 0.47. These results are consistent with there being a subgroup of patients who benefit from the CBTi but when we analyse on an ITT basis the average impact across the target population is somewhat reduced (from around a mean change of ‘4’ to a mean change of ‘3’ in FES-I score).

Longitudinal analysis of Falls Efficacy Scale–International version

To provide a more detailed picture of how FES-I scores changed over time, a further analysis, utilising data from all four time points (baseline, 8 weeks, 6 months and 12 months) was conducted. The distribution of time to completion is shown in Figure 28.

FIGURE 28.

Time to completion of FES-I.

The four distinct CBTis of bars correspond to visits at baseline and 8 weeks, 6 months and 1 year post randomisation. Time in years was included in this longitudinal analysis as a continuous variable.

• Scores at the follow-up visits were analysed using a mixed-model approach. All models included two random effects, variation between:

  • participants

  • follow-up visits within participants.

Model 1 included the following fixed effects:

• baseline FES-I scores

• pain when walking (stratification variable, binary indicator)

• referral for strength and balance training (stratification variable, binary indicator)

• gender (stratification variable, binary indicator)

• site (two binary indicators: Galleries Day Unit and North Tyneside Falls Prevention Service)

• trend over time

• study arm (an impact of the CBTi).

Additionally, model 2 included an interaction between study arm and time (equivalently to fitting a separate trend over time for intervention and control patients).

The parameter estimates corresponding to model 1 are given in Table 29.

Sigma_u and sigma_e are the estimated SD of the between-participant and within-participant random effects, respectively. This model suggests that the annual trend does not differ significantly from zero. The annual change in FES-I score is –0.78 (95% CI –1.80 to 0.23). In this model the estimated impact of the intervention corresponds with an average difference in FES-I scores between study arms over the three follow-up time points; this was –3.54 (95% CI –4.96 to 2.12).

In model 2, we fit separate trends over time for patients in the control and intervention arms. The parameter estimates corresponding to this model are given in Table 30.

The estimated annual trend in the control arm is –0.68 (95% CI –2.09 to 0.73) and the estimated annual trend in the intervention arm is –0.90 (95% CI –2.36 to 0.57). The difference in these trends is –0.22 (95% CI –2.26 to 1.81) and does not differ significantly from zero. There is no evidence that rate of change in FES-I differed between control and intervention patients. In this model the CBTi coefficient corresponds with the difference between the study arms at a single point in time (because we have non-parallel regression lines corresponding to the two study arms). The trend variables have been coded in such a way that this coefficient represents the difference between arms 12 months post randomisation.

As the interaction between time and CBTi is not significant, applying the principle of parsimony, model 1 is the preferred model. The mean impact of the CBTi over the period of follow-up is a reduction of 3.54 (95% CI 2.12 to 4.96).

Numbers of falls and injuries sustained

The number of falls and related injuries is summarised in Table 31.

The distribution of the number of falls reported by participants was positively skewed (Figure 29).

FIGURE 29.

Distribution of falls.

Of the 347 participants who returned at least one diary, 154 (44.4%; 95% CI 39.1% to 50.0%) reported no falls at all. One participant reported a total of 73 falls.

Sixty-one per cent of men and 53% of women reported at least one fall (relative risk = 1.13; 95% CI 0.93 to 1.38). Fifty-eight per cent of those reporting pain when walking at baseline and 54% of those reporting no pain when walking reported at least one fall (relative risk = 1.06; 95% CI 0.88 to 1.28). Fifty-eight per cent of those not referred for strength and balance training and 55% of those who were referred for strength and balance training reported at least one fall (relative risk = 1.06; 95% CI 0.87 to 1.29). The proportion of patients reporting a fall was broadly similar across the three sites (Falls and Syncope Service 55.3%, Galleries Day Unit 47.4%, North Tyneside Falls Prevention Service, 57.3%).

The number of falls was analysed using a negative binomial regression model adjusted for randomisation strata (site, pain when walking, referral for strength and balance training, and gender) and the number of diaries returned (as an exposure variable). The reference model is shown in Table 32.

The estimated impact of the CBTi on the likelihood of falling corresponds to an incidence risk ratio of 0.88 (95% CI 0.61 to 1.27). The point estimate of 0.88 suggests a slight reduction in the risk of falling but the 95% CI includes one; the difference between arms is not statistically significant. One patient reported 73 falls. Excluding this patient from the analysis changed the estimated impact of CBT to an incidence risk ratio of 1.02 (95% CI 0.71 to 1.46).

The above analysis is based on the 347 participants who returned at least one diary. To look at the potential impact of missing diaries, the analysis was repeated making the assumption that for those patients who returned at least one diary that they had no falls for the missing months. The corresponding estimate of the incidence rate ratio (IRR) was 0.74 (95% CI 0.51 to 1.06).

Likelihood of a participant reporting at least one fall

The likelihood of a participant reporting at least one fall was analysed using logistic regression. The modelling strategy (adjusting for strata and the number of diaries returned) is as reported above.

The estimated impact of the CBTi based on the 347 participants who returned at least one diary adjusted for site, gender, pain when walking, referral for strength and balance training, and the number of diaries returned corresponded with an odds ratio (OR) of 1.17 (95% CI 0.75 to 1.81).

Sensitivity analysis, accounting for missing data, provided the following estimates. Assumption that:

1. there were no falls corresponding to the missing diaries for the 347 participants who returned at least one diary. OR 1.10 (95% CI 0.72 to 1.69)

2. patients who did not return any diaries at all did not fall OR 0.92 (95% CI 0.62 to 1.36)

3. patients who did not return any diaries at all did fall OR 1.29 (95% CI 0.86 to 1.92).

Options (2) and (3) represent the most extreme scenarios. In each case, the 95% CI includes ‘1’; there is no evidence that the CBTi had an impact on the likelihood of reporting one or more falls.

A total of 97 falls resulted in a reported soft tissue injury. The number of soft tissue injuries reported by individual participants ranged from zero to four, with the vast majority (80%) of participants reporting no such injuries. Only four (4.1%; 95% CI 1.1% to 10.2%) of these injuries required medical intervention. A small number of falls (10) resulted in a reported fracture. One participant reported three falls resulting in a fracture; a second participant reported two falls, both of which resulted in a fracture. Five participants reported a single fracture resulting from a fall. The CBTi had no impact on the incidence of injurious falls.

Serious adverse events

Forty-eight participants (11.6%, 95% CI 8.6% to 15.0%) experienced one or more serious adverse events. A total of sixty-nine serious adverse events were recorded. Thirty-four participants experienced a single event; 10 participants, two events; three participants, three events; and one participant experienced six events. Most of the events (89.9%) were categorised as either ‘moderate’ or ‘severe’ (Table 34).

Just over 20% of serious adverse events related to a fall suffered by the participant. Three of the adverse events culminated with the death of the participant and these events had the following descriptions:

• acute intestinal obstruction and incarcerated incisional hernia: raised number of white blood cells, raised level of C-reactive protein, low calcium and magnesium levels, and very raised creatinine level

• pneumonia and pulmonary oedema – died in hospital

• stroke – rushed to hospital but died in Casualty at 12.40.

The first of the events listed pertained to a participant who was randomised to CBTi but happened prior to any of the scheduled CBTi sessions. The other two events involved participants in the control arm. None of these deaths is believed to be causally related to participation in the study.

There is no evidence that the serious adverse event rate differed between study arm (Table 35).

The mean number of reported events were:

• 0.20 (95% CI 0.12 to 0.30) for participants in the control arm

• 0.13 (95% CI 0.08 to 0.20) for participants the CBTi arm.

The difference in means between study arms = –0.07 (95% CI –0.18 to 0.04). Adjusting for differences between strata using negative binomial regression, the estimated IRR = 0.61 (95% CI 0.32 to 1.16); there is no evidence of a difference in the likelihood of reporting serious adverse events.

One hundred and seventy-five participants (42.2%, 95% CI 37.3% to 47.1%) reported at least one adverse event. A total of 514 adverse events were recorded. Most of the events (90.9%) were either ‘possibly’ or ‘definitely’ fall related. None of the events was deemed to be ‘possibly’, ‘probably’ or ‘definitely’ attributed to the receipt of CBT. Sixteen events (3.1%) were classified as ‘severe’; two-thirds of the events were classified as ‘mild’.

The mean number of reported non-serious adverse events were:

• 1.30 (95% CI 0.94 to 1.71) for participants in the control arm

• 1.18 (95% CI 0.86 to 1.54) for participants in the CBTi arm.

The difference in means between study arms = –0.12 (95% CI –0.60 to 0.42). Adjusting for differences between strata using negative binomial regression, the estimated IRR = 0.91 (95% CI 0.63 to 1.33); there is no evidence of a difference in the likelihood of reporting adverse events.

Considering all adverse events (both serious and non-serious) the corresponding IRR was 0.88 (95% CI 0.63 to 1.23). CBTi does not appear to be a particularly high-risk intervention.

Data collection and consent

Data collection involved ethnographic methods, including semistructured interviews, informal discussions and non-participant observation (captured in field notes). Therapy sessions (notes summaries and audio recordings) provided data to assess the content and delivery of the CBTi. Participants in the process evaluation included staff who were working in sites from which clients were recruited (hereafter ‘professionals’); STRIDE therapists who were involved in delivering the CBTi; clients who were participating in the intervention arm of the trial and family members (hereafter ‘caregivers’); and members of the study team and TSC. The data collected for the process evaluation are summarised in Table 36. The average duration of each episode of data collection was 49 minutes for interviews, 14 minutes for informal discussions, 143 minutes for observations and 49 minutes for meetings. Overall, 75% of data collected was recorded (although not all was transcribed). Field notes were also made to support analysis (75 pages in Phase I and 268 pages in Phase II). All data collection for the process evaluation was undertaken by CB, with the exception of one focus group with the STRIDE therapists, which was carried out by a student as part of her master’s dissertation.

Semistructured interviews

Interviews with all participants explored perceptions of key factors that might impede or facilitate the CBTi being effective and workable in practice. The topic guides were loosely structured around NPT,38 with ample scope for participants to raise issues that they saw as important (see Appendix 5 for an overview of topic guides).

Professionals were interviewed about the service provided, their views of the CBTi, and their experiences of identifying and recruiting clients to take part in the study. Two professionals interviewed in Phase I of the study were interviewed again towards the end of the study to assess whether or not their perceptions of the CBTi had changed.

Each STRIDE therapist was interviewed two or three times to explore how their interpretations and evaluations of the intervention evolved across the duration of the study. For efficiency, their reflections towards the end of the study were obtained through two focus groups: one conducted by CB and VD, and the other by a master’s student supervised by VD.

In Phase II (the RCT; see Chapters 3 and 4), interviews with clients and caregivers focused on their experiences of the intervention, the extent to which the techniques had been embedded into daily life, and the ‘fit’ of the intervention with other services (such as strength and balance training or the in-house rehabilitation programme provided by site 2 (see Table 5). Phase II clients were recruited from the database of trial participants using ethics committee approved recruitment and consent processes (see below). Maximum variation sampling was used (based on gender, age, degree of fear of falling, STRIDE therapist seen, and STRIDE therapist’s perception of key factors influencing activity restriction). Caregivers were recruited via clients who were interviewed.

Members of the study team were interviewed to aid understanding of the conduct of the study, or to explore key issues as they emerged. Some key team members were interviewed multiple times.

Informal discussion

Informal discussions had not been included in early versions of the protocol (but emerged as a key source of data for the study). These discussions typically arose in response to an event or issue arising in the conduct of the study. These contemporaneous discussions helped us to understand and manage emerging issues. They also allowed a more in-depth exploration of issues that were raised in formal meetings from different perspectives.

Non-participant observation

Routine practice was observed at participating sites to gain an understanding of the fit between the CBTi and existing service provision. This included observation of initial assessments, follow-up appointments and interventions provided directly by, or through, the service (see Tables 4 and 5).

A range of activities involving the STRIDE therapists was also observed, including initial training; individual and group supervision; and a small number of sessions with clients. Finally, meetings focusing on intervention development, study management and ad hoc meetings to discuss specific issues (e.g. client recruitment) were also observed with appropriate consent (see Consent). Observation was supplemented with recordings and documentation (e.g. training materials) when available.

Content and delivery of the cognitive–behavioural therapy intervention

The STRIDE therapists summarised their work with each client who completed the study on a structured form (see Appendix 6 for three completed examples). This included details of the formulation (predisposing, precipitating and perpetuating factors); perceived barriers to activity; focus of the intervention; goals; planned treatment strategies; a rating of client engagement with the CBTi; perceived outcomes of the intervention from the perspective of the STRIDE therapist; and a narrative summary. Further data on the content and delivery of the CBTi were obtained through non-participant observation of therapy sessions. Additionally, CB listened to, and made field notes on, a purposive sample of recorded therapy sessions (including some with clients who dropped out of the CBTi).

Consent

All participants in interviews, informal discussions and meetings were provided with a Participant Information Sheet and asked to provide written informed consent prior to participation. Verbal consent was sought prior to each subsequent episode of data collection. The consent form that clients completed for the RCT included permission to audio-record the CBTi sessions. Professionals sought verbal consent from clients and any companions for (non-recorded) observation of clinic appointments and interventions.

Variations to protocol

Some variations to the process evaluation protocol (versions 1–4) were required. Fewer follow-up interviews to assess changes in perceptions over time were conducted than intended. In terms of professionals, this was because the addition of extra study sites necessitated interviews with a larger sample of professionals across the three sites. In relation to clients, early follow-up interviews indicated that they yielded few additional data to justify the additional respondent burden that these posed. Most clients were therefore interviewed once, but at different points along the intervention trajectory. We interviewed fewer caregivers than anticipated, as their involvement in the CBTi proved to be the exception rather than the rule.

Delivery of cognitive–behavioural therapy intervention by non-specialist STRIDE therapists

A key aim of the study was to explore the feasibility of training HCAs to deliver the CBTi to older people. This section describes the training and supervision provided. The recruitment of the STRIDE therapists is described later (in the section on conduct and delivery of the RCT); briefly, the posts were advertised and 32 people applied, most of whom were graduates. Three people were appointed to share two full-time posts, two of whom had a background in health care.

Peer support

Peer support was identified by the STRIDE therapists as a key adjunct to the formal supervision that was provided by the psychologists. This tended to be on an ad hoc basis, although the three STRIDE therapists aimed to meet weekly, usually for 2–3 hours, particularly at the beginning of their work. None of these meetings was observed, but the STRIDE therapists described them as covering a wide range of areas, including self-audit; discussing concepts; talking through alternative strategies for use with specific clients; and general issues relating to the conduct of the research. The process of self-audit was illustrated through a discussion captured as part of a group supervision session:

We did like a mini thing between the three of us one day, where we analysed each other’s assessment to see if we came up with the same formulation as was written down . . . In all three, we all came up with pretty much the same rationale from the information provided on the assessment.

ST1, group supervision, 23 January 2013

The frequency of peer support varied throughout the study depending on workload, working arrangements and team dynamics. Different STRIDE therapists were seen as having different strengths, for example the STRIDE therapist with a physiotherapy background was seen as having particular skills and knowledge that was relevant to clients who had difficulties in managing pain.

In addition to support relating to working with clients, the meetings also focused on establishing and owning the infrastructure to facilitate delivery of the CBTi. The psychologists highlighted the extent to which the STRIDE therapists gained a sense of involvement and ownership of the CBTi:

They were doing so much problem solving during those meetings. They were thinking locked briefcases, transport, continuation sheets, forms, referral. They were thinking through all these things as – as a group in a way we should’ve thought through. But they were coming up with it on the hoof and taking ownership of how they were going to do it. And that – and again that helped them bond.

R4, informal discussion, 5 July 2012

Multidimensional nature of fear of falling

The Phase I interviews to inform the development of the CBTi indicated that fear of falling was a more complex construct than had previously been recognised and three distinct types of clients were identified (see Figures 1–3). Through the delivery of the CBTi to clients, the multidimensional nature of fear of falling was further revealed and even clients with apparently similar presentation responded differently to the CBTi:

There’s a lot more issues coming out than what we first thought I think, we had the pain, the fear and the ones that were just getting on with it. I know we had that to begin with but it’s much more complicated than that because you could have two patients with pain issues but completely different and they can do completely differently.

ST3, interview, 4 September 2012

As part of the initial formulation, the STRIDE therapists identified the factors that were thought to have precipitated fear of falling. A total of 391 precipitating factors were identified by 132 clients (four identified no precipitating factors and data were missing for two clients). After recoding the data to avoid double counting of similar factors, 317 unique factors were identified (a mean of 2.3 per client); 31% of clients identified three precipitating factors and 11% identified four or more.

The most common precipitating event was a fall, blackout or problems with balance (reported by 81% of clients). Other common precipitating factors were bereavement (21%) and life events, such as family disputes, retirement and caring responsibilities (17%). Clients also identified a wide range of other medical problems that contributed to their current difficulties (these are described below).

Clients therefore presented with a range of challenging medical and psychosocial conditions (some examples are provided in Box 3). The psychologists acknowledged that it would be challenging for any therapist to work with the complex clients recruited to the study, many of whom were described as experiencing ‘existential narrowing’ (R4) as a result of poor health, bereavement and social isolation. In this context it proved difficult for the STRIDE therapists to embed their newly acquired skills:

It can be really quite difficult when you’re actually trying to implement something that you think you’ve got the hang of now but actually the clients don’t fit that kind of you know they’re not doing what they’re supposed to you know kind of they’re not saying what they’re supposed to, they’re not presenting as we’ve been told they were going to be, and that can be quite, well very unsettling.

R4/R12, discussion, 24 October 2012

Faced with this complexity, it was crucial for the STRIDE therapists to be flexible and able to adapt their approach. Understandably, this was challenging at the outset of the study, although the STRIDE therapists grew in confidence over time. As it was clear that the factors restricting activities were rarely limited to fear of falling, the STRIDE therapists shifted from a focus purely on fear of falling to a broader focus on the range of barriers that led to clients limiting their lives and avoiding certain situations or activities:

It started off with like, fear of falling, that’s what we originally aimed at looking, but it started becoming like more, what is causing you to avoid different things [. . .].

And it didn’t matter so much as what necessarily caused the avoidance if we could reverse that cycle and that improved that situation.

Focus group 2, 17 June 2014

Working with clients with complex and unmet medical needs

Although site 3 aimed to prevent falls by reviewing and addressing risk factors of patients identified through screening in primary care, patients attending sites 1 and 2 typically had more complex medical histories and were perceived as more frail (as suggested by the difference in some of the baseline characteristics reported in Table 12). Precipitating factors relating to medical problems were recoded into three groups: musculoskeletal problems (including knee and hip replacement, arthritis; reported by 51% of clients); cardiovascular and respiratory problems (23% of clients); and other medical problems (e.g. diabetes, cancer; 38% of clients). As would be anticipated with an older client group, only a minority of clients (20%) did not describe medical problems in any of these categories, and 29% identified problems across two or three of these categories. STRIDE therapists and clients agreed that the CBTi was most effective after any medical issues had been resolved, as it enabled clients to maximise the benefits of their improved health by tackling any residual loss of confidence and encouraging them to try new activities:

If you work on that physical strength, definitely the confidence comes back. But [. . .] the variety of applications of that confidence is quicker because you’re getting different ideas of how to apply it.

C5969 interview, 29 January 2014

Professionals in sites 1 and 2 similarly suggested that greater integration of different interventions would be beneficial:

I just think it’s a missed opportunity, probably for interventions to be joined up [. . .] if somebody’s having ongoing medical intervention it may be that, that that’s not as joined up.

P20 interview, 24 March 2014

Not all complex medical conditions were optimally managed, with some clients having significant unmet needs. The STRIDE therapists were frustrated by the lack of appropriate care provided to some clients, particularly when they felt that this was creating dependency and reducing quality of life. This led to the emergence of an advocacy role to expedite appropriate interventions; this usually involved encouraging clients to seek further medical help but sometimes the STRIDE therapists intervened on a client’s behalf:

We’re signposting patients before they go into the GP practices and telling them what they need to tell the GPs because in seven minutes they’re getting nothing [. . .] we’re doing a lot of like almost skills rehearsal, ‘right the GP needs to know this’.

ST2, interview, 19 June 2013

The advocacy role occasionally taken by the STRIDE therapists was valued by staff in participating sites, as it enabled them to address any deterioration or emerging problems promptly.

The multiple medical issues faced by many clients led to a growing recognition by the STRIDE therapists that further falls were (often) inevitable. This in turn prompted a subtle shift to a more explicit focus on falls preparation whereby the STRIDE therapists aimed to minimise the psychological and physical impact of falls by making clients more robust, increasing positive safety behaviours (e.g. carrying a mobile phone) and providing advice on getting up after a fall:

Well I’m still worried about it [falling] because I’ve got Parkinson’s and I shake a lot but I was like given a lot of confidence when I actually did fall just by what she said ‘well just sit there and take some deep breaths and then try to get up and make sure you’re all right’ and that sort of thing. I’m still confident with that.

C4465, interview, 26 June 2014

Fit of cognitive–behavioural therapy intervention with existing interventions

All sites either provided in-house interventions or referred clients to appropriate community resources (e.g. Staying Steady or strength and balance classes; see Table 5). Given the varied nature of existing interventions available to clients, we explored the fit between the CBTi and these interventions. Clients interviewed for the process evaluation generally valued all of the interventions that they received. The physiotherapist at site 1 similarly saw the CBTi as being complementary to existing provision:

. . . quite often what we’re able to do is very clinic based. So I can give people exercises to do. I can do a certain amount of goal setting with them, but I can’t go out and see them at home [. . .] So I think it might be quite a nice opportunity to make some of the clinic-based stuff a little bit more meaningful.

Interview with P20, 24 March 2014

However, other professionals sometimes saw the CBTi as potentially overlapping or conflicting with existing interventions. This was particularly the case when goals related to deconditioning and clients were doing activities, such as sit to stand exercises and practising going up and down a single step. Although the STRIDE therapists considered the exercises as graded activities, as these movements were an integral component towards achieving the broader goal, clients typically perceived them as ‘exercises’ and described them as such when asked about the CBTi by professionals. Unsurprisingly, professionals at site 2 were concerned that the STRIDE therapists were giving patients additional exercises that were outside the scope of their rehabilitation programme. The study team tried to address this by arranging for a STRIDE therapist to attend a team meeting at site 2 but this did not entirely dispel their concerns about the CBTi:

So that’s why we went over there and talked to them, and told them what we were doing and how we’d gone about it a different way, and for a lot of them they were very open in that, in that meeting and for some of them, you got the impression they were sceptical. [. . .] I was very careful not to give exercises. I asked about ‘Have you been given exercises from your knee classes? What are you doing with the falls clinic? Could you do them a little bit more? Could that feed into what we’re doing?’ But a lot of it was breaking down say, sitting to standing, or people wanting to get on and off the floor to do their garden. I could break it down, because of my movement background and they might see that as being – it’s exercise but it was functional. It was very much functional, but then you did feel as though you were sort of treading a line.

ST2, focus group 1, 2 July 2014

A more general issue for the STRIDE therapists was how to manage conflicting advice. This arose most frequently in relation to the use of walking aids and care after a hip or knee replacement. All of the STRIDE therapists worked with clients who had internalised advice given after hip or knee replacement surgery and changed their behaviour permanently as a result (e.g. by never kneeling or by holding on to bannisters on the stairs). The STRIDE therapists were aware of the potential conflict between the CBTi and advice given by clinicians:

They are being told by the NHS ‘don’t do this any more; and don’t do that any more’ and ‘that would be dangerous’.

And we come in.

And we’re coming in saying ‘no, try it’.

Informal discussion ST2/ST3, 7 March 2014

The extent to which clients were willing to try the STRIDE therapists’ advice partly depended on the extent to which it was congruent with their own wishes and preferences. For example, one client who was dependent on a walking stick following a stroke wanted to return to the running track (having been a runner all of his life). The initial goal set with the STRIDE therapist was to walk round the track once or twice using his stick:

Anyway, the first time I went round, I walked round twelve times. I just put my stick down, that was it, walking round. Not just walking, speed, speed walking [. . .] The first two laps, I walked gently round. Then I’d, you know, get in my STRIDE, get faster and faster and faster. Then I’ll do ten fast laps, roughly three, four hundred metres in three, three and half minutes, or three and a quarter minutes. Which is what I’ve been doing all my life, this. Done running all my life. I felt, God, that’s great. So from that day on, I had the stick with us, all the time but I never, ever use it.

C4458 interview, 2 July 2014

Although this is an extreme example, it illustrates the ways in which clients could underestimate their physical abilities through becoming dependent on walking aids. Strategies that the STRIDE therapists might use to deal with potentially conflicting advice were explored in a group supervision session:

R4 asked whether they had tried to change clients’ behaviour. Although the STRIDE therapists had sometimes tried to tackle the issue, ‘but they’ve still got that ‘the doctor has said’. R4 suggested that one approach would be to say ‘ok your doctor was right, but that was for the acute period’. ST3 said that she had tried this approach and said that it had seemed to work. R4 elaborated that they could continue by saying: ‘that’s for the recovery period, it is absolutely the right thing to do, but in fact what you might be doing now is inadvertently keeping your knees a little bit weak . . . so agree with the doctors’ advice, so you’re not setting up an opposition between you and the doctor but just translate it, point out the difference between acute and ongoing’.

Field notes of group supervision, 19 September 2013

Other tensions between the CBTi and existing interventions related to the underlying ethos of services. This was most evident in site 2, which seemed more risk averse than the other sites. Their emphasis on safety was evident in the assessment documentation (which recorded whether or not a risk assessment was needed in relation to a range of areas, including moving and handling, medication, mental status); observations of interactions between staff and patients; and comments made by clients in therapy sessions and informal discussions and meetings with the STRIDE therapists:

. . . at the end of the assessment, a member of staff commented that it was ‘really, really safe here’. She then elaborated that two people usually walk patients down to the rehabilitation room. If patients are unsteady and are with just one member of staff, they will use a wheelchair to get them from the waiting room to the rehabilitation room.

Field notes of observation at site 2, 29 October 2013

But they tell you at the falls clinic never do exercises where you’re stretching yourself, only do them within your own limits.

C4661, session 2 with ST2, 10 December 2013

STRIDE therapist uncertainty over underlying rationales and specific techniques

The complexity of the clients inevitably challenged the STRIDE therapists in the light of their relatively limited training. Three specific areas in which the STRIDE therapists’ skills could have been developed further were the ability to transmit CBT principles to clients; tuning into client motivation; and setting SMART (specific, measurable, achievable, realistic and time-limited) goals. The psychologists anticipated that the STRIDE therapists would revert to existing patterns of interacting with clients in situations in which they were anxious or uncertain about new techniques. Two examples of such patterns were being directive when clients did not seem to be taking an active role in the CBTi, and focusing on exercise-based goals (particularly for the STRIDE therapist with a background in physiotherapy, as this was her area of expertise).

Setting specific, measurable, achievable, realistic and time-limited goals

Although the identification of valued goals that could be worked towards during the initial eight sessions was a central component of the CBTi, the ability of the STRIDE therapists to work collaboratively with clients and identify SMART goals was varied. The STRIDE therapists had no difficulties in explaining the principle of breaking down goals:

ST3 then reiterates that next time ‘we’ll come up with some goals that you want to achieve over the eight weeks’. She explains that they will ‘break them down into bite-sized chunks, or baby steps that we’ll work on over the weeks to build up to the long term goals’. She draws a parallel with running a marathon, developing a training plan and building up the distance over time.

Field notes of observation of session 1, ST3/C1521, 9 September 2013

As in the above example, the example of running a marathon was frequently used by the STRIDE therapists; more appropriate examples may have been more helpful (particularly as client feedback on the manual indicated that they valued examples with which they could identify). The importance of setting SMART goals and exploring obstacles was emphasised in supervision:

R4 highlighted the importance of defining clear homework goals; and signposted the STRIDE therapists to work on implementation intentions which indicated that people were more likely to follow through if there was a clear plan in terms of when, how often etc. [. . .] R4 went back to the issue of verbally negotiating the goal and ensuring that goals were SMART; he also suggested that they should explore potential obstacles to the goal.

Field notes of group supervision, 19 September 2012

These principles, however, were not always put into practice, and the STRIDE therapists could not always see how to break a goal into tasks:

Sometimes it’s helpful breaking these up, but some things, things don’t apply to it. So, because it’ll be like, you know, they’re like, ‘I want to arrange . . . you know, I want to go to the theatre and stuff’, okay, so it’s like, ‘Book a seat at the theatre’.

ST1, interview, 11 September 2013

In the above example, exploring implementation intentions and potential obstacles could have helped to identify specific tasks that the client could have worked on (e.g. anxieties about travel, difficulties hearing on the telephone, worries about delaying people in adjoining seats if they are not able to negotiate their way out of the seats quickly).

Identifying goals with clients who were ‘stoics’ (i.e. clients with fear of falling but no experiential avoidance; see Figure 3) was particularly difficult as:

They were really pretty much doing the things, they just needed that bit of support and guidance really in to what they wanted to do. And there’s only so much you can do with those people.

ST1, interview, 23 August 2012

This led to the emergence of a ‘supportive’ role, for which the STRIDE therapists primarily encouraged clients to continue with their existing activities rather than identifying new goals and applying CBT techniques.

Despite the strong emphasis in the training on working collaboratively with clients, as already described, when faced with clients who did not seem to ‘get’ the intervention or who seemed to lack motivation, the STRIDE therapists tended to become more directive and suggest goals. These were typically informed either by the process of assessment and formulation or through observation:

As she walked to the door I said ‘you’re touching all the walls’. I said ‘maybe we could look at creating a little programme where you improved your balance in the house’.

ST2, group supervision, 16 January 2013

Goals were also circumscribed by the physical abilities of clients and this contributed to the shift towards more exercise-based goals. The physiological knowledge of the STRIDE therapist with a background in physiotherapy was valued by the psychologists, as it complemented their knowledge base:

By nature of our training we think everything is psychological, so we forget that if you have been stuck at home for five years you’re probably not that fit, so when you do go out the house your heart rate is going to go up a lot because you’re not fit and you’re going to interpret that as an anxiety and you’re just going to start the whole panic cycle, so actually having that kind of physiological complement to that where you address the physiological deconditioning in tandem with that.

R4/R12 discussion, 7 May 2014

Reflecting the varied health and social circumstances of the clients, there was wide variation in the scope of the goals identified. For example, goals ranged from ‘picking something up off the floor’ (C1514) to ‘going to live in London for a month to learn a foreign language’ (C5072). Review of the goals recorded in the STRIDE therapist summaries highlights the non-specific nature of many goals. A total of 234 goals were identified by 129 clients (seven clients had no goals and data were missing for two clients). Of the 234 goals, context was specified for half (49%), duration for 14% and frequency for 8%. Reviewing goals and homework tasks recorded in case notes during supervision might have helped the STRIDE therapists to refine their ability to set SMART goals. The goals were coded into nine categories (Figure 30); walking-related goals were most common, with relatively few goals relating to social activities.

FIGURE 30.

Client goals for CBTi.

As part of the CBTi, it was intended that clients would grade the perceived difficulty of their goals at the outset of therapy, approximately midway through the sessions and in the eighth session. These ratings would indicate the extent to which the CBTi enabled clients to meet their individual goals and supplement the formal outcome measures. There was, however, significant variability between the STRIDE therapists in the extent to which these ratings were completed (the percentage of clients having ratings for at least one goal at all three time points ranged from 25% to 71%; p< 0.01; chi-squared). The STRIDE therapists found it difficult to elicit ratings from some clients, as they often preferred to describe their progress or difficulties verbally rather than attaching a numeric value. For the 118 goals for which ratings were available at the start of therapy and in the eighth session, the majority were rated as having reduced in difficulty (78.8%), with only a minority either remaining at the same level of difficulty (14.4%) or being more difficult (6.8%). The difficulties in consistently obtaining ratings from clients suggest that further training and discussion of eliciting ratings was needed.

Client perspectives of cognitive–behavioural therapy intervention and the STRIDE therapists

Client interviews suggested that they conceptualised the intervention in terms of four domains: increasing self-awareness; learning practical techniques; companionship; and ‘not for me’ (Table 40). Although some clients perceived the intervention to focus exclusively on a single domain, others recognised two or more of these domains. Their perspective on the intervention also reflected the extent to which they took an active or passive role in the CBTi (see Table 40).

Within the category of self-awareness there was a distinction between clients who viewed the CBTi as an opportunity to learn more about themselves and gain insight into the factors contributing to their fear of falling and those who took a more passive role and relied on the STRIDE therapist to ‘make you think about it’ (C4158). In terms of practical techniques, some clients understood the intervention in terms of identifying targets and breaking these into manageable activities, whereas others perceived the CBTi as essentially exercise based. The latter clients, together with those who conceptualised the CBTi as primarily about companionship, tended to take a more passive role.

A final group of clients perceived the CBTi as ‘not for me’; these clients tended to withdraw either from CBTi or from the RCT altogether. Previous studies of falls prevention strategies have highlighted that older adults may perceive strategies as ‘better for others than for me’. 100 Clients who were already managing their fear of falling by continuing to be active despite their fear of falling (‘stoics’) tended to see the CBTi as superfluous, as they had no experiential avoidance. Although some of these clients continued with the intervention, with the STRIDE therapist taking a supportive role, others chose to drop out (see Table 40). A small number of clients did not identify as having lost confidence and therefore did not see the CBTi as relevant (this is consistent with dropout rates being associated with lower FES-I scores; see Table 22):

So we got to the rationale part and I said well ‘it sounds as if you’ve lost a little bit of your confidence’, ‘no I’m a very confident person, I don’t think this is for me’, really shut down, and I was like ‘OK well this is not for everybody, do you want to keep hold of the manual?’ ‘No, no I’m not interested’.

ST3, group supervision, 5 September 2013

Other clients who thought the intervention was ‘not for me’ tended to have a wider sense of resignation about their lives and capabilities and were not interested in pursuing change:

It’s no good doing things that are out of your reach; you’ve got to be feasible.

C5371/ST3, withdrew after session 3, 23 January 2014

Finally, a number of clients withdrew from the CBTi for reasons unrelated to the intervention including life events (e.g. bereavement), poor health and ongoing medical investigations or treatment:

She ended up withdrawing from the CBT side of things because it wasn’t appropriate for her at that time because she had far too much that needed to be done before true CBT could be carried out.

ST1, focus group 2, 17 June 2014

Dropout rates provide some indication of acceptability of the CBTi. Overall, 65.7% of clients allocated to CBTi completed the study (all of whom completed all nine sessions of the CBTi); 8.1% withdrew from the intervention but continued to complete outcome measures and 26.2% fully withdrew. Of those who withdrew from CBTi, 12.9% did so before their first CBT session, and 21.4% during the intervention after attending an average of 3.0 (SD 2.5) sessions.

Relatively few caregivers were directly involved with the CBTi; when they provided support to clients this mainly took place outside formal sessions. The involvement of caregivers centred on providing encouragement:

I have fell a couple of times like I told ST3 and it did knock us back and my husband after reading the manual opened the front door, put us out, closed the front door and locked it and said ‘you’ve done it before, go 3 doors up and come back’ and that’s the way I did it. I started from the beginning again.

PC14, interview, 13 September 2012

Independent of their understandings of the CBTi, clients consistently valued their interactions with the STRIDE therapists and the small number of caregivers interviewed similarly emphasised the interpersonal skills of the STRIDE therapists. In describing these interactions, comments related to four domains: the development of a therapeutic alliance; negotiation; support and challenge; and technical skills/advice (Table 41). The development of a therapeutic alliance is key to the success of psychological interventions101 and the STRIDE therapists proved skilled at developing positive relationships with clients. The challenges in working collaboratively with clients to identify goals was highlighted earlier in this chapter; while most clients reported that the process of developing goals and identifying tasks was shared, some of the phrasing used suggests that the STRIDE therapists sometimes took a more directive approach:

She [STRIDE therapist] used to spend an hour just doing exercises and talking about how you feel, how your confidence feels, setting yourself goals, like going to a football match. I used to be right into that and she said ‘well get one of your sons to take you’.

C4465 interview, 26 June 2014 (author‘s emphasis in bold text)

I can truly say that if she set me an objective the first week, we would discuss it the next time. She came and discussed it but she never let it go, until she could write it off and I knew that, so I did have to try.

C5969 interview, 29 January 2014 (author‘s emphasis in bold text)

It was clear, however, that clients did not find this approach oppressive or too directive. This suggests that the STRIDE therapists modified their style in terms of working collaboratively or being more directive according to the extent to which clients took an active role in identifying goals and/or tasks. The support/challenge model highlights the importance of judging the optimal levels of both support and challenge. 98 The interviews indicated that clients felt able to modify tasks that seemed too challenging, and were fully supported by the STRIDE therapists in doing so. Some challenges seemed (from the perspective of the process evaluator) too great; for example, one client reported having been physically sick prior to going out in the dark for the first time for years; another client had been terrified during a 10-minute walk down a hill (which took her 45 minutes to complete). Both of these clients, however, reported feeling an immense sense of pride and achievement on successfully completing these challenges.

Although professionals in participating sites received relatively little feedback from patients about the intervention, any comments tended to focus on the interpersonal aspects of the CBTi and were consistent with the interview data:

I mean people have come and said, you know, ‘That nice ST3 comes out to see me’ or, ‘Isn’t ST1 lovely, she comes to see me.’ And, and so they’ll tell me more about, I think, the interaction that they’ve had, the interpersonal side of things.

P20 interview, 24 March 2014

Perceived benefits of cognitive–behavioural therapy intervention

A range of positive outcomes from the intervention were identified by clients, caregivers and STRIDE therapists (Table 42). Although some clients reported specific changes, others found it difficult to explain the outcomes of the CBTi:

All I know is how I feel and I feel different. I feel good and I don’t think ‘oh I’d like to go to such as such – Oh I couldn’t manage that’; now ‘I’d like to go such and such, right we’ll go’. It is nebulous, it’s very difficult to explain.

C3028 interview, ST1, 8 November 2013

No negative outcomes of the CBTi were reported, even when clients had fallen during the intervention period. Caregivers generally identified similar outcomes, although some felt the client was exaggerating the benefits. Some caregivers had directly benefited from the CBTi, either because of reduced anxiety or concern about the client, or because of improvements in the client’s mood (see Table 42).

Professionals working in participating sites occasionally gained some insight into client experiences of the intervention during follow-up appointments:

The ones that have been in the treatment arm all seem to be, they have engaged I think and they’ve, all the ones that come back anyway that see us, they seem to be very positive, they’ve all been doing a bit of exercise as well as whatever the CBT aspect is and most of them seem to have benefited, I haven’t had anybody who has said it’s been a negative experience at all.

P4 interview, 19 June 2014

Internalising the principles of cognitive–behavioural therapy intervention

A key aim of the intervention was to equip clients with the skills to apply CBTi principles to new contexts and become, in a sense, their own therapists. Although available data provide only limited insight into the extent to which the principles were internalised and the factors that facilitated or inhibited clients’ abilities to generalise the CBTi to new challenges, it was clear that clients varied in the extent to which they had embedded the underlying principles. For some clients, the role of the STRIDE therapist in providing encouragement seemed crucial and, when such support was withdrawn at the end of the eight sessions, they were not necessarily able to sustain changes on their own:

There are patients where you’ve gone back and it’s been like, ‘So how have you done?’ ‘Well, I did it for a bit.’ ‘So what stopped you? What curtailed it?’ ‘Oh, well I just stopped doing it as much.’ I think it’s – and they’re like, ‘Well you weren’t here’.

ST1, focus group 1, 2 July 2014

Although some clients continued with the exercises suggested by the STRIDE therapist, they did not always have the ability to identify new goals and break these into manageable (and relevant) steps. For example, when one client was asked in an interview whether he had any current goals, his response was ‘to be able to go out by myself and maybe have a pint in the pub’ (C4465, ST2). When asked about how he planned to achieve this, he explained:

. . . just actually getting around in the house and being more independent, our [name of son] still comes every morning and me other son comes at teatime and before I go to bed one of them will ring us.

C4465 interview, ST2, 26 June 2014

When the interviewer tried to clarify how this related to the goal, the client explained:

Oh I’ve got miles to go yet; it’s in [village], it’s about a mile and a half away, 2 mile, so I don’t think I’ll ever achieve that one.

C4465 interview, ST2, 26 June 2014

Applying the principles of the CBTi to going out during the winter was challenging. Some clients simply reported that they would not venture outdoors in poor weather, but one client explained how she would use the techniques learned:

If it’s just snowing I’m fine. But once it starts getting a little bit slippy, I just refuse to go out, you know? I just haven’t got that confidence for that yet, you know?

C3514 interview, ST3, 10 June 2013

I would judge on the day how I feel, what the weather is like, but I’d also take into account now I’ve had these sessions with ST2 ‘well really wait a minute how can we get around this and make it better, what if it’s going to be icy for a week you’re not going to get out, I’ve got to get out, how am I going to get out?’ So yes, I would be starting to think again ‘what’s the best way?’ Like my footwear I’ve got special things I put on over my shoes and walking you know I look where I’m going. I mean even when it’s been really bad [. . .] I’m thinking you cannot be wrapped up in cotton wool all the time, you cannot. So yes it has made a difference because I would break it down again.

PC07 interview, ST2, 1 November 2012

The manual was a valued resource for some clients who referred back to it to enable them to maintain progress despite setbacks:

The way I did it was I sort of like cut my brain in half and had a conversation with the left side of my head and the right side of my head. When the left side of my head said ‘no you were going to fall’ I took on board, if it makes sense to you, the manual and convinced myself I wasn’t going to fall because the manual said positive thinking and that’s what I did and I used to have this conversation in my head and I mean I was telling my sister and she looked at us as though I was going crackers but it does work, it does.

PC14 interview, ST3, 13 September 2012

Increased awareness of the ways in which fear of falling (often alongside physical impairments and ill health) had limited their activities often helped motivate clients to maintain activities:

I do sometimes think, ‘I’m not going back there.’ And the coffee morning, not last Saturday but the Saturday before, was that wet day and I’m sitting there [. . .] and I’m thinking, ‘I’m not going. Oh I’m not going to go. I’m not going to get soaking wet and not go. No, I’m not going to bother.’ And then I just thought, ‘Aha, get yourself organised. Go in the car.’ And that was it, so I went.

C5969 interview, ST3, 29 January 2014

Practical organisation of cognitive–behavioural therapy intervention

The perspectives of all stakeholders are included in this section. Few quotations are included in this section due to space limitations; instead an overview of the advantages/disadvantages of existing arrangements and suggested changes is provided in Table 43.

Several suggestions made by clients were already in place. For example, the STRIDE therapists occasionally accompanied clients with tasks; this was seen as a legitimate activity by the psychologists, particularly in the initial stages:

I think if people can do it on their own that’s you know it’s even better but we want safe experiences, people to have safe experiences, we don’t want failure experiences, so if it means going out with a STRIDE therapist then I would certainly encourage them.

R4/R12, discussion 19 June 2013

Although some clients suggested more flexibility in the timing and duration of sessions, this was already offered by the STRIDE therapists, as and when needed. Although the therapy nominally comprised eight sessions at weekly intervals, in reality the sessions were scheduled to fit in with clients’ other commitments (reflected in the average duration of the initial eight sessions; see Chapter 4). This meant that the 8-week assessment, designed to capture immediate postintervention outcomes, frequently took place before the CBTi was complete (see Chapter 4). The duration of sessions was also flexible; some clients (typically those who were well motivated and/or for whom the STRIDE therapists had a supportive role) had sessions of only 15–20 minutes once the intervention was under way. The flexibility in delivering the CBTi reflects the constraints of working in a ‘real world’ setting. Many of our clients were active, but many more were unwell with multiple comorbidities, and hospital and GP appointments, and hence found it difficult to commit to a rigid timetable of weekly sessions.

There was a strong preference for home visits among clients and, despite the additional costs in terms of travelling time, the STRIDE therapists also thought home visits were appropriate for this client group (particularly at the beginning of the intervention):

Well initially one of the things that you’re fighting against is going out, so you wouldn’t go to a clinic [. . .] it would be an easy cop out not to go. There’s no way would you not be in for somebody coming to your home. But you would use any excuse [not to go to a clinic], the cold, the rain, too sunny, too hot. You would use any excuse.

C5969 interview, ST3, 29 January 2014

From the STRIDE therapists’ perspectives, offering an intervention in primary care seemed more appropriate than basing a service in secondary care or specialist falls services. This would allow the early identification of clients with fear of falling (or barriers to activities) and could avoid some of the problems of implementing CBTi at the same time as addressing other medical problems:

Or even offer a service within a GP practice – if you have any concerns about your confidence, about your walking kind of book an appointment with dadadadada. You know, that may be the best way to start as such to actually having someone within your primary as supposed to your tertiary and secondary’s and stuff. Or even have sort of more person bouncing between GP practices to GP practice spending a day in each one.

ST1, focus group 2, 17 June 2014

As highlighted in previous sections, other suggested changes included a pre-assessment interview, assessment of client motivation to change, ensuring that any medical conditions were optimally managed prior to the start of the CBTi and better integration with existing interventions, in particular combining CBTi with specialist input relating to exercises.

Ideally you should do them [rehabilitation and CBTi] combined; doing one before the other doesn’t quite work as well I would have thought. Obviously we didn’t have the opportunity to really do it properly but I think if you combined some of the stuff we did in CBT with like physio rehab on the patient . . .

Like strength and balance classes.

. . . would work really well.

Focus group 2, 17 June 2014

Professionals’ views of cognitive–behavioural therapy intervention

None of the sites offered specific interventions targeting fear of falling. The views of professionals in site 3 on the CBTi have already been described (see Chapter 2). Senior staff in site 2 hoped that the CBTi would complement their existing rehabilitation programme by directly addressing fear of falling:

Fear of falling wasn’t really changing or was increasing so that’s where this comes in because we were hoping that the addition of the talk through the CBT would possibly help us.

P21/P22 interview, 28 November 2013

Although at the outset of the study, some professionals subscribed to the view that fear of falling would automatically decrease with improved physical strength, their views seemed to have changed by the end of the study:

There are some people still I see with very severe fear of falling and it’s a great shame that there doesn’t seem anything to do other than offer them physiotherapy in the vain hope that this is somehow going to improve their confidence.

P4 interview, 19 June 2014 (author‘s emphasis in bold text)

Despite the lack of robust feedback, staff at sites 1 and 2 expressed an interest in providing CBTi after the end of the study, for example by training existing staff to deliver CBT alongside their existing rehabilitation programme or by employing one of the STRIDE therapists on a sessional basis:

I think the useful part is thinking about our own healthcare assistants being trained to do it; they do a lot of the rehab and so they build up quite a relationship with patients so in that way I think it would be useful, very useful.

P21/P22 interview, 28 November 2013

In site 1, where there were perceived to be ‘pockets’ of CBT within the Trust, it was suggested that a ‘generic CBT team’ could be developed who could work into different hospital departments and clinics as needed:

. . . that means that you’re not just relying on one person so if they then leave or go somewhere [. . .] you’ve got a bit of succession planning and you’ve got a bit of a group.

P4, interview, 19 June 2014

Managing a perpetual state of flux

The study faced a number of challenges throughout, including delays with research passports, disagreements over the funding and banding of the HCAs to deliver the intervention, threats to the continued funding of the initial (single) site and recruitment difficulties which resulted in a threat to the continued funding of the study. In addition, as with most studies, changes to the protocol and sites resulted in a number of ethical amendments. The STRIDE therapists also had to cope with a number of practical challenges including not having an office or access to computers when they were first appointed. The timing of these various events meant that the study team was faced with an almost continuous stream of challenges; although these challenges are by no means unique, the number and pace of them was particularly relentless within the present study. Although morale generally remained high among core staff (i.e. those most heavily involved in the study on a day-to-day basis), there were moments at which morale dipped either individually or (less frequently) collectively:

I think there was a real loss of morale 2 weeks ago, everybody was thinking about different jobs, they were all thinking everything was ‘going down the pan’ as they put it, so even if they were invested in the therapy, they weren’t invested in the study because they didn’t believe it was working.

R4, informal discussion, 3 May 2013

As with previous research,102 delays with research passports impacted on the study either by delaying data collection by staff already in post or by delaying the appointment of temporary staff to facilitate recruitment. The process of adding the study to CB’s existing research passport took 7 months, meaning that fieldwork and analysis relating to Phase I of the study had to be condensed into a very short period. Even when additional staff were recruited to temporary posts to facilitate recruitment and/or baseline assessments, some took up other posts because of delays in obtaining their research passports. Similar difficulties were experienced in providing support with recruitment in site 2:

R30 had been keen to help on a voluntary basis with recruitment in site 2 but had not been able to start until December despite having agreed to help out in August. These delays had occurred despite having supportive human resources and research and development departments.

Field notes of TSC meeting, 16 January 2014

The recruitment of the STRIDE therapists proved more complex than anticipated. The planned grading of the posts (band 2) was contested and following extensive consultations the posts were advertised at band 4 because of the relative clinical independence of the posts. Owing to factors outside the control of the study team, the posts were advertised as CTAs based within the CRF. A generic CTA job description was used with the only permitted modification being the addition of a single sentence (on p. 5) explaining that the purpose of the post was to deliver a CBT intervention to older people. As a result, one candidate assumed that CBT would just be one part of the role and that the post would be based ‘. . . in a clinic, possibly taking blood’ (ST2 interview, 27 June 2012) and a second candidate thought ‘it was going to be like a drugs trial or something like that’ (ST3 informal discussion, 2 July 2012).

The implications for the study and the generalisability of the findings relate to whether the candidates applying for a CTA post differ systematically from those who would have applied for a HCA post (as originally intended). Several professionals queried the extent to which the STRIDE therapists were typical of HCAs:

Is ST1 a typical representative of a Band 4? I don’t know, it seems to me she’s quite exceptional [. . .] she’s got great, far greater skills I think, personal skills and engagement and again professionalism and reliability and not off sick [. . .] you didn’t really have typical people.

P4, interview, 19 June 2014

Despite their reservations, professionals recognised that it was becoming increasingly common for people with degrees to work as HCAs in the current economic climate. Indeed nearly all of the applicants for the posts were graduates:

The health-care assistants I’ve come across in previous, in private physio practices are almost all graduates with things like sports science degrees [. . .] that might just be the current demographic where there’s a lot of people who want to go into medicine or physio or sort of health-related sciences and can’t get a vocational training job.

Given the nature of the post – delivering a CBT intervention – it is likely that the post would have attracted recent psychology graduates who were interested in pursuing a career in health care. The fact that two of our STRIDE therapists were from this background suggests that the people whom we appointed were not untypical. The third STRIDE therapist had a background in physiotherapy and was studying for a master’s degree in clinical research.

The study was initially funded as a single site study. The funding of this site came under threat on three occasions during the lifetime of the study and, eventually, funding was withdrawn and the clinic closed at the end of the study recruitment period. Negotiations with the commissioners meant that we succeeded in delaying the closure of the site by reducing clinic hours for the last few months of the study. The recruitment of two further sites had significant implications for the team in terms of research governance, travelling time and in understanding how the intervention ‘fitted’ with the service already provided at each of the three sites.

Recruitment to the trial

Low recruitment rates led to a monitoring meeting with the funders in February 2013 (month 9 of the 32-month data collection period) and the potential withdrawal of study funding. The level of the core team’s investment in the study was highlighted by the intense efforts made to reach recruitment targets to ensure continued funding. This took its toll on team members both physically and emotionally and led to reallocation of resources to ensure the continued viability of the study. An agreement by the funder that the power of the study could be reduced from 90% to 80% enabled the required sample size to be reduced from 580 to 412. Even with this reduced target, the issue of recruitment remained challenging until the end of the study when we succeeded in achieving the recruitment target the day before recruitment closed. A range of factors contributed to our difficulties in meeting our recruitment target, none of which is unique to the present study. 103 An overview of the challenges and strategies implemented to address these is provided in Table 44. The strategies used to enhance recruitment were informed by Patient and Public Involvement (PPI) and client interviews.

Although using external staff provided by the CRF to introduce the study to eligible clients was successful, the study team was not always informed which clinics were being covered by CRF staff or when CRF staff were on leave. This resulted in incomplete coverage of clinics. Clinic staff were similarly unaware of when CRF cover was being provided and so were unsure whether or not they should discuss the study with patients. Improved communication enabled most of these issues to be resolved. Successful changes to the recruitment strategies resulted in an influx of work and backlogs in clients waiting to be contacted by the RAs for consent and baseline assessment, which required further shifts in resources to enable clients to be seen and randomised with minimal delay.

In order to meet recruitment targets, the STRIDE therapists became involved in recruitment. This had a number of benefits – chief among them the successful recruitment of the required number of clients for the study. Other benefits for the STRIDE therapists were an increased awareness of participating sites and more insight into the client pathway into the study. There were also wider benefits for the study, as the direct contact with the STRIDE therapists increased professionals’ confidence in the intervention:

So I knew that they were nice, nice isn’t the word, professional people, so that makes a difference because if you have somebody who is going to see somebody who is disinterested or not bothered or whatever then you don’t, then that would have an effect on their perception of our unit, but I didn’t, I knew that they weren’t like that and I knew that they you know they’re always good with their appointments and you know they turn up, they’re reliable aren’t they?

P4, interview, 19 June 2014

Capturing the outcomes of cognitive–behavioural therapy intervention

While there is a comprehensive evidence base surrounding the use of the FES-I in the evaluation of fear of falling,29,31,56,104 professionals in all sites expressed concerns about the use of the FES-I as the primary outcome measure for the trial. Sites 2 and 3 had before-and-after FES-I scores for patients who had completed the in-house rehabilitation programme, and strength and balance training, respectively. Neither of these interventions had resulted in an improvement in FES-I scores, leading staff to question whether or not the FES-I would capture any changes resulting from the CBTi. One professional additionally queried the interpretation of any changes in FES-I score:

They’re grading how confident they feel on a range of activities which may or not be important to them. So I think it’s very difficult to say, ‘Well I’ve gone from four to two on sweeping the floor’ you know, that’s good, but what does it actually mean to that person?

P20, interview, 24 March 2014

One problem for the RCT was the need to specify the outcome measures in the bid prior to the preliminary therapeutic interviews which provided an understanding of the factors involved in fear of falling and the likely outcomes of an intervention:

We need to know what sorts of things they think they are likely to change before we choose

R4, informal discussion, 13 July 2011

The interviews with clients confirmed that their underlying fear of falling might not have changed, despite a range of other positive self-reported outcomes (see Table 44). The psychologists and STRIDE therapists similarly identified a wide range of outcomes that might not be captured by the FES-I:

Some people’s lives have certainly blossomed as a consequence of it and you know that may only be the fact that you know they’re obviously walking up the stairs now within their house, or they’re actually able to walk to the kitchen and so on without you know to going out you know independently around town on you know on metros and are walking on the promenade or whatever.

R4/R12, interview, 7 May 2014

Making sense of cognitive–behavioural therapy intervention

For stakeholders to invest the necessary time and effort in making the CBTi ‘work’, the intervention needed to have coherence, that is to ‘make sense’ to the different stakeholders involved in the study. The achievement of this varied both across and within stakeholder groups. For the STRIDE therapists the work of ‘making sense’ of CBTi was initially hindered by the greater than anticipated diversity of clients and slow recruitment. Through the process of enacting CBTi over time, supported by the initial training and ongoing supervision, the STRIDE therapists’ understandings of the intervention developed, although they remained less confident with some aspects of the CBT approach. For clients achieving coherence at a conceptual level could be difficult and they had varying understandings of what the intervention involved. For some, the intervention came to make sense through the process of ‘doing’ CBTi; others, however, never ‘got’ the intervention, which could limit their ability to internalise and generalise the underlying principles of CBTi. Some professionals were uncertain about the relevance of psychological therapy to clients attending for falls assessment and the team had to work to address these concerns during the intervention development phase to ensure that staff would be comfortable recruiting clinic patients into the trial (see Chapter 2). This process was more challenging in site 2, where no members of the study team were directly involved in service provision and there was a risk averse culture. The research team shared a general understanding of the CBTi and the study as a whole, although at times different understandings about study processes (e.g. data collection) among team members became apparent, and needed to be resolved.

Investing in cognitive–behavioural therapy intervention

Engagement and commitment to making the CBTi ‘work’ – cognitive participation – was generally high. The STRIDE therapists were employed to learn the skills required for delivering the CBTi, but thoroughly embraced their role and became central to problem-solving. The training gave therapists legitimacy, despite raising their awareness of challenges in working with the client group. Over time, the therapists became increasingly engaged in the practical organisation of therapy, ‘taking ownership’ of the CBTi, as well as some study processes and documentation. Clients themselves engaged variably with the CBTi. Although the STRIDE therapists utilised a range of strategies to promote clients’ engagement with therapy, their efforts were not successful with all clients. The engagement of professionals, who had a key role in recruiting participants into the trial, was variable across the sites. The core research team were highly motivated to complete the study, solving problems as they presented, and actively engaging in additional roles and tasks where required to achieve study milestones (e.g. recruitment).

Enacting cognitive–behavioural therapy intervention

The collective action work around conducting the CBTi, and the trial as the supporting structure for it, presented challenges for all stakeholders. The successful conduct of both the CBTi and the study were dependent on the resources, arrangements, relationships and skills that were required to make the intervention work. For the STRIDE therapists, a series of practical problems (e.g. not having dedicated office space, logistical barriers to travelling to appointments) initially impacted on their capacity to efficiently complete their work but were resolved over time. Their work was facilitated by the support provided by the team (and each other) for ongoing skill development, and by the high level of confidence from the team in the work they did with clients. The extent to which clients were actively involved with the therapy varied widely; although some enthusiastically set themselves challenges and made significant progress in a short space of time, others relied on the STRIDE therapist to provide impetus. This variation seemed to reflect differences in personal approaches and motivation, perceived benefits of the CBTi and the complex and multiple comorbidities that they experienced. Professionals contributed variably to recruiting participants to the trial, for a range of reasons relating to both individual (e.g. attitudes towards therapy) and structural/cultural (e.g. risk averse) factors. The core research team worked hard to address problems that arose in conducting the study, and although highly motivated, experienced challenges in resolving some issues

Appraising and adapting cognitive–behavioural therapy intervention

A final NPT mechanism, key to understanding the embedding of the CBTi, is the capacity to appraise the value of the intervention and scope for making improvements and adapting it the local contexts – reflexive monitoring. The STRIDE therapists demonstrated a high degree of reflectiveness and adaptation throughout the delivery of the CBTi, to the extent that some changes in the focus of the intervention raised questions over the identity of the CBTi. The extent to which clients internalised the skills that they had learned and applied these to new situations following the completion of therapy was variable. For some clients, increased self-awareness, together with an understanding of basic CBT techniques (such as graded exposure) enabled them to sustain changes. Others, however, found it difficult to generalise CBT principles to new situations without the support of the STRIDE therapists. Professionals at the sites had much more limited capacity to make appraisals of any impact of the CBTi, as there were no formal mechanisms established for staff to receive feedback about clients following therapy. Some received informal feedback from patients who were clients in the study, at times this was positive but in some instances this generated professionals’ anxiety about increased risk of falling. As noted already, the core research team were continually engaged in problem-solving and ‘troubleshooting’ as the study progressed. However, a higher degree of anticipation of problems and forward planning may have facilitated the conduct of the study, and reflections on learning from the process to apply in further studies have been made by the team.

Cognitive–behavioural therapy intervention costs

The costs of the intervention (CBTi) were microcosted and estimated on a per-patient basis for those in the intervention arm of the trial. The intervention costs comprised two components: the costs of (1) training and supporting three HCAs who delivered the CBTi and (2) its delivery. The costs of training the HCAs who delivered the intervention included a 5-day programme that was delivered by two psychologists (costed as clinical psychologists). During the delivery of the intervention, each HCA received a 1-hour fortnightly one-to-one supervision meeting (with NSa) plus a 1-hour monthly team meeting (with VD). The costs of training were apportioned to each individual taking part in the intervention arm of the trial.

Valuation of NHS and informal care giving resource use

With the exception of the CBTi intervention, patient-level resource use relating to falls and mobility, including all NHS and informal caregiver costs, were identified and measured using a resource utilisation form (see Appendix 7) self-completed by participants in both arms of the trial at 6 and 12 months after randomisation. The use of the resource utilisation form allowed us to divide NHS resource use into secondary care, primary care, community care and ambulance journeys. It also allowed the collation of informal care giving time. Secondary care resources included inpatient stays, accident and emergency (A&E) visits and outpatient visits. Primary and community care resources included general practice (GP) and home visits, practice nurse visits at the surgery, and district nurse visits in the home. With respect to the use of ambulances, the number of trips was recorded. Informal caregiving time comprised the average time per week spent by family and/or friends who were helping participants with activities that they would have usually been able to undertake if they had not experienced problems with mobility or a fall.

For each trial participant, all components of treatment costs stratified by category of resource use were computed by multiplying units of resource use by their unit costs. These were then summed over all resource-use categories to obtain a total annual cost for each participant both from a NHS perspective and a NHS and caregiver perspective. This was then used to generate the average cost per patient in each arm of the trial.

Unit cost sources included NHS reference costs and Unit Costs of Health and Social Care. 108 The unit costs utilised for each category of resource use can be found in Tables 45 and 46 and are expressed in 2013–14 UK pounds sterling (£). 108

Sensitivity analysis

In addition to the stochastic sensitivity analysis, which was undertaken to allow presentation of the level of variance around outcome measures included in the CEA and CUA, sensitivity analysis was also performed to gauge the impact of varying key assumptions and/or parameter values in the base-case analysis. Specifically the cost of an inpatient admission was varied. This was conducted to more accurately capture the impact of length of stay on total costs as a few participants had very high lengths of stay. For this analysis, rather than use a single tariff for inpatient admissions, we separated the hospitalisation costs and the treatment costs by utilising the relevant non-elective long stay reference cost for length of stay then multiplying this by the average length of stay for the national tariff. These costs were then subtracted from the national tariff to estimate the treatment costs without any length of stay component. Total costs per patient were calculated by multiplying the treatment cost by the number of hospital admissions and multiplying their length of stay by the non-elective long-stay reference cost relevant to the national tariff.

Missing data

Missing data can be separated into:

1. Item-level missingness Data missing as a result of blank returns on particular items of questionnaires that had been returned partially completed.

2. Missing because of questionnaire non-response Data missing as a result of dropout.

Item level missingness for resource use and costs was dealt with by using a simple imputation which involved imputing the lower and upper interquartile range for participant’s use of each item of the resource utilisation form when there were missing data. We used this range as the mean value was very skewed because of a small number of participants who used very high quantities of medical services and the majority of participants who responded to the questionnaires had little resource use. We used a simple imputation method to interpolate missing HRQoL data at 8 weeks and 6 months by utilising the weighted average of each participants’ adjacent observations at each time point.

Missing cost and QALY data resulting from dropout were imputed using multiple imputation to allow for comparison and test the robustness of the base-case results. The regression analyses used to impute missing data included the following variables measured at baseline: age, sex, gait and balance score, handgrip strength, functional reach, falls efficacy, concern about falling (0–10 scale), pain when walking, participation score, loneliness score and site. In addition, the number of sessions of the CBTi completed was included as a predictor variable. The QALYs and costs were imputed at 12 months using regression models with the above variables included as predictors. Each of these methods was undertaken separately for each study group and both groups together (with group added as a predictor variable). In each case, 50 data sets were imputed and a pooled estimate of the impact of the intervention at 1 year across data sets was based on an analysis of covariance model with adjustments for baseline (with QALY data) and randomisation strata.

Data completeness

Data were missing either because questionnaires were not fully completed at all time points or because of patient dropout, as presented in Chapter 4. A total of 415 participants were randomised: 205 to the control arm and 210 to the intervention arm. A participant was included in the deterministic cost analysis and categorised as complete data if they had returned either a 6- or 12-month resource utilisation form or both of these. This gave the maximum numbers for analysis. The number of participants included for analysis was 169 (82%) in the control arm and 155 (74%) in the intervention arm. Thus, a total of 324 (78%) participants who had returned information on resource use either at 6 or 12 months’ follow-up were included in the analysis. Missing data within the resource-use questionnaire varied across questions (4.6–9.9%). Complete EQ-5D (5L) data to estimate QALYs for which patients completed the questionnaires at each of the four time points were available for 112 (61%) of patients in the control arm and 85 (46%) of patients in the intervention arm. The number of responses returned with completed EQ-5D (5L) questionnaires by trial arm are shown in Figures 31 and 32.

FIGURE 31.

Completion of EQ-5D (5L) data at baseline and follow-ups in the intervention arm.

FIGURE 32.

Completion of EQ-5D (5L) data at baseline and follow-ups in the control arm.

The figures suggest that the loss to follow-up in the intervention arm was greatest at 8 weeks and 6 months compared with the control arm. Those who completed the EQ-5D (5L) questionnaires at 6 months, in the intervention arm, appeared to be no less likely to be lost to follow-up than those who completed the 6-month follow-up in the control arm.

Complete SF-6D data to estimate QALYs for which participants completed the SF-36 at each of the four time points were available for 84 (45%) of participants in the control arm and 69 (36%) of participants in the intervention arm. The number of responses returned with completed SF-36 questionnaires by trial arm is shown in Figures 33 and 34. These completion rates are lower than those for the EQ-5D (5L), which might be expected given the higher response burden associated with the SF-36.

FIGURE 33.

Completion of SF-36 data at baseline and follow-ups in the intervention arm.

FIGURE 34.

Completion of SF-36 data at baseline and follow-ups in the control arm

As in the case of EQ-5D (5L), the majority of cases were lost to follow-up before month 12, with more losses occurring in the intervention arm.

Complete cost and FES-I outcome data were available for 310 (75%) patients: 159 (78%) in the control arm and 151 (72%) in the intervention arm. Complete cost and EQ-5D (5L) QALY data were available for 187 (45%) patients: 79 (39%) in the control arm and 108 (51%) in the intervention arm. Complete cost and SF-6D QALY data were available for 152 (37%) of patients: 69 (34%) in the control arm and 83 (40%) in the intervention arm. The base-case cost-effectiveness analyses used complete cost and outcome data (n = 310 for FES-I; n = 187 for EQ-5D (5L) QALYs; n = 152 for SF-6D QALYs), but we also include results for all available data, where possible, to maximise transparency of results.

Costs and resource use

Effectiveness: quality-adjusted life-years

The QALY gains for CBTi compared with usual care only are presented with utility values from both EQ-5D (5L) and SF-6D. As noted above, the results presented in this section are based on only those participants with complete HRQoL responses at all four time points (n = 197 for EQ-5D (5L); n = 153 for SF-6D).

The model used to estimate a 1-year QALY was two-twelfths of any utility gain (or loss) made between baseline and 8 weeks plus four-twelfths of any gain (or loss) between 8 weeks and 6 months plus six-twelfths of any gain (or loss) between 6 and 12 months (see Equation 1, above). In each case sex, study site and baseline utility were controlled for.

where u_0,i is baseline utility for the ‘i th’ participant; u_2,i, u_6,i and u_12,i are the utilities at 8 weeks, 6 months and 12 months; s_1i and s_2i are the dummy variable for the different sites; f_i is the dummy representing the gender of the ‘i th’ participant; and CBT_i is the intervention dummy.

Incremental cost-effectiveness

These are based on a data set of complete cases both in terms of costs and outcomes (n = 310 for FES-I; n = 187 for EQ-5D (5L) QALYs; n = 152 for SF-6D QALYs).

Incremental cost-effectiveness ratio based on Falls Efficacy Scale–International version

The results show that, using FES-I as an outcome, on average, CBTi was more costly and more effective than usual care alone (Table 52). The incremental cost per unit reduction in FES-I for CBTi compared with usual care, only from a NHS perspective, was £24. The incremental cost per reduction in FES-I for CBTi compared with usual care, from a NHS and caregiver perspective, was £71.

TABLE 52 - Cost-effectiveness analysis (n = 310) using FES-I: CBTi vs. usual care alone

Intervention	Cost (£)	Comparison of FES-I at 12 months adjusted for baseline	Incremental cost per unit reduction in FES-I (£)	Probability that the intervention is cost-effective for different threshold values for different threshold values for society’s WTP for a unit reduction in FES-I:
				£0	£1000	£10,000
NHS perspective
CBTi	704	5.67	24	0.33	1	1
Usual care only	624	2.28		0.67	0	0
NHS and caregiver perspective
CBTi	2154	5.67	71	0.29	1	1
Usual care only	1915	2.28		0.71	0	0

The deterministic results do not illustrate the imprecision surrounding estimates but these can be illustrated using the results of the bootstrapping. The estimated likelihood of treatment, akin to the 1 – p-value in a one-sided statistical test produced from the results of the bootstrapping, are also shown in Table 52.

Graphically, the results of bootstrapping are shown for FES-I (Figure 35) and cost-effectiveness of CBTi being cost-effective compared with usual care only. This figure illustrates the statistical imprecision surrounding estimates of costs but the majority of iterations produced from the bootstrapping analysis are in the north-east and south-east quadrants, which illustrates that CBTi has a more beneficial impact on FES-I than usual care alone. The CEAC presented in Figure 36 is the graphical presentation of the data presented in Table 52 on the likelihood that each treatment is cost-effective. This figure highlights that as society’s WTP for a unit reduction in FES-I increases, there is an increase in the likelihood that CBTi and usual care would be considered cost-effective.

FIGURE 35.

Cost-effectiveness plane showing bootstrapped replicates of the ICER for NHS costs and FES-I.

FIGURE 36.

Cost-effectiveness acceptability curve for NHS costs and FES-I, showing the probability that the intervention is cost-effective at different WTP thresholds.

Figures 37 and 38 show the analysis from the NHS and caregiver perspective. These figures illustrate shows broadly similar results to those when the perspective was of the NHS alone.

FIGURE 37.

Cost-effectiveness plane showing bootstrapped replicates of the ICER for NHS and caregiver costs and FES-I.

FIGURE 38.

Cost-effectiveness acceptability curve for NHS and caregiver costs and FES-I showing the probability that the intervention is cost-effective at different WTP thresholds.

Incremental cost per quality adjusted life-year gained based on European Quality of Life-5 Dimensions (5 Level) responses

The results in Table 53 show that, using the EQ-5D (5L) QALY, from the NHS perspective, CBTi was, on average, more costly and less effective than usual care alone. From this cost perspective, CBTi was dominated, on average, by usual care. However, from a NHS and caregiver perspective, CBTi was less costly but less effective than usual care. Thus, the incremental cost per QALY gain for usual care compared with CBTi and usual care from a NHS and caregiver perspective was £3250. The difference in finding is caused by differences between the two analyses in which participants were able to contribute complete data.

TABLE 53 - Cost-effectiveness analysis (n = 187) using EQ-5D (5L) QALYS: CBTi vs. usual care alone

Intervention	Cost (£)	Change in EQ-5D (5L) from baseline	Incremental cost per QALY (£)	Probability that the intervention is cost-effective for different threshold values for different threshold values for society’s WTP for a QALY:
				£0	£1000	£30,000
NHS perspective
CBTi	764	–0.001	Dominated	0.36	0.35	0.24
Usual care only	670	0.003		0.64	0.65	0.76
NHS and caregiver perspective
CBTi	2017	–0.001		0.49	0.49	0.40
Usual care only	2030	0.003	3250	0.51	0.51	0.60

Figures 39 and 40 graphically show the results of the stochastic analysis from a NHS perspective. Figure 39 illustrates that the majority of iterations produced from the bootstrapping analysis are in the north-west and south-west quadrants, and that CBTi is likely to have has a less beneficial effect on QALYs than usual care alone. The corresponding CEAC is presented in Figure 40. As this figure illustrates, there is, approximately, a 70% chance that usual care alone would be considered cost-effective over the range of threshold values for society’s WTP for a QALY presented.

FIGURE 39.

Cost-effectiveness plane showing bootstrapped replicates of the ICER for NHS costs and EQ-5D (5L) QALYs.

FIGURE 40.

Cost-effectiveness acceptability curve for NHS costs and EQ-5D (5L) QALYs, showing the probability that the intervention is cost-effective at different WTP thresholds.

Figures 41 and 42 show the same analysis from the NHS and caregiver perspective. Figure 41 illustrates the statistical imprecision surrounding estimates of costs but the majority of iterations produced from the bootstrapping analysis are in the North West and South West quadrants and this clearly illustrates that CBTi is likely to be less effective compared with usual care alone. The CEAC presented in Figure 42 highlights the increase in the likelihood that usual care alone would be considered cost-effective as society’s WTP for a QALY gain increases.

FIGURE 41.

Cost-effectiveness plane showing bootstrapped replicates of the ICER for NHS and caregiver costs and EQ-5D (5L) QALYs.

FIGURE 42.

Cost-effectiveness acceptability curve for NHS and caregiver costs and EQ-5D (5L) QALYs showing the probability that the intervention is cost-effective at different WTP thresholds.

Incremental cost per quality-assisted life-year gained-based Short Form questionnaire-6 Dimensions data

The results in Table 54 show that, when using SF-6D QALYs as an outcome, and from a NHS perspective, CBTi was, on average, more costly and more effective than usual care alone. The point estimate of the incremental cost per QALY for CBTi compared with usual care only from a NHS perspective was £38,000. Table 54 also shows that from the NHS perspective and caregiver perspective, CBTi was, on average, less costly and more effective using the QALY than usual care alone. From this cost perspective, usual care alone was on average therefore dominated by CBTi and usual care.

TABLE 54 - Cost-effectiveness analysis (n = 152) using SF-6D QALYs: CBTi vs. usual care alone

Intervention	Cost (£)	Change in QALYs	Incremental cost per QALY (£)	Probability that the intervention is cost-effective for different threshold values for different threshold values for society’s WTP for a QALY:
				£0	£1000	£30,000
NHS perspective
CBTi	771	0.021	38,000	0.40	0.40	0.39
Usual care only	695	0.019		0.60	0.60	0.61
NHS and caregiver perspective
CBTi	1913	0.021		0.67	0.67	0.66
Usual care only	2155	0.019	Dominated	0.33	0.33	0.34

The statistical imprecision surrounding NHS costs, QALYs and incremental cost per QALY is illustrated in Figure 43, which shows that the majority of iterations are in the north-east and north-west quadrants, meaning that the intervention group is likely to be more expensive than usual care alone. The CEAC presented in Figure 44 highlights, as society’s WTP for a QALY, the likelihood usual care alone would be considered cost-effective over a range of threshold values.

FIGURE 43.

Cost-effectiveness plane showing bootstrapped replicates of the ICER for NHS costs and SF-6D QALYs.

FIGURE 44.

Cost-effectiveness acceptability curve for NHS costs and SF-6D QALYs, showing the probability that the intervention is cost-effective at different WTP thresholds.

The results for the NHS and caregiver perspective are shown in detail in Figures 45 and 46. These results show that usual care is unlikely to be cost-effective over the range of threshold values for society’s WTP for a QALY presented.

FIGURE 45.

Cost-effectiveness plane showing bootstrapped replicates of the ICER for NHS and caregiver costs and SF-6D QALYs.

FIGURE 46.

Cost-effectiveness acceptability curve for NHS and caregiver costs and SF-6D QALYs, showing the probability that the intervention is cost-effective at different WTP thresholds.

Allowance for uncertainty

In conjunction with our stochastic sensitivity analyses, we also conducted deterministic sensitivity analyses to explore uncertainty in the base-case analysis.

Implications of missing data

Summary of cost-effectiveness results

Using FES-I as an outcome measure, the base-case analysis suggests that CBTi is more costly and more effective than usual care alone. Stochastic sensitivity analysis suggests that when society’s WTP for a unit change in FES-I was > £20, the probability of CBTi and usual care being cost-effective would approach ‘1’. The analogous threshold for NHS and caregiver perspective was £58. These base-case results were insensitive to changes explored in all of the sensitivity analysis, including multiple imputation.

When QALYs derived from EQ-5D (5L) were taken as the outcome measure, the results from a NHS costing perspective suggest that CBTi was, on average, more expensive and less effective than usual care alone. CBTi is therefore dominated by usual care and cannot be considered as the most cost-effective treatment option. However, from a NHS and caregiver costing perspective, CBTi was cheaper than, but less effective than, usual care. However, stochastic sensitivity analysis suggested that the intervention is unlikely to be cost-effective over a range of threshold values for society’s WTP for a QALY. These base-case results were insensitive to the sensitivity analyses, including the multiple imputation, for which the intervention is dominated by usual care from both costing perspectives.

When QALYs were based on the SF-6D scores, from a NHS perspective costing perspective the intervention was more costly but more effective than the control. From a NHS and caregiver perspective, although the intervention was more effective, it was less costly than the control. However, results from the CEAC suggest that it is unlikely that CBTi would be cost-effective at typical values that society might be willing to pay from a NHS costing perspective. Only when costed from a NHS and caregiver perspective could the intervention be considered to be cost-effective. The results of the sensitivity analyses, and, in particular, the multiple imputation, suggested that CBTi was dominated by usual care and cannot be considered as the most cost-effective treatment option.

Recruitment

The level of skill and autonomy required by the HCAs meant that employment at NHS Band 4 or above would be vital, with recruitment taking into account understanding and experience of the client group, empathy, team working and communication and problem-solving skills.

Training

Health-care assistants needed to learn complex new skills in order to deliver the CBTi. Initial training was short, with ongoing supervision by the psychologists. The training was generally felt to be adequate by the psychologists and the HCAs, and could usefully be adopted as core training for potential HCAs delivering the CBTi. Key areas to address in future training include an emphasis on the rationale for the CBTi before moving on to more detailed techniques to ensure that trainees understand the intervention in a more holistic way rather than focusing on a more protocol-driven, technical approach. One of the most difficult skills for the HCAs to master was how to introduce CBT to clients, many of whom were sceptical of ‘talking treatments’. Greater emphasis on modelling, role playing and practising the introduction of CBT should facilitate client understanding of, and engagement with, such an intervention at the earliest stages of treatment.

The use of ‘real play’, for which trainees use their own issues and problems to explore CBT techniques, was key in helping trainees start embedding their skills while providing insight into the experience of being a client at the start of their training. This approach further facilitates learning, with the use of CBT techniques during training enabling trainees to address their own anxieties and thereby increasing their confidence in the intervention. Soft skills, such as empathy and collaborative working should be more explicitly worked on, alongside training on the support and challenge model. Working towards valued goals identified by patients is an important part of effective and collaborative therapy. Although goal setting was covered in the initial training, data suggest that detailed review of patient goals in supervision would help to embed the principles of SMART goal setting and ensure genuinely individualised therapy.

Finally, top-up training should be available to trainees to address any skill gaps and provide opportunities to practise skills at the point at which they are being used. Initial training provides an excellent grounding, but there is a great deal of new material delivered in a relatively short space of time; reinforcement over time is vital.

Supervision

The HCAs’ ability to embed and develop their CBTi skills was facilitated by individual and group supervision sessions and the HCAs’ self-initiated peer support sessions. The ‘real play’ techniques, highlighted above, are powerful tools in the training process and need to be appropriately resourced and supervised.

Consideration of the role and availability of additional multidisciplinary support is helpful for trainees, given the complex needs of many of their client group. Very few of the clients who received the CBTi would have considered or been eligible for help from existing psychological services despite the NHS IAPT (Improving Access to Psychological Therapies) programme. This was exemplified by our HCAs’ growing health advocacy role, through which individual clients’ new or worsening intercurrent illnesses (e.g. increasing confusion, blackouts, chest pain, injurious falls) required medical and other input that would otherwise have been left to a much later stage. The study has graphically highlighted the extent to which older people’s lives are curtailed by medical and psychosocial events that are frequently unreported to health and social care professionals.