Discussion

This study was conducted as part of a commissioned call by the National Institute for Health Research Public Health Research funding stream to evaluate interventions in community organisations to improve health and well-being and reduce health inequalities. We were funded to undertake a pilot study owing to the uncertainties of conducting a randomised trial in people with ID. This chapter will review each of the study objectives against the progression criteria. We will discuss the barriers that we encountered and make suggestions about how these barriers could be overcome. We will also discuss the strengths and limitations of our study, and implications for future research.

Progression criteria

We listed a number of criteria that would be considered in evaluating whether or not the trial had been successful:

• At least 70% (35) of participants are recruited from the 50 potentially eligible individuals who are approached.
• At least 70% of participants with ID in the intervention group are successfully matched to a volunteer (befriender) and they meet for a minimum of 10 meetings.
• The dropout rate of participants with ID in both groups, post randomisation, is < 30% at 6 months’ follow-up.
• The intervention and trial procedures are considered to be acceptable by volunteers and individuals with ID.
• If the target recruitment rate and number of matched pairs is between 50% and 69%, or the dropout rate is between 31% and 50%, we would consider if these could be rectified by modifying the protocol. If this was the case, a full study could still be considered. If the recruitment rates were < 50% and the dropout rates were > 50%, then a full trial would not be considered feasible.

Recruitment

Participants with intellectual disability

Recruitment of participants with ID and volunteers was the most important feasibility objective and the most challenging. Our recruitment period was 6 months. We recruited 16 participants with ID (our target was at least 40 participants, based on 50 participants being approached). We therefore recruited 40% of our target in relation to people with ID. In our progression criteria, we stated that a recruitment target below 50% would indicate that our trial was not feasible.

Challenges in recruiting to RCTs have been demonstrated across a range of different populations and settings. Only 56% of studies funded by the Health Technology Assessment programme have achieved their target sample size, which demonstrates the importance of conducting feasibility studies first.67 Recruitment of participants with ID to clinical trials is considered to be particularly challenging, with a number of RCTs failing to recruit.68,69 Reasons for this include communication difficulties, the need to access participants through gatekeepers, issues related to consent and capacity, and limited knowledge of the value of RCTs among carers.70–72 However, there is also evidence for the increasing acceptability of RCTs.73

One of the reasons for the lower than expected success in recruitment to our study was that we significantly overestimated recruitment of participants via the befriending schemes. Initial consultation with the schemes (particularly the Suffolk scheme) appeared to indicate that there were a large number of people who were waiting to be matched (figures suggested that there were 100 people waiting to be matched). However, the Suffolk scheme thought that it was unethical to randomise participants as it would undermine their relationship with their service users, and offered individuals the option of entering the study or befriending through their usual route. As a consequence, most people decided not to enrol in the study. This severely affected the recruitment of participants to the study.

Conducting research outside the NHS had several challenges. During the study, the Suffolk befriending scheme experienced funding cuts, which led to the two volunteer co-ordinators involved in the study being employed on a part-time basis when they were formerly full time. They had responsibility for managing existing befriending relationships that were not part of the study as well as assisting the research team with recruitment. These changes undoubtedly had an impact on the effectiveness of recruitment to the study, as intervention costs (e.g. staff costs) were not covered by the grant funding. We also did not have access to support from clinical studies officers from the Clinical Research Network to assist with recruitment from the befriending schemes, as they were not NHS services.

A review of barriers and facilitators in the evaluation of evidence-based interventions in third-sector organisations found that lack of financial resources and knowledge of appropriate research methodology and outcome indicators were the main barriers preventing service evaluation. Facilitators included getting appropriate support, an organisational culture that supported research and the motivation to be accountable to stakeholders.74 In our study, we worked collaboratively with the befriending services to provide the expertise in research methods, and the services were motivated to engage in research and to demonstrate an evidence base for their service. However, we were not able to provide financial support. In a recent national survey of befriending services for people with ID in the UK, carried out by the research team, staff reported that funding and the recruitment of volunteers were the major challenges encountered by their services.75

Priebe et al.37 were able to recruit participants with psychosis into their study, possibly because they received research funding to deliver the intervention, which gave them more control over the recruitment of volunteers and the intervention itself. However, there are clearly financial and implementation implications of using this approach as, should the befriending intervention be effective, it would not be sustainable without continued funding.

Another issue that may have affected recruitment was the eligibility criteria. This was fed back to us at a PPI meeting with individuals with ID. They felt that the study should have been open to all people with ID and commented that some people who are depressed may not want to admit that they are depressed, or may find it more difficult to take part in this type of study because of their depression. When participants were approached by clinicians to take part in the study from ID services in NELFT, we received feedback that many were reluctant to take part because they were afraid of meeting someone new, or that they had anxieties about going out. It appears that those people who most needed the intervention (e.g. those who were socially isolated and had few opportunities to access the community) were often the people who were less willing to take part. Of the 24 referrals, only 16 met the eligibility criteria (66.7%), and it is likely that more referrals could have been received if the eligibility criteria were less rigid, for example in relation to meeting a threshold score on the GDS-LD. In addition, depression and depressive symptoms may be under-recognised in people with ID, as many people may have low-grade, chronic symptoms. Although we set an inclusion criterion of a GDS-LD score of 5, which is below the threshold for a diagnosis of depression (score of 12), the mean score at baseline in both groups was well above the cut-off value for depression (18.6 in the intervention group and 21.4 in the control group), suggesting that the majority of participants were experiencing episodes of moderate depression and were more impaired than expected. If individuals with significant depressive symptoms were being approached to take part in the study, it is not surprising that there was a reluctance to take part, and that the number of individuals being approached was smaller than anticipated. This indicates that there was possible confusion about the eligibility criteria, which may have contributed to the low recruitment rate.

The 6-month study recruitment period was possibly overambitious. A volunteer co-ordinator who was part of the PPI group and runs ‘the Gig Buddy Scheme’ in Oxford thought that a 12-month recruitment period would have been more realistic based on their own experiences.

Suggestions for improving recruitment

The PPI group made suggestions that recruitment via peers may have helped by reducing anxieties that people with ID may have about talking to professionals. They also made suggestions to broaden the eligibility criteria and to remove the need for individuals to have depressive symptoms. Another approach to improving recruitment that we considered but were not able to implement was an extension to the recruitment period, which was not granted, including approaching additional befriending services to take part, increasing the number of NHS sites from which referrals were received (including Patient Identification Sites) and involving the Clinical Research Network in recruiting participants from NHS sites. Another source of recruitment that we considered was voluntary services, such as advocacy organisations.

Recruitment of volunteers

Over the 6-month period, we recruited 10 volunteers (our target was to match all participants randomised to the intervention group with a volunteer, and we were therefore aiming to recruit at least 20 volunteers).

The participating befriending schemes were responsible for recruiting volunteers; they recruited through their websites and other volunteering sites, local newspapers, and colleges and universities. We were aware that recruitment of volunteers was potentially going to be challenging, particularly in the Suffolk area, which is a more rural location and has fewer transportation links than London. Therefore, access to transportation (e.g. a car) may have had an impact on the willingness of volunteers to take up this role.

Attempts were made to bolster recruitment through the research team advertising the study and benefits of volunteering to undergraduate psychology students at University College London. There was a lot of initial interest from students, but many were deterred by the duration of the intervention and the location of the befriendees who lived on the outskirts of London, with students preferring to work with people living in central London. We had limited control over the recruitment of volunteers as this process was managed by the befriending services. However, feedback from the befriending services in the qualitative interviews suggests that more support could have been provided by the research team in advertising the study at volunteer recruitment event days.

Suggestions for improving recruitment of volunteers

Understanding the motivation of volunteers may enable targeted recruitment based on clear messages and incentives. The volunteers who participated in the study cited career progression, enjoyment and good use of spare time as motivations for taking part. Students in health-related professions may volunteer because it gives them work experience that may help them to progress in their career.26 The recruitment of students from colleges and institutions nearer to the location of befriendees could help to overcome recruitment issues. The PPI group also suggested that volunteers may be apprehensive about working with people with IDs if they have not had prior contact. Inviting volunteers to a social activity group with people with ID beforehand may help to alleviate their anxieties and provide an opportunity to meet new people.

Retention of participants and volunteers

The retention of participants with ID who had been randomised was excellent. No participant dropped out of the study, and they all agreed to complete the outcome assessments at baseline and follow-up, apart from one participant who completed only a few measures at follow-up.

Retention of volunteers was fairly good. Of the 10 volunteers who participated in the study, only two dropped out (retention of 80%). The remaining eight volunteers completed the baseline and follow-up assessments. We therefore met our minimum criterion of < 30% dropout rate from the study.

Matching of participants and volunteers

We matched six out of eight participants in the intervention group with a volunteer (75%), which met our minimum criterion of 70%. One participant could not be matched because of geographical distance and another because of a deterioration in their mental health. One participant had to be rematched with another volunteer after the first volunteer dropped out of the study. No volunteers were matched with more than one participant. Four pairs continued their relationship beyond the 6-month period.

Adverse effects and unintended consequences

Befriending was not associated with any serious adverse events for people with ID. However, we were unable to complete the 12-month assessment and, therefore, it was not possible to monitor the impact of relationships ending on the well-being of individuals beyond the 6-month period. Four out of the six matched pairs had decided to continue their relationship beyond 6 months and the other two pairs met on only a few occasions. There was a positive outcome for one participant, who was supported to report an incident of mistreatment by a paid carer at their supported living placement. The participant may not have reported this incident without the support they received from the volunteer.

However, there were unintended consequences for volunteers. One volunteer experienced conflict with members of staff at their befriendee’s supported living placement, which arose because of a misunderstanding about the volunteer’s role. This consequently led to the volunteer becoming distressed and deciding to drop out of the study. Another volunteer was unable to continue with their relationship with the befriendee because the befriendee’s mother was not satisfied with the match. The volunteer was disappointed with this situation, particularly as the befriendee was not involved in making the decision. In both situations, the volunteer co-ordinator attempted to resolve the dispute and provided emotional support to the volunteers. Similar conflicts with staff and carers were reported in our survey of befriending services and volunteers in the UK,75 suggesting that this is not an uncommon issue encountered by volunteers. More support should be provided to volunteers to negotiate such challenging situations to prevent relationships breaking down unnecessarily, and service users with ID should be empowered to make decisions about whether or not they wish to continue with relationships, rather than these decisions being made by carers.

Adherence to the intervention

There was good adherence to the training by volunteers, and most volunteers engaged with the supervision process. The pairs were expected to meet for a minimum of 60 minutes per meeting, but this was exceeded (with an average meeting duration of 118 minutes). We set a minimum target of 10 meetings between the befriender and the participant. The average number of meetings was almost 12 (11.8) and, based on this, the minimum target was met. However, only four of the six pairs (66.7%) actually met ≥ 10 times. In Priebe et al.’s37 RCT of befriending in people with psychosis, only half the participants met their predefined threshold of at least 13 meetings over 1 year, suggesting that implementation may be an issue in befriending trials. In our survey of befriending services for people with ID, we found that there was a large variation in the frequency of meetings, with the majority of pairs meeting for a period ranging from every 2 weeks to every other month.75 In our study, we also stipulated that the majority of the sessions should be in the community, which did occur, but the range of activities was limited and mainly consisted of visiting cafes and restaurants, and going for walks. Trips to museums/galleries and to the cinema occurred on only a few occasions and there were no activities involving recreational activities, such as swimming or going to the gym. These findings are consistent with our survey of befriending services, which found that the most frequently reported activities involved going to cafes and restaurants, followed by visits to parks and outdoor spaces.75 One possible explanation is that individuals with ID did not always have sufficient funds to cover activities (this was raised by several volunteers), which might have prevented pairs from engaging in more novel activities, or it may be related to difficulties in motivating individuals who have depressive symptoms. Spending time indoors and having ‘a conversation’ was also frequently reported, highlighting the role of volunteers in providing emotional support. No particular types of activities were regarded as being more pleasurable, indicating that spending time together may be more important than the nature of the activity.

Acceptability of study procedures

Volunteers were satisfied with the trial processes and procedures and thought that the questionnaires were relevant and of the right length. Some participants with ID expressed their frustration at being allocated to the control group. They thought that some of the information was difficult to understand and that the questionnaires were too long.

Acceptability of the intervention

Based on the qualitative interviews, overall, the participants and volunteers valued the opportunity to have a befriender and to take part in the study. This, alongside the very high retention rates of participants in the study and the fact that four out of the six matched pairs had plans to continue their relationship beyond 6 months, suggests that the intervention may have been acceptable, at least to those who took part in the study. We did not conduct interviews with individuals who did not want to take part in the study, and their views about the intervention may have been different. The low recruitment rate may also indicate that the intervention was not widely acceptable to people with ID who had symptoms of depression because they were anxious about meeting someone they were not familiar with or were not motivated to engage in new activities.

There were some issues with the intervention. Volunteers commented on the need for more flexibility in relation to the frequency of visits, with a preference for visits every 2 weeks rather than weekly because of other commitments. Flexibility in supervision arrangements and the nature of contacts (e.g. permitting contact via social media) was also suggested. A recent qualitative study of the experiences of befriendees and befrienders who had participated in a RCT of befriending76 highlighted the importance of having a flexible approach in relation to the role of the befriender and the nature of the intervention, and that befriending was an authentic relationship that developed over a period of time.77 Burn et al.77 also discovered that befriending was a considerable commitment for volunteers and necessitated balancing a number of competing commitments, which was also reported in our study. The mutual benefits of befriending was a positive finding in our trial, which indicated that we should not assume that only participants with ID will benefit from the intervention.

Another highlighted issue, which may affect the acceptability of the intervention, was the expectations of carers, who had a strong influence on aspects of the intervention, such as whether or not a match should continue (this led to one participant being left unmatched after one session) and conflict with another volunteer over their selection of activities. As discussed above, this led to two volunteers dropping out. Carers misunderstood the role of volunteers and thought that they should assume the same responsibilities as a support worker. Guidelines on how to work with carers, and clear discussions with carers about the roles and responsibilities of volunteers need to be considered. Financial issues were raised by a number of volunteers, particularly during supervision. On occasions, volunteers had to pay for the participants’ expenses during visits, as carers did not realise that these expenses were not covered. Good communication between carers, volunteers and the befriending service was considered to be essential.

Measurement of outcomes

The lower than expected recruitment rate resulted in a small sample size and unbalanced groups, despite randomisation. However, our exploratory analysis of outcome revealed a mean four-point reduction in the GDS-LD score in the intervention group, after adjusting for the baseline score, which equates to a standardised effect size of 0.5 SDs. The 95% CIs around this difference in GDS-LD score (95% CI –11.2 to 3.2 SDs) suggest that, although this is a non-significant result, an important clinical difference could be found in a main trial. However, there were baseline differences in the characteristics between the two groups and, given the small sample size and that only six participants in the intervention group actually received the intervention (a further two received one or two sessions), it may well be a chance finding. Nevertheless, the findings are similar to that of a feasibility study of mentoring in 29 people with ID and a matched comparison group. They found a reduction in GDS-LD (proxy) scores in the intervention group, with an effect size of 0.51, and this approached significance (p = 0.10).41 This suggests that befriending may reduce depressive symptoms, but that our results should be treated with considerable caution because of the small sample size.

Previous studies of befriending in other populations have found small effect sizes in reducing depressive symptoms.21 However, one trial of a ‘visiting service’ for elderly widows found that participants who had high levels of social isolation, two or more physical illnesses or low levels of education (no education or primary school only) responded better to the intervention, with statistically significant effect sizes of between 0.54 and 0.63 for a reduction in depressive symptoms.76 This group shares characteristics with people with ID, suggesting that the benefits of befriending may be greater for more disadvantaged groups.

Health economic evaluation

Data collection methods for a future economic evaluation were developed, and their feasibility was tested during this study. There was a good return of the resource use questionnaire from carers. Data on intervention costs were obtained from befriending services. Participants completed the EQ-5D-Y descriptive system and the EQ VAS, but there were important discrepancies between the scores, which could make calculation of QALYs for a cost–utility analysis less reliable. The differences may have been due to a number of reasons, such as poor understanding of instructions on how to complete the EQ VAS or because the EQ-5D-Y is not an appropriate measure for adults with ID (it has not been validated for use in this group) and may not capture relevant aspects of health that are perceived to be important. In the absence of a validated measure of health-related quality of life in this population, perhaps the EQ-5D-Y should not be used in isolation to assess health-related quality-of-life outcomes in trials involving adults with ID, and should be used in conjunction with a proxy measure.

The results of our study show that the intervention has a probability of 35% of being cost-effective. These results are conservative and should be interpreted with caution because of the small sample size and other issues that affected the study.

Strengths of the study

• To the best of our knowledge, this is the first pilot study to investigate the feasibility and acceptability of conducting a RCT of one-to-one befriending by volunteers for people with ID and symptoms of depression. Other studies used a non-randomised design or did not have a control group.40,41
• The befriending intervention was delivered by community befriending schemes. This reflected actual practice, as the majority of befriending services are not funded by the NHS and may receive only partial or no funding from local authorities.
• Although some attempt was made to standardise the intervention through the recommendation of weekly visits (each lasting at least 60 minutes, over a period of 6 months), there was room for flexibility in terms of how the matched pair spent their time and, therefore, the trial was pragmatic in its approach.
• The mixed-methods process evaluation provided useful insights into the perspectives of the participants and data on adherence, which could potentially inform changes to the intervention in a future study.
• The trial included an economic evaluation to assess the feasibility of data collection methods and some preliminary data on resource use.

Limitations of the study

• The study recruited only 40% of the target sample of people with ID, which suggests that recruitment is not feasible without changes to our approach.
• Some of the recruitment issues identified could be rectified through better planning in a full trial; for example, there were clearly issues relating to poor communication between the study team and the befriending services. We could have been more proactive in our engagement with the befriending services and made more sustained efforts to ensure that they understood and were in full agreement with the study design and randomisation process, rather than assuming that they were on board. They provided feedback indicating that they would have appreciated support with the recruitment of volunteers through events hosted by the study team, and this is another aspect that could be improved.
• We were unable to implement significant changes to the recruitment strategies owing to delays in the study being set up and our request for an extension to the study not being approved.
• We were unable to complete a 12-month follow-up because of delays in the set-up of the study. We therefore do not have data on adverse effects and unintended consequences beyond the 6-month period, including the impact of relationships ending on depressive symptoms and the well-being of participants. This information was not available at the 6-month follow-up as several of the pairs were planning to continue their relationship beyond 6 months.
• Proxy outcome measures (e.g. for depressive symptoms) were not used, and they may have been helpful, alongside the self-report measures, in establishing whether or not the intervention had positive outcomes. Some of the measures used in the study had a limited number of data on psychometric properties.
• We did not collect demographic data on car ownership and whether volunteers were working full time or part time, which might be helpful in understanding the characteristics of volunteers who are likely to take part in such a study.
• We could not mobilise PPI in the study as well as we intended.

Amendments to the logic model

Based on our experiences of conducting this study, we have proposed a few changes to the logic model (see Appendix 5). These include the need for additional inputs and resources, such as clear eligibility criteria and a comprehensive recruitment strategy in relation to people with ID, and the need for individuals with ID to have access to sufficient funds to be able to access activities in the community. Additional enablers include ensuring that carers are aware of the roles and responsibilities of volunteers, and the need for good communication between the befriending services, carers and volunteers to reduce the likelihood of potential conflict.

Implications and recommendations for future research

Although our results suggest that a fully powered RCT would not be feasible, it is possible to rectify these issues through alternative recruitment strategies, including recruiting participants through multiple NHS trusts and recruiting volunteers from multiple befriending services, or the provision of adequate resources to fund the intervention outside community befriending services. A future study should explore these alternative recruitment strategies.

There are challenges in conducting research with voluntary organisations as they often lack the necessary resources to recruit participants and implement the intervention, and researchers may have limited influence over these aspects of the research. Circumstances may also change during the course of the research project, owing to financial insecurity associated with external funding sources. The importance and value of collaborating with voluntary organisations in research should be encouraged.

Conducting research involving participants with depressive symptoms (many of whom had a clinical diagnosis of depression) proved to be challenging owing to difficulties in motivating depressed individuals to engage in novel activities. It is not clear whether the low recruitment rate was because the intervention was not acceptable to individuals with depressive symptoms or because the recruitment strategy was limited. This requires further exploration. For example, individuals with ID and depressive symptoms could be surveyed about whether or not they would be willing to engage in a befriending intervention. A future study on befriending in people with ID could have broader eligibility criteria and not be restricted to participants with depressive symptoms or other symptoms of mental ill health, and have a focus on evaluating whether or not it leads to improvements in well-being and social outcomes in the short and long term.

One of the issues elucidated by this study was the need for improved communication between all parties involved in the befriending relationship. Poor communication and misunderstandings between carers and volunteers led to relationships breaking down. It is paramount that all befriending services develop guidelines on how to work more effectively with carers.

Future research studies could also explore the use of peer support, which is defined as the mutual receiving and giving of emotional, social or practical support between individuals with shared lived experience.78 It has been shown to have positive outcomes in studies of people with mental illness.79 Although the use of a peer support model or a ‘buddy scheme’ in relation to people with ID has been used in a number of settings, research has been limited to only a few studies. One pilot study suggests that peer support may be useful in facilitating supported employment in young people with ID80 and another suggests that having a ‘buddy’ may help young people transition to a new school or college.81 Peer support could be used more broadly to deliver changes in social outcomes and may be more acceptable to some individuals with ID, as they may find it easier to relate to peers than lay volunteers.

Future studies should give careful consideration to the design of the study, such as the use of non-randomised study designs in evaluating befriending.37 For example, a prospective longitudinal study that involves the collection of data from a large number of services, and takes into account key confounders, could provide data on longer-term benefits.

A further recommendation is the need for effective involvement of patients and the public in the study design and in the recruitment of participants to identify and overcome key barriers to recruitment. Recruitment approaches could involve people with ID publicising the study (through the use of film clips and social media) and directly recruiting participants by explaining the study to their peers and describing personal experiences of befriending and how it has benefited them.

Conclusion

This pilot RCT of one-to-one befriending in people with ID and depressive symptoms failed to recruit a sufficient number of participants. Therefore, a future, larger-scale RCT is unlikely to be feasible unless there are substantial changes to the recruitment of participants with ID involving multicentre recruitment from NHS and social care services.