Behavioural activation versus guided self-help for depression in adults with learning disabilities: the BeatIt RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 10/104/34. The contractual start date was in January 2013. The draft report began editorial review in March 2017 and was accepted for publication in March 2018. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Christopher Williams reports personal fees from Taylor and Francis (Abingdon, UK) and from Five Areas Ltd (Clydebank, UK) outside the submitted work and is president of the British Association for Behavioural and Cognitive Psychotherapies (BABCP). Sally-Ann Cooper reports grants from the National Institute for Health Research Health Technology Assessment programme during the conduct of the study.

Copyright statement

© Queen’s Printer and Controller of HMSO 2018. This work was produced by Jahoda et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2018 Queen’s Printer and Controller of HMSO

Depression and learning disabilities

Depression is common and places a significant burden on health services. For example, in 2011/12 the cost of prescribed antidepressant medication alone was £31.4M in Scotland1 and in 2015 it was £284.7M in England. 2 From 2008–12, of all psychiatric hospital discharges, mood disorders were the most common discharge diagnosis in women. 3 Depression is clearly a major public health challenge: it is the third leading contributor to the global burden of disease, and is expected to rise; the World Health Organization predicts that it will be the second leading contributor to the global burden of disease by 2020. 4

The term ‘learning disability’ refers to people who have significant impairments of both intellectual and functional ability, with age at onset occurring before adulthood. A significant proportion of the UK population has learning disabilities. Approximately 2% of adults and 3.5% of children have an intelligence quotient (IQ) of < 70. 5,6 Individuals with learning disabilities have higher levels of mental ill health than the general population, with a point prevalence of 40% for adults. 7

Depression is at least as common in adults with learning disabilities as in the general population, with a point prevalence of ≈5%. 8,9 Indeed, depression is the most common type of mental ill health experienced by adults with learning disabilities;7 anxiety disorders are also common, with a point prevalence of ≈4%. 10 Depression is more enduring than for the general population,11 suggesting that it is either a more severe condition, or more poorly managed. For example, a study with a British cohort found that adults with learning disabilities were four times more likely than the population without a learning disability to meet criteria for chronic depression over a 28-year period. 12 The US 2000 incident cohort with learning disabilities has been calculated to have lifetime costs (in excess of costs for people without learning disabilities) of US$44.1B. 13 Poorly addressed depression makes a clear contribution to these costs. Hence, as well as the human suffering that depression brings to people with learning disabilities, their families, local communities and society more widely, inadequately managed depression is a financial burden.

Psychological therapies for depression

In recent years there have been important innovations in the treatment of depression. A number of high-intensity psychosocial interventions are as effective as, and longer lasting than, medications in the treatment of non-psychotic depression. 14 This was confirmed in a recent individual patient-level meta-analysis with over 1700 patients treated in randomised controlled trials (RCTs). 15 The National Institute for Health and Care Excellence (NICE) recommends cognitive–behavioural therapy (CBT) to treat mild to moderate depression,16 and high-intensity forms of CBT delivered by mental health experts are recommended to treat moderate and severe levels of depression. 17 There is now an increasing emphasis on low-intensity delivery, as used in the Improving Access to Psychological Therapies (IAPT) programme in England. These include written self-help books, computerised CBT and self-help groups. 18–22 The evidence base for these approaches continues to grow, given the attraction of therapies that are less resource intensive.

In behavioural activation, the focus is on behaviour more than on cognition, emphasising engagement with potential environmental reinforcers. Behavioural activation also takes account of valued activities,23 emphasising the importance of purposeful routine activities such as household chores and self-care, as well as achievement, pleasure and closeness to others. Avoidance is a key target for change, with the aim of breaking the vicious cycle linked to mood and activity, whereby reduced activity lowers mood. In turn, the worse people feel, the more withdrawn they become. Behavioural activation has been shown to be at least as effective as antidepressant medications, and superior or non-inferior to CBT, placebo pills and treatment as usual (TAU) among patients with more severe depression,24–27 with effects as lasting as CBT following treatment termination. 28 Behavioural activation appears to be less complicated to learn than CBT, it is possible to train non-specialist nurses to deliver it26 and it can be delivered as a high-intensity or low-intensity therapy. 29,30

Psychological therapies for depression in adults with learning disabilities

Although psychological therapies have become established first-line interventions for depression in the general population, this has not been the case for adults with learning disabilities, owing to the additional complexities involved in making these interventions accessible to adults with cognitive and verbal communication impairments. Arguably, psychological therapies are more advisable than pharmacotherapy for adults with learning disabilities than for the general population, as limitations in verbal communication skills reduce their ability to report and describe adverse effects of drugs, which can be further disabling or potentially have serious health effects. Perhaps more importantly, people with learning disabilities should have the opportunity to access effective psychological therapies just like anyone else. Awareness of the inequity in provision of psychological therapies has grown, but there remain considerable limitations in the existing evidence base, and in its implementation. This was recently synthesised by the NICE guideline31 on mental health problems in people with learning disabilities. A key point is the need for modifications to the treatment interventions, depending on the type and extent of need of each adult with learning disabilities. 32

In 2016, NICE identified that the only available evidence on psychological interventions for depression in people with learning disabilities was for CBT, adapted for people with learning disabilities. Only three RCTs (total participants, n = 130)33–35 and three controlled before-and-after studies (total participants, n = 130)36–38 were identified that explored the use of CBT for the treatment, or prevention, of depression in adults with learning disabilities. These included only participants with mild, or mild and moderate, learning disabilities. Only two studies reported outcomes beyond the immediate end of treatment. 35,36 Although the trials had variations in the adaptations made to the intervention – and suffered from inadequate power as a result of the small size (being feasibility or pilot studies), leading to imprecise estimates – NICE concluded that CBT may result in a clinically meaningful reduction in depressive symptoms over TAU at 38 weeks’ follow-up. However, the combined evidence was assessed, using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach, as being of very low quality. One feasibility RCT with 32 participants also evaluated the cost-effectiveness (from the perspective of the NHS and social care services) of individual CBT constituting 16 weekly 1-hour sessions. 35 The cost difference between the two arms was £5650 but given the small size of the study, it is unclear whether or not CBT is a cost-effective option for people with mild to moderate learning disabilities. NICE recommended adapted CBT for depression be considered for adults with mild/moderate learning disabilities, but were not confident enough in the evidence to make a strong recommendation.

There are possible advantages to considering the use of behavioural activation for people with learning disabilities, given that it is less complicated to learn than CBT and can be delivered by non-specialists. However, the only study on behavioural activation with people with learning disabilities and depression was the feasibility study we undertook to inform the design of the study reported here. It was a pre–post trial with 22 participants recruited over a 12-month period, at one site. 39 Outcomes showed evidence of positive change on depressive symptoms for those able to self-report on the Glasgow Depression Scale40 pre- and post-intervention, and at 3 months’ follow-up after end of treatment. We therefore used this study to inform the design of the trial reported here.

Trial objectives

Study design

This was the first large-scale RCT of an adapted individual psychological therapy for people with learning disabilities and a mental health problem. The design was a multicentre, single-blind RCT of adapted behavioural activation compared with an adapted, guided self-help intervention. The behavioural activation intervention was referred to as ‘BeatIt’, and the guided self-help intervention was called ‘StepUp’.

To ensure that it would be possible to recruit to the study, there was an internal pilot phase. Consequently, there were two phases of data collection:

• Phase 1 – there was an initial 7-month internal pilot phase in Scotland, in which the criterion for success was to recruit a minimum of 20 participants (approximately three per month).

• Phase 2 – following the successful completion of the internal pilot phase, the study sites in England and Wales were opened and recruitment continued at all three sites for an additional 11 months.

The trial was supplemented with an economic evaluation to consider the cost-effectiveness of providing the intervention compared with the attention control (see Chapter 6). There was also a qualitative study to explore the views and experiences of participants, their supporters and therapists who took part in the trial (see Chapter 7). The trial protocol has been published. 41 A description of all approved changes to the original protocol that were submitted are shown in Table 1.

Ethics approval and research governance

Multicentre approval was granted by the West of Scotland Research Ethics Committee 3. Research and development approval was granted in all study sites: Cumbria Partnership NHS Foundation Trust, Lancashire Care NHS Foundation Trust, Betsi Cadwaladr University Health Board, South Staffordshire and Shropshire Healthcare NHS Foundation Trust, NHS Lanarkshire, NHS Ayrshire and Arran, and NHS Greater Glasgow and Clyde. The International Standard Randomised Controlled Trial Number (ISRCTN) reference for the study is ISRCTN09753005.

Participants

A multipoint recruitment strategy was adopted with the aim of recruiting participants with mild to moderate learning disabilities and clinical depression across three study sites: (1) Scotland (Greater Glasgow and Clyde, Lanarkshire and Ayrshire), (2) England (Cumbria and Lancashire) and (3) North Wales (Betsi Cadwaladr University Health Board). Participants were also recruited from Shropshire and Staffordshire specialist learning disability services. For pragmatic reasons, Shropshire and Staffordshire were administered by the North Wales site. The recruitment sites comprised both rural and urban areas.

Potential participants were screened by research assistants and considered eligible for recruitment if they met all of the following inclusion criteria and did not meet any of the exclusion criteria.

Recruitment procedure

Although a multipoint recruitment strategy was adopted, most participants were recruited through specialist community health teams for people with learning disabilities. A smaller number of participants were recruited through social work colleagues, third sector organisations working alongside the community teams and IAPT services in Lancashire, which are open to anyone with mental health problems.

The approach to recruitment was for members of the research team to meet with members of the community teams or other organisations and provide an explanation about the study and suitable participants. This was a critical task, as few people with learning disabilities self-refer or are given psychological help for depression. Most receive help because their difficulties have become a problem for someone else. Referrals to health services are often for behavioural difficulties such as anger management problems or anxiety disorders that are proving disruptive or difficult for others to manage or support. Therefore, when describing who might be suitable participants, members of the research team highlighted that it would be important to bear in mind individuals whose depressive symptoms might be overshadowed by other presenting problems. Members of the community teams or other organisations were then able to identify potential participants, provide them with a brief explanation of the study and give them an information pack. These packs contained a letter, information sheets and a Freepost envelope that they could return if they were interested in finding out more about the study. It was suggested to participants that they might find it useful to discuss the study information with a friend or supporter.

Adults with learning disabilities are often supported by several people. For example, they may have multiple paid carers working in shifts or different supporters in their home and day-centre environments. This means that the potential participant may not have known who they could discuss the study with. This could have created a situation in which information sheets went missing before potential participants were able to discuss the study with supporters and make an informed decision about whether or not they wanted to participate. To take account of this, as in previous studies, the member of staff who gave out the information sheet was asked to notify a NHS secretary in the learning disabilities team (who was independent of the research study) that an information sheet had been handed out. After 2 weeks, if no tear off reply slip had been received, the NHS secretary contacted the individual once, by telephone, to check that they still had the information pack. If the information pack had gone missing, a second information pack was sent out.

On receiving a reply slip, a member of the research team arranged to meet the potential participant at their home or another convenient location for them. The participant was also asked if they would like someone to support them when they met to discuss the research project. When the member of the research team met with the participant, they talked through the information sheet and invited questions. If the participant was satisfied with the responses obtained, they would then be invited to participate in the study.

Advice about the information sheets and the approach taken to recruitment was provided by the Trial Steering Committee. The views of the Committee members with intellectual disabilities and a family member proved particularly helpful.

Informed consent

Individuals who chose to take part in the study were asked to complete a written consent form by members of the research team. The consent form was read to the individual with learning disabilities, they were asked to sign it, and this was witnessed by a carer or another individual who was independent of the study. Those who did not have the capacity to consent to participate were excluded from the study. The researchers all received training on assessing capacity to consent in adults with learning disabilities based on the relevant UK legislation45,46 and established best practice.

Randomisation, concealment and blinding

Participants who provided informed consent were screened and provided baseline data before being randomised. The flow diagram for the study is shown in Figure 1. Individuals were allocated to the BeatIt arm or the StepUp arm in a 1 : 1 ratio, using a blocked randomisation within each study centre. Mixed block sizes of length four and six were used at random. The randomisation was stratified by study centre and the use of antidepressants. At the design stage, several potential stratification variables were considered but ultimately not used, including the use by participants of other drugs that may have some mood-stabilising properties and are commonly prescribed in this population. For example, an estimated 25% of the population have comorbid epilepsy and may be taking carbamazepine, sodium valproate, lamotrigine or pindolol. Changes in the prescription of antidepressants and other mood-stabilising drugs were monitored over the duration of the study.

FIGURE 1.

Flow chart showing BeatIt study design and participant follow-up. BLESID, Bangor Life Events Schedule for Intellectual Disabilities; BPI-S, Behavior Problems Inventory for Individuals with Intellectual Disabilities – Short Form; CSRI, Client Service Receipt Inventory; EDSE, Emotional Difficulties Self-Efficacy; EQ-5D-Y, EuroQol-5 Dimensions – youth version; GAS-ID, Glasgow Anxiety Scale for people with an Intellectual Disability; GDS-LD, Glasgow Depression Scale for People with Learning Disabilities; ICI, Index of Community Involvement; IDDS, Intellectual Disabilities Depression Scale; IPDL, Index of Participation in Domestic Life; RA, research assistant; SSQ3, Social Support Questionnaire – three questions.

To conceal the allocation of participants from the research team, researchers randomised each participant to a treatment arm using an automated system run by the Robertson Centre for Biostatistics (RCB). The system did not reveal the random allocation to the researcher but notified the study coordinator, who then contacted the clinicians to arrange subsequent treatment visits. Thus, the researchers collecting outcome data remained unaware of the study arm to which participants had been assigned.

Interventions

Fidelity to the intervention

Data management

The trial was conducted in accordance with the principles of Good Clinical Practice (GCP) to ensure quality of data and protection of participants. These matters were carefully discussed with members of the Trial Steering Committee; members of the Committee with learning disabilities requested to run through the measures with the researchers in Scotland. When doing so, the Committee members offered helpful advice about ensuring that the process of data collection was engaging and accessible for participants. Data were collected by trained researchers and health professionals, using detailed standard operating procedures. Any inconsistencies were investigated and resolved in a timely manner. Careful quality checks were made before data were submitted to RCB. Once data had been submitted, further independent quality checks were performed by RCB before database lock. After database lock, the final statistical analyses were carried out.

Data collection and blinding

As shown in Figure 1, data were collected at three time points:

1. baseline

2. time 1 – 4 months post randomisation (post intervention)

3. time 2 – 12 months post randomisation (follow-up/maintenance).

The data were collected on an outreach basis, face to face, and this ordinarily occurred at a participant’s home, unless another location that offered privacy was preferable. As data were also collected from carers, arrangements were made to have separate meetings with both carers and participants when the researcher visited. There was one additional data collection point at 8 months via telephone with the carer alone, to chart any changes in the participant’s medication use and receipt of services.

Both treatments were of similar duration and included a support person. Moreover, both groups of participants were told that they were joining the BeatIt trial, rather than being told that they were being allocated to the BeatIt or StepUp intervention. These key similarities helped prevent research assistants from becoming unblinded if participants made reference to their therapy sessions during follow-up meetings. In addition, none of the researchers conducted qualitative interviews or was involved in the fidelity checks for participants from whom they collected outcome data. Qualitative interviews and therapy fidelity ratings were carried out by a researcher other than the researcher who gathered the outcome data from the participant concerned.

Measures

Sample size

In the first 18 months of an earlier pre–post trial of BeatIt, the mean reduction in GDS-LD39 scores at 3 months’ follow-up was 8.50 points [standard deviation (SD) 5.24 points].

The present study was powered to detect a mean between-group difference of 0.6 SD units, or 3.14 points on the GDS-LD.

If the BeatIt group in this study could achieve an 8.5-point improvement in GDS-LD scores at 12 months, then this allows for the StepUp group showing 5.36-point improvement over the same time period (i.e. 63% of the improvement in the BeatIt group).

Alternatively, this allows for a small improvement in the StepUp group, in conjunction with a large short-term improvement in the BeatIt group, followed by some regression. For example, if the BeatIt group show an improvement from baseline to 12 months of 6 points, then the study would be powered to detect a difference if the mean improvement in the StepUp group was 2.86 points.

To have 90% power to detect this difference, the study required 60 participants in each arm to provide outcome data at 12 months post randomisation. The primary analysis was to be an analysis of covariance, adjusting for baseline GDS-LD score, which would have the power to detect smaller intervention effects, depending on the level of correlation in scores over time.

There were no data to inform the effect of clustering of outcomes for participants with learning disabilities seen by each therapist. The assumption was made that each therapist would work with an average of nine participants (i.e. several part-time therapists at each site) and an intraclass correlation of 0.025 was assumed, resulting in the sample size being increased by 20% to 72 per group, or 144 in total. A recruitment target of 166 participants allowed for a loss to follow-up of ≤ 13.3%. A meta-analysis of research with the general population66 found a post-intervention effect size on self-reported depression symptoms of behavioural activation therapy versus supportive therapy of 0.75. These designs were similar to our own attention control design. However, they did not report data regarding long-term follow-up in comparison with supportive therapy. The effects relative to brief psychotherapy were 0.56 post intervention and 0.50 after an average follow-up of 4 months, suggesting that the effects of behavioural activation therapy might persist for some time. Our follow-up at 12 months post randomisation was approximately 9 months post intervention, so we would be able to detect differences between groups only if they persisted over a longer time frame than usually studied. Therefore, we believed that an effect size for sample size estimation purposes of 0.60 was realistic given the results of this meta-analysis for behavioural activation versus supportive therapy, and this would also be considered to be of ‘moderate’ size and thus meaningful from a clinical perspective for an individual therapeutic intervention.

Statistical analyses

The statistical analyses were specified in a statistical analysis plan (SAP) (see Appendix 2), which was approved before database lock.

The primary analysis compared GDS-LD scores at 12 months post randomisation between intervention groups, adjusting for baseline GDS-LD scores, study centre and use of antidepressants at baseline within a mixed-effects linear regression model, including therapist as a random effect. Similar methods were applied to the primary outcome measure at the immediate post-intervention assessment (4 months post randomisation) and to secondary outcome measures at each assessment point. These models were used to estimate between-group differences at each follow-up time point, and to estimate mean changes from baseline within each intervention group. Repeated measures analyses, adjusting for stratification factors, were also applied to each outcome measure. These models were used to confirm the results of analyses looking at each follow-up time point separately, as well as estimating mean changes in outcomes between 4 and 12 months. In general, analyses were carried out using the available data. However, analyses of the primary outcome were repeated using multiple imputation, to assess whether or not the results were sensitive to missing data. To impute missing values at each time point, prediction models were based on age, antidepressant use and any previous or subsequent measurements of the primary outcome; prediction models did not include randomised group. Models for the primary outcome were extended to explore the effects of baseline characteristics, including the stratification factors, chronicity of depressive symptoms, life events and history of previous psychological intervention. The moderating effects of these factors were explored by using appropriately constructed interaction terms within linear regression models. These moderation analyses were exploratory only and designed to inform future translation of the intervention into routine clinical practice. Selected analyses were repeated using a per-protocol population of participants who attended at least eight BeatIt therapy sessions or at least six StepUp sessions.

Additional exploratory analyses were outlined in the SAP as potential avenues for future investigation. These analyses have not yet been carried out and are not included in this report.

Flow of participants into the trial

Figure 2 reports the flow of participants into the trial. In total, 934 information packs were sent out, from which 233 reply slips were received (four of these slips indicated that the potential participant was not interested in taking part). In total, 186 people gave their consent to take part in the study, and at that point their eligibility was assessed. Of these 186 people, 19 people were excluded, most as a result of not meeting depression criteria (eight people) or the IQ being above the threshold for inclusion in the trial (six people). An additional three people withdrew before randomisation, and three people were randomised but no therapist was available.

FIGURE 2.

Participant flow diagram. Data were removed from analysis. (a) Intention-to-treat population; and (b) per-protocol population. SSSFT, South Staffordshire and Shropshire Healthcare NHS Foundation Trust. a, Includes one participant initially allocated to StepUp in error.

This resulted in 161 people being randomised into the trial. Six people were lost to follow-up and 14 people withdrew during the trial, resulting in a total of 141 participants who completed the trial. Sixty-nine of these 141 participants were recruited in Scotland, 38 in England and 34 in Wales.

Table 4 shows that two-thirds of the participants were referred by community nurses, psychologists and psychiatrists who worked in community teams for people with learning disabilities.

Participants at baseline

Seventy-seven participants were randomised to receive StepUp and 84 participants were randomised to receive BeatIt. The imbalance of seven participants was a result of using a stratified randomisation scheme. Participants were randomised in six strata defined by recruitment site (Scotland, England, Wales) and whether or not participants were taking antidepressants at baseline (yes/no). Within these strata, participants were randomised in blocks of length four (two to each arm) and six (three to each arm), with block lengths occurring at random. Therefore, within each stratum, the imbalance between treatment groups could be at most three in either direction (if the study stopped midway through a block of six allocations, and that block had consisted of three allocations to one group by three allocations to the other group). Across the six strata, at the end of the trial, the balance of randomisations were as follows:

• Scotland, taking antidepressants: 29 to BeatIt and 27 to StepUp; difference = +2

• Scotland, not taking antidepressants: 10 to BeatIt and eight to StepUp; difference = +2

• England, taking antidepressants: 12 to BeatIt and 13 to StepUp; difference = –1

• England, not taking antidepressants: 12 to BeatIt and nine to StepUp; difference = +3

• Wales, taking antidepressants: 12 to BeatIt and 11 to StepUp; difference = +1

• Wales, not taking antidepressants: nine to BeatIt, and nine to StepUp; difference = 0.

Table 5 describes demographic and health characteristics of the 161 participants who began the trial at baseline, broken down across the two therapy arms. Overall, a slight majority of participants were female, with a mean age of 40 years. Most participants were white and single. The mean Full Scale IQ was 58.34 for StepUp participants and 55.44 for BeatIt participants.

A majority of participants (58.4% of StepUp participants and 65.5% of BeatIt participants) had some degree of visual impairment, with smaller proportions (10.4% of StepUp participants; 23.8% of BeatIt participants) having some degree of hearing impairment. Approximately one-quarter of participants (26.0% of StepUp participants; 22.6% of BeatIt participants) had mobility problems.

Table 6 describes selected circumstances of the 161 participants who began the trial at baseline, broken down by therapy arm. In terms of deciles of neighbourhood deprivation (where 1 is the most deprived and 10 is the least deprived) based on participants’ postcodes, participants on average were living in slightly more deprived postcodes than the median. In terms of negative life events, participants in both therapy arms had experienced on average more than one negatively experienced life event in the 12 months before baseline (mean 1.36 events for StepUp participants; mean 1.48 events for BeatIt participants).

Service receipt in the 4 months prior to baseline

Over two-thirds of participants (68.8% of StepUp participants; 70.2% of BeatIt participants) were receiving support from services at least daily. According to the PSS scale, participants were receiving social support from an average of fewer than two family members (1.32 for StepUp participants; 1.98 for BeatIt participants) and social support from an average of approximately three non-family members (3.30 for StepUp participants; 2.92 for BeatIt participants).

Approximately one-fifth of participants had previously received therapy for depression (18.2% of StepUp participants; 20.2% of BeatIt participants). Approximately two-thirds of participants were being prescribed antidepressant medication (66.2% of StepUp participants; 63.1% of BeatIt participants). In terms of mood-stabilising medication, 19.5% of StepUp participants and 13.1% of BeatIt participants were prescribed this medication. It was also established that 11.7% of StepUp participants and 4.8% of BeatIt participants were taking antiepileptic medication.

Table 7 describes the services used by participants in the 4 months prior to baseline, broken down by therapy type. This was collected using the Client Service Receipt Inventory (CSRI) and gives some indication of the context of participants’ lives as they entered the trial.

In cases in which < 5% of participants at baseline had received a particular service in the 4 months before baseline (eight participants or fewer), full data are excluded from Table 7. Specific services used by eight or fewer participants in total in the 4 months before baseline were sheltered work (used by three people in total); drop-in centre (four people); learning disabilities centre (zero people); psychiatric intensive care ward (zero people); acute psychiatric ward (one person); psychiatric rehabilitation ward (one person); general medical ward (seven people); other hospital specialties (two people); day hospital (two people); psychiatric outpatient visits (five people); community mental health team member (one person); healthcare assistant (three people); speech therapist (one person); art/drama/music therapist (one person); alternative therapist (two people); social work assistant (two people); and audiologist (five people).

Table 7 shows 25 indicators of service use, used by nine or more participants in the trial in the 4 months prior to baseline. For specific types of daytime activity, the number of weeks a service was used in the 4 months prior to baseline, the mean number of hours per week the service was used, and the number of people who used the service are presented. For hospital services and community-based services, the mean number of contacts with the service in the 4 months prior to baseline and the number of people having contact with the service are presented.

There was a statistically significant difference between BeatIt and StepUp participants in one of these indicators – StepUp participants had been attending adult education for more weeks than BeatIt participants.

In terms of daytime activities, across both therapy arms, 23 participants had attended a day centre in the 4 months prior to baseline, 31 participants had engaged in voluntary work, 14 participants had attended adult education and 45 participants had attended a social club. For those participants that had engaged in each daytime activity, participants spent 18–21 hours per week attending a day centre (mean 18.4 hours for StepUp participants; mean 20.8 hours for BeatIt participants), 5 hours per week engaged in voluntary work (mean 4.7 hours for StepUp participants; mean 5.3 hours for BeatIt participants), 5–9 hours per week attending adult education (mean 4.9 hours for StepUp participants; mean 8.6 hours for BeatIt participants), and 3 hours per week attending a social club (mean 3.3 hours for both StepUp and BeatIt participants).

In terms of contact with hospital services, across both therapy arms, 29 participants had attended other hospital outpatient visits in the 4 months prior to baseline, with StepUp participants having a mean of 5.3 contacts and BeatIt participants having a mean of 2.9 contacts. Sixteen participants had visited accident and emergency (A&E) in the 4 months prior to baseline, with StepUp participants having a mean of 1.6 contacts with A&E and BeatIt participants having 1 contact with A&E.

In terms of contacts with community-based services in the 4 months prior to baseline, most participants (94 across both therapy arms) had had contact with a GP, with a mean of 3.2 contacts for StepUp participants and 3.1 contacts for BeatIt participants).

Across both therapy arms 54 participants had had contact with a community psychiatrist (mean 1.8 contacts for StepUp participants; mean 2.4 contacts for BeatIt participants), 31 participants had had contact with a psychologist (mean 5.4 contacts for StepUp participants; mean 5.1 contacts for BeatIt participants), and nine participants had had contact with a community psychiatric nurse (mean 5.3 contacts for StepUp participants; mean 8.2 contacts for BeatIt participants).

In total, 56 participants had had contact with a learning disability nurse (mean 5.9 contacts for StepUp participants; mean 6.2 contacts for BeatIt participants), 23 participants had had contact with another community nurse (mean 28.8 contacts for StepUp participants; mean 3.3 contacts for BeatIt participants), 12 participants had had contact with a physiotherapist (mean 6.3 contacts for StepUp participants; mean 4.6 contacts for BeatIt participants) and 15 participants had had contact with an occupational therapist (mean 6.1 contacts for StepUp participants; mean 5.4 contacts for BeatIt participants).

Across both therapy arms, 57 participants had had contact with a social worker or care manager in the 4 months prior to baseline (mean 3.2 contacts for StepUp participants; mean 3.7 contacts for BeatIt participants), 47 participants had had contact with a home help or home care worker (mean 80.3 contacts for StepUp participants; mean 84.3 contacts for BeatIt participants), and 19 participants had had contact with an advocate or counsellor (mean 5.3 contacts for StepUp participants; mean 4.7 contacts for BeatIt participants).

Overall, 62 participants had had contact with a dentist (mean 2.0 contacts for StepUp participants; mean 1.5 contacts for BeatIt participants), 23 had had contact with an optician (mean 1.5 contacts for StepUp participants; mean 1.1 contacts for BeatIt participants), and 28 participants had had contact with a chiropodist (mean 1.8 contacts for StepUp participants; mean 2.3 contacts for BeatIt participants). In total, 12 participants had had contact with employment services or a job centre in the 4 months prior to baseline (mean 5.6 contacts for StepUp participants; mean 6.6 contacts for BeatIt participants).

Carers at baseline

Table 8 describes characteristics of the carers of the 161 participants who began the trial, broken down by therapy arm. Carers were those who had a significant caring responsibility for the participant and knew the participant sufficiently well to complete the informant measures in the study. A carer could also be the supporter for the participant in the therapy, but this was not a requirement, and a participant could have several people in their lives with significant caring roles.

Table 8 shows that approximately three-quarters of carers were female (74.0% of carers of StepUp participants; 75.0% of carers of BeatIt participants), with an average age of 47 years. Almost all carers were white. In terms of their relationship to the participant, over one-quarter were family members (28.6% of carers of participants in each therapy arm), over two-fifths were workers (41.6% of carers of StepUp participants; 41.7% of carers of BeatIt participants), and over one-quarter were ‘other’ carers (29.9% of carers of StepUp participants; 29.8% of carers of BeatIt participants). In terms of average weekly contact between the carer and the participant, approximately two-thirds of carers had less than daily contact with participants (64.9% of carers of StepUp participants; 67.9% of carers of BeatIt participants); a small proportion had daily contact with participants (11.7% of StepUp participants; 9.5% of BeatIt participants); and almost one-quarter of carers were living with the participant (23.4% of carers of StepUp participants; 22.6% of carers of BeatIt participants).

Training and supervision

The therapy was provided by 76 therapists. Sixteen (21.1%) therapists were men (BeatIt, n = 8 men; StepUp, n = 8 men) and 60 (78.9%) were women (BeatIt, n = 35 women; StepUp, n = 25 women), with a mean age of 40.5 years (SD 10.5 years) [BeatIt therapist mean age = 40.4 years (SD 11.4 years); StepUp therapist mean age = 41.8 years (SD 9.7 years)]. There were no significant differences in sex [χ² = 0.36, degrees of freedom (df) = 1, not significant] or age (t = 0.55, df = 73, not significant) of therapist. Fifty (65.8%) therapists were nurses (BeatIt, n = 28 nurses; StepUp, n = 22 nurses), 13 (17.1%) were psychology assistants or trainees (BeatIt, n = 9 assistants or trainees; StepUp, n = 4 assistants or trainees), 4 (5.3%) were occupational therapists (BeatIt, n = 1 occupational therapist; StepUp, n = 3 occupational therapists) and 6 (7.9%) were CBT therapists (BeatIt, n = 3 CBT therapists; StepUp, n = 3 CBT therapists). There were three therapists whose profession was not recorded. There was no significant difference in distribution of professions between the arms of the study (χ² = 2.6, df = 3, not significant). Forty-six (60.5%) therapists had worked with people with learning disabilities for > 5 years (BeatIt, n = 26 therapists; StepUp, n = 20 therapists), 10 (13.2%) for 1 to 5 years (BeatIt, n = 7 therapists; StepUp, n = 3 therapists) and 18 (23.7%) for < 1 year (BeatIt, n = 10 therapists; StepUp, n = 8 therapists). These data were not available for two therapists. There was no significant difference between the arms of the study in the distribution of the duration that therapists had been working with people who have learning disabilities (χ² = 0.68, df = 2, not significant). Twenty-nine (38.2%) of the therapists had no previous therapy training (BeatIt, n = 14 therapists; StepUp, n = 15 therapists); 13 (17.1%) had attended an accredited training course (BeatIt, n = 6 therapists; StepUp, n = 7 therapists); 10 (13.2%) had received ‘in-house’ training (BeatIt, n = 4 therapists; StepUp, n = 6 therapists), 10 (13.2%) had received training as part of their professional training (BeatIt, n = 5 therapists; StepUp, n = 5 therapists) and 4 (4.8%) had received training from other sources (BeatIt, n = 4 therapists; StepUp, n = 0 therapists). These data were not available for 10 therapists. There was no significant difference between the arms of the study in the distribution of previous therapy training (χ² = 4.38, df = 4, not significant).

The manual stipulated that there should be one supervision meeting for every two therapy sessions. All supervisors in the study were clinical psychologists. Therapists in the BeatIt arm received a mean of 6.4 supervision sessions (SD 0.37 sessions, range 1–13 sessions), and those in the StepUp arm had a mean of 4.7 supervision sessions (SD 0.29 sessions, range 1–10 sessions). However, these data can be misleading, as not all participants received the full number of sessions (as a result of drop-out etc.), and preferred levels of supervision were actually defined in terms of the ratio of supervision meetings to therapy sessions. Excluding participants who received no sessions, the mean ratio of supervision meetings to therapy sessions delivered was 1.9 (SD 1.36). The expected ratio was 2.0, and a mean of < 2.0 indicates a slightly higher than expected level of supervision. Ninety-eight (73.1%) cases met the criterion of at least one supervision meeting for every two therapy sessions, whereas 36 (25.5%) did not. The lowest ratio of supervision to therapy was 4.0, which occurred in one BeatIt case in which only three supervision meetings were recorded for 12 therapy sessions.

The measurement of fidelity during the course of the intervention

There were 144 fidelity recordings available for analysis. These constituted recordings for 93 participants, 49 from StepUp and 44 from BeatIt, with 51 participants having two recordings available. There are multiple reasons why recordings were not always completed: 21 participants refused consent for recording, 39 participants did not provide recordings because they withdrew from the study before being recorded, 12 recordings were not made because the participant or carer refused to have the recording made on the day and two recordings were lost in transfer or as a result of other technical difficulties with recording equipment. The baseline scores for GDS-LD, IDDS and BPI-S Aggressive Behaviour subscale and age for the 93 participants for whom fidelity recordings were available were compared with the 68 participants for whom they were not, and there were no statistically significant differences. There was also no association between the availability of fidelity recording and the sex of the participants. A second researcher rated 48 recordings to determine inter-rater reliability; the intraclass correlation for agreement was 0.76 (df = 43; p < 0.001).

Therapy quality and non-specific factors

When looking at the first available recording from each of the 93 participants who provided recordings, the 10-item fidelity scale had a mean total score of 33.1 points (SD 4.1 points, range 18–39 points). Table 9 shows that there was a very similar pattern of findings for both the BeatIt and StepUp arms. The highest scoring items were those relating to empathy, warmth and not including therapy activity that was not part of the manual; the lowest scoring items were for including the client and supporter in the process of setting the session agenda and asking for feedback at the end of the session. The scale had an alpha of 0.76 and a mean-adjusted item-total correlation of 0.42 (SD 0.1, range 0.31–0.58). The stability of fidelity was 0.64, estimated through a Pearson’s correlation of first and second fidelity ratings for the 51 participants with two recordings. There was no significant difference in the fidelity scores between the two arms (BeatIt: mean = 32.3, SD 4.1; StepUp: mean = 33.9, SD 3.9; t = 1.5, not significant).

Covering key elements/activities in individual therapy sessions

For the session-specific technical ratings, each session could gain a score that ranged from 3, if all elements were conformed to, to 0 if none of the elements was conformed to. For BeatIt, 33 sessions obtained a score of 3, 10 sessions obtained a score of 2, one session obtained a score of 1 and no sessions received a score of 0. In the StepUp arm, 42 sessions obtained a score of 3, seven sessions obtained a score of 2 and no sessions received a score of 1 or 0. The differences in distribution of scores between the groups is not significant (Fisher’s exact test p = 0.23). For the purpose of further analysis, the data were collapsed across both arms and treated as dichotomous, with 75 sessions gaining a full score of 3 and 18 obtaining a score of < 3. The 75 sessions with a score of 3 had a mean therapy quality scale score of 33.7 (SD 3.6); the 18 sessions with a score of < 3 had a mean score of 33.1 (SD 4.7). This is a statistically significant difference (t = 2.5, df = 91; p = 0.014), showing a significant relationship between two measures of fidelity.

Delivery and consistency of support

Patients randomised to the BeatIt arm of the trial began their therapy sessions a median of 36 days (interquartile range 28–49 days) after randomisation, compared with 27 days (interquartile range 21–42 days) for those randomised to StepUp; this was a statistically significant difference (p = 0.003, Mann–Whitney U-test). This difference was because the BeatIt intervention began with a session for the supporters alone, to provide them with information about therapy and their role in sessions.

There was excellent attendance of the sessions, with participants in the BeatIt arm attending an average of 9.9 sessions (SD 3.2 sessions) and StepUp participants attending an average of 7.1 sessions (SD 2.1 sessions). Sixty-six per cent of the BeatIt therapy sessions were delivered as scheduled and 77% of the StepUp sessions were delivered as planned.

For BeatIt, 38 (51%) participants had the same supporter throughout their therapy, 16 (23%) had two supporters, 10 (14%) had three supporters, 1 (1%) had four supporters and 2 (3%) had five supporters. For StepUp, 42 (61%) participants had one supporter throughout their therapy, 16 (23%) had two supporters, 6 (9%) had three supporters, 1 (1%) had four supporters and 2 (3%) had five supporters.

Primary outcome

A reduction in the GDS-LD total score represents an improvement in symptoms of depression. In the analyses reported here, a positive intervention effect indicates a greater reduction in depression symptoms in the BeatIt group compared with the StepUp group. The GDS-LD scores post intervention (4 months after randomisation) and at follow-up (12 months after randomisation – the primary end point) were analysed in adjusted mixed-effects linear regression models (Table 10) that adjusted for baseline GDS-LD scores, therapist effects (as a random effect), study centre and baseline antidepressant use.

There were no statistically significant unadjusted or adjusted group differences in the effects of the StepUp or BeatIt interventions on GDS-LD scores at 12 months (primary end point) or 4 months (p-value column of Table 10).

The same analysis models were also applied to the changes in GDS-LD scores between 4 and 12 months to examine whether or not the treatments may have had different effects on GDS-LD scores from post intervention to follow-up. There was no overall change in GDS-LD scores from 4 to 12 months [estimate –0.42, 95% confidence interval (CI) –2.30 to 1.47; p = 0.663] and no interaction between treatment group and data collection point (4 vs. 12 months) (estimate 0.56, 95% CI –2.15 to 3.27; p = 0.685).

All of the above reported analyses were repeated using multiple imputation for missing data. Although the estimates for effects changed slightly, the overall pattern of findings was identical.

Secondary, within-group analyses (see Table 10) showed that participants in both treatment groups reported significant reductions in GDS-LD scores at both 4 and 12 months. The analysis of 4- to 12-month change summarised above suggests that these changes from baseline are accounted for mainly by reductions in GDS-LD scores from baseline to post intervention (4 months), with no change from post intervention (4 months) to follow-up (12 months).

Secondary outcomes

Exploratory moderator analyses

For the primary outcome measure, measured using the GDS-LD, exploratory moderator analyses were conducted on a number of variables at baseline for outcomes at 4 months and 12 months. These analyses investigated whether or not there were main effects at 4 months and 12 months related to the baseline variable, and whether there were interaction effects at 4 months and 12 months between the baseline variable and the therapy arm.

Table 28 presents the exploratory moderator analyses for the following baseline variables: participant age, gender, Full Scale IQ, expectations of change and ABS-RC2 domains (personal self-sufficiency, community self-sufficiency, personal social responsibility, self-direction, responsibility and socialisation), study site, neighbourhood deprivation, relationship of carer to participant, level of service support for participant, BLESID impact of negative life events in the 12 months before baseline, whether or not the participant had received any previous psychological therapy for depression, prescribed usage of antidepressants and prescribed usage of mood-stabilising medication.

Out of 18 potential predictor variables of change in GDS-LD score from baseline to 4 months, there were two statistically significant main effects. For participants with higher ABS-RC2 socialisation scores at baseline, there were bigger reductions in GDS-LD scores at 4 months. There was also a main effect of study site: for participants in Wales there were bigger reductions in GDS-LD scores at 4 months compared with participants in Scotland. There were no interaction effects of baseline variables by therapy type for change in GDS-LD scores at 4 months.

Out of 18 potential predictor variables of change in GDS-LD scores from baseline to 12 months, there were two statistically significant main effects. For participants with lower ABS-RC2 personal self-sufficiency scores at baseline, there were bigger reductions in GDS-LD scores at 12 months. For participants with higher expectations of change at baseline, there were bigger reductions in GDS-LD scores at 12 months. There was also one interaction effect of study site: in Scotland there were bigger decreases in GDS-LD scores from baseline to 12 months for participants in the BeatIt group, whereas in Wales there were bigger decreases in GDS-LD scores from baseline to 12 months for participants in the StepUp group.

Per-protocol analyses

Per-protocol analyses were also conducted; however, there were no differences in the pattern of results from those reported in this chapter. Given that there were virtually no treatment group differences found that could be examined in terms of the effects of a per-protocol approach, these analyses are not described further here.

Adverse events

Following randomisation, there were 19 adverse events reported by 15 participants allocated to StepUp and 21 adverse events reported by 18 participants allocated to BeatIt (Table 29).

All 24 serious adverse events reported during the trial were hospitalisations.

Five participants in the StepUp group were admitted to a mental health ward, of whom three presented with suicidal ideation. Four participants in the BeatIt group were admitted to a mental health ward, of whom two presented with suicidal ideation.

There was one reported episode of attempted suicide by overdose during the trial. The participant had been randomised into the study but had not yet started therapy sessions. The participant was subsequently withdrawn based on previously undisclosed suicide attempts.

Introduction

This chapter outlines the health economics analysis of BeatIt. The objective of the analysis was to evaluate whether or not BeatIt is cost-effective compared with a guided self-help intervention (StepUp), as determined by an incremental cost per quality-adjusted life-year (QALY).

Methods

Results

Outcomes

Results of the EQ-5D-Y at each level, dimension and visit are presented in Table 36. For the mobility dimension, the proportion of participants who experienced ‘no problems walking about’ increased from 70.0% to 80.7%, and the proportion reporting ‘a lot of problems walking about’ decreased from 16.4% at baseline to 8.9% at the 12-month visit. For the self-care (‘looking after myself’) dimension, the proportion of participants reporting ‘no problems washing or dressing myself’ increased from 69.3% at baseline to 76.3% at the 12-month follow-up period, and the proportion reporting ‘a lot’ of problems decreased from 13.6% at baseline to 8.9% at follow-up. Although for the usual activities (hobbies, sports, etc.) dimension, the proportion of participants reporting ‘no problems’ decreased from 80.0% to 76.3% over the 12-month follow-up period, the proportion reporting ‘a lot’ of problems doing usual activities increased from 7.86% to 10.4%. The pain or discomfort domain returned to the previous trends seen in the mobility and ‘looking after myself’ dimensions: the proportion of participants reporting no pain or discomfort increased from 62.8% to 77.8% over the trial period, and those experiencing ‘a lot of pain or discomfort’ decreased from 11.4% to 7.4%. The final dimension of ‘feeling worried, sad or unhappy’ is arguably the most important one in this trial, with the proportion of participants reporting that they were ‘not worried, sad or unhappy’ increasing from 42.1% to 55.6%, and the proportion reporting that they felt ‘very worried, sad or unhappy’ decreasing from 20.7% to 13.3%.

TABLE 36 - The EQ-5D-Y results by dimension

EQ-5D-Y dimension	n (%)
	Baseline	4 months	12 months
Mobility
Level 1 (no problems)	98 (70.0)	103 (76.3)	109 (80.7)
Level 2 (some problems)	19 (13.6)	17 (13.6)	14 (10.4)
Level 3 (a lot of problems)	23 (16.4)	15 (11.1)	12 (8.9)
Looking after myself
Level 1 (no problems)	97 (69.3)	99 (73.3)	103 (76.3)
Level 2 (some problems)	24 (17.1)	25 (18.5)	20 (14.8)
Level 3 (a lot of problems)	19 (13.6)	11 (8.2)	12 (8.9)
Usual activities
Level 1 (no problems)	112 (80.0)	101 (74.8)	103 (76.3)
Level 2 (some problems)	17 (12.1)	24 (17.8)	18 (13.3)
Level 3 (a lot of problems)	11 (7.86)	10 (7.4)	14 (10.4)
Pain or discomfort
Level 1 (no pain)	88 (62.8)	92 (68.2)	105 (77.8)
Level 2 (some pain)	36 (25.7)	29 (21.5)	20 (14.8)
Level 3 (a lot of pain)	16 (11.4)	14 (10.4)	10 (7.4)
Feeling worried, sad or unhappy
Level 1 (not worried, sad or unhappy)	59 (42.1)	79 (58.5)	75 (55.6)
Level 2 (a bit worried, sad or unhappy)	52 (37.1)	44 (32.6)	42 (31.1)
Level 3 (very worried, sad or unhappy)	29 (20.7)	12 (8.9)	18 (13.3)

The EQ-5D-Y results for utilities at baseline and 4 and 12 months post randomisation and QALYs for the 12-month trial period are presented in Table 37.

TABLE 37 - EQ-5D-Y utilities and QALYs

Time point	Trial arm				Difference
	StepUp		BeatIt
	Mean (SD)	n (%)	Mean (SD)	n (%)	Mean (SE)	95% CI
Baseline	0.638 (0.374)	72 (98.6)	0.474 (0.438)	68 (100.0)	0.163 (0.069)	0.0270 to 0.300
4 months	0.701 (0.358)	71 (97.3)	0.637 (0.412)	64 (94.1)	0.064 (0.067)	–0.069 to 0.196
12 months	0.697 (0.349)	70 (95.9)	0.675 (0.410)	65 (95.6)	0.022 (0.066)	–0.109 to 0.152
QALYs	0.691 (0.287)	68 (93.2)	0.628 (0.361)	61 (89.7)	0.063 (0.058)	–0.052 to 0.178

It can be seen that there is no significant difference between arms in utilities at 4 months or 12 months; however, at baseline there is a significant difference of 0.163 (SE 0.069), with BeatIt participants reporting significantly lower QoL than StepUp participants. The results of the EQ-5D-Y scores are illustrated as a profile over time in Figure 3. This clearly illustrates the difference between the groups at baseline and 4 and 12 months. The area between the two lines represents the QALYs gained over the 12-month follow-up period between treatments. Figure 3 also presents the 95% CI (as bars) around StepUp and BeatIt EQ-5D-Y scores at baseline and 4 and 12 months. The scores and 95% CIs for both trial arms converge towards the 12-month follow-up point.

FIGURE 3.

The EQ-5D-Y scores over time, by group (all available data).

Uncertainty

The complete-case and multiple imputation analyses both show no statistical differences between mean costs or QALYs. To explore the robustness of these results, 1000 non-parametric bootstrapping samples (unadjusted) were taken from the observed data and the resultant cost-effectiveness pairs are used in this exploratory analysis.

The cost-effectiveness plane presented in Figure 4 illustrates the 1000 bootstrap samples, showing the mean difference in costs and QALYs between the StepUp and BeatIt interventions. It can be seen that the samples fall into all four quadrants of the cost-effectiveness plane, showing that the difference in mean costs and QALYs can favour either intervention and that there is no evidence to show that one intervention is more cost-effective than the other.

FIGURE 4.

Cost-effectiveness plane showing mean difference in costs and QALYs between StepUp and BeatIt.

The likelihood of cost-effectiveness at various thresholds is illustrated in the cost-effectiveness acceptability curve presented in Figure 5.

FIGURE 5.

Cost-effectiveness acceptability curve for the BeatIt intervention.

Figure 5 illustrates that at a willingness-to-pay (WTP) threshold of £0, there is a 41.7% likelihood of BeatIt being cost-effective compared with StepUp, and at the current NICE WTP threshold of £30,000, there is a 41.1% chance of BeatIt being cost-effective compared with StepUp. Conversely, StepUp has a 58.9% likelihood of being cost-effective at the £30,000 WTP threshold. The results in Figure 5 show that BeatIt is not likely to be cost-effective compared with StepUp at any WTP threshold: at £120,000 the likelihood of BeatIt being cost-effective compared with StepUp is 44.1%. This reflects the huge uncertainty in the results.

A further exploratory bootstrapping exercise has been carried out using only the intervention costs; we are certain that BeatIt is more costly than StepUp to deliver, but there is no significant difference in the other cost data collected. The results of the exploratory bootstrapping are presented on a cost-effectiveness plane in Figure 6.

FIGURE 6.

Cost-effectiveness plane, for intervention costs only.

It can be seen from Figure 6 that, although we can be certain that BeatIt is costlier than StepUp to deliver, we cannot be certain that the QALYs differ between arms, as is the same for the primary outcome of GDS-LD score. The 95% CI for the ICER ranges from –£307,000 to £211,000, indicating that at no point does the ICER fall below £211,000 for BeatIt compared with StepUp, which is not considered cost-effective.

These exploratory bootstrapping results have also been presented on a cost-effectiveness acceptability curve in Figure 7.

FIGURE 7.

Cost-effectiveness acceptability curve, for intervention costs only.

It can be seen from Figure 7 that at a WTP threshold of £0, there is an 0% likelihood of BeatIt being cost-effective compared with StepUp; however, as the WTP threshold increases, the probability of BeatIt being cost-effective compared with StepUp increases. At the current NICE WTP threshold of £30,000 there is a 29.5% chance of BeatIt being cost-effective compared with StepUp.

Introduction

When examining the cost-effectiveness of psychological interventions for mental health problems in people with learning disabilities, it is important to seek to understand the experiences of people centrally involved in the intervention alongside the quantitative data. This may illuminate facilitators of and barriers to change and offer insights about how to implement interventions effectively. This component of the study garnered qualitative data separately from people with learning disabilities who had received therapy, supporters in therapy sessions and therapists (all across both types of intervention) to understand the experiences of the therapy process from each perspective.

Previous research asking people with learning disabilities about their experiences of undergoing psychological therapy has found that people valued specific techniques used as part of the particular intervention they were receiving, but placed more emphasis on the non-specific aspects of their therapy such as the therapeutic relationship, being listened to and respected by the therapist, and being able to talk about their lives. 74,75 People with learning disabilities participating in a group intervention found it challenging to talk about their individual feelings in a group context. 75

Having a supporter (whether a family member or a paid carer) with the person with learning disabilities during at least part of a psychological intervention has been suggested as an important adjustment to the routine therapy process, to assist people undergoing the therapy with motivation and activities between therapy sessions. 76–78 Research on the impact of the presence of a member of staff as a supporter during psychological interventions with people with learning disabilities is mixed;74,79 however, research asking supporters directly about their experiences is lacking.

There has also been very little research concerning the experience of therapists trained to deliver manualised psychological interventions to people with learning disabilities. Stimpson et al. 80 explored the views of lay therapists who ran anger management groups as part of a trial of a group anger management intervention for people with learning disabilities. Although the therapists had some trepidation about taking on the role, they were largely positive about running the groups. Obtaining the views of more experienced professionals, trained to deliver individual psychological interventions, has the potential to provide important insights that would help with the wider roll-out of the interventions by services.

Method

Findings

Findings from the qualitative analyses are outlined in terms of themes, subthemes and specific issues within subthemes. As the overall thematic structure derived from framework analysis was the same for people with learning disabilities and supporters, these findings will be presented together. The thematic structure for therapists was different from that for people with learning disabilities and supporters; therefore, findings from therapists will be presented separately.

Content analysis of therapists’ logs

The therapy logs that were completed after each session by the therapists contained a section in which they were asked to identify barriers that they faced when delivering the therapy. Barriers recorded on all of the therapy logs completed by therapists were extracted and the content analysed.

For the BeatIt intervention, the most frequently reported barriers were participants’ lack of interest in activities (reported at least once for 61 participants), a lack of resources between sessions (reported at least once for 35 participants), the non-completion of diaries (reported at least once for 28 participants), issues with supporters in sessions (reported at least once for 26 participants) and a lack of continuity with carers (reported at least once for 25 participants). For the StepUp intervention, the most frequently reported barriers were distractibility (reported at least once for 38 participants), understanding of materials (reported at least once for 16 participants), booklets not seeming relevant (reported at least once for 13 participants) and lack of continuity with carers (reported at least once for 11 participants).

The BeatIt therapists also recorded the participants’ planned homework activities on the therapy logs. All these planned homework activities were also extracted from the logs and the content was analysed. Of these, 51% were day-time activities, 26% were domestic or daily living activities, 7% were social activities and 16% were preparatory activities.

Summary

The accounts of people with learning disabilities undergoing therapy, their supporters and the therapists delivering the therapies increase our understanding of how people experienced the interventions in this study.

People with learning disabilities and their supporters clearly described the circumstances they were in before therapy and their perceived need for an intervention, focused on the person’s depression. Both people with learning disabilities and their supporters reported a range of feelings and perspectives about starting therapy, from nervousness and uncertainty to an attitude that it was worth trying the intervention, particularly in the absence of other mental health interventions.

For both interventions, people with learning disabilities, supporters and therapists generally (but not universally) described good working relationships being quickly established, and all agreed on the importance of these therapeutic relationships. When supporters were committed, positive, reliable and had time, their advocacy/scaffolding role within sessions and encouragement/reminders between therapy sessions were highly valued by people with learning disabilities and therapists.

Most people with learning disabilities, supporters and therapists generally viewed the materials in either intervention as accessible and useful, although more adaptability and flexibility in their use were also recommended.

In addition to these non-specific factors, people with learning disabilities, supporters and therapists clearly described specific aspects of each intervention. People with learning disabilities receiving the BeatIt intervention described important aspects of the therapy as having the time and space to talk and be listened to, generating ideas about activities and making plans, building their confidence, and understanding the materials and paperwork used within the therapy. Supporters described having a liaison function and encouraging/reminding the person to engage in activities between sessions. Therapists discussed being able to tailor aspects of the intervention towards the circumstances and interests of the person.

For people with learning disabilities receiving the StepUp intervention, important aspects of the therapy were the process of using the self-help booklets and making them relevant to themselves. Supporters thought that the intervention gave them new insights into the person they were supporting and gave them a greater understanding of depression more generally. Therapists also discussed how the intervention could be made personally relevant, and how they helped the person to solve problems that he/she faced and to think differently about his/her situation. They felt that the intervention provided useful tools for the person beyond the end of therapy.

The highly structured, manualised and time-limited nature of both interventions was valued by supporters and therapists. Although both these groups mentioned wanting to adapt and individualise interventions more to suit individuals’ needs, therapists mentioned the importance of a clear structure in keeping them focused. People with learning disabilities reported feeling sadness and disappointment when the therapy came to an end, although some reluctantly accepted that this had to happen.

Some people with learning disabilities and supporters reported continuing changes beyond the end of therapy, with materials (often individually adapted) continuing to be used and activities and coping strategies being maintained in accordance with the specific intervention. Others reported that these had not been maintained at the time of the interviews but mentioned that the materials would be used when they needed them.

Both people with learning disabilities and supporters involved in both types of intervention reported similar positive effects of the therapy on people’s lives. People with learning disabilities mentioned improvements in their mood, a more positive sense of self and more independence, more active lives and better relationships, and bullying or victimisation to have stopped. Supporters described people being happier and more confident, talking more and showing improved social skills, engaging in more activities, being more empowered and having better problem-solving skills. They also reported that people had better relationships with friends, family and with themselves (the supporters).

Primary objective

There were no statistically significant differences between the StepUp and BeatIt groups in GDS-LD scores at the 12-month primary outcome point. Therefore, there is no evidence that BeatIt is effective when compared with StepUp as an active control condition.

Although this was not designed as a non-inferiority study, an important consideration is that the study resulted in the outcomes of two active interventions being compared (see Chapter 7). Similarly designed trials conducted with the general population provide supportive evidence to add to the available armoury of effective interventions for use in depression. For example, the Cost and Outcome of Behavioural Activation versus Cognitive Behavioural Therapy for Depression (COBRA)27 non-inferiority trial promotes behavioural activation as being as equally effective as CBT. People with learning disabilities and their therapists are in a different position: there are no existing trials showing any psychological interventions to be effective in the treatment of depression in people with learning disabilities, including no high-quality evidence base for CBT. Ours is the first large-scale RCT conducted on psychological interventions for depression, and, other than our own feasibility study,39 is the only RCT of any size concerning behavioural activation with adults with learning disabilities. This is also the only trial involving an adapted, guided self-help treatment for depression for adults with learning disabilities. On the main outcome measure, participants in our trial had marked improvements in depressive symptoms in both conditions.

So how do we interpret these findings? We cannot conclude that either intervention is effective. However, both BeatIt and StepUp recipients reported considerable improvements in their symptoms of depression at 4 months post randomisation (following end of treatment), and this was sustained for both groups 12 months post randomisation. These findings could be explained by a regression of scores to the mean over time and a range of non-specific effects of receiving therapy and being involved in a clinical trial, but improvement in depression could suggest a successful response to both interventions, given the following factors:

• The ‘unsuccessful’ previous treatment histories of the participants. At baseline, 66% of the participants with learning disabilities who were allocated to the StepUp intervention, and 63% who were allocated to the BeatIt intervention, were taking antidepressants. A total of 19% of the participants allocated to the StepUp intervention were taking mood-stabilising drugs (primarily for the management of epilepsy, but still with a theoretically mood-stabilising effect), and 13% allocated to the BeatIt intervention were taking mood-stabilising drugs [with 69% taking drugs of one or both of these classes (i.e. antidepressants and mood-stabilising drugs)]. In addition, 18% of the participants allocated to the StepUp intervention and 20% of the participants allocated to the BeatIt intervention had received prior psychological therapy (not specifically defined). Despite these ‘interventions’, all participants were depressed at baseline. This is suggestive of a view that these adults might have been at the severe and/or more enduring end of the depression spectrum.

• The natural history of depression in the population with learning disabilities. It is apparent from previously published literature that people with learning disabilities are more likely to have enduring depression than those without learning difficulties. For example, reports from the longitudinal UK 1958 birth cohort found that at 33 years of age, adults with learning disabilities had significantly higher depressive scores than the adults without learning disabilities,84 and 15% of the cohort with learning disabilities, compared with only 3% of those without, had chronic depression (high depressive scores at ages 23, 33 and 43 years). 11 Similar evidence comes from the longitudinal study of the UK 1946 birth cohort. When aged 15, 36, and 43 years, individuals with learning disabilities were found to have significantly more depression and anxiety symptoms than those without learning disabilities,12 suggesting a more severe and enduring pattern of depressive illness.

• Six previous small pilot trials of psychological interventions for depression in people with mild/moderate learning disabilities,33–38 reviewed by NICE,31 have reported depression data from TAU conditions. For participants in TAU in all six studies, there was virtually no change in depression scores from baseline to follow-up, including up to 11 months after baseline.

Although it cannot be discounted, it seems unlikely that the pattern of improvements in self-reported depressive symptoms in both BeatIt and StepUp groups – improvements by 4 months that were sustained at 12 months – could be just as a result of a natural history of spontaneous recovery of symptoms; however, participation in the trial may have had an effect.

Alongside these findings from the study, there were no related serious adverse events reported, suggesting that neither intervention carries any notable risk to participants.

Secondary objectives

Other psychological outcomes followed a similar pattern of results to the primary outcome. Again, there was no evidence that the BeatIt intervention is effective when compared with the StepUp intervention.

For carer-reported depression in the participants, self-reported anxiety and carer-reported aggressive behaviour, there were no treatment group differences at the 12-month follow-up (primary end point) or at 4 months post baseline. Within-treatment group analyses showed significant reductions from baseline in carer-reported depression and self-reported anxiety at both 4 and 12 months, associated primarily with reductions from baseline to 4 months that were maintained to follow-up. Similar analyses for carer-reported aggression showed a reduction from baseline to 4 months, but only for the BeatIt group.

Considering QoL outcomes together (self-reported health-related QoL, community and domestic activity measures, PSS), there were again no treatment group differences at the 12-month follow-up (primary outcome) or at 4 months after baseline. The one exception was increased participation in domestic and leisure activities among the BeatIt group participants at 12-month follow-up compared with the StepUp group participants. Although self-reported health-related QoL (measured using the EQ-5D) followed a similar pattern of within-group changes (improvements compared with baseline in both treatment groups), other QoL indicators demonstrated a more mixed pattern of results, with some showing no change from baseline, some a positive change in the BeatIt group, and some a positive change in the StepUp group.

Two of the three indicators of participants’ adaptive behaviour skills showed no treatment group differences at either the 4- or 12-month follow-ups, but BeatIt participants had higher socialisation scores at 4 months only (i.e. not at the 12-month follow-up primary end point). Within-group analyses suggested generally increased adaptive skills in the two treatment groups in comparison with baseline, including added increases following therapy up to the 12-month follow-up point.

Carers’ sense of self-efficacy in supporting adults with learning disabilities and depression increased within both treatment groups from baseline, but again there were no treatment group differences at either 4 or 12 months. Carers’ relationships with participants were almost universally lacking evidence of negativity on the basis of analysis of FMSSs, showing very little change, and analysis models could not be appropriately fitted.

Exploratory analyses of predictors and potential moderators of outcome suggested that the two treatments worked equally well across a variety of participant characteristics. Notably, IQ scores were unrelated to outcome. One interesting finding was that the participants’ positive expectation of change at the outset predicted a better outcome.

Health economics

The results of the economic evaluation indicate that there was no difference in outcomes between arms as measured by changes in mean QALYs between baseline and the 12-month follow-up, which is reflected by the lack of difference between arms in the primary outcome of change to GDS-LD score.

Although there was no significant difference in mean total costs between arms, this masks a highly significant difference in intervention costs. An alternative interpretation, therefore, is that the increased intervention cost in the BeatIt arm did not result in any significant cost savings in non-intervention-related costs.

The uncertainty around the complete-case and multiple imputation analyses results was explored using bootstrapping techniques. The results of the bootstrapping exercise confirmed the results: there was no evidence to suggest either intervention was cost-effective compared with the other if the focus was on the total costs. However, if non-significant, non-intervention costs were excluded, the bootstrap analysis confirmed that the BeatIt intervention never entered the cost-effectiveness range.

A review by Romeo and Molosankwe85 in 2010 found a paucity of economic evaluations carried out on people with learning disabilities. They found 10 papers reporting economic studies in learning disability, 4 of which reported costs and outcomes, and the areas covered included service use, community living and challenging behaviour. None of these studies had a cost–utility analysis, in which the outcome is preference-based, such as a QALY. In 2011, Ekers et al. 86 reported on a small pragmatic RCT (n = 47 participants) comparing behavioural activation with TAU for depression in adults with learning difficulties. Ekers et al. 86 found that behavioural activation resulted in a gain of 0.20 QALYs, producing an ICER of £5756 per QALY, which is considered cost-effective using NICE’s current threshold of £20,000 to £30,000 per QALY. 87 More recently, a report was published by Willner et al. 68 in 2013 exploring how costs are affected by challenging behaviour in people with learning disabilities. The participants were randomised to either a CBT manualised intervention or TAU ‘waiting list’ control arm. Willner et al. 68 concluded that there were no statistically significant differences in costs between groups, and because of lack of power, the study was not able to detect economic differences between groups. We have not found any previous studies looking at the cost-effectiveness of depression treatments for people with learning disabilities.

The strengths of this study are that < 5% of data were missing in all questionnaires for resource use and QoL, and cost-effectiveness results are consistent across all analyses carried out.

Limitations of the study include the imbalance in EQ-5D-Y utilities at baseline, with the BeatIt group participants reporting significantly lower QoL than those in the StepUp group. The generic EQ-5D-Y may not be sufficiently sensitive for this population. We found a small association between GDS-LD and EQ-5D-Y scores, but note that the analysis suggests a high amount of unexplained variability.

It was possible to administer the EQ-5D-Y to a population with learning difficulties. However, there is no tariff available to value the EQ-5D-Y for children and adolescents, or for those with a learning disability. Future research could include investigating whether or not there is a need for a learning-disability tariff.

In summary, given the lack of differences between the BeatIt and StepUp groups at the primary end point (12-month follow-up), it is not surprising that there was also no economic evidence to suggest that BeatIt may be more cost-effective than StepUp. No differences in resource use were found at 12 months, although the BeatIt intervention did cost more to deliver than the StepUp intervention. Overall, the vast majority of the support costs for participants in both treatment groups were not related to the treatments themselves: intervention costs were approximately 4%–6.5% of the total support costs.

Qualitative study

Results of the qualitative research with participants, supporters and therapists were in concert with the quantitative findings. In particular, both treatments were perceived as active interventions, and both treatments were valued in terms of their structure, content and perceived impact/outcomes. The qualitative data from supporters did offer additional insight into the impact of taking part in therapy on carer–participant relationships. Supporters reported understanding more about the person and/or his/her depression following either treatment, and reported more positive relationships with participants following either treatment.

Delivery of the interventions

Data from multiple sources across the study are consistent in showing that both therapies were conducted to the criteria set out in the protocol, and that participants, supporters and therapists viewed both interventions positively.

From fidelity ratings of therapy sessions, therapists across the BeatIt and StepUp interventions demonstrated similar high levels of fidelity to non-specific aspects of both interventions, and also demonstrated high fidelity to specific aspects of the intervention that they were delivering. This high level of therapist fidelity was achieved with a large cohort of 84 therapists who generally saw very few participants as part of the trial. From the qualitative interviews and focus groups, participants, supporters and therapists in both interventions agreed that good therapeutic working relationships were established quickly and maintained throughout therapy, and were an important component of the clinical effectiveness of either intervention.

In the qualitative study, both participants and supporters sometimes talked about feeling nervous and uncertain about starting either intervention, yet they were also generally positive because there was a perceived absence of other therapeutic options available for people with learning disabilities and mental health problems or depression. From therapist logs, participant adherence to both interventions, in terms of attending therapy sessions, was high and for most participants the presence of a supporter in therapy sessions was relatively reliable and stable. Supervision was also delivered to therapists in both interventions approximately once every two therapy sessions, as specified in the protocol.

The majority of therapists in the study (67.1%) were community learning disability nurses. Most therapists (63.1%) were experienced in working with people with learning disabilities (having experience of ≥ 5 years) and most therapists (60.7%) had received some previous therapy training.

In the qualitative interviews, community learning disability nurses valued the structure, clear focus and goals, and clinical supervision – which were seen as contrasts to their usual working practices – of both interventions. The highly structured, manualised and time-limited nature of both interventions was also valued by the supporters in the qualitative interviews. People with learning disabilities in qualitative interviews reported feeling sadness and disappointment when the therapy came to an end, although some reluctantly accepted that this had to happen. For some supporters and therapists there was a tension between the utility of a highly structured, manualised intervention and a desire to individualise the structure of the intervention to focus more on the issues facing specific participants with whom they were working.

In terms of adaptations to the interventions to make them more accessible to people with learning disabilities, both interventions used extensively adapted materials. Supporters and therapists generally reported that these adapted materials were accessible and useful. Nevertheless, participants, supporters and therapists all suggested specific changes to the materials in each intervention – some of the main changes being put forward were about making the materials more accessible and flexible to suit individual participants’ circumstances and preferences. Some supporters from both interventions reported that they had continued to use therapy materials, which they had personalised for the individual they were supporting.

A further adaptation to both interventions was to involve a supporter in therapy sessions. In the qualitative study, participants, supporters and therapists generally (although not universally) agreed that when supporters were committed, positive, reliable and had time, their advocacy/scaffolding role within sessions and encouragement/reminders between therapy sessions were highly valued and an important component of the clinical effectiveness of either intervention. Obtaining reliable, consistent and effective supporter involvement was problematic for a minority, as reported in therapist logs.

In terms of specific therapy factors, the participants in the BeatIt group valued having the time and space to talk and be listened to, generating ideas about activities and making plans, and building their confidence. Supporters described having a liaison function, and encouraging/reminding the person to engage in activities between sessions. Therapists discussed being able to tailor aspects of the intervention towards the circumstances and interests of the person. In therapist logs, barriers specific to the BeatIt intervention included a lack of resources to support participants to engage in activities and the non-completion of diaries between therapy sessions.

For the StepUp intervention, participants described using the self-help booklets and making them relevant. Supporters thought that the intervention gave them new insights into the person they were supporting and a greater understanding of depression. Therapists discussed how the intervention helped the person solve problems and think differently about their situation, and felt that the intervention provided useful tools for the person beyond the end of therapy. In therapist logs, barriers specific to the StepUp intervention included participants getting distracted, participants’ understanding of the materials and some of the booklets not being relevant for the individual participant.

Although participants and supporters described some aspects specific to each intervention, across both interventions participants and supporters reported similar positive impacts of either therapy on people’s lives, including improvements in their mood, a more positive sense of self, more independence, more active lives and better relationships with others.

Summary of findings

Primary and secondary outcomes, economic data and qualitative results all clearly demonstrate that there was no evidence for the BeatIt intervention being more effective than the StepUp intervention.

Based on these results, we cannot conclude that either intervention is effective. However, patients in both arms of the trial showed improvements in outcomes during the intervention phase that were maintained at 12 months; this may indicate that both the BeatIt and StepUp interventions are beneficial. There is a need for the clinical effectiveness of these (or similar) interventions to be compared with TAU in future RCT evaluations.

Conduct of the trial

Important insights and data were also obtained concerning the delivery of the trial. The ratio of trial information packs sent out to initial responses received was 4–5 : 1. Working through intermediaries to recruit participants was one of the challenges of conducting the trial. The main source of participants was from community learning disability teams, with smaller numbers of participants recruited from mainstream IAPT services and care providers. One drawback was that some professionals in IAPT services appeared to lack knowledge of people with learning disabilities and mistakenly referred people with specific learning difficulties such as dyslexia to the trial. Differences in the main referral sources between England, Wales and Scotland is likely to have reflected the different service patterns and practices. This highlights that a multipoint recruitment strategy, responsive to different service contexts, is vital in a multicentre trial.

We successfully recruited and randomised 161 participants into the trial and informed consent was obtained with the use of carefully developed accessible materials. There was good adherence and retention: 86% of participants in the BeatIt arm completed therapy (a minimum of 8 sessions), 82% of participants in the StepUp arm completed therapy (a minimum of 6 sessions) and 141 (87.6%) participants were followed up for 12 months. Community nurses and other professionals working in specialist community learning disability teams and IAPT services were trained as therapists to deliver the interventions. However, competing work demands meant that therapists had limited time available to deliver therapy as part of the trial and they were generally able to work with only one or two participants at any one point in time. This could pose a problem if therapists lacked capacity to take on new participants, as a therapist for each arm of the trial needed to be available at a location before a participant could be randomised to that location. Consequently, to ensure that there were sufficient numbers of therapists to allow new participants to be randomised in a timely fashion, a far higher number of therapists needed to be trained than originally anticipated: 95 in total. In the end, 76 of the trained therapists were assigned participants to work with. Importantly, it proved feasible to deliver the training to this large number of therapists. Supervisors were also recruited from local services and trained to supervise the delivery of the interventions.

The two manuals developed for the interventions provided comprehensive guidance and support materials for the therapists and it proved possible to deliver the therapies, at scale, with good fidelity to the manuals. Audio-recordings of therapy sessions were rated for fidelity. Although the inter-rater agreement for the fidelity ratings was good, it proved challenging to obtain audio-recordings, because of a reluctance on the part of some participants and therapists to consent to the audio-recording of sessions.

Overall, the successful delivery of the BeatIt trial contradicts a widely held contention that it is not possible to recruit to a large-scale RCT of an individual psychological therapy for people who have a learning disability. 88–90 The lessons learned may have important implications for the conduct of future trials of psychological therapies involving people with learning disabilities.

Limitations

The major limitation of the trial was the absence of a TAU condition, which would have been required to establish the clinical effectiveness of the BeatIt intervention. This was the original design suggested by the research team but was rejected by the funder, who asked for an attention control group to be recruited instead. The decision was made to use guided self-help (StepUp) as the attention control arm because it was seen as a less person-centred and more rigid approach with a different set of active ingredients, that is, psychoeducation rather than activity based. It was also chosen because it was an ethical option, as guided self-help is a NICE-recommended91 intervention for general adult mental health and we had a package available, with carefully adapted materials for people with learning disabilities. The StepUp intervention was also designed to involve a supporter in therapy sessions, thereby incorporating a key element of attention for which we wanted to control. The research team and funder-approved protocol failed to anticipate the ability of the therapists to personalise the materials or the agency shown by the participants and their supporters to follow through on ideas and plans made in therapy sessions.

Although, in retrospect, comparing TAU with the BeatIt intervention would have been a preferable design in terms of examining the clinical effectiveness of the BeatIt intervention, there remains the question of whether or not this would have made recruitment to the trial more challenging. Participants randomised to the trial knew that they would be receiving one of two interventions, making participation in the trial a more attractive proposition. Moreover, working through intermediaries meant that the researchers were not only having to recruit the participants themselves but were also having to make the case to carers or professionals working with the individual with learning disabilities that the study was worthwhile. Carers and professionals might have been less likely to engage with the recruitment process knowing that the person they supported or worked with may not have received an intervention.

There were a greater number of withdrawals from the BeatIt arm (n = 12) of the study than the StepUp arm (n = 2). This may have reflected the greater number of BeatIt sessions and the more active nature of the intervention, in which the aim was to support the participants and supporters to engage in scheduled activities between sessions. However, it is also worth noting that three of the BeatIt withdrawals happened before the participants entered therapy and another BeatIt participant withdrew when they were contacted for the 12-month follow-up visit, after the intervention had finished.

Lessons for trials with people who have learning disabilities

Our feasibility study helped to inform careful advanced planning of recruitment to our study, with a further opportunity to refine the strategy during the internal pilot. Consequently, when implementing our RCT, we achieved our target recruitment within the planned time frame. This was despite the fact that recruitment of people with learning disabilities to participate in trials is notoriously difficult. For example, a Health Technology Assessment (HTA) programme-funded study92 on antipsychotic medication for adults with learning disabilities had substantial recruitment problems: only 72% of the total number of intended participants were recruited, despite an increase in the recruitment time from 2 to 4 years and extending recruitment outside the UK. In addition, a HTA programme-funded study93 on antipsychotic drug withdrawal in adults with learning disabilities closed at the internal pilot stage because of recruitment problems. Even in general population trials, recruitment is often a challenge. A study94 of 114 UK trials funded by the Medical Research Council and the HTA programme found that only 31% of trials recruited successfully, whereas 45% recruited < 80% of the intended number of participants, and more than half required an extension. A Cochrane review synthesised the evidence around strategies to improve recruitment to RCTs,95 and concluded that triallists should include evaluations of their recruitment strategies in their trials, and funders should support this because the number of interventions that have been rigorously evaluated in the context of a real trial is low.

The challenges of recruitment in learning disability studies include a smaller population base and need for multiple sites, the randomisation process itself, the need to recruit through third-party intermediaries and via multiple sources to engage with enough people for the message to be passed on to potential participants, concerns by paid carers of the legalities of supporting consent to participate and the need for information to be accessible for the potential participants. 92,96–98 We have demonstrated that these potential obstacles can be successfully navigated and overcome across three countries of the UK, now significantly divergent in terms of the provision of health and social care services. This requires careful planning and ensuring that adequate human resources are available to devote to the time-consuming recruitment process, which is expensive but essential when conducting trials within this population. Finding and training sufficient numbers of therapists to allow participants to be randomised in a timely fashion to either arm of the trial also emerged as a significant consideration. Established links with existing services proved vital in obtaining the numbers of therapists required. An advantage of this strategy was that the trial demonstrated the delivery of the interventions by relevant professionals in routine clinical practice.

Clinical implications and future directions

There have been no previous large trials of psychological interventions for adults with learning disabilities. Hence, the lack of difference between the BeatIt and StepUp interventions means it is too early to say if both interventions are equally effective, or not effective.

Recruitment to our study was mostly from practitioners in community learning disability teams; the participants were already receiving considerable input from team members, including community nurses, psychologists and psychiatrists, and a higher proportion of participants in this study were receiving antidepressants than in previous population-based studies of adults with learning disabilities. 99 Despite this, the qualitative interviews on supporters’ expectations before the study identified that supporters perceived that there was nothing else available specifically to help with the depression being experienced by the person whom they supported. Participants were depressed at entry to the study, and there was an overall group-level clinical improvement associated with both the BeatIt and StepUp interventions. The interventions were also relatively low cost, in the context of the wider ongoing support that the participants were receiving.

It proved difficult to recruit via IAPT services in England (there is no direct equivalent in Scotland or Wales), and of the potential 14 participants who were referred, only 9 were suitable to be randomised. There appeared to be a poor understanding of learning disability by IAPT services and this is consistent with other research findings. 100 Perhaps this is in part explained by the lack of any psychological interventions for treatment of depression specifically recommended for use with adults with learning disabilities. However, it does indicate a training need within IAPT, and the need to support research and development initiatives to make IAPT services more accessible to people with learning disabilities. 101

Difficulties with assessment also arose in the study. For example, it was challenging to measure meaningful change in activity for people with learning disabilities, given that small and statistically insignificant increases may have considerable personal salience. Furthermore, it is also worth noting that the mean baseline GDS-LD40 score (16.7 points) for the study participants, although in the published clinically significant range, was lower than the mean scores reported by the 19 participants involved in the original development of the measure (23.4 points). This suggests that it might be important for further work to be carried out on concerning the clinical cut-off scores for the GDS-LD.

The next step in the research agenda will be to better understand whether or not the BeatIt and StepUp interventions could be effective with adults who have learning disabilities, exploring ethical trial designs to do so. Given the very similar outcomes found for both the BeatIt and StepUp interventions in the present study, it may be possible to examine the clinical effectiveness of both interventions in one trial, compared with TAU. The participants randomised to the intervention arm could be offered a choice of either the BeatIt or StepUp intervention, allowing a comparison of both interventions with TAU. This would be an ethical and cost-effective design.

It is also important to note that this study was conducted with people with mild/moderate learning disabilities. Adults with severe/profound learning disabilities also experience depression. The communicative and cognitive demands of CBT-based interventions place such interventions out of the reach of adults with severe/profound learning disabilities. These adults are also more at risk of adverse effects of pharmacotherapy, which can be both disabling and potentially serious, because of limited or no verbal communication skills to describe their symptoms of adverse effects, hence late and more severe presentations of such effects. NICE31 made a specific recommendation that research should be undertaken with people with severe learning disabilities to determine the clinical and cost-effectiveness of psychosocial interventions to treat mental health problems such as depression. We strongly endorse this, and consider further modification of the BeatIt intervention to be ideally suited for, and a high priority to trial with, adults with severe/profound learning disabilities.

Contributions of authors

Professor Andrew Jahoda was the chief investigator, contributed to the design of the study and was part of the writing team.

Professor Richard Hastings was a co-applicant, contributed to the design of the study and was part of the writing team.

Professor Chris Hatton was a co-applicant, contributed to the design of the study and was part of the writing team.

Professor Sally-Ann Cooper was a co-applicant, contributed to the design of the study and was part of the writing team.

Ms Nicola McMeekin was the health economist working on the study and part of the writing team.

Professor Dave Dagnan was a co-applicant, contributed to the design of the study and was part of the writing team.

Dr Kim Appleton was the trial manager and part of the writing team.

Ms Katie Scott was a researcher on the project and contributed to the report.

Ms Lauren Fulton was a researcher on the project and part of the writing team.

Professor Robert Jones was a co-applicant, contributed to the design of the project and reviewed the report.

Dr Alex McConnachie was a co-applicant and the senior trial statistician and reviewed and contributed to the report.

Ms Rachel Zhang is a statistician who worked on the trial and contributed to this report.

Ms Rosie Knight was a researcher on the trial and contributed to the report.

Ms Dawn Knowles was a researcher on the trial and contributed to the report.

Professor Christopher Williams was a co-applicant who contributed to the design of the study and to this report.

Professor Andy Briggs was a co-applicant who designed the health economic component of the study and reviewed the report.

Professor Craig Melville was a co-applicant, contributed to the design of the study and was part of the writing team.

Publications

Jahoda A, Melville C, Cooper SA, Hastings R, Briggs A, Dagnan D, et al. BEAT-IT: comparing a behavioural activation treatment for depression in adults with intellectual disabilities with an attention control: study protocol for a randomised controlled trial. Trials 2015;16:595.

Jahoda A, Hastings R, Hatton C, Cooper SA, Dagnan D, Zhang R, et al. Comparison of behavioural activation with guided self-help for treatment of depression in adults with intellectual disabilities: a randomised controlled trial. Lancet Psychiatry 2017;4:909–19

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.

Glasgow Depression Scale for people with a Learning Disability

The GDS-LD is a series of 20 self-rating questions asking a participant about their experiences in the previous week.

For each question, there are three possible answers: (1) never/no, (2) sometimes and (3) always/a lot, each being scored 0, 1 or 2, resulting in a range of possible scores from 0 to 40.

Glasgow Anxiety Scale for people with an Intellectual Disability

The GAS-ID is a series of 27 questions asking a participant about their experiences in the previous week.

It is divided into three categories: (1) worries (10 questions), (2) specific fears (9 questions) and (3) physiological symptoms (8 questions). For each question there are three possible answers: (1) never/no, (2) sometimes and (3) always/a lot, each being scored 0, 1 or 2, resulting in a range of possible scores from 0 to 54.

The Index of Community Involvement

The ICI is a list of 48 activities divided into three categories: (1) domestic and leisure (8 activities), (2) social (11 activities) and (3) community (19 activities). Participants are asked which activities they took part in over the previous month and how often they took part in those activities.

Each item is scored from 0 (never did activity) to 5 (daily) based on frequency of participation. There is a range of possible scores from 0 to 240.

Emotional Difficulties Self-Efficacy scale

Supporters are asked four questions about their responses to emotional difficulties, experienced by the participants whom they support, over the previous month. Each item is scored from 1 (not at all confident) to 7 (very confident). There is a range of possible scores from 4 to 28.

Behavior Problems Inventory for Individuals with Intellectual Disabilities – Short Form

The BPI-S has a list of 10 aggressive behaviours. Supporters were asked how often in the previous month they observed the participants whom they support display each behaviour. Each item is scored from 0 (never) to 5 (more than once per hour). There is a range of possible scores from 0 to 50.

Intellectual Disabilities Depression Scale

The IDDS consists of questions on 38 behaviours. The supporters were asked how often the participants whom they support experienced each of the behaviours, over the previous 4 weeks. Each item is scored from 0 (never) to 6 (always). There is a range of possible scores from 0 to 228.

Adaptive Behavior Scale at screening visit

The Adaptive Behavior Scale (ABS) questionnaire is adapted from the ABS-RC2. It is directed to the supporter at the screening visit as a series of 39 questions from six categories: (1) personal self-sufficiency, (2) community self-sufficiency, (3) social responsibility, (4) self-direction, (5) responsibility and (6) socialisation.

Supporters were asked to rate the participants whom they support on each of the items.

Adaptive Behavior Scale at the 4- and 12-month visits

At the 4- and 12-month visits, the supporters were asked to rate those whom they support in three of the six categories measured at screening:

1. Self-direction (five items), with a range of possible total scores from 0 to 23.

2. Responsibility (three items), with a range of possible scores from 0 to 10.

3. Socialisation (seven items), with a range of possible total scores from 0 to 26.

Bangor Life Events Schedule for Intellectual Disabilities

The BLESID consists of 24 significant life events from five categories. It is used to determine the life events that participants have experienced over the previous 12 months, the frequency of these events and the impact the events have had on the participants’ lives.

Each event receives a score of 0 (did not happen), 1 (happened once), or 2 (more than once) for frequency. If an event occurred, it then receives a score of –1 (bad), 0 (in the middle) or 1 (good) for impact.

Social Support Questionnaire – three questions

This was previously referred to as PSS.

The participant was asked to name people who they can count on to help them in each of the three ways described. They were then asked how happy they are with the support they receive.

Their satisfaction with the help they receive is given a score of 3 (very happy), 2 (quite happy) or 1 (not happy).

A randomised controlled trial comparing a behavioural activation treatment for depression in adults with learning disabilities with an attention control (BeatIt) statistical analysis plan

Introduction

Statistical analysis plan

Analyses

Per-protocol analyses

The analyses described in Baseline characteristics, Primary outcome and Secondary outcomes will be repeated in a per-protocol analysis. The analysis population included will be those who have attended at least eight BeatIt treatment sessions or six attention control sessions.

Additional analyses

The analyses in the previous sections constitute the main SAP analyses, and will be programmed and validated prior to database lock and release of the randomisation codes to the study statisticians. The following analyses will be carried out after database lock.

Potential mediation effects will be explored through extensions to the primary analysis regression model. Changes in outcome measures to 4 months post randomisation will be explored as mediators of effects to follow-up at 12 months post randomisation. Attendance at therapy sessions and therapist-rated session data will be explored as potential mediators of short- and long-term intervention effect differences.

In addition, sensitivity analyses for the secondary outcomes will be analysed by the same method as the primary outcome outlined in Sensitivity analyses.

Depending on the results of the main analyses, complier-average causal effect (CACE) analysis may be explored to account for missed therapy sessions.

The scope of these analyses will be agreed with the study team after review of the main SAP analysis results and documented prior to carrying out any additional analyses.

Document history

This is v1.0 of the SAP for the final analysis of the BeatIt study, dated 23 November 2016, and is the initial creation of this document.

Tables

Dummy tables will be produced during the development of the statistical analysis programs for review and feedback. Approval of the content of the final statistical outputs will be a requirement for database lock.

Figures

Dummy figures will be produced during the development of the statistical analysis programs for review and feedback. Approval of the content of the final statistical outputs will be a requirement for database lock.

Listings

No formal data listings will be produced as part of the final statistical outputs. Raw and derived datasets will be made available to the study team after the final statistical outputs have been produced.