This internal pilot trial demonstrated that the trial team could recruit 0.74 participants per day over a 3-month period and collect valid primary outcome data for 80% of randomised participants, rates which fulfilled pre-specified success criteria. It also established that local charitable providers do not necessarily have the resources to match service demand that exceeds what they would usually respond to. The trial was closed early when targets for recruitment and training of volunteer interventionists, built into the subcontract between the University and the service provider, were unmet, leaving 55/78 (70%) of participants randomised to the intervention unable to receive it more than 3 months after randomisation.
The adjusted, between-arm mean difference (9.5; 95% CI, 4.5 to 14.5 points) in the primary outcome (SF-36 MH score at 6 months) is within the 5 to 10 point range defined by the instrument developers as “clinically and socially relevant” , but must be treated with caution. The lack of statistical power, the small number of intervention arm participants who received the intervention per protocol, the method of recruitment which was reliant upon individuals opting in and the possibility of resentful demoralisation  of control arm participants not blinded to their allocation, all make this estimate of effect problematic for decision-making purposes.
The participant consent rate was 1.6% of those approached about the study, which is in line with rates achieved in other studies evaluating preventive interventions in middle-aged and older populations within the region [14, 42]. The same studies also confirm that opt-in recruitment for the purposes of research studies using targeted mail-outs is more reliable than direct referral by health and social care professionals or third sector agencies, if potentially less externally valid. Levels of participant self-efficacy that are high and levels of loneliness that are low, when compared with population norms [43, 44], and low uptake rate presents the possibility of participant bias, where the target population for preventive services does not get involved in the RCTs which evaluate them [45, 46]. The 20% attrition rate is at the margin of acceptability ; comorbidity, exhaustion, and respondent burden frequently result in rates of 20% and 30% in RCTs evaluating health promotion interventions in community-dwelling older people .
A strength of our study is that volunteers received standardised training and delivered an intervention that is manualised and therefore more reproducible than most interventions intended to ameliorate social isolation or loneliness . In line with the Medical Research Council Framework , a process evaluation, nested within the trial to assess intervention fidelity and quality of implementation, will be published in the NIHR monograph series. The problems experienced by the service provider with the recruitment and retention of volunteers are well-documented in research papers on volunteer management. The retention of volunteers can be affected by personal and organisational factors. The evidence that demographic and psychometric variables are associated with volunteer adherence to work programmes is often weak or contradictory; the factors most strongly associated with retention are educational achievement, prior voluntary experience and life-course stability [36, 37]. Policies which require recipients of state benefits to be available for paid work, or which sanction participation by those perceived as ‘workshy’ are thought by some researchers to be creating barriers to sustained volunteering . Two volunteers dropped out of our programme due to pressure to take paid work. Organisational factors are often cited as drivers of volunteer attrition [36, 51, 52]. Sufficient support to ensure that volunteers are comfortable with their role and its procedures [36, 52–55], with a professional volunteer co-ordinator dedicated wholly to the programme, is essential . Volunteers also frequently cite ongoing training as a motivation for programme adherence [36, 53, 54, 56]. More generally, congruence between the goals and ideals of the volunteers and those of the voluntary sector organisations for which they work are thought to promote adherence . The only trial of which we are aware which documented recruitment and attrition of volunteers to deliver a befriending intervention recorded 60/124 (48%) of those expressing interest completing training and 49/60 (82% of those trained) delivering the intervention, compared to 11/42 (26%) and 3/11 (27%), respectively, in our study . Reasons for volunteer attrition were not reported. Unlike in our study, volunteers were recruited and hosted by more than one type of organisation, and resource was available to employ dedicated volunteer co-ordinators, jointly managed by the ‘host’ organisations and the research team. Those wishing to recruit rapidly and retain large numbers of volunteers to telephone befriending programmes should consider using either charitable providers in multiple population centres or outsourcing the work of volunteer recruitment and management to commercial or (where available) state providers.
Our trial adds to tentative evidence that community befriending interventions may be effective in the preservation of good mental health. One systematic review found that, compared with usual care or no treatment, befriending demonstrated small but significant effects on self-reported symptoms of depression in nine studies, five of which followed up for 12 months or more . However, the results should be interpreted with caution as the review authors acknowledged the possibility of publication bias in their work, and only half of the included studies evaluated befriending by lay volunteers, as in our trial. A second systematic review evaluated interventions to reduce loneliness and social isolation in older people , including two randomised evaluations of telephone interventions [16, 60], and one of a combination one-to-one/group programme like that evaluated in this paper . Whilst the quality of most of the included studies was poor, the review concluded that effective interventions had a theoretical basis and offered “social activity and/or support within a group format” . A third systematic review synthesised the results of RCTs evaluating four strategies to reduce loneliness and social isolation . The conclusion drawn was that social cognitive training interventions yielded greater effect sizes than trials of interventions to enhance social support, improve social skills, or increase opportunities for social interaction, and is unlikely to have applicability to volunteer-led interventions. Its formal analysis, showing that a group-based format was not found to be an effective modifier, is likely to be an artefact of the number of professionally-delivered one-to-one interventions included in the review.
The need for well-conducted studies evaluating theoretically informed, manualised interventions to alleviate loneliness and reduce social isolation in older people remains . Services commissioned especially for research studies are likely to encounter the same issues with matching service supply to demand as demonstrated through our study . The natural tendency of many local voluntary sector organisations may be to deliberately regulate demand for their services based on the resources available to them, resulting in ‘trickle’ recruitment of both volunteers and clients based on self-referral . Ideally services with well-established and effective processes for volunteer recruitment and management systems should be identified to ensure the feasibility of large-scale evaluation. There may also be an argument for non-randomised evaluations if there are significant concerns that trial participants are systematically different from those who would take up the offer of an intervention outside the context of a randomised trial.