Putting Life in Years (PLINY): a pilot randomised controlled trial and mixed-methods process evaluation of telephone friendship groups for community living older adults.

Participants

Two main methods were used to identify potentially eligible study candidates. We worked with an existing research cohort that is following the lives of 20,000 adults in the area over a period of 10 years and includes individuals who have signalled a willingness to be contacted about further research. Between June 2011 and July 2011 we sent letters with a postage-paid response card and a candidate leaflet to 528 participants in the cohort aged ≥ 75 years. We also invited general practices to help identify potentially eligible study candidates. Between June 2011 and December 2012, 18 general practices sent letters to 9051 patients. The letters included the same candidate leaflet and an invitation to complete a postage-paid response card to express an interest in the study. Response cards were returned to the recruiting site (University of Sheffield, Sheffield, UK).

A pack containing the same candidate leaflet and postage-paid response card was also given to workers in services across the city that were likely to come into contact with older people with the aim of asking them to identify potential participants. In many instances, researchers personally delivered packs and spent time with workers explaining the aims of the study and what it entailed. The Community Intermediate Care Service (NHS) was provided with 500 packs, the city council’s main library received 50 packs and the mobile library service received 200 packs. In addition, two A3 and 30 A4 posters were provided for display. The Community Access and Reablement Service (CARS) was given 100 packs and the local Meals on Wheels service received 120 packs. In addition, 200 packs were given to an extra care scheme (housing with care services available if or when required), a local housing association and the local churches council for community care. Research assistants distributed 200 packs at community events in the locality including a Lifewise event, Regenerate RISE (Reaching the ISolated Elderly) and a local well-being festival (150 packs). Two referrer information sheets and 100 packs were sent to the Allied Healthcare Group; one referrer information sheet and a study leaflet were sent to the Older People’s Partnership Board and distributed to its network (22 May 2012); and five packs each were given to nine Healthy Living Pharmacies.

An unknown number of packs was also sent to relevant public and voluntary sector outlets: the local Expert Elders Network, the local Pensioners Action Group, the local Wellbeing Consortium, Age Well, a victim support group, an older adults community mental health team and a black and minority ethnic community mental health development worker.

Information about the study was circulated to local community and media outlets including the city council’s Help Yourself web page and the local newspaper.

Research assistants telephoned all potential candidates who had returned a response card. A number of candidates telephoned the research team directly. Research assistants checked initial eligibility during the telephone call, for example age and living situation. Research assistants arranged to visit those who were identified as being potentially eligible and interested in finding out more about the study. Appointments were arranged approximately 5 days after the telephone call to allow sufficient time for the candidates to receive and read the Participant Information Sheet, which was posted out (or e-mailed on request) by the research assistants. The Participant Information Sheet was reviewed by the lay representative on the TMG as part of the submission of essential documents to the REC. Research assistants visited potentially eligible candidates in their own home to conduct a screening visit. Those eligible to join the study were aged ≥ 75 years; had good cognitive function, defined as having a six-item Cognitive Impairment Test (6CIT) score of ≤ 7; were living independently (including those who were co-resident with others) or in sheltered extra care housing; and were able to understand and converse in English. The exclusion criteria were (1) the inability to use a telephone effectively with appropriate assistive technology; (2) living in a residential/nursing care home; and (3) already receiving a telephone intervention.

Written informed consent was obtained by research assistants either at the screening visit or at a separate visit if additional time was required to make a decision whether or not to participate. Research assistants administered the 6CIT and calculated the score during the visit. Candidates who were ineligible because of a 6CIT score > 7 were subsequently contacted by a clinically qualified member of the research team and told that they were not eligible to be involved and advised to contact their doctor. A letter containing the score was sent to the candidate. For candidates who were eligible, the research assistant taking consent and administering the baseline questionnaires informed another member of the research team (research assistant or trial manager) of the screening identifier so that they could randomise and inform participants of their allocation. On allocation, and before they were contacted by a volunteer, participants allocated to receive telephone friendship (TF) were sent a ‘question and answer’ document about TF groups by the research team (see Appendix 2) and were advised that the service provider’s volunteer would contact them. The research assistants or the study manager informed the service provider of intervention participants by letter. Initially, this was carried out each time a participant was allocated to receive TF. However, the research team and the service provider subsequently agreed to wait until six participants (sufficient to make a group) had been allocated before forwarding details to the service provider.

Participants were able to withdraw from active participation in the study on request. Individuals who withdrew from the intervention were not replaced. Written consent was obtained to share information with the NHS Health and Social Care Information Centre and other central UK NHS bodies to check participants’ health status and help minimise the risk of telephoning or writing to participants who died before follow-up. Both study arms received postal updates on the study at 2 and 4 months after randomisation.

Interventions

The intervention design is detailed in Appendix 3.

Candidates were screened as described in the previous section. Those who consented were randomly allocated to one of two groups (see Randomisaton and blinding):

1. TF group calls provided through the voluntary (charitable) sector

2. a control group who received usual health and social care following randomisation.

The aim of the intervention was to increase contact between individuals with the intention of forming new acquaintances and friendships. By improving perceptions of companionship and support the aim was to reduce perceived isolation and improve participants’ sense of confidence and mental well-being. The intervention was designed by MC and built on the findings of a previous study which suggested that group calls, following one-to-one befriending, may help older adults to share interests. 24,25

The interventions were delivered by trained volunteer facilitators. Volunteers were recruited by the service provider. Volunteers had no previous experience of protocolised befriending or facilitating conversations, either face to face or by telephone.

The one-to-one individual intervention consisted of up to six calls between each participant and a volunteer befriender. The purpose of the one-to-one calls was to support the participant and prepare him or her for the group conversations. One-to-one calls were brief (10–20 minute) friendly conversations that were held each week for a duration of 6 weeks, beginning with familiarisation and everyday conversation and moving towards a focus on the group calls including topics of interest and supporting the participants with concerns about starting group sessions. Volunteer befrienders telephoned participants using the Community Network’s teleconferencing system. Although not designed for one-to-one calls, the use of the system enabled cost-free calls for participants and volunteers. A detailed description of the training is provided in Appendix 4.

Objectives

The primary objective of the main study, a parallel-group RCT, was to determine whether mental well-being, as measured by the Short Form questionnaire-36 items (SF-36) (mental health dimension) 6 months after randomisation, is significantly increased in participants allocated to receive the TF group intervention compared with participants allocated to a control group (receiving only contact by card/letter at months 2, 4, 8 and 10 with no further contact other than follow-up assessment).

Secondary objectives were to:

1. Identify, using qualitative methods, the psychosocial and environmental factors, as well as implementation issues, that may mediate or modify the effectiveness of the intervention, specifically voluntary sector readiness to take forward new forms of services, the best modes of delivery of telephone support/friendship, how volunteers (facilitators) can be supported and retained, and the extent to which fidelity of the intervention is maintained within and across the participating organisations.

2. To determine any lasting impact on mental well-being by repeat measurement with all participants 12 months after baseline assessment.

3. To examine whether there is any significant improvement in the intervention arm compared with the standard care arm in the physical dimension of the SF-36 at 6 months and 12 months following baseline assessment.

4. To measure the extent of use of health and social care and community facilities by participants over time to determine whether the intervention is cost-effective compared with standard care.

Outcomes

Table 2 shows the timing of the assessments and interventions. All baseline assessments and interventions were carried out in participants’ homes using the case report form (see Appendix 5). Follow-up assessment at 6 months post randomisation was carried out by telephone (unless a home visit was indicated). The primary end point was the level of mental well-being at 6 months post randomisation using the SF-36 mental health dimension. Secondary end points were:

1. other dimensions of the SF-36 to measure all aspects of health including physical health32

2. the Patient Health Questionnaire – nine questions (PHQ-9)33

3. the European Quality of Life-5 Dimensions (EQ-5D) score (for health economic analysis)34

4. the General Perceived Self-Efficacy Scale (GSE) score35

5. the de Jong Gierveld Loneliness Scale score36

6. Office for National Statistics (ONS) well-being measure37

7. a health and social care resource use questionnaire to collect participants’ use of health, social care and community services (for health economic analysis). 38

All primary and secondary outcomes were measured at 6 months post randomisation.

Sample size

For the purposes of sample size estimation the primary outcome was the mean SF-36 mental health dimension score at 6 months post randomisation. The SF-36 mental health dimension is scored on a scale from 0 (poor) to 100 (good health). A previous general population survey of residents demonstrated that the SF-36 can successfully be used as an outcome measure for community-dwelling residents aged ≥ 75 years, with a response rate of 82% being achieved. 39 From this general population survey of 3084 community residents, the mean SF-36 mental health score was 68.3, with a standard deviation (SD) of 19.9. 39

The developers of the SF-36 have suggested that differences between treatment groups of between 5 and 10 points on the 100-point scale can be regarded as ‘clinically and socially relevant’. 40 We assumed a SD of 20 points for the SF-36 mental health dimension at 6 months post randomisation and that a mean difference in mental health score of ≥ 8 points between the intervention group and the control group is the smallest difference that can be regarded as clinically and practically important.

Assuming that a mean difference of ≥ 8 points on the SF-36 mental health dimension between the intervention group and the control group is the smallest difference of clinical and practical importance that is worth detecting, then with 248 subjects (124 intervention, 124 control) the trial was originally determined to have 90% power to detect this mean difference or greater as statistically significant at the 5% (two-sided) significance level using a two independent samples t-test. We assumed a correlation of 0.50 between the baseline and the 6-month SF-36 mental health scores. However, the telephone befriending intervention is a group or facilitator-led intervention. Therefore, the success of the intervention may depend on the volunteer facilitator delivering it so that the outcomes of the participants in the same group with the same volunteer facilitator may be clustered. We therefore assumed an average cluster size of six participants per telephone befriending group and an intracluster correlation (ICC) of 0.04 so the design effect is 1.28. With these assumptions and 99 participants per group, the power of the analysis was reduced to 80% to detect a mean difference of ≥ 8 points in the 6-month SF-36 mental health score. If 20% of the participants drop out and are lost to follow-up then we would have needed to recruit and randomise 124 participants per arm (248 in total).

Randomisation and blinding

Eligible participants were randomised to one of the two arms by the trial manager or a research assistant through a centralised web-based randomisation service provided through the Clinical Trials Research Unit (CTRU). The randomisation sequence was generated in advance by a CTRU statistician, not by the trial team. There were no stratification factors in the randomisation sequence. A sequence of treatment/intervention assignments was randomly permuted in blocks of varying size to ensure that enough participants were allocated evenly to each arm of the trial. Participants, outcome assessors and the trial manager were not blind to treatment allocation because of the practical nature of the intervention. All outcomes were self-reported using validated questionnaires (except for sociodemographics and health and social care resource use, which were assessed using bespoke instruments). Trial statisticians and the principal investigator were blinded to the treatment allocation codes until after the final analysis. Data presented to the trial steering committee (TSC) and the TMG did not identify treatment allocations.

Statistical methods

Economic analysis

The case report form (see Appendix 5) included questions about the use of primary and secondary care health services, social care services and voluntary and private sector services.

The following elements were planned for the health economics analysis:

1. Costing of the TF service.

2. Costing of participants’ health, social care and voluntary service use during the trial.

3. Cost-effectiveness analysis using a range of outcome measures and a cost–utility analysis using the EQ-5D. The resulting cost per quality-adjusted life-year (QALY) would be compared with the NICE threshold of £20,000–30,000 per QALY gained.

4. An exploratory analysis of participants’ willingness to pay for a TF scheme.

Because of the early closure of the trial for reasons outlined in Chapter 4 (see Assessment of study feasibility), a high proportion of participants allocated to the intervention arm did not receive the intervention and it was therefore not appropriate to conduct the planned health economics analysis. Frequency tables for participants’ service use are presented in Chapter 4 (see Health and social care resource use); differences between the intervention group and the control group at follow-up should be interpreted with caution.

Background

This report is concordant with Consolidated Criteria for Reporting Qualitative Research (COREQ) guidelines for reporting qualitative research. 47 The purpose of the qualitative research was to evaluate the impact of TF groups on older people as well as their perceived advantages and disadvantages in terms of well-being. The objective was an assessment of the acceptability and appropriateness of the intervention for preventing loneliness and maintaining good mental health. Some aspects of the fidelity assessment (e.g. views on the receipt and enactment of the intervention) were also informed by the qualitative research (see Methods for the fidelity assessment).

Methods for the participant interviews

The aim of the qualitative research was to explore to what extent older people considered TF groups to have made an impact on their well-being. A topic guide (see Appendix 6) was used to undertake semistructured interviews. This was based on a previous study25 and was tailored by MC and RG-W in line with the secondary end points for the trial. The topic guide was not piloted but was reviewed by the lay representative on the TMG as part of the review of essential documents submitted for ethical approval.

The topic guide covered questions regarding participants’ needs and expectations of telephone befriending, its impact on their health and well-being and accessibility and acceptability of the telephone discussion. It also inquired after participants’ experiences of the volunteer facilitator and whether they felt that the telephone discussions were or were not a good way to give them the support that they needed.

All participants allocated to the telephone befriending intervention and provided with a volunteer facilitator were invited (by telephone) to participate in a semistructured interview (24 individuals). We interviewed all 19 participants who volunteered to take part. Reasons for declining a research interview were not elicited but were recorded if volunteered by the participant. One female research associate (RG-W) and one female research assistant (RD) performed the interviews in April 2013. RG-W had studied qualitative research techniques as part of her MA in Research Methods and had experience of in-depth and semistructured interviews. RD had experience of qualitative interviews but was a novice in terms of qualitative analysis. Neither of the interviewers delivered the intervention to the interviewees. RD visited some interviewees to collect baseline data and/or collected 6-month follow-up data by telephone for the main trial. Interviewees would have known that the interviewers were on the research team and were from the University of Sheffield and may have associated them with the volunteers delivering the intervention and/or with delivery of the intervention. The interviewers were asked to withhold their own opinions, personal goals and characteristics and to reiterate the purpose of the research and that this interview was separate from the intervention. No repeat interviews were undertaken or field notes taken.

The interviews lasted between 14 and 63 minutes (median 29 minutes) and were conducted face to face in a place selected by the participants. All were conducted in participants’ homes. Written consent for audio recording was obtained when participants entered the study. For all but three interviews no one was present except for the participant and the researcher. Three interviews were interrupted by (1) a participant’s daughter (bringing a drink), (2) a participant’s cleaner and (3) a visitor. Sociodemographic data were collected from participants as part of the main trial and are reported in Chapter 4. Interviews were digitally recorded and transcribed verbatim. Transcripts were not returned to the participants for comment or correction.

Data analysis commenced during the data collection period using a constant comparative method to identify themes and where interviews and analysis each informed the other. Data analysis of the transcripts was conducted using NVivo 9 (QSR International, Warrington, UK). We used a ‘framework’ approach to analysis in which a priori and emergent themes were identified using the following stages: familiarisation, identifying a thematic framework, indexing, mapping and interpretation (charting was not undertaken). 48 A priori themes of interest were acceptability and accessibility of the group telephone discussion; subthemes were derived inductively through familiarisation with the transcripts. 48,49 Results were used to explore factors that may have mediated and moderated the intervention and contributed towards the findings of the trial and to identify any other emerging issues or factors that may have influenced the uptake of the intervention and which had not previously been documented. 50 It is unlikely that data saturation was achieved. 51 Participants were not asked to provide feedback on the identified themes.

Initially, we indexed transcripts using our own thematic framework (see Appendix 7); this was later supplemented with codes based on concepts from the group dynamics literature, relating to characteristics of functioning groups, specifically group cohesion and disclosure. 31

Methods for the volunteer interviews

The aim of the qualitative research was also to explore the experiences of volunteer facilitators in delivering the intervention. A semistructured topic guide was used with themes including the accessibility and acceptability of delivering befriending by telephone and the motivations of volunteers to take part and their experiences of facilitating group discussions, including their perceptions of participant benefit. All volunteers who remained in contact with the service provider were invited to take part in a semistructured interview. This included volunteers who dropped out before, mid and post completion of facilitator training, resulting in a sample of three. Two had completed the one-to-one and group telephone call phases of the intervention and one dropped out during the group facilitator training. We did not elicit reasons for declining an interview. RG-W performed the interviews in April 2013 and had a previous relationship with all volunteers having attended volunteer induction and one-to-one training sessions to support the service provider (by request). Interviewed volunteers knew that the interviewer was on the research team and from the University of Sheffield and not from the service provider. The interviewer was asked to withhold her own opinions and to reiterate the purpose of the research and that the interview was separate from the intervention and the service provider. No repeat interviews were undertaken or field notes taken.

The interviews lasted between 18 and 59 minutes (median 43 minutes) and were conducted face to face in a place selected by participants. Two interviews were conducted at the home of the volunteers and one was conducted at the University of Sheffield. For all interviews no one was present except for the participant and the researcher. Sociodemographic data were not collected for volunteers. A topic guide (see Appendix 6) was used. This was based on the secondary end points of the study and was informed by some elements of the fidelity framework, for instance whether the group experienced conflict or followed ground rules (see Methods for the fidelity assessment). The topic guide was not piloted.

Interviews were digitally recorded and transcribed verbatim. Transcripts were not returned to the participants for comment or correction. Data analysis of transcripts was conducted by RG-W by hand using a constant comparative method to identify themes. Analysis and interpretation followed relevant themes from the qualitative research framework developed from the participant interviews (see Chapter 5, Results of the participant interviews). Results were used to explore potential explanations for the quantitative findings and identify other emerging issues or factors influencing volunteer-led interventions. The final outcome was a synthesis of coded data and subthemes including those relevant to the fidelity assessment.

Telephone befriending design

To assess comparable ‘treatment dose’, the number, frequency and duration of one-to-one and group telephone contacts were established. The minimum number of one-to-one contacts was recommended as three on the basis that some participants would need more one-to-one contacts to be sufficiently confident to join the group discussions. A maximum of six one-to-one contacts was set. A maximum of 12 group telephone contacts was established and a minimum (in terms of treatment dose) was set at nine (of 12) group calls. The frequency of all telephone contact was weekly.

Telephone befriending training content assessment and methods

Volunteers were trained by Community Network’s group facilitation skills trainer. Attendance at training sessions was monitored by register taken by the single trainer who trained all volunteer facilitators. The training was delivered over the telephone to a number of trainees (maximum five) over four 1-hour sessions. The session content focused on providing skills and techniques to enable the facilitator to support the group to work well as a group and fulfil its purpose (see Methods for the main trial, Interventions).

The trial manager (RG-W) developed a fidelity checklist based on the standard training delivered to all volunteers who facilitate telephone discussion via the Community Networks’ teleconferencing system. This was reviewed by the content expert (MC) and the Community Network trainer to ensure that core components were included and that materials and practice delivered by the trainer could be assessed for consistency across groups. The checklist also included components to assess volunteer facilitator skill acquisition. The checklist was piloted by RG-W and MC using a sample of audio recordings, with modifications made where necessary (see Appendix 8, Training content checklist).

A purposive sample of training sessions was audio recorded across and within the training groups. RG-W and a research assistant (LN) used the training content checklist to assess the content and delivery techniques conveyed to trainee (volunteer) facilitators and facilitator skill acquisition. Checklists were completed and scored separately by the two coders and scores were compared. All scores were reviewed to ensure consistency in interpretation of the checklist items, with areas of dispute discussed and agreement reached by consensus. When agreement could not be reached the original observations, and therefore scores, remained the same. Median scores were calculated to provide an overall fidelity score for each training group and show the degree of consistency in the content delivered to trainee facilitators.

Treatment fidelity assessment and methods

To ensure that the criteria for treatment fidelity were met, those delivering the group befriending intervention (volunteer facilitators) were assessed for adherence to the intervention protocol across the 12 weeks. The assessment of treatment fidelity by volunteer facilitators used an intervention delivery checklist developed by RG-W based on the techniques delivered in the facilitator skills training and on a one-to-one training manual provided to volunteers (see Appendix 4). The training fidelity checklist was reviewed by MC and the Community Network trainer with modifications made where appropriate (see Appendix 8, Intervention delivery checklist).

According to the facilitation skills training, as groups develop, the level of input required from the volunteer facilitator should diminish over time. Therefore, the sample of audio recordings included three time points (weeks 1, 6 and 12) to assess the degree to which volunteers used their acquired skills and adhered to the intervention protocol during delivery. The checklists were designed to take into account variation in the content of sessions and the fact that, if some scenarios did not arise, volunteer facilitators could not be expected to demonstrate the appropriate response.

Telephone befriending delivery

Attendance at all planned sessions was recorded through call registers completed by volunteer facilitators at every session during both the one-to-one phase and the group phase. Volunteer facilitators recorded any difficulties with the delivery of the intervention protocol on the call registers. Issues arising during delivery of the intervention were noted by the service provider and forwarded to the research team. The challenges of implementation and barriers to uptake were examined with a convenience sample of volunteer facilitators (see Methods for the qualitative research and Chapter 5, Results of the volunteer interviews).

A purposive sample of audio recordings of group sessions was used to assess the match with the intervention protocol in terms of the content and techniques delivered and the extent to which volunteer facilitators enabled choice and decision-making. Note that the protocol (see Appendix 1) also incorrectly refers to the concept of intervention ‘drift’, which implies a trend away from intervention fidelity known to exist at baseline, something not established in this study. Samples were taken at three time points – weeks 1, 6 and 12 – to examine intervention delivery and volunteer facilitator skills and receipt of the intervention and enactment by participants. Checklists were completed and scored separately by the two observers and scores were compared. Coders reviewed scores to ensure consistency in interpretation of checklist items, with areas of dispute discussed and agreement reached by consensus. When agreement could not be reached the original observations, and therefore scores, remained the same. Median scores were calculated to provide an overall percentage score for each facilitated group.

Telephone befriending receipt and enactment

Unlike formal behaviour change interventions, such as cognitive–behavioural therapy or motivational interviewing, the PLINY intervention does not attempt to tightly regulate behaviour outside the delivery setting. The intervention attempts to reduce the discrepancy or (mis-)match between the quality and quantity of existing relationships and relationship expectations. 55 The intervention, through facilitated dialogue between participants, is intended to create a safe environment in which social contact can improve perceptions of available companionship and support (see Appendix 3). It follows that, if the volunteer facilitators are delivering the intervention per protocol, then the group is ‘working’ well and participants should not be exhibiting problem behaviours known to inhibit successful group experiences. For this reason, we limit our assessment of participant receipt and enactment to evidence of their performance as part of a friendship group. To try and identify whether participants found the group intervention to be appropriate, acceptable and beneficial (see Methods for the qualitative research and Chapter 5, Results of the participant interviews), we specifically reviewed interview transcripts with participants and volunteer facilitators for evidence of characteristics which indicated that the groups were in a transitional phase, towards functioning well as a group:31

• defensiveness and resistance

• conflict

• confrontation.

Other characteristics – anxiety, the struggle for control, challenges to the group leader and the leader’s reactions to resistance – were not seen as relevant in this intervention, as the level of disclosure, anxiety and resistance within befriending groups was anticipated to be lower than in a therapeutic group.

Interview transcripts and audio recordings were also reviewed for the following problem behaviours that are counterproductive to group functioning:31

• silence and lack of participation

• monopolistic behaviour

• hostile behaviour

• dependency

• acting superior

• socialising (before the end of the programme)

• ‘band-aiding’ (e.g. try to sooth/lessen pain when someone is upset).

Other behaviours – storytelling, questioning, giving advice, intellectualising, emotionalising (dwelling on getting in touch with their feelings) – were not seen as problematic in this intervention.

Conditional funding

It was originally intended to use multiple service providers to deliver the intervention. This was not possible for the following reasons. We considered it likely that the delivery at scale of a manualised intervention by volunteers would need the stable base offered by formal training and monitoring by experienced volunteer co-ordinators. For this, it was essential to secure funding. The funding was secured from a national charity on condition that we would use the money to deliver the intervention through one or more of its local branches only; we were unable to use other organisations to deliver the intervention. We looked at the viability of recruiting other branches of the charity to deliver the intervention. Other branches did express an interest but were unable to provide the intervention to participants recruited in the urban centre where the study was ongoing. Each branch of the charity was restricted through its constitution to serve the needs of its (bounded) local population. No branch could provide volunteers to work outside its geographical area. The research team was not adequately resourced to work in other geographical areas, which, in participant recruitment terms, was unnecessary in a conurbation with an estimated population of over half a million people.

An overview of the funding made available by the national charity is provided in Table 3. Detailed breakdowns follow of the resource for (1) the service provider, the local branch of the national charity, to recruit, train and mentor volunteer befriending facilitators (Table 4) and (2) a specialist trainer in group facilitation to support the manualisation of the intervention, provide advice on assessing its fidelity and train the volunteer facilitators (Table 5).

The contract with the service provider

Contractual negotiations with the service provider ran between 20 October 2011 and 14 June 2012 when the contract was signed. The service provider was contracted to:

1. Identify and recruit suitable volunteers for the role of volunteer/facilitator for the delivery of the PLINY research intervention, including carrying out Criminal Records Bureau (CRB) checks.

2. Ensure that volunteers are oriented to working with older people and willing to deliver telephone befriending, are trained to carry out one-to-one calls in line with the PLINY research intervention and are ready to receive training in telephone befriending facilitation (to be delivered by a third party) in the numbers and by the dates shown in Table 6.

3. Ensure that volunteers take responsibility for scheduling and (subject to participant adherence) delivery of up to six one-to-one and 12 group telephone sessions for each person recruited to the PLINY research study and randomised to the TF group.

4. Ensure that there are sufficient volunteers to provide cover in the event of volunteer facilitator absence or discontinuation.

5. Provide ongoing ‘mentoring’ to volunteers, in line with the service provider’s policies and procedures and the PLINY research intervention, to ensure a point of contact and support.

6. Provide regular (at least monthly) updates to the research team on levels of volunteer recruitment and retention and feed back information, including the one-to-one and group call registers, to inform the research.

7. Alert the research team at the earliest opportunity if a participant wishes to withdraw or is unable to participate in the intervention (TF groups), with reasons recorded (if provided by the participant).

Item (3), the delegation of first contact and scheduling of calls, might not be considered best practice for sustaining a volunteer befriending service. For instance, a Delphi survey of volunteer co-ordinators managing befriending services found general agreement that they should be managed either by a full-time or a part-time project co-ordinator. 25 The volunteer co-ordinators also agreed that it was essential to have a monitoring system in place (p. 51). 24 We were unable to broker such an arrangement within the available finances.

Marketing

Activity by the service provider to promote the volunteer opportunity included its website (news archive, 25 September 2012; accessed 10 May 2013), the Northern Community Assembly website (field note, 21 November 2012), the local Wellbeing Consortium (field note, 20 November 2012), a local newspaper and a range of community and voluntary networks and organisations available in the locality, which we have not named to preserve the anonymity of the service provider (field note, 14 June 2012; TMG, 15 November 2013). The service provider reported that potential volunteers referred to them by other agencies (e.g. Jobcentre Plus) were often not suitable for the facilitator role (TMG, 19 September 2012).

Suggestions for additional strategies to promote the volunteer role in the locality were made by the study team (e.g. TMG, 15 November 2013).

Training

Training sessions for the group intervention required a minimum of four volunteers for the training group to be feasible. The service provider identified an initial group of six volunteers early in the project (TMG, 20 February 2012) and scheduled one-to-one training for them in March 2012. The charity reported a number of implementation issues including matching the availability of volunteers to training dates (TMG, 20 February 2012). They also found that retaining volunteers between recruitment and training was difficult and required more resources than anticipated (TMG, 20 August 2012). It should be noted that the first group of volunteers (n = 4 in two groups) received induction and one-to-one calls training from the service provider in March 2012, 2 months before the scheduled start of participant recruitment. In fact, participant recruitment did not commence until June 2012, 1 month late, because of delays in contracting. A lower than anticipated response to the initial recruitment strategy (direct mail out to participants of a population cohort) meant a further delay before the research team had recruited and randomised the six intervention-arm participants needed for a group. According to the service provider, this delay caused the attrition of several existing volunteers (see Figure 1). At a time when the rate of participant recruitment was starting to increase, the service provider advised the study team that it was not actively recruiting volunteers (TMG, 19 September 2012) because there was an insufficient number of randomised participants. Instead, the service provider was waiting for candidate volunteers to approach them in response to advertisements.

Once the research team had managed to increase the rate of participant recruitment through general practice mail-outs, the service provider experienced repeated difficulties identifying volunteers to fill facilitator training groups. As a result, the first two training sessions (May 2012) contained only two genuine volunteers; to make the training viable, the service provider’s staff and members of the study team – who did not intend to deliver the intervention – made up the places to make the training viable. A finite training budget meant that running sessions with insufficient numbers of genuine volunteers was not sustainable. As a result, we agreed that the four (ideally five) places on training sessions scheduled for some time in the future had to be filled by a certain date – the ‘book by’ date – or they would be cancelled. ‘Book by’ dates were arranged with the group facilitator trainer to assist the service provider as it reported (TMG, 14 June 2012) practical difficulties in co-ordinating volunteers at the times and pace required by (1) the trial, which had a window of 1 year to recruit 248 participants to test the effectiveness of a public health intervention, which had to be rolled out at scale, and (2) the group training (four 1-hour telephone sessions on different days). The service provider did not always confirm whether sufficient volunteers had been identified by the ‘book by’ date despite reminders from the trainer/study manager (e-mail and telephone, 16 November 2012).

The total number of volunteers group trained between 17 May 2012 and 22 October 2012 was 11, instead of the 20 who should have been trained. Two trained volunteers were not available to take on a group; one was on a student placement with the service provider and needed to return to full-time education and one was available for only 1 day per week, having assumed that they could make befriending calls in the evening. Three training sessions (during which 15 more volunteers should have been trained) were cancelled between August 2012 and January 2013 because of a lack of take-up. Three volunteers facilitated four groups (n = 24) to completion between September 2012 and May 2013 (with up to 6 weeks one-to-one befriending beforehand). One group received one-to-one befriending from a fourth volunteer facilitator who dropped out before the group stage. An existing volunteer took over for the group calls stage (see Monitoring volunteers). The number of days that volunteers ‘survived’ in the project (from completing group training to the day that they dropped out) ranged from 12 to 118 (mean 62 days).

Monitoring volunteers

Feedback from volunteers was collected by the service provider and reported to the TMG and, when relevant, to other volunteers delivering the service. The study team also captured implantation issues during set-up and recruitment in field notes.

The service provider was responsible for providing ongoing ‘mentoring’ to volunteers, in line with its existing policies and procedures relating to volunteers and the intervention protocol, to ensure a point of contact and support for the volunteers whilst they were delivering the TF service. The charity provided a summary of the project in its induction pack together with copies of the one-to-one training manual (field notes, 20 March 2012, 15 June 2012, 9 October 2012). Volunteers often contacted the study team with enquiries about what to do in certain circumstances, for instance if participants missed calls and the facilitator had been only able to contact one (of six) participants in the first week (field note, 30 October 2012), they were going on holiday (field note, 17 September 2012) or if they experienced technical difficulties with audio recording calls (field note, 29 September 12) (see also Boundaries between research and service delivery). The reasons why volunteers contacted the study team rather than the service provider are considered in Boundaries between research and service delivery and Chapter 5 (see Results of the volunteer interviews).

Volunteers reported difficulties in contacting participants to arrange the initial and subsequent one-to-one telephone calls (see Chapter 5, Results of the volunteer interviews). Volunteers reported that it would be better to make calls in the early evening and that some participants had also reported this. however, to safeguard participants using the service the provider did not permit volunteers to make calls before 0900 or after 1700 from Monday to Friday. This resulted in one volunteer dropping out (see Figure 1).

In reviewing volunteer identification and recruitment (field note, 5 December 2012), the service provider identified three issues that it felt were impacting on the recruitment and retention of volunteers: (1) existing volunteers found it difficult to contact participants for the one-to-one calls as they were ‘socially active’, resulting in the volunteers being reluctant to take on another group; (2) there was a time delay between volunteers being trained and actually delivering the service; and (3) there was a lack of introductions by the service provider between the volunteers and the ‘participants’. The service provider decided that volunteers would introduce themselves to study participants at the first contact (one-to-one call) and be responsible for scheduling one-to-one and group calls (field note, 22 March 2012). One of the volunteer co-ordinators stated that, for their face-to-face visiting service, volunteers often want to be introduced to clients by the service provider staff and felt that this may have contributed to the difficulty in recruiting or retaining volunteers (field note, 2 January 2012).

One volunteer (out of 10) who completed all training dropped out during intervention delivery (between completion of the one-to-one calls and the start of the group calls) because of ill-health. Attempts by the service provider to contact two trained volunteers failed and, because of the already limited pool of volunteers, the only facilitators available were those already running a friendship group. The service provider reported contacting participants of the group to let them know and asked the existing volunteers (n = 2) if they would be willing to take on the group (field note, 5 December 2012). One volunteer agreed but did not run the groups concurrently, which resulted in a delay in the group calls starting. Some participants were difficult to contact (field note, 14 February 2012) and at least one member of the group was not contacted (see Chapter 5, Results of volunteer interviews).

In accordance with the intervention design, the service provider asked volunteer facilitators to discuss with participants in their groups whether they would like to volunteer to run their own group when they had finished their own involvement, thereby contributing to the notion of older people helping other older people. Feedback on responses was not provided.

Boundaries between research and service delivery

Boundaries between the research and the service being investigated were blurred in some instances. Participants called the study team to advise that they would miss the week’s call; volunteers called the study team directly if they were unsure about what to do (see Monitoring volunteers). For instance, one volunteer enquired about what information participants had received about ‘how it all works’ (field note, 26 September 2012). Another enquired about how the group could exchange contact details so that members could meet up (field note, 25 March 2013). The service provider also referred volunteers to the study team for information, for example one volunteer enquired about the Christmas period (VF03, female, field note, 3 December 2012) and one volunteer (not interviewed, female) contacted the team because she lacked confidence in making the calls via the Community Network system (field note, 2 November 2013). Additional training was provided to this volunteer by the study team (field note, 8 November 2012).

Main results

Figure 2 shows the participant flow diagram; 157 participants were consented and randomised to the intervention group (n = 78) or the control group (n = 79).

FIGURE 2.

Participant flow diagram. SYC, South Yorkshire cohort. a, Three were ineligible, 21 were not interested in the intervention, 11 did not have enough time to participate, two were not willing to complete the questionnaires, two needed more information, one was not willing to take part if they did not receive the intervention, one preferred not to give a reason and 47 stated other reasons. b, 17 failed the 6CIT, one was already receiving telephone services and one was unable to use the telephone. c, One did not want to take part and one wanted to discuss the study with a family member. d, One was withdrawn by the chief investigator because of a protocol violation relating to eligibility, two withdrew consent shortly after allocation [one was unhappy with the involvement of (service provider) and one did not feel the study was for them], five withdrew consent at the point of arranging the 6-month follow-up (one because of ill-health, one was no longer unhappy and so did not want to take part, one was unhappy with the intervention – at this point they had not received any calls, one was unhappy with being left uninformed about the lack of intervention and one withdrew because of personal or family issues) and one was not contactable (at least six attempts were made by telephone and a reminder letter was sent). e, One withdrew consent at the point of arranging the 6-month follow-up (unhappy with allocation to the study arm), one was not contactable (no dial tone; letter and e-mail reminders sent), one was on holiday for 4 weeks and two refused (one did not want to answer the same questions as they did not feel any different and one felt too ill to answer any questions). f, Two no longer wanted the intervention (one was too busy and one thought that the intervention was not for them) and one did not give any reason for withdrawal from the intervention. g, Assigned to TF group 5 but the intervention was not delivered because the volunteer facilitator dropped out.

Table 7 shows the baseline characteristics of the 157 randomised participants. Overall, the two randomised groups were well matched with respect to baseline demographic characteristics. Tables 8 and 9 show the baseline participant-reported outcome scores for the SF-36, EQ-5D, PHQ-9, de Jong Gierveld Loneliness Scale, ONS well-being outcome and the GSE. Again, the two randomised groups were well matched with respect to baseline quality of life scores.

Only 70 out of the 157 randomised participants were followed up for 6 months post randomisation. These participants were part of the initial internal pilot study to assess the feasibility of the full RCT. Table 10 shows the baseline demographic characteristics of these 70 randomised participants. The two groups were reasonably well matched with respect to baseline demographic characteristics, although there is some evidence that slightly more women were randomised to receive the intervention. Tables 11 and 12 show the mean baseline participant-reported outcome scores for these 70 participants. The mean participant-reported outcome scores were broadly similar for both groups, although there is some evidence that the intervention group had a slightly better quality of life at baseline, as measured by the SF-36 mental health, general health and vitality dimensions.

By 6 months post randomisation, 56 participants had valid primary outcome data (SF-36 mental health dimension). There were 30 responders and five non-responders at 6 months in the control group and 26 responders and nine non-responders in the intervention group. Tables 13 and 14 show the baseline demographic and quality of life characteristics of these four groups of participants respectively.

At 6 months post randomisation, the SF-36 mental health mean (SD) scores were 77.5 (18.4) in the intervention group and 70.7 (21.2) in the control group, a mean difference of 6.5 (95% CI –3.0 to 16.0); after adjusting for age, sex and baseline score the mean difference was 9.5 (95% CI 4.5 to 14.5) (Table 15). The estimated ICC from the marginal model for the primary outcome was –0.06, indicating little if any clustering by facilitator; the correlation between baseline and 6-month mental health scores was 0.78.

The developers of the SF-36 have suggested that differences between treatment groups of between 5 and 10 points on the 100-point scale can be regarded as ‘clinically and socially relevant’. 40 For the original sample size calculation we assumed that a mean difference in SF-36 mental health dimension scores of ≥ 8 points at 6 months post randomisation between the intervention group and the control group is the smallest difference that can be regarded as clinically and practically important. The point estimates for the primary outcome and their associated CIs suggest that the likely effect of the telephone befriending intervention is within a clinically relevant range. However, Figure 3 shows how the SF-36 mental health scores change over time in the two randomised groups; over the 6 month follow-up period there is no change in scores in the intervention group but there is a decline or deterioration in scores in the control group.

FIGURE 3.

Mean SF-36 mental health dimension scores over time by randomised group.

Table 15 also shows that for the secondary participant-reported outcomes, such as the other dimensions of the SF-36, the differences in quality of life favoured the intervention group. For five dimensions of the SF-36 (role – physical, bodily pain, social functioning, physical component score and mental component score), after adjustment for baseline score, age and sex, the CI excluded zero, suggesting a non-zero effect (Figure 4).

FIGURE 4.

Forest plot of mean differences in 6-month post-randomisation SF-36 outcomes between groups, adjusted for age, sex and baseline score (n = 56).

There were no differences in mean scores between the intervention group and the control group for the other participant-reported outcomes, except for the ONS well-being total score.

The attrition rate at 6 months was 20%, with only 80% (56/70) of participants having valid primary outcome data; however, the results for the primary outcome were robust to missing data. Table 16 shows the results of a sensitivity analysis of the primary outcome, the mean 6-month post-randomisation SF-36 mental health dimension score, using a variety of methods to impute missing primary outcome data. All imputation methods produced similar results (Figures 5 and 6). Figure 5 shows the sensitivity analysis with adjustment for covariates (baseline score, age and sex) and Figure 6 shows the results without adjustment for any covariates. The estimated treatment effects from the various sensitivity analyses performed produced broadly similar results, suggesting that the results are fairly robust.

FIGURE 5.

Forest plot of sensitivity analysis of mean difference in 6-month post-randomisation SF-36 mental health dimension scores between groups, adjusted for age, sex and baseline score. PMM, predictive mean matching.

FIGURE 6.

Forest plot of sensitivity analysis of mean difference in 6-month post-randomisation SF-36 mental health dimension scores between groups. PMM, predictive mean matching.

A per-protocol data set was defined as all participants in the control group and participants in the intervention group who completed ≥ 75% of the group telephone calls over the 12 weeks of the group intervention (the one-to-one telephone calls with a volunteer were not included in the definition of ‘per protocol’). This means that if a TF group completed 12 group telephone calls, individuals were part of the per-protocol data set if they were present for the duration of nine or more of the calls. Only 35% (9/26) of the participants in the intervention group who had valid 6-month outcome data had completed ≥ 75% of the group telephone calls over the 12 weeks of the group intervention. Table 16 and Figure 7 show the results for the per-protocol sample for the primary outcome only. At 6 months post randomisation, the mean (SD) SF-36 mental health dimension scores were 73.9 (17.5) for the nine participants in the intervention group and 70.7 (21.2) for the control group, a mean difference of 3.2 (95% CI –5.2 to 11.6); after adjusting for age, sex and baseline score, the mean difference was 8.0 (95% CI 3.3 to 12.7). The full results for the per-protocol population are reported in Appendix 9.

FIGURE 7.

Mean SF-36 mental health dimension scores over time for the ITT and per-protocol samples.

Hospital use

Medication

Day service use

Community service use

The sample

Twenty-four participants out of 78 (pilot study n = 35) randomly allocated to the research intervention arm were allocated to a volunteer over the course of the study. We attempted to interview all 24. One participant was withdrawn from the study (because of protocol non-compliance) before being invited to an interview. Of the remaining 23 participants, two did not consent to an interview (one had withdrawn after the first one-to-one call and one had withdrawn before the calls started), one could not be contacted (no dial tone, e-mail sent) and one declined to take part in an interview (no reason given). Those participants allocated to the intervention but who were not allocated a volunteer facilitator and were not followed up were not invited to take part in an interview.

Nineteen (83% of the target) participants in the intervention arm participated in semistructured interviews between March 2013 and April 2013. The baseline characteristics of the interview population compared well with the baseline characteristics of the trial participants who did not take part in interviews. Those interviewed were slightly older than those not interviewed, were more likely to live alone (79% vs. 63%) and scored slightly higher for mental health function at baseline. Conversely, those interviewed scored slightly lower for physical function at baseline (see Appendix 10).

Prior states

Acceptability

Twelve participants described TF in positive terms such as ‘enjoyable’, ‘pleasant’ or ‘quite good’. Nine participants described the process as being easy. A further three participants described their experience as ‘interesting’. Some of these participants found the group to be an acceptable vehicle for sharing life experiences:

It was interesting, it was good to meet other people [er] and find out about them and how they ticked and how they, you know, had a life before, it was good, I enjoyed it.

006, female

Conversely, some participants felt that the one-to-one volunteer befriending was acceptable but that the group was not for them:

You don’t know what to talk about when you don’t know people and you don’t know what they want to talk about.

015, female

Some participants described a level of initial apprehension, which diminished over time as they began to recognise the voice that was speaking and found that they had enough in common with other group members. A minority of participants found friendships based solely on telephone conversation inferior to those outside the group that also involved face-to-face communication:

Well it’s an unusual sort of, I mean if I want to telephone friends I do, you know, of course talking to people who you don’t know and you haven’t seen, that’s different.

038, female

the thing I found difficult first was that you were talking to people you’d never seen, I think visual things are so important, and therefore when you are talking in a group, you’re listening politely to what they say, and you don’t know really when to come in.

065, male

it was strange . . . you listen to one person and then you listen to another person and then another, I just found it difficult.

015, female

Group dynamics

The dynamics of each group were in part influenced by the volunteer facilitator; coded content was often also coded under the heading ‘acceptability’. Three key subthemes were identified within group dynamics: cohesiveness, disclosure and peers.

Value

The sample

We attempted to interview all volunteers who remained in contact with the service provider (n = 8) about their experiences. This included volunteers who dropped out before, midway through and post completion of the group facilitation skills training. Three volunteers consented to an interview: two had completed delivery of the intervention and one had dropped out during the group facilitator training.

The volunteers participated in semistructured interviews between February 2013 and April 2013. Sociodemographic data were not collected for volunteers; however, those who were inducted by the service provider were aged from their 20s to their 70s. The sample included one volunteer of working age (working part-time) and two who were retired.

We used the same macrolevel themes for the volunteer interviews as for the participant interviews and most material could be categorised under the ‘prior states’ and ‘technical’ headings. The results of the volunteer interviews are presented first and the analysis of intervention fidelity is presented in Results of the fidelity assessment.

Prior states and expectations

All facilitators said that they had volunteered because of a desire to help others and because the telephone facilitator role sounded interesting. Two volunteers said that they wanted to get something out of volunteering themselves (VF01, VF03). One volunteer expected the role to involve more face-to-face interaction with participants (VF02). Having dropped out, they also reflected on the type of role that they would have preferred, such as working in a day centre where they would meet people. Another volunteer described being apprehensive about making group calls and being especially worried about conflict arising in the group (VF03). Volunteers expected TF groups to consist of participants who were isolated, lonely and housebound but found that there was more variation (see Chapter 3, Monitoring volunteers).

Technical

Study design

A minimum and maximum number, duration and frequency of telephone contacts were established as part of the intervention design (see Chapter 2, Interventions and Telephone befriending design). Nine participants received the per-protocol minimum number of calls of nine (out of 12). Table 29 summarises the number of group calls made by the volunteer facilitators and shows the time period over which calls took place. Two volunteer facilitators conducted 12 discussions with each taking place 1 week apart for 12 weeks. The two other groups, facilitated by the same volunteer (consecutively), completed 12 calls over a longer period of time.

Assessment of volunteer experience and competence (training)

A set of qualities and skills required of volunteers to enable them to carry out the volunteer befriender/facilitator role were agreed prior to the service provider promoting the opportunity in the locality. Volunteers were screened by the service provider, which included a criminal records check with the UK Disclosure and Barring Service to assess their suitability for the role. All volunteers had some experience of communicating either face to face or by telephone in their current or previous employment. The majority of the volunteers had some experience of volunteering although none had done so by telephone.

Assessment of training content delivered by the trainer

Assessment of group intervention delivery by volunteer facilitators

Participant engagement in group calls (receipt)

Interview transcripts from the qualitative interviews and audio recordings of observed group discussions were reviewed for problem behaviours associated with group cohesion. 31 One volunteer facilitator mentioned an incidence of monopolistic behaviour in her group (VF03). One incidence of silence and lack of participation was self-reported by a participant (R002). There was no evidence of band-aiding, acting superior, hostile behaviour, dependency or socialising outside of group calls before the end of the programme.

The intervention delivery checklist included four fidelity items that assessed group members’ participation in calls: observing ground rules, introducing topics, showing support and commitment (see Appendix 8, Intervention delivery checklist). The role of the facilitator was to make it easy for participants to join together as a group to discuss whatever they liked. The volunteers were asked to use their skills to facilitate discussion, monitor development of the group and allow the group to ‘perform’ as a group. The extent to which participants were able to ‘perform’ in the group is closely related to volunteer facilitator performance, for instance if the volunteer facilitator did not discuss and agree ground rules then the participants could not be expected to comply with them. This was accounted for in the scoring of the checklist.

Table 32 shows participant fidelity scores for each time point by group and overall percentage fidelity based on median observed scores.

The small sample of observed groups prevents the use of formal statistical tests to investigate the association between low facilitator fidelity and low participant fidelity, but some observations on apparent association may be worthwhile. Participants in group 2 were observed as having the lowest overall percentage fidelity; the volunteer facilitator score for this group was also the lowest across the groups sampled. Similarly, the highest volunteer facilitator and participant fidelity scores observed across groups were both in group 4.

Interview transcripts and audio recordings were also reviewed for relevant characteristics associated with the transitional phase, which the group must recognise and deal with to progress to the working stage (see Chapter 2, Methods for the fidelity assessment, Telephone befriending receipt and enactment). 32 One incidence of confrontation was reported (see Results of the volunteer interviews, Technical, Facilitator performance) and one incidence of conflict was observed (group 4, week 1). There was no evidence of defensiveness and resistance.

In summary, the assessment of participant fidelity indicated little evidence of groups moving through the behaviours expected in the transitional phase of group formation, which group dynamics theorists think of as necessary in the early stages for a group to become high functioning. In lay terms, the group members are acting in ways that could be characterised as reserved and tentative rather than open and expressive. These findings almost certainly relate to both the generally poor intervention fidelity of the volunteer facilitators and the fact that the groups were time-limited, discouraging emotional investment.

Strengths and limitations of the randomised controlled trial

This report presents the results of a pilot trial phase of a discontinued study. As such, it would be inadvisable to use the results as a guide to the potential mental health benefits achievable through befriending programmes; any estimates of beneficial effect presented in the results may be unrealistic because of the limited sample size. The report does, however, demonstrate that recruitment to, and retention in, such a trial is possible.

Recruitment to trials evaluating preventative interventions is known to be problematic, with typically 1–5% of those screened being randomised compared with 20–26% in trials evaluating therapeutic interventions according to one overview. 59 Our consent rate was 1.6% with 147 participants recruited from targeted mail-outs and 10 from referral by health and social care professionals, the voluntary sector or advertisements. These figures are consistent with other trials of health promotion interventions using mass mail-outs for participant recruitment in the same geographical region; these studies also show recruitment through health and social care referral to be wholly inadequate for timely participant accrual. 60,61 It follows that the low take-up rate need not indicate poor acceptability of either the study procedures or the intervention, although it may raise questions whether telephone befriending can be understood as a public health intervention. 62 Similarly, attrition in RCTs evaluating preventative measures in community-dwelling older people is frequently between 20% and 30% because of comorbidity, exhaustion and respondent burden. 63 Our study performed well in limiting attrition to 20%.

At least one RCT evaluating professionally led befriending groups for older people has been published. 64 To our knowledge, this is the first RCT that has sought to evaluate a volunteer-led group intervention co-ordinated by the voluntary sector, or one mediated by telephone rather than face to face. However, there are three aspects of the trial’s external validity of which readers should be aware before attempting to generalise from our results. These are the unique use of a hybrid (one-to-one followed by group) intervention (see Other studies evaluating interventions to address loneliness); intervention delivery by a local franchise of a charity with a federal structure (see Other studies involving volunteer provision); and the character of the randomised population, which we consider now.

Trial participants who were interviewed for the qualitative substudy did not generally identify a need to take part in a telephone group, instead expressing a wish to help with the research or to help other older people who were genuinely lonely or isolated (see Chapter 5, Results of the participant interviews, Prior states). There are a variety of ways in which we can interpret this finding. First, it is possible that the study was subject to a severe form of ‘volunteer bias’ or ‘self-selection bias’, defined as systematic error deriving from the sample containing only those people who are willing to participate in the research study and not those who would otherwise have sought help from the service. 65 This hypothesis is given weight by the distribution of scores for the de Jong Gierveld Loneliness Scale and the GSE. The mean (SD) de Jong Gierveld Loneliness Scale score for intervention participants in our study was 3.6 (3.2), on a scale in which a score of ≥ 3 indicates loneliness and a score of ≥ 9 indicates severe loneliness. 66 de Jong Gierveld reports the Dutch population norm for this scale as 4.9, somewhat higher than the mean in our study. GSE mean (SD) scores for intervention participants at baseline [32.9 (5.1)] are also higher than the population mean reported in a multinational psychometric study [29.5 (5.32)]. 67 It is relatively well documented that ‘joiners’ – those with a propensity to establish and sustain group membership – are psychologically healthier than those who are not ‘joiners’. 68 There were indications through intervention delivery and reiterated during interviews that some participants were well supported and socially engaged (see Chapter 5, Results of volunteer interviews, Technical). However, we must allow for the possibility that a number of participants were genuinely help-seeking but unwilling to define themselves as such, anticipating that they would be viewed negatively by the interviewer (social desirability bias69 or unacceptable disease bias65). Similar claims, that participation in research is altruistic and not help-seeking, have been noted in other studies in which the medical condition (e.g. depression) or risk factor (e.g. loneliness) is in some way stigmatised. 70,71 Although this may appear to have serious consequences for the external validity of the research, there is some evidence that this dynamic does characterise routine clinical practice too. For instance, a strength of group psychotherapy is that people gain a sense of self-worth by being part of the group and from feeling of benefit to others rather than being a burden. 57 For these reasons we might understand the expression of altruism and denial of need as not only inevitable in research but also a likely response to service receipt.

Sustained attempts were made between September 2011 and March 2012 to engage members of the public in the development and implementation of the intervention, beyond the support offered by lay representatives on the TMG and TSC. These attempts were unsuccessful and may have adversely affected the quality of the project outputs. That said, the intervention was based on the findings of consultations with those who had delivered and received telephone befriending based in other settings and so the design and implementation of the intervention were not without lay perspectives.

Strengths and limitations of the qualitative research

A qualitative research component was integrated into the study to better understand the success or failure of the delivery of the intervention. 72,73 The design and conduct of the qualitative research met the recommended standards for conduct and reporting, with qualitative researchers, independent from the interventionists, contributing throughout with clearly stated aims. 74 Although we made every effort to interview as many as possible of those who received the intervention, the small numbers available mean that it is unlikely that descriptive or theoretical saturation was achieved. Many of the reported themes identified echo findings from existing related literature, but we cannot overlook the possibility that new descriptive codes, categories or themes might have emerged with the analysis of further data.

Strengths and limitations of the fidelity assessment

A major strength of this study is the detail presented on the development of the intervention and its adaptation for delivery through a particular teleconference provider under the co-ordination of a host charity in one locality. The content of and adherence to the volunteer training programme and the intervention delivered to participants are described in detail. Standardised manuals, guidelines and training were all provided to ensure that each volunteer received the same support to deliver the intervention. This level of description of the complexity of the intervention is in line with the Medical Research Council framework75 and is beyond that available for most interventions intended to ameliorate social isolation or loneliness. 53

Although the service provider initially agreed to give volunteers ongoing support after training, in the end they had a ‘hands-off’ approach to co-ordination, without any monitoring of volunteer facilitators or feeding back to volunteer facilitators on the quality of their intervention delivery. Given the resource and knowledge constraints under which devolved charities operate, this model is likely to be generalisable for third-sector organisations that operate as small independent ‘franchise’ organisations. Charities that operate a command and control model may be better resourced to undertake the ongoing support and continuous quality improvement work necessary to facilitate intervention adherence and prevent intervention ‘drift’, a decline over time in the fidelity with which the intervention is delivered, when initial competence is established. 54

The fidelity substudy, embedded within the trial, used a framework based on that developed by the Behaviour Change Consortium. 54 Although this system worked well for evaluating training and delivery, it proved problematic for assessing the receipt and enactment of the intervention, the mechanics of which were aimed at the management of group dynamics within 1-hour group sessions (see Chapter 2, Treatment fidelity assessment and methods) rather than at changing individual clinical outcomes over the long term. Enactment is one facet of a more general concept of responsiveness, which our qualitative research tried to assess in terms of which participants viewed the intervention as being of relevance to them. The results suggested that some participants were not fully engaged in the group discussions. In part, this response may be an effect of the nesting of the intervention within a research study (see Chapter 3, Boundaries between research and service delivery). The denial of need by many participants (see Chapter 5, Results of the participant interviews, Prior states) as well as the occasional claims that group cohesiveness could be low (see Chapter 5, Results of the participant interviews, Group dynamics) may suggest that the participants were unresponsive to the intervention, regardless of intervention fidelity achieved by the volunteers. These findings provide an important frame through which the positive but uncertain quantitative results should be viewed.

A shortcoming of the fidelity substudy was that limited data were collected on the content delivered by volunteers during the one-to-one calls. The study used Community Network’s system to record a sample of one-to-one and group conversations. One volunteer reported to the study manager having technical difficulties recording one-to-one calls, and Community Network reported difficulties in extracting the recordings from the system as volunteers were able to arrange calls at any time rather than them being set up through an operator (Community Network)-managed call. The group intervention delivery checklists will not have captured some volunteer behaviour during one-to-one calls that may have contravened the protocol and/or impacted on outcomes. For example, we would not be aware if, during these one-to-one discussions, volunteer facilitators had introduced participants to group ground rules, which are supposed to be agreed by all participants at the outset but were not evidenced through the recorded group calls in most cases.

Perspectives on the intervention from the group therapy intervention literature

We have already identified that a strength of the research intervention was that it required a protocolised training programme for its volunteer facilitators, which informed them about how to manage group dynamics and the practice of group facilitation. The format of the training itself enabled trainee facilitators to receive a group experience as a participant. Because of resource constraints, the training programme was relatively limited compared with that provided for trainee psychotherapists, who would typically be able to observe experienced group therapists at work and receive close supervision when facilitating their maiden group. 57 The practice of psychotherapy is not necessarily a good analogue for what is achievable or desirable in interventions to enhance social support or increase opportunities for social interaction (see Other studies evaluating interventions to address loneliness). Nonetheless, our sense is that, without ongoing training and monitoring, 4 hours of training was insufficient, given the fidelity scores (see Perspectives on the intervention from the group therapy intervention literature and Other befriending intervention studies). In the face of public health challenges there is a temptation to think that any service is better than none. 76,77 Although group interventions have the power to transform the health and potential of individuals, it is well documented that groups that are hastily thrown together or led by someone without proper training may be ‘more damaging than beneficial’ (p. 5). 31 Although a small number of participants were dissatisfied with the intervention and others expressed disappointment at how the groups ended, the study did not find any evidence of harm. Nonetheless, the ongoing monitoring of group characteristics and processes throughout an intervention is the sine qua non of a successful group intervention and our research casts doubt that volunteers with 4 hours of training in group facilitation are properly prepared for the task. 31,57,58

Good mental health or well-being outcomes and participant adherence are typically associated with good group cohesion, which in turn is made possible through the promotion of group bonding during early sessions. 31,57 Group cohesion is linked to an individual’s motivations and feelings about the group, the extent to which the members of the group feel similar to one another and whether the group has a goal or objective. 78 The early closure of the trial means that there are too few data to make any definitive remarks about the success of the intervention in promoting group cohesiveness. However, participant interviews indicated that one group, which expressed high levels of mutual interest and reported flowing conversation, maintained all six group members throughout the programme and seemed to be the most cohesive of the four groups. Participants from the smaller groups discussed members leaving the group and missing calls, leaving the group with less to talk about (see Chapter 5, Results of the participant interviews, Group dynamics). Although optimal group size is a function of session and programme duration, there seems to be some agreement in the literature that the group size for adults should not drop below five or six participants with one facilitator. 31,57,79 The opportunities to engage with others are reduced as the group size diminishes, with passive behaviour and negative group image being frequent consequences. For these reasons, some theorists advocate replacing group members or combining small groups in preference to persisting with insufficient membership. 57 These techniques were deliberately eschewed in the design of the PLINY intervention as the introduction of new members can adversely affect the promotion of group cohesion. 31 Face-to-face group therapists can address this issue by setting the group size at nine or ten, the level at which a group can withstand likely attrition, especially if closed to new participants and time limited. 57 However, we are uncertain whether a group of this size can be sustained with a group telephone intervention.

Elsewhere in the literature, group cohesiveness is associated with high self-disclosure. 57 No participant from any group expressed high levels of self-disclosure in our study, with some participants indicating the difficulties in developing real friendships without face-to-face contact (see Results of the participant interviews, Value). This finding is confirmed in other qualitative research that has explicitly compared telephone with face-to-face befriending for older people in the sighted population. 80 However, it is worth noting that a recently completed study found that older people with visual impairment often prefer technology-based communication (social media, e-mail, telephone). The lack of visual cues did not give them additional communication difficulties but, rather, made them feel equal with those with whom they were communicating through electronic media (Patrick Okonji, Northumbria University, 2013, personal communication).

The intervention was designed so that, when possible, participants with ‘similar interests’ would be placed in the same group. This was not possible in practice because of resource constraints but also because of restrictions imposed by the design of the intervention itself. More specifically, one-to-one calls had to be initiated by a volunteer facilitator before interests could be established. It was felt that continuity of volunteer contact between the one-to-one and the group phases was more important than the matching of interests, which would have necessitated passing participants between volunteers to form a group with specific shared interests. The service provider decided to allocate six participants sequentially to each group, delegating first contact with this group of participants to the next available volunteer. Participants would have waited longer for a group to start if they were to be matched to others with similar interests. ‘Matching’ participants on the basis of similar interests is, in general, less critical to successful group composition in personal growth interventions. 31 Some research suggests that older people tend to place a higher value on social skills and a friendly disposition than on shared interests. 80

Group cohesion can be increased or challenged by members socialising outside scheduled sessions. 31 The original intention was that our research intervention should encourage contact among participants in the hope that they would establish acquaintances and friendships that could extend beyond the 12-week group programme, which would include the initiation of independent calls during the 12 weeks of facilitated calls. However, the policy of the teleconference provider, Community Network, discouraged contact outside friendship group calls, to maintain confidentiality and ensure that participants do not share personal information. Volunteer facilitators were therefore trained to discourage contact and were advised to introduce confidentiality as a ‘minimum’ ground rule. The facilitator training also highlighted the potential impact of such contact on the dynamics of the groups, for instance cliques may form or some participants may become isolated within the group. Community Network advises host organisations to decide whether to permit sharing of information among users of the service and to manage this process, recommending that written permission for sharing telephone numbers is obtained ideally at the end of group calls to limit adverse impacts on the group. From what we are able to ascertain from the interviews with the volunteer facilitators and participants and observations recorded in field notes, adherence with Community Network policy was uneven. On the one hand, information seems to have been shared between participants and/or the volunteer facilitator without express permission, against the intervention protocol (see Chapter 5, Results of the fidelity assessment, Assessment of group intervention delivery by volunteer facilitators). On the other hand, some participants who wanted to continue with contact after the group ended were left disappointed (see Chapter 5, Results of the participant interviews, Acceptability).

Other befriending intervention studies

One recent non-systematic review has highlighted the absence of, and need for more, RCTs that, like our own, incorporate standardised quality of life measures. 53 We are aware of one systematic review of RCTs of interventions focused on community befriending, the searches for which were updated in April 2008. 81 Compared with usual care or no treatment, befriending demonstrated a small but statistically significant effect on self-reported symptoms of depression in nine studies with follow-ups of < 12 months [standardised mean difference (SMD) 0.27, 95% CI 0.48 to 0.06, from nine studies) and > 12 months (SMD 0.18, 95% CI 0.05 to 0.32, from five studies). These results should be interpreted with caution for at least two reasons. First, the researchers used a funnel plot, a graph developed to assess the risk of publication bias, to show that studies with negative results may remain unpublished. Second, only half of the studies in the systematic review involved befriending by lay volunteers, as in our study, with the remainder evaluating provision by various professionals. The content, duration, frequency and intensity of the befriending programmes are not explicit in the data abstraction tables but, based on the findings of the other review,53 interventions are unlikely to be well described in the original articles.

One well-described UK RCT, the Befriending and Cost of Caring (BECCA) study, evaluated the effectiveness of volunteer-led befriending for carers of people with dementia, successfully recruiting 236 participants to the trial. 82 As with our study the befriending schemes were established for the purposes of the trial and were nested within charitable organisations with experience of supporting volunteers undertaking befriending. Like the PLINY trial, the BECCA project worked with a national charity to set up local networks of volunteer befrienders. Unlike the PLINY trial, the BECCA team partnered with organisations other than local branches of the partner national charity. The BECCA team also found the resource to employ dedicated volunteer co-ordinators in each locale, jointly managed by the operations manager from the ‘host’ organisations and the research team. In contrast, the part-time volunteer co-ordinator in the PLINY project was an existing member of staff at the host organisation, with existing commitments and responsibilities. BECCA volunteers received 12 hours of training and were supported by the organisations throughout. Volunteers and carers were matched and introduced by the befriender co-ordinator. Training content was similar to that received by volunteers in the PLINY trial; however, PLINY volunteers made their own introductions to participants during the first one-to-one call. The boundaries for volunteers – for instance signposting, rather than giving advice – although slightly different from those of the PLINY trial, were clearly set out in the training. The BECCA trial reported that 60 out of 124 (48%) volunteers who expressed an interest completed training and that 49 out of these 60 (82% of those trained) delivered the intervention; this compared with 10 out of 42 (24%) and three out of 10 (30%), respectively, in our study.

Other studies involving volunteer provision

Our study shows that the ability of a local franchise of a national charity to recruit and retain volunteers (interventionists) will not always be adequate for the rapid roll-out of a public health intervention at scale during a trial. The literature on volunteering is dominated by correlates studies, proposing psychometric or demographic factors that characterise those who volunteer and who adhere to volunteer roles, with comparatively few articles describing the frequency and duration of volunteer involvement in the population, recruitment techniques or organisational factors that improve volunteer recruitment and retention. 83

As we have demonstrated, volunteer recruitment represents a significant management challenge, particularly for smaller organisations, even when a part-time paid volunteer co-ordinator exists, unless that post is dedicated to the intervention. The limited evidence on best practice, summarised by an up-to-date systematic review,84 indicates that direct contact by representatives of the organisation and word of mouth are the most common and probably the most effective methods of recruiting volunteers, with advertising providing a supporting role. 85 From this, as well as from personal communications with representatives of Age UK National, we might conclude that the right methods of recruitment were used but that the task of recruitment that was set for the service provider was too onerous, given the rates of volunteer recruitment required for the success of our study.

The literature divides the variables that affect the retention of volunteers once recruited into categories of personal and organisational factors. 84,86 There is only weak or contradictory evidence for an association between volunteer retention and demographic or psychometric variables. Personal factors that do appear to predict adherence to volunteer programmes include higher levels of education, previous experience as a volunteer and ‘stability or continuity in the life course’, including a volunteer’s changing relationships with the geographical locality, higher education, their family and the labour market. The requirement by the UK government for those receiving state benefits to be available for paid work, along with the increasingly tough sanctions for those perceived as ‘work-shy’, are thought to be making volunteering increasingly difficult for many who were previously in a position to do so. 87 In our study, two volunteers dropped out, one before and one after training, because of finding full-time work or additional part-time work. The service provider also reported having to take on volunteers referred by Jobcentre Plus, an executive agency of the UK Department for Work and Pensions, but that these volunteers were unsuitable for delivery of the study intervention, being extrinsically motivated.

This is not the only area of interaction between the state and the voluntary sector that might make recruitment and retention of volunteers difficult. Two trained volunteers cited discomfort with the programme itself as a reason for them discontinuing their involvement. One expressed a belief that funding for the research was inappropriately taking funds away from health services. The other felt that the research was trying to prove that services that might otherwise be delivered by health and social care professionals could be delivered ‘on the cheap’ by using volunteers. By the time the PLINY intervention was being delivered it had become difficult to talk about voluntary work in the UK without reference to the 2010 government’s Big Society policy, a stated objective of which is ‘a society where people come together to solve problems and improve life for themselves and their communities’. 88 This policy has had its detractors in the academic literature with some seeing it as ineffective89–94 and others going further, accusing it of providing rhetorical cover for an agenda of spending cuts, privatisation and regressive taxation that will increase inequality,95–98 not least because small-scale voluntary sector organisations are thought to have been most seriously compromised by the deficit reduction programme. 99 Our study provides limited evidence that some individuals who might previously have volunteered are deterred by an association with a political agenda that they eschew, which adds to an ongoing debate about whether such policies are eroding confidence in the third sector. 100,101 More generally, voluntary sector organisations committed to delivering a service using volunteers should seek to establish congruence between the goals and ideals of the volunteers and those of the organisation. 86

A number of organisational factors thought to influence volunteer retention are more in the control of voluntary sector organisations. Negative experiences related to organisational factors are commonly cited reasons for volunteer attrition. 86,102–104 Field notes from our study recorded instances when volunteers reported frustration at being unable to reach the volunteer co-ordinator at the service provider, which was reflected in an over-reliance on the study team to answer queries or solve problems associated with intervention delivery (see Chapter 3, Boundaries between research and service delivery). Commitment is known to be positively related to the reception of sufficient support to ensure that volunteers are comfortable with their role and its procedures. 85,86,104–106 A related factor that is known to be a key motivator in retaining volunteers is the availability of ongoing training. Volunteers often cite the availability of ongoing training whilst they are delivering an intervention as a motivation for adherence to a programme. 85,86,105,107 We have already commented on the lack of ongoing training to increase the confidence of volunteers and to maintain intervention adherence; this is a consequence of constrained costs and is a clear weakness of the intervention evaluated in our study (see Perspectives on the intervention from the group therapy intervention literature). When resources allow and volunteers are enthusiastic, volunteer befriending programmes should conduct training booster sessions, conduct in vivo observations or record and review sessions, conduct weekly supervision and hold periodic meetings with, or allow easy access for questions to, trainers for the sake of both intervention fidelity and volunteer retention. 54,57

Participant recruitment rates to primary prevention trials are known to be low compared with those in therapy trials59 and it seems likely that the slow start to participant recruitment in the PLINY trial had an impact on volunteer retention (see Chapter 3, Recruitment and retention of volunteers). Rapid allocation of volunteers to participants on a large scale is possible in established, centrally organised national services such as Talk and Support, the telephone befriending scheme run by the Royal National Institute for the Blind (RNIB), which aims to match a client with a TF group within 4 weeks. 108 Our study suggests that the capacity and degree of flexibility within a local franchise of a national charity to respond to variations in client demand for a newly established service is apparently limited. Talk and Support facilitators usually start facilitating a group within 2 weeks of completing training (Mark Berkeley, RNIB, 24 June 2013, personal communication). Our study suggests that volunteers may be unwilling to wait a period of weeks following training for their role to start (in this case, for there to be sufficient clients to form a friendship group). This indicates a key threat to the successful implementation of volunteer-led interventions co-ordinated by the voluntary sector and to their evaluation in RCTs, which will be met only by more committed and supportive management by volunteer co-ordinators and by the availability of ongoing training.

Other studies evaluating interventions to address loneliness

One of the key aims of the study was to examine the benefits of an intervention that may allay loneliness. Since the start of our study, two systematic reviews of RCTS evaluating interventions to reduce social isolation have been conducted. The first, by Dickens et al. ,109 aimed to evaluate interventions designed to reduce social isolation and loneliness in older people. The review included studies of group and one-to-one interventions, with 79% and 55%, respectively, reporting at least one improved participant outcome. The review found two studies that involved telephone interventions. A quasi-experimental study that examined 1-hour group telephone support for blind community-dwelling older people in the USA reported reduced loneliness and an increased number of social activities at 8 weeks. 110 The intervention is described as ‘short-term problem-solving and discussion of coping methods’. The second study reported that one-to-one telephone support for low-income housing residents in the USA did not result in a reduction in the level of self-reported loneliness compared with no intervention. 26 Only one study included in this review combined one-to-one approaches with group approaches to combat social isolation; this was in caregivers to people with Alzheimer’s disease in the USA. 111 The reviewers concluded that the evidence base needs to be improved by carrying out well-conducted studies. They did indicate that ‘common characteristics of effective interventions may include having a theoretical basis, and offering social activity and/or support within a group format’. The reviewers also noted that interventions that include older people as ‘active participants also appeared more likely to be effective’ (p. 20). 109 These primary research studies, all of which were classified by the reviewers as at high risk of bias,109 confirm that the combination approach to intervention (one-to-one plus group befriending) was relatively novel and that our study is one of the better designed and conducted of its type.

The second systematic review, by Masi et al. ,112 evaluated four primary intervention strategies. Four trials evaluating social cognitive training interventions showed greater effect sizes than 12 trials of interventions to enhance social support, two trials of interventions to improve social skills and two trials of interventions to increase opportunities for social interaction. Potential modifiers of effect, such as group-based format or the use of technology, were not found to be statistically significant. The reviewers concluded that ‘simply bringing lonely people together may not result in new friendships because the thoughts and behaviours of lonely individuals make them less attractive to one another as relationship partners’ (p. 257). 113 On the other hand, ‘correcting maladaptive social cognition (for instance through CBT [cognitive–behavioural therapy]-style interventions) offers the best chance for reducing loneliness’ (p. 259). 113 This review, then, provides a question mark over the value of befriending interventions that do not involve an active psychological component intended to induce behaviour change. This said, it is far from clear that cognitive–behavioural approaches can address precursors of loneliness such as bereavement,113 and our research team shares the concerns of those who counsel against the medicalisation of commonplace experiences such as loneliness. 114,115 Even if one accepts the research findings of Masi et al. ,112 financial constraints and policy decisions by state health providers mean that access to trained cognitive–behavioural therapists is limited and available only within stepped-care models for those with more serious mental health problems. 116 Given the ability of volunteers to faithfully deliver quite basic group facilitation, when trained within typical voluntary sector financial constraints, it seems unlikely that widespread and systematic access to more psychologically sophisticated behaviour change interventions is feasible for the alleviation of loneliness, even if they do have an impact.

Previous studies evaluating interventions to prevent loneliness and social isolation have been criticised for failing to publish data on organisational set-up and implementation costs. 53,109 Although we were unable to conduct a full cost-effectiveness analysis as part of our study, a strength of our report is that it does include cost impact data.

Service development and implementation

In common with the findings from previous research,53,117 we found that the successful implementation of a service that aims to reduce loneliness requires strong partnership arrangements with commissioners and other providers. The scale of operation of individual third-sector organisations, such as the local provider involved in this study, means that, for any intervention to be delivered to more than a small number of individuals, partnership and cross-agency working is essential.

To demonstrate future sustainability, funding to deliver the intervention had to be identified from organisations that might feasibly provide such a service. Identifying funding for an intervention to meet needs that currently fall outside the remit of existing statutory service provision proved challenging, with third-sector organisations being the only avenue available. This provides some indication of the extent of change that is necessary within both statutory and non-statutory services to meet the current and future needs of an ageing population. The difficulty of identifying people who are lonely and isolated was a factor in recruitment to this study but is also an important consideration for services that aim to meet their needs.

Our findings supports those of a previous Delphi survey which indicated that the sustainability and scalability of telephone befriending services is predicated on sufficient time and resources being dedicated to administer the service. 24 The results of this research demonstrate the deleterious impact of insufficient resources being allocated to volunteer recruitment and ongoing support. After the study closed it emerged that the number of volunteers required for this study equated to the existing total volunteer workforce of the service provider. Therefore, capacity to support the volunteers would have had to double to take into account the needs of the study. The problems resulting from over-reliance on the good will of one or two individuals to scale up services was evident.

The intervention delivered through this study necessitated volunteer participation in mandatory training to prepare them to deliver specific skills. This level of demand can negatively impact on volunteer turnover, with retention then demanding a higher level of resources from the host organisation. 119 Additionally, delivering services that are evidence based is arguably more likely to involve structured and possibly specialised training. This offers greater opportunities for a volunteer workforce who may be looking to improve skills and gain valuable experience for paid work but also places greater resource demands on the host, particularly when a high turnover of volunteers exists. Therefore, the sustainability of such a workforce and what it might deliver can be fragile, emphasising the need to determine the true resource implications of using volunteers to deliver services.

Research

The intervention designed, implemented and evaluated in this study was intended for lonely and isolated older people, particularly the housebound. The process of recruitment confirmed the difficulty of reaching out to these individuals. Substantial efforts were made to involve relevant health and social care professionals in the process of recruitment. However, this was unsuccessful and few people were identified through this route for potential participation. The reasons for lack of recruitment through front-line professionals can be postulated, even though we were not able to pursue this, for example negative connotations of randomisation, lack of time to explain the study, low on list of priorities. In ideal circumstances the study timeline would provide scope to be able to ameliorate such barriers, and resources would be available to be able to recruit through communities, which can be a successful strategy. Nevertheless, in this instance it proved necessary to use mass mail-outs to potential recruits through GP surgeries to meet the target numbers in the available time. Questions remain regarding how to effectively reach those in need of such an intervention. This is important given the indications of benefit that this research identified.

The need to take time to embed such an intervention into day-to-day practice before subjecting it to rigorous evaluation has been illustrated. Other studies that comprise the development and evaluation of a complex intervention can involve further pilot work or a multistage programme with clear interim progression criteria. This is beneficial when there are a large number of factors involved in both service delivery and the associated research. Quite early in this study it became clear that the demands of the research eroded the already limited capacity of the service provider to organise intervention delivery. Also, study requirements involved protocolisation and randomisation, which is counter to the philosophy of reduced bureaucracy and creating community capital that can exist within charitable organisations. 117 Conversely, it was evident that the idiosyncrasies of the delivery site challenged study progression.

This research demonstrated that a definitive RCT comparing telephone befriending with no telephone befriending is feasible. However, overall success was dependent on two recruitment targets, the first being for older people as participants and the second being for volunteers who were suitable to be trained and supported in a facilitation role. Future research involving volunteers needs to take account of this dual recruitment target. A number of other issues with the implementation of such a trial are worthy of mention to inform the design of future studies.

Was it realistic to expect a franchised third-sector provider to introduce a new intervention and support individuals to deliver it at scale in the manner attempted through this study? Our results would indicate not in the short to medium term. Two other UK voluntary sector organisations are successful in achieving larger-scale delivery of telephone befriending but they have been able to build this capacity over many years and both have paid staff dedicated to running the service and supporting the volunteer workforce. The findings of this study suggest that reduced ambition is necessary to match the capabilities of the programmes under evaluation and there might be benefit in including research to determine the service models that can lead to successful delivery of community-based interventions within constrained finances. However, if funders and researchers deem a full-scale pragmatic trial an appropriate step, then it should be multicentre and, to minimise the risk of intervention failure, use established services already operating at scale, when intervention funding allows. Additionally, there are issues regarding the extent to which intervention components should be standardised and the degree of local adaptation that is acceptable to meet the real-world challenges of delivery.

Social isolation and loneliness

The intervention draws on de Jong Gierveld’s loneliness model,28 which is based on a cognitive theoretical approach to loneliness. Characteristic of this approach to loneliness is the emphasis on the discrepancy between what one wants in terms of interpersonal affection and intimacy and what one has; the greater the discrepancy, the greater the loneliness. Importantly, it differentiates between social loneliness and emotional loneliness. Social loneliness refers to the absence of a broader engaging social network of friends and acquaintances, whereas emotional loneliness relates to the absence of an intimate companion.

See http://home.fsw.vu.nl/TG.van.Tilburg/manual_loneliness_scale_1999.htm (accessed 20 June 2013).

Some recent literature distinguishes two aspects of isolation: ‘social disconnectedness’ – a lack of contacts with others and indicated by situational factors, for example small network sizes – and ‘perceived isolation’ – the subjective experience of a shortfall in one’s social resources, for example companionship and support.

The intervention will impact on social and emotional loneliness by reducing the discrepancy between the older person’s relationship expectations (or standards) and his or her cognitive evaluation of the (mis)match between the quality and quantity of existing relationships. 55 The intervention will reduce perceived isolation by improving perceptions of companionship and support.

Self-efficacy refers to the ability to gain and maintain a belief in personal competence or control in achieving various aspects of well-being. The higher a person’s self-efficacy is with respect to obtaining external resources, the more likely it is that the person will undertake those activities and apply the effort needed to do so. 14 A low sense of self-efficacy has been shown to indicate low self-esteem and the potential for depression and feelings of anxiety and helplessness, whereas high levels of self-efficacy will encourage more investment in terms of individual effort and persistence and, ultimately, generate improvements in quality of life.

A declining sense of self-efficacy may often stem more from disuse and negative cultural expectations than from biological ageing and can set in motion self-perpetuating processes that result in lower cognitive and behavioural functioning. Monotonous environments that require little independent thought or judgement diminish the quality of functioning, whereas intellectually challenging ones enhance it (Bandura, 1993). 131 Research suggests that there is a strong association between high self-efficacy and less feelings of loneliness. 14

Bandura’s theory of self efficacy29 describes individuals’ self-belief in coping with adversity in various domains of functioning, such as persistence in the face of barriers and recovery from setbacks, and involves four main approaches: mastery, vicarious (observed) experience, verbal persuasion and perception of affective and physiological states.

Social learning theory states that people’s perceptions of their capabilities (i.e. self-efficacy) affect their behaviour, thinking and emotional reactions in stressful situations. It has been suggested that social learning theory can inform interpretations of behaviour and cognitive change in support/self-help groups. Other people, viewed as role models, can elicit and enhance experiential knowledge based on first-hand experience (p. 51). 132 Social learning theory suggests that there are four requirements for people to learn and adapt behaviour: attention, retention (remembering what one observed), reproduction (ability to reproduce the behaviour) and motivation (good reason) to want to adopt the behaviour. Although the aim of the intervention is to reduce feelings of loneliness, which may be linked to self-efficacy, it is unlikely that we will be able to measure sustainable changes in self-efficacy over the study period. We should also be clear that the purpose of the group discussions on the telephone is not about setting and achieving goals or major behaviour change but about reducing loneliness and social isolation and improving participants’ sense of confidence and mental well-being (in some cases reducing depression).

Before the conversations start

The outline of the six weekly conversations (the intervention) will be described to the potential participants and they will be offered the opportunity to ask questions, raise concerns and make comments. Older people who may benefit from this intervention are recruited through GP practices in ▪. As far as possible, volunteers (befrienders) will be matched to the participants on the basis of common interests and any particular wishes made by the participants. However, befrienders are trained to facilitate conversations such that the older person’s interests are always prioritised. This will avoid any delay in establishing the one-to-one intervention.

The first stage: detail

The first stage, six one-to-one telephone conversations, will take the format of brief (10–20 minutes) friendly conversations about regular everyday events. Three stages are suggested, which will support the older person and also prepare them for the group conversations. The three stages are (1) familiarisation – getting to know each other’s interests and the older person’s expectations; (2) everyday conversation – recent events, sharing experiences, focusing on the positives, but also allowing space for ‘chat’ about the older person’s health, feelings, worries, etc.; and (3) everyday conversation continues with added focus on developing specific conversation topics, such as resources and accessing services, current affairs, television programmes, grandparenting, special interests (gardening, music, etc.). The older person will be encouraged to raise these topics in the ensuing group sessions.

The one-to-one sessions are not intended as professional counselling sessions. Should this need become apparent (expressed by the older person), the volunteer will, without breaking confidentiality, raise it with the volunteer co-ordinator and an individualised solution will be proposed (referring him or her back to the recruiting charity, providing him or her with a named contact for support).

It is important that the older person who is the recipient of the telephone call is in control of the topics discussed. However, the befriending volunteer will guide the conversations through the three stages, which may not always happen neatly in this sequence but are likely to be covered over the six sessions.

At the end of each conversation, the volunteer and the older person agree the time and date when the next conversation will take place.

Should an individual not wish to move on to participate in group conversations, the host charity will be asked if it has the capacity to continue providing a one-to-one befriending service for the individual. This option will not be offered at the start of the programme as an alternative to the group conversation. It will be considered only in those situations in which an individual expressly states that he or she does not wish to continue in a group.

Stage 2

The second stage consists of one group conversation per week over 12 weeks. Older people who have taken part in six one-to-one conversations are ideally allocated to groups in which, on the basis of the previous conversations, they are most likely to have common interests. The size of the groups will range from six to eight individuals and the sessions will last between 30 and 60 minutes.

The trained facilitator will be provided with a handbook on group facilitation. The role of the facilitator will be to make it easy for the group to achieve its purpose. He or she will do this by:

• creating and maintaining a safe and stimulating space for discussion

• enabling the group to listen and respond to other members of the group

• ensuring the interests of one or two individuals do not dominate the group

• ensuring that all participants have an opportunity to contribute

• ensuring that the discussions are not ‘destructive’ or damaging to individuals in the group and

• constantly monitoring and fostering the purpose of the group as well as the relationships and individuals within it.

Before the meetings start, the facilitator will communicate with the participants to determine a mutually agreeable time for the meetings. Conversations should ideally take place on the same day of the week and at the same time. Participants will receive written information about teleconferencing and:

• what to do if they are unable to participate in one conversation

• what happens if they do not answer when the facilitator phones

• the acceptability of initiating between-call contact with other participants

• the confidentiality of what is discussed in the group

• boundaries between what is/is not discussed

• what happens if a group member expresses the need for help with regard to his or her health

• how the facilitator will manage disagreements during the conversations

• what happens if there are technical problems, such as how to reconnect if they are accidentally disconnected or have to leave for part of the meeting

• key contacts

• the right to withdraw from the study.

These points will be discussed with the group during the first meeting. Depending on the composition of the group, the conversation session structure may develop in one of two ways:

1. as a subject interest group (e.g. current affairs, local history, music), who will discuss the same subject matter each time or

2. a friendship support group, in which members agree the topic to be discussed at each session.

Cattan et al. ’s23 systematic review suggests that both types of group (subject interest and friendship support) may be effective. The review also highlighted that improving self-esteem and internal locus of control (elements of self-efficacy) could enhance the effectiveness of an intervention to reduce feelings of loneliness.