1.1 Review aims

This review aimed to outline the key characteristics of befriending interventions for adults with intellectual disabilities, to explore and synthesise the psychological and social outcomes of such interventions, and to identify future research directions required to advance the evidence base.

2.1 Search strategy

The review was conducted following preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines (Page et al., 2021). Systematic searches of the PsycINFO, MEDLINE and Web of Science databases were conducted in October 2021. Search terms were developed through scoping searches, review of published search strategies, and consultation with a subject specific librarian. Multiple diverse search terms for ‘befriending’ were utilised to retrieve all relevant articles due to an expectation of there being nuanced differences between befriending and other similar interventions. Search terms relating to (1) intellectual disabilities; (2) befriending; and (3) adults, were combined using an ‘AND’ Boolean operator. Searches and article selection were limited to peer-reviewed journals. See Table A1 in Appendix A for a full list of the search terms used in each database.

2.2 Eligibility

For the purpose of this review, befriending was defined as a one-to-one ‘friend-like’, emotionally supportive relationship, with a commitment over time, organised and supported by an external organisation, and where one party was deemed likely to benefit. We distinguished this from mentoring (which typically had more focus on pre-determined goals, training or teaching and was often related to particular transitions e.g., school, university, workplace), peer-support (where someone with intellectual disabilities supports another person with intellectual disabilities) or friendship (a more private, spontaneous relationship where the relevant parties would otherwise have met).

The inclusion criteria were (a) studies reporting findings on adults described as having intellectual disabilities and/or autism; (b) studies focused upon befriending interventions as defined above; (c) primary studies with any type of design, including quantitative, qualitative or mixed methods. Exclusion criteria were (a) studies not relating to befriending as defined, for example group interventions, peer-support, mentoring or one-off support; (b) studies focusing on friendship with a paid worker or family member rather than a volunteer; (c) secondary studies such as systematic reviews or meta-analyses; (d) discussion papers or meeting abstracts.

2.3 Study selection

References retrieved from the database searches were exported to EndNote X9 and systematically de-duplicated (Bramer et al., 2016). The titles and abstracts of remaining articles were screened for eligibility by the first author and the reference lists of these screened studies were manually searched for any additional references. Full texts for the remaining references were retrieved and reviewed for inclusion by the first and last authors. Discrepancies in decisions on whether to include certain studies were resolved through discussion

2.4 Data extraction and analysis

A data extraction form was developed based on the research questions, and data from each of the included studies was extracted. Narrative synthesis (Popay et al., 2006) was used to analyse the identified papers. This included tabulating outcomes, comparing differences and quality appraising each of the studies. The major findings were grouped into categories which were refined over the course of the synthesis through ongoing comparison and discussion.

2.5 Quality appraisal

The QualSyst tool (Kmet et al., 2004) was used to critically appraise the included papers. QualSyst provides a systematic, reproducible and quantitative means of assessing research quality across a broad range of study designs. It sets out separate criteria for assessing qualitative and quantitative methods (see Table B1 in Appendix B for an overview of checklist items). Studies were rated against the relevant checklist(s), scoring either yes (2), partial (1), no (0) or n/a for each item, dependent upon the extent to which they fulfilled the respective criterion. To increase the reliability of ratings, the first and last authors initially scored five studies independently before discussing any discrepancies in the scoring decisions. The first author then completed the rating of the remaining studies. For each study, a final summary score was calculated by summing the total score across relevant items and dividing by the total possible score. These summary scores were used as an overall indicator of the relative quality of the studies.

3.1 Critical appraisal of included studies

There was a broad range of quality appraisal ratings across the papers (see Table 2). The summary scores of the nine studies employing qualitative approaches ranged from 0.2 to 0.95, averaging 0.75. Scores were highest against the ‘question/objective description’ criterion (item 1). Performance against the ‘use of verification procedures’ criterion (item 8) was more mixed, with five studies fully satisfying the criterion requirements, and four studies not achieving them at all. ‘Reflexivity of account’ (item 10) scored lowest overall, with no studies explicitly addressing the impact of the authors' own personal characteristics. Of the five studies using quantitative measures, summary scores ranged from 0.6 to 1.0, averaging 0.82. These studies had high scores against the ‘method of subject selection or information sources description’ criterion (item 3), with clearly outlined sampling strategies. Scores were lowest for ‘controlling for confounding’ (item 12), with the two relevant studies (Ali et al., 2021; Hughes & Walden, 1999) deemed to have incompletely controlled for confounding factors.

Whilst Kmet et al. (2004) did not specify cut-off thresholds, the authors debated whether to include the two studies with low appraisal ratings (Hughes & Walden, 1999; Mavropoulou, 2007). As this paper aimed to comprehensively review the existing evidence and its limitations, it was decided to include the studies. Study limitations and quality rating scores have been provided alongside the data synthesis to provide an indication of the validity/reliability of the findings.

3.2 Key characteristics of befriending interventions

The studies provided limited detail as to the characteristics of the external organisations supporting the befriending relationships. Whilst each organisation provided a befriending service to people with intellectual disabilities, whether these were disability specific organisations or mainstream community organisations was not always explicitly stated by study authors, with descriptors such as ‘community befriending services’ (Ali et al., 2021) or ‘charitable organisation[s] administering a befriending scheme for adults with learning disabilities’ (Southby, 2019) insufficient to understand if the organisation was specifically focused upon disability or intellectual disability. In some studies, the befriending was provided by pre-existing organisations catering to people with intellectual disabilities (Hardman & Clark, 2006; Jameson, 1998; Tse et al., 2021), and in others the provision was set up specifically at the time of the study (Fyffe & Raskin, 2015; Green et al., 1995).

3.3 Reported psychological and social outcomes of befriending

Key findings relating to the psychological and social outcomes of befriending interventions were collated from the 11 studies, compared and grouped into different outcome categories (see Table 5).

4.1 Key findings

This review aimed to synthesise the literature on the characteristics and outcomes of befriending interventions for adults with intellectual disabilities. Eleven studies were included, with various study designs and focuses. Befriending schemes employed similar recruitment and matching methods and aimed to foster friendships and increase community participation though approaches to fostering social inclusion differed between studies. Few interventions had strict parameters around frequency or duration of contact, though some did specify that activities should be community-based. Activities undertaken were broad ranging, including home- and community-based activities, with casual leisure activities such as visiting cafes/restaurants, walking or going to the movies most popular.

For befriendees, the most frequently reported outcomes were increased participation in the community and making new friends, suggesting the primary aims of befriending services are often being achieved. However, the degree of true social inclusion fostered by the activities undertaken was questioned, and some substitution of befrienders for existing social network members was reported. Similarly, whilst befriending can positively affect a befriendee's mood, it also presents a risk of negatively affecting emotional wellbeing, particularly where the ending of a befriending relationship is not adequately managed.

Befriending appears to lead to increased confidence and independence for befriendees, which may contribute to the mixed findings on new experiences also reported, with befriendees feeling more confident contributing to decision making and voicing their preferences for doing familiar, repeated activities rather than always striving for novel experiences. Information on decision making in befriending relationships is not commonly reported so this link remains unclear. This review's findings on befriendee outcomes support existing research with mental health populations which suggest befriending promotes community participation (Mitchell & Pistrang, 2011) and increases befriendees' number of social contacts (Priebe et al., 2020).

For befrienders, whilst certain challenges such as negotiating the friendship/professional relationship balance are reported, individual outcomes seem more universally positive overall. Befrienders gained knowledge and new experiences, felt a sense of giving back and expanded their social communities through taking part in befriending. This echoes the experiences reported by studies looking at volunteer experiences of mental health befriending (Cassidy et al., 2019; Toner et al., 2018) and may highlight the power dynamic between befriendee (‘deemed likely to benefit’ from befriending) and befriender who acts as a volunteer, therefore having more control over the experience and the relationship. Befriending has a broader impact upon carers, providing respite for family carers and reducing pressure on other services to organise activities.

4.2 Limitations of the evidence base

Despite the popularity of befriending schemes, this systematic review (with its intentionally broad search terms) only identified 11 studies. Similar to reviews of befriending in other populations (Siette et al., 2017) and of other interventions promoting social participation for people with intellectual disabilities (Howarth et al., 2016), this review suggests that befriending does have positive outcomes but that a stronger evidence base is required to inform policy and practice. On an individual basis, the majority of the 11 studies reviewed appeared of relatively high quality, with the qualitative and quantitative studies scoring QualSyst averages of 0.75 and 0.82, respectively. However, the studies were methodologically diverse and mostly qualitative and exploratory in nature, with those utilising quantitative approaches limited in their data analysis due to small sample sizes. Another key limitation of the evidence base is the limited input of people with intellectual disabilities into study design and conduct. Ali et al.'s RCT protocol (Ali et al., 2020) described consultation with befriending scheme participants with and without intellectual disabilities during planning stages, and plans to engage a ‘public and patient involvement’ group to advise on materials, attend study management meetings and contribute to data collection and dissemination. Heslop (2005) also acknowledged the contribution of an advisory group to her study, though the extent of their involvement was not explicated. None of the other studies identified in this review mentioned any input from people with intellectual disabilities. Lack of direct input from people with intellectual disabilities into the research questions, study designs or data collection and analysis stages, leads to a lack of representation and an imbalance of power in the conduct of research.

4.3 Limitations of the current review

Whilst the last author independently screened full text articles and rated a third of the studies using the quality appraisal tool, the rest of the review was conducted by the first author, increasing the risk of bias in the synthesis and interpretation of results. Though no English language filter was applied on database searches, only search terms in English were used, and grey literature was excluded which may have reduced the comprehensiveness of the review. Whilst the search terms used were intentionally broad this may have increased heterogeneity and the complexity of the narrative synthesis.

4.4 Implications for future research and practice

This review highlights the limited body of evidence pertaining to the effectiveness and impact of befriending interventions for people with intellectual disabilities. Following on from Ali et al.'s (2021) pilot RCT, future studies could employ broader eligibility criteria or focus upon a broader range of outcomes. In practice, UK-based befriending services have increased outcome measure collection and evaluation over the past 15 years (Tse et al., 2021). Extending this by use of standardised, validated measures across befriending services could create opportunities for robust longitudinal studies. Tracking the long-term effects of befriending, (particularly for those befriendees who have experienced the endings of befriending relationships) is vital.

This review further indicates a need for clarity around social network substitution and whether expanding social networks may come at the expense of important existing relationships. Work into understanding the potential value of diversifying existing network composition, rather than simply increasing network size would be valuable. The extent to which community participation through befriending equates to or supports social inclusion should also be further explored, with reference to the importance of setting (e.g., mainstream community versus disability specific versus private), degree of involvement in any given activity, and the degree to which interpersonal relationships are necessary to facilitate this. There is a need for further exploratory analysis of the mechanisms of change and optimal methods of delivery of befriending in this population, and for more research that hears the voices of people with intellectual disabilities directly, both as participants and as contributors to research design and conduct. In including these voices directly, research could consider what adults with intellectual disabilities themselves prioritise in terms of the desired outcomes of any befriending intervention and how this links with understandings of social inclusion.