The nature and quality of friendship for older adults with an intellectual disability in Ireland

1.1 Social inclusion for people with intellectual disabilities

Historically, there has been a lack of consensus and clarity in the literature about the meaning and use of terms such as inclusion, participation, integration and belonging (Amado et al., 2013; Overmars-Marx et al., 2014). In recent years, social and community participation has been framed using the ‘activities and participation’ domain of the World Health Organization's (WHO) International Classification of Functioning, Disability and Health (ICF), within the context of a biopsychosocial model of disability (McCausland et al., 2016, 2016a; Verdonschot et al., 2009; World Health Organization, 2001). Social inclusion has variously been defined as comprising social roles and participation in community (Abbott & McConkey, 2006; Cobigo et al., 2012; Simplican et al., 2015), as relationships including reciprocal relationships (Abbott & McConkey, 2006; Cobigo et al., 2012; Hall, 2009; Simplican et al., 2015), and as belonging (Cobigo et al., 2012; Hall, 2009; Overmars-Marx et al., 2014). It has been identified as a subjective and dynamic social construct which is not fixed but may vary depending on individual and environmental factors (Cobigo et al., 2012; Overmars-Marx et al., 2014), and which may create acceptance and opportunity (Abbott & McConkey, 2006).

There are further similarities between these social concepts and concepts of belonging. For people with ID, belonging has been identified as ‘a subjective feeling of value and respect derived from a reciprocal relationship to an external referent that is built on a foundation of shared experiences, beliefs or personal characteristics’ (Mahar et al., 2013). Similarly, belonging has been viewed by adults with ID with regard to place, as being part of a community, as having relationships and as identity (Strnadová et al., 2018); and by young people with ID as engaging with similar people, having social relationships, negotiating meaningful roles and navigating norms and expectations (Renwick et al., 2019). However, developing belonging in community for people with intellectual disabilities may be a complex and lengthy process, which requires the building of trust, engagement with a range of stakeholders, and creating connections and alliances in the community (Power, 2013).

Regardless of the model or concept used, and perhaps stemming from our innate human need to belong (Baumeister & Leary, 1995), interpersonal relationships including friendships are central to all interpretations of the social lives of people with ID, be that through a lens of social inclusion, participation or belonging.

1.2 Friendship for people with intellectual disabilities

Having friends predicts improved outcomes in almost every area of life (Friedman & Rizzolo, 2018). People with ID across the life course tend to have smaller social networks and fewer friends than others (McCausland et al., 2016; Merrells et al., 2019; White & Forrester-Jones, 2019). Differences in how social networks are measured, including whether or not paid staff are counted, have resulted in considerable variations (McCausland et al., 2018b). When staff were excluded, network sizes of just two people have been reported for adults with ID by proxy respondents (Robertson et al., 2001), and five people for adults with profound intellectual and multiple disabilities (PIMD) (Kamstra et al., 2015), whereas a network size of 22 people when staff were included was identified by self-reporting adults with ID (Forrester-Jones et al., 2006). This same number (22) was later replicated for adolescents with ID using the same method and compared to a network size of 62 for typically developing peers. The largest differences related to friendships and acquaintances, including social contacts from ‘club/church’ (ID mean = 1.35; non-ID mean = 10.72), ‘other friends’ (ID mean = 5.88; non-ID mean = 14.94) and ‘social acquaintances’ (ID mean = 0.47; non-ID mean = 13.39) (White & Forrester-Jones, 2019).

Forrester-Jones et al. (2006) found that just 11% of all network members were social acquaintances and other friends, with a further 25% being other service users with ID. Kamstra et al. (2015) identified just 11% of the network members of adults with PIMD who were friends or potential friends (8.9% included friends without a disability and other contacts; 2.2% people with ID). Emerson and McVilly (2004) reported a median of two friendship activities with friends with ID during a 4-week period, and zero activities with friends without ID. All of these studies used proxy informants. A mix of self and proxy informants included by McCausland et al. (2018b) reported that more than four in ten (43%) older people with ID had no friends outside their own home. These findings emphasise the limited social networks that people with ID have, with few social contacts and activities for many outside of staff, family and peers with intellectual disability.

A recent US study found the majority of adults with IDD (84%) had friends, but fewer were satisfied with the number of friends they had (56%), and less than half (47%) were satisfied with the amount of contact with their friends (Friedman & Rizzolo, 2018). Amongst young people with ID, friends and friendship were important but many felt lonely and wanted more or deeper friendships (Robinson et al., 2018). Young adults with ID felt excluded within mainstream community and faced difficulty developing and maintaining friendships with peers with ID (Merrells et al., 2019). Younger people with ID began to use social media to interact with friends, but did so less than peers without ID and still had much smaller networks overall (White & Forrester-Jones, 2019).

1.3 Type and quality of friendship

Friendship has been defined in terms of joint leisure activities and both practical and emotional support; while moral character, effective communication and understanding were identified as valued traits in friends (McVilly et al., 2006). People with ID have identified a range of friends including peers with ID, co-workers with and without disability, family and support workers; reciprocity is a key determinant of friendship, including with support workers (Callus, 2017). However, perceived similarity and equality was also identified as important for friendship, and people with ID have noted greater comfort with other people with ID and the possibility of ‘true friendship’ through a shared sense of identity (McVilly et al., 2006). Similarly, a shared identity or perceived similarity has also been identified as a basis for developing the bonds and belonging of community (Anderson, 1983; Cohen, 1985; Delanty, 2003; McCausland, Brennan, McCallion, & McCarron, 2016b; Sarason, 1974) and noted above as important to concepts of belonging (Mahar et al., 2013; Renwick et al., 2019). Younger people with ID described friendship in terms of companionship, shared interests and enjoyment in fellowship, with closeness and reciprocity important qualities (Sigstad, 2017).

The status of paid support staff as friends of people with ID has been questioned. While people with ID commonly consider paid staff to be their friends, this may be problematic given high staff turnover and a lack of reciprocity (Friedman & Rizzolo, 2018). Having things always done for you and to you cannot foster reciprocity and should not be expected to in a professional relationship (Forrester-Jones et al., 2006). A tension may exist between providing professional loving care that is likely based on a trusting relationship, and maintaining a separation between the professional and private in staff relationships with people with ID (Hermsen et al., 2014). Others have noted a power imbalance in such relationships (Cummins & Lau, 2003) and how the longevity of staff relationships may be affected by employment-related changes that do not affect other friendships (Bigby, 2008). However, similar dynamics and challenges may also be present in friendships that arise in workplace contexts within the general population (Schawbel, 2018).

Nonetheless, it is evident from previous studies that many people with ID consider paid staff to be their friends and that paid staff fulfil a variety of roles that may be considered as friendship activities. While staff relationships may be less reciprocal, staff are nonetheless the main providers of emotional as well as practical support for people with ID (Forrester-Jones et al., 2006). Older adults with ID identified support staff as the person most involved in their social activities, and staff were by far the most commonly cited confidant amongst self-reporting respondents (McCausland et al., 2018b). Further study of the same dataset found that confiding in support staff was protective against loneliness, but confiding in friends and others was not (Wormald et al., 2019). The quality of contact with professionals was identified as greatly important to the overall well-being of people with ID (van Asselt-Goverts et al., 2015).

1.4 Factors associated with friendship

Multiple factors from the personal to the environmental and macro-level variables influence the degree of friendship, inclusion and participation for people with ID (Gauthier-Boudreault et al., 2019; McCausland et al., 2016, 2016a; Simplican et al., 2015). For example, having non-resident friends was associated with people with ID exercising more choice in everyday and key life decisions (McCausland et al., 2018a). People in competitive and supported employment were more likely to have (non-staff and non-family) friends than those in segregated occupational settings (Friedman & Rizzolo, 2018). However, three categories of factors in particular have been consistently associated with friendship, networks and inclusion for people with ID: individual ability and support needs; residential setting and type of accommodation; and type of support. While these factors are to a large extent interrelated—for example, with individual needs influencing type support, or accommodation type being influenced by support type—the following sections will examine each factor separately.

2.1 Population & procedures

Data were drawn from wave 3 of the Intellectual Disability Supplement to the Irish Longitudinal Study on Ageing (IDS-TILDA), a longitudinal cohort study of adults with ID aged 40 years and above living in Ireland. Forty was used as the inclusion age in this study, compared to 50 years commonly used in ageing studies of the general population, because people with an intellectual disability tend to present with age-related health conditions at an earlier age than the general population (Burke et al., 2014). An original sample of 753 eligible people was randomly drawn from the National Intellectual Disability Database (NIDD) and consented to take part in the study. Ethical approval for the study was granted by the Faculty of Health Sciences Research Ethics Committee at Trinity College, Dublin. Ethical approval was reaffirmed at the start of each wave of data collection, every three years. Wave 3 data were collected between 2016 and 2017 from a total of 609 remaining participants, using a combination of self-reported, supported and proxy interviews. Data collection included a computer-assisted personal interview (CAPI) and a self-completed pre-interview questionnaire (PIQ). CAPI interviews were face-to-face interviews lasting approximately two hours completed directly with participants with intellectual disability, with or without a support person as required, or by proxy by someone who knew the participant well (at least 6 months). PIQs were posted to participants/support persons in advance of a CAPI, completed by them, and collected by the field researcher who called to conduct the face-to-face interview. Table 1 provides a profile of the sample at Wave 3. Level of intellectual disability was based on a formal diagnosis of intellectual disability obtained from the disability services supporting participants in Wave 1 of the study.

2.2 Measures

Table 2 outlines the measures used in this study. Responses to ‘Do you have friends?’ (Yes = 543; sometimes = 17; no = 46) and ‘Do you have a best friend?’ (Yes = 296; sometimes = 21; no = 243) were recoded for analysis into binary yes (yes; sometimes) and no responses. ‘Best friend’ was the concept used to identify a friendship that was particularly close/strong, as subjectively understood by respondents. This item was one of five questions used by Mehling and Tassé (2014) as indicators of overall quality of social relationships, in an analysis of the National Core Indicator (NCI) Survey. The concept of a best friend confers a particular exclusivity and strength in comparison to other friendships (Branje et al., 2007) and is important enough that the lack of such a friend may be considered a deficit in developing adolescents (Richey & Richey, 1980) and associated with increased risk including loneliness (Rubin et al., 2007); while close or strong friendships have also been highlighted as important by people with intellectual disability (Robinson et al., 2018; Sigstad, 2017). The friendship intimacy scale (Mendelson & Aboud, 1999) and friendship closeness scale (Bukowski et al., 1994) were both adapted to reduce the number of response options to three.

Of the 609 participants, 560 answered the question ‘Do you have a best friend?’ A range of variables were tested for their bivariate association with this dependent variable. Table 3 below summarises the counts and missing values of variables included in the multivariate analysis. ‘Communication difficulty’ was based on a question adapted from the National Disability Survey (Central Statistics Office, 2008): Do you have any difficulty speaking or making yourself understood when speaking? Other variables in Table 3 were developed by IDS-TILDA. The Satisfaction with Life measure is from the survey by Diener et al. (1985). Challenging behaviour score was based on the Behaviour Problem Inventory-Short Form (BPI-S) (Rojahn et al., 2012). Community belonging was measured with the question: Do you feel part of your community? (Carpiano & Hystad, 2011). The Rapid Assessment of Physical Activity (RAPA) measured activity levels (Topolski et al., 2006). Other variables in Table 3 were developed by IDS-TILDA. An additional self-reported measure, ‘friendship quality’ (n = 167) was calculated as the sum of the two adapted friendship scales (Bukowski et al., 1994; Mendelson & Aboud, 1999).

2.3 Analysis

Descriptive data were generated for measures outlined in Table 2 above. Fifteen independent variables were included in the multivariate analysis, including gender which was not significantly associated with having a best friend but retained for demographic purposes. Of these, three had more than 20% missing (challenging behaviour, SWLS and loneliness). SWLS and loneliness were self-report only measures and to include these variables in a regression would result in the model being fitted to the sub-sample that self-reported only. It was therefore decided to construct the analyses using decision trees which allow for the inclusion of all missing values.

CHAID (Chi-square Automatic Interaction Detector) analysis is an algorithm used for discovering relationships between a categorical response variable and other categorical predictor variables, while controlling for each predictor variable. It is useful when looking for patterns in datasets with lots of categorical variables and is a convenient way of summarising the data as the relationships can be easily visualised. Because CHAID uses multiway splits by default, it needs rather large sample sizes to work effectively, since with small sample sizes the respondent groups can quickly become too small for reliable analysis. One important advantage of CHAID over alternatives such as multiple regression is that it is non-parametric.

A few factors influenced the decision to use CHAID for further examination of the data. Firstly, the convenience of CHAID modelling is that all variables, without consideration to violating any assumptions, may be entered into the model. Therefore, potential factors which are theoretically and practically sound may be included in CHAID modelling. Secondly, CHAID modelling can handle many variables. Thirdly, the CHAID algorithm has a unique method of handling missing values, whereby no values are discarded. Instead, in the tree growth and classification, the user-missing values of nominal independent variables are treated as ordinary values and are included in the analysis as a single combined category.

3.1 Rates and types of friendship

Overall, 92.4% (560/606) reported that they had friends. Co-resident friends were the most reported (71.8%, 402/560), while almost two-thirds of those with friends also reported key worker/support staff friends (62.5%, 350/560) and non-resident friends (62.0%, 347/560).

Of the 560 individuals who had friends, 56.6% (317/560) said they had a ‘best friend’. Amongst those with a best friend, the most common type by far was a best friend with an intellectual disability (63.2%, 187/296) (see Table 4). Most best friends were non-resident (60.5%, 179/296), with a substantial minority of best friends co-residing with respondents (39.5%, 117/296).

3.2 Factors associated with having a ‘best friend’

Variables outlined in Table 3 were entered into the CHAID model to identify factors associated with having a best friend (Figure 1). This model had an overall classification accuracy of 64.3% with its ability to correctly classify ‘having a best friend’ of 82.3%.

Figure 1 shows that, at the first level (the ‘trunk’) we have all respondents who answered the question ‘Do you have a best friend?’ (560/609). As we progress down the tree to the first branch, we identify the factor with the greatest impact on having a best friend, while the effects of other independent variables were controlled. In this case, the type of respondent, self-reporting versus proxy, was the strongest predictor of having a best friend. A total of 41.6% (91/219) of proxy respondents had a best friend, compared to 66.3% (226/341) of those who self-reported (i.e. self-reporting alone or with some assistance). For proxy-only respondents, the next best predictor of having a best friend was communication (i.e. difficulty speaking). Within this group of proxy respondents, 54.7% (35/64) of those with no/some communication difficulty had a best friend, compared to just 36.1% (56/127) of those with poor communication (cannot do at all/much difficulty). For those who self-reported, challenging behaviour was the next most important variable. Amongst self-reporters, 72.4% (155/214) of those with no challenging behaviour or who did not complete the scale had a best friend. This decreased to 55.9% (71/127) for those with challenging behaviour.

In summary, of the 15 variables included in the model, respondent type (proxy versus self-reported) was the strongest predictor of having a best friend. Then, amongst self-reporting respondents, challenging behaviour was the strongest predictor of having a best friend, while communication ability was the strongest predictor amongst proxy respondents.

3.3 Friendship quality

Of the 317 people who have a best friend, 167 answered the self-reported friendship quality scales. Types of best friend, reported in Table 4 above, were grouped into three categories for further analysis: Friend with an intellectual disability (63.2%, 187/296), Carer/service provider (15.9%,47/296) and Other (including family, work colleague and other friend) (21.0%,62/296). The CHAID analysis in Figure 2 shows that, for those who self-reported, type of friend was the most important and only significant predictor of friendship quality in the model. Those with a best friend who was a ‘carer/service provider’ or ‘other’ (family, work colleague or other) had a higher average friendship quality score (mean = 27.04, SD = 3.1) than those whose best friend had an intellectual disability (mean = 24.8, SD = 4.56).

Due to the small sample size, friendship quality was also analysed using a regression model. A total of seven cases were excluded due to missing values so the model was fitted to a total of 160 cases using forward selection. This method provides an initial screening of the candidate variables when a large group of variables exists. It starts with no variables in the model, testing the addition of each variable using a chosen model-fit criterion, adding the variable (if any) whose inclusion gives the most statistically significant improvement of the fit, and repeating this process until no additional variable significantly improves the model (Table 5).

The final model included two significant variables (p < .05), friendship type and satisfaction with life (SWLS), with an R-squared of 0.075. The reference category for the dummy variable type of friend was chosen to be friend with an intellectual disability, and for the SWLS, it was set as less than satisfied. Those whose best friend was a carer/service provider and were satisfied with life had the highest predicted mean friendship quality score (mean = 27.3), followed by those whose best friend type was ‘other’ (i.e. non peer, family, service or work-related friends) and were satisfied with life (mean = 26.9). Those who had a best friend with an intellectual disability and were less than satisfied with life had the lowest predicted mean friendship quality score (23.2).

The regression model, therefore, confirmed lower friendship quality for best friends with ID compared to other types of best friends, as previously found in the decision-tree analysis.

4.1 Rates and types of friendship

Having friends predicts improved outcomes in almost every area of life (Friedman & Rizzolo, 2018). Therefore, it was encouraging that the large majority of respondents (92.4%, 560/606) had friends; although just over half of these had a best friend (56.6%, 317/560; or 52% of the full sample (317/609)). This is broadly comparable to rates of friendship found in previous studies (Dolva et al., 2019; Friedman & Rizzolo, 2018). The finding that almost half of respondents did not have a friendship close enough to call a ‘best friend’ may also reflect previous findings of dissatisfaction with existing levels of friends and a desire for more friends (Friedman & Rizzolo, 2018; Robinson et al., 2018), difficulties reported with developing and maintaining friendship (Merrells et al., 2019) and smaller networks including fewer friends than found in the general population (White & Forrester-Jones, 2019). In developmental psychology, approximately three-quarters of young children have reciprocally nominated best friends (Rubin et al., 2007). Data from National Core Indicators (NCI) Survey in the United States, from a sample of adults aged 18 and above with intellectual and developmental disabilities (IDD), found an average rate of friendship (79%) lower than rates found here—however, these excluded family and staff friendships. Rates for ‘best friend’ (including family or staff) for adults of all ages with all types IDD were higher than those found here, with an average of 74% across all states (National Core Indicators, 2020). Differences here may be due to age profiles of the samples, especially considering that the rate of best friends for the youngest group in the IDS-TILDA sample (aged 40–49 years) was, at 73.4%, much closer to the NCI rate (McCausland et al., 2017).

Respondents reported having a range of different types of friends, as found in other studies (Callus, 2017). Co-resident friends were the most common type of friend, and best friends with ID were more prevalent than other types of best friend, which may support earlier findings of greater comfort and possibility for true friendship with peers with ID (McVilly et al., 2006). However, such findings also raise questions about choice and the opportunity for friendship for this population, when a majority of respondents previously reported no choice in where they lived (75.4%) or who they lived with (85.5%) (McCarron et al., 2011).

That co-resident friends were more common than non-resident friends, but non-resident best friends were more common than co-resident best friends, may support the idea that real or close friendship is facilitated by having greater opportunity to make friends—that is for those who are able to get out and socialise more. In other words, almost everyone had a friend and these were most commonly co-resident friends, but only half of the sample, predominantly those with better ability (including self-reporters and people with less communication and behaviour support needs), had a best friend, and these were most commonly non-resident. The shared identity thesis for true or lasting friendship (McVilly et al., 2006) may also be supported by the finding that best friends with ID were more prevalent than other types of best friend. However, this support may depend on opportunities for those with non-resident best friends to meet and form friendships with people who did not have ID, which was unknown in this study.

The literature suggests that people with ID have limited opportunity to meet and develop friendships with people who do not have ID (Dolva et al., 2019; Forrester-Jones et al., 2006; Overmars-Marx et al., 2017), meaning the high rate of non-resident best friends with ID may also to some extent be associated with limited choice and opportunity. Where opportunities and support are provided for people with ID to meet and get to know people without identified disabilities, who for example might share a similar interest or live in the same neighbourhood, other studies have shown there is potential for friendship between people with and without ID (Overmars-Marx et al., 2018; Stancliffe et al., 2015).

Findings here also confirm that support staff play an important social role in the lives of older people with ID. Almost two-thirds of respondents had key worker/support staff friends, the same proportion who had non-resident friends. In addition, carer/service provider was the second most common best friend reported (after friend with ID), identified by almost one in six respondents who had a best friend. These findings add to the literature which show how prevalent and important a social role staff play in the lives of people with ID (van Asselt-Goverts et al., 2015; Bigby, 2008; Forrester-Jones et al., 2006; Friedman & Rizzolo, 2018; McCausland et al., 2018b; Wormald et al., 2019). These findings persist despite misgivings about the inclusion of support staff as ‘friends’ to people with ID who they are paid to support (Bigby, 2008; Cummins & Lau, 2003; Forrester-Jones et al., 2006; Friedman & Rizzolo, 2018; Hermsen et al., 2014), perhaps reflecting a current reality that is short of ideal regarding the breadth and depth of friendships for people with ID. Ultimately, opportunity and choice is a key factor in shaping friendship for people with ID, who will continue to rely heavily on staff friendships until they are adequately supported in social opportunities to be able to choose friendships with other groups.

4.2 Factors associated with having a best friend

Self-reporting respondents were much more likely than proxy respondents to have a best friend. Proxy respondents (57.6%, 141/245) were significantly more likely to have a severe-profound level of ID than self-reporters (7.0%, 22/314; p < .001). Compared to self-reporters, they were almost twice as likely to report fair-poor physical health (19.4% versus 10.7%, p < .05) and almost three times as likely to report fair-poor mental health (28.7% versus 11.6%, p < .001). These and other factors related to higher support needs are likely to impact an individual's ability to engage in social and friendship activities (Felce & Emerson, 2001; Friedman & Rizzolo, 2018; Kamstra, van der Putten, Post, et al., 2015; Mansell & Beadle-Brown, 2009; Robinson et al., 2018) and may further explain a reduced rate of having a best friend.

Another potential factor may be differences arising due to the nature of self- and proxy-reporting. The literature has been inconclusive about the existence or extent of a bias or inconsistencies based on respondent type (Perry, 2004; Stancliffe, 2000), with some identifying consistent reporting (Schalock & Keith, 1993; Stancliffe, 1999) and others identifying differences (Claes et al., 2012; Rapley et al., 1998; Stancliffe, 1995). Consistency between self-reported and proxy responses is lower on subjective measures and higher on objective measures (Emerson et al., 2013; Perry & Felce, 2002). Measures may be taken by researchers to enhance concordance (Perkins, 2007), and IDS-TILDA attempted to minimise any potential for inconsistencies in response types with a standard that proxies had to know the individual with ID for at least 6 months. Nevertheless, the subjectiveness of the appraisal of someone as the individual's best friend may have influenced proxy responses.

A second question arising from differences in rates of best friends is whether environment and support needs mean someone is more or less likely to have a best friend. The literature shows that residence in smaller community-based accommodation is better for social inclusion (Emerson, 2004; Noonan Walsh et al., 2007), that setting characteristics are a more important determinant than individual characteristics (Emerson & McVilly, 2004), and that type of support within these settings is the key determinant of inclusion (Amado et al., 2013; Friedman & Rizzolo, 2018; Kamstra et al., 2019; Kozma et al., 2009; McConkey et al., 2019; McVilly et al., 2006; Noonan Walsh et al., 2010; Overmars-Marx et al., 2017; Talman et al., 2019). While the current study does not include data on the type or extent of support available to respondents, or what gaps exist, residence type was a broad indicator. While two-thirds of proxy respondents lived in congregated/campus settings (67.4%, 174/258), three-quarters of self-reporters lived in community settings (73.4%, 256/349). These findings should be treated with caution as they are likely confounded by other variables such as level of ID, but, in Ireland, it is established that those with greater support needs continue to reside in congregated settings (Health Service Executive, 2011). The results here suggest that within congregated environments needs around friendship are not being met, to the extent that most do not have close friends.

A third question is whether the difference between cohorts in having a best friend is related to the understanding or interpretation of the proxy. Almost two-thirds of proxy respondents with more severe communication difficulties (63.9%, 99/155) did not have a best friend, compared with fewer than half of those with less severe communication difficulties (45.3%, 29/64). However, it may be that communication between the individual with ID and their support personnel is also a factor. Fewer proxy respondents having a best friend may be associated with their communication needs and how these are inadequately supported, or they may have a best friend but have difficulty communicating this to their proxy—that is the proxy is unaware of the individual's belief that they have a best friend.

These findings support the recent literature which found that communication ability is associated with having a best friend (Dolva et al., 2019), that staff knowing the friendship preferences of those they support increases their chances of having friends (Friedman & Rizzolo, 2018), that staff perceptions and understanding of the individuals they support is important to participation and connecting with others (Kamstra et al., 2019; Talman et al., 2019), and that continuity and familiarity of support staff enhances inclusion (Friedman & Rizzolo, 2018; McCausland et al., 2019). This suggests that better communication support for those with such difficulties is required, including training for staff to better understand, interpret and support people with significant communication difficulties to meet people and to form and maintain friendships.

Amongst self-reporting respondents, two-thirds of whom had a best friend, those with no challenging behaviour were more likely to have a best friend (72.4%, 155/214) than those who had challenging behaviour (55.9%, 71/127). These findings supports previous studies which identified the social impact of behavioural difficulties including reduced social inclusion and network size (Emerson, 2001; Matson et al., 2011; McCausland et al., 2019; McVilly et al., 2006). Public and community perceptions of challenging behaviour, or even the potential for challenging behaviour, may also present a barrier to inclusion (Overmars-Marx et al., 2018). Once again, the findings suggest that better support for people with ID to develop and maintain friendships is needed, in this case support with challenging behaviours that impact on their chances of having close friends. Psychotropic medicine and applied behaviour analysis are amongst the more common treatments for challenging behaviour (Matson et al., 2011). Other strategies to support inclusion for those with behavioural support needs include maintaining familiar support staff who know the individual well (McCausland et al., 2019) and educating individuals with ID about inappropriate behaviour (Overmars-Marx et al., 2017).

4.3 Friendship quality

In the model exploring friendship quality, which combined scales of friendship closeness (Bukowski et al., 1994) and intimacy (Mendelson & Aboud, 1999), the analysis found that relationships with best friends with ID were of significantly lower quality than relationships with other best friends (including service providers, family and work colleagues).

At face value, this finding may contradict the literature which found that perceived similarity and equality in a friend was important to people with ID creating comfort with such peers and the possibility of true friendships (McVilly et al., 2006). However, the same study, and others, also identified generic traits of friends/friendship that may be more difficult to find and maintain such as moral character, communication and support (McVilly et al., 2006), companionship, shared interests and closeness (Callus, 2017), as well as reciprocity (Callus, 2017; Sigstad, 2017).

Findings regarding friendship quality are revealing given that best friends with ID were the most common type, accounting for almost two-thirds of reported best friends (63.2%, 187/296). So our findings here imply that older people with ID are more likely to have friends with ID, and according to the literature, these may be more likely to last than other friendships; but the quality of these relationships may be poorer than relationships with friends who do not have ID. Taken together, these findings may support our previous assessment that choice and opportunity are critical to the development of friendships for people with ID, and that older people with ID in Ireland are currently denied that choice. This may result in developing friendships more out of convenience than anything else, such as shared interests or beliefs, with a potentially negatively impact on their friendship quality where this is the case. The relatively small r-squared in the regression model that was run to confirm the initial CHAID analysis suggests the complexity of friendship quality and the existence of multiple other important factors.

4.4 Study limitations

‘Friends’ and ‘best friend’ were not defined for participants, and interpretation was thus based on individuals’ understanding or self-definition of these concepts. It is possible that some respondents had more than one best friend or a group of close friends and may have therefore responded ‘no’ to having ‘a best friend’, thereby underestimating the rate of best friends.

5.1 Supporting friendship and inclusion

While there is some support of the potential for Information and Communication Technologies (ICTs) to help people with ID develop or maintain friendships (Caton & Chapman, 2016), the evidence to date suggests that we remain some way short of this potential being realised. In practice, developing social inclusion through use of ICTs is dependent on having the opportunity and skills required to access and use these technologies (Mäkinen, 2006), which is more difficult for people with ID (Balandin & Molka-Danielsen, 2015; McCausland et al., 2017), and thus requires additional supports to realise their potential (Barlott et al., 2019). Recent evidence suggests that a younger generation of people with ID may be making greater use of social media to interact with friends; however, they used technology to interact with friends less than peers without ID, and their networks were not grown by technology use (White & Forrester-Jones, 2019).

Other studies highlighted success and potential for increased inclusion using strategies including local community Interventions based on shared interests, active support and volunteer mentoring (Craig & Bigby, 2015; Stancliffe et al., 2015; Wilson et al., 2015), or promoting encounters in public spaces with other members of their community (Bigby & Wiesel, 2015, 2019; Bould et al., 2018). However, simply promoting contact may not be effective, and success may depend on facilitating places, community members and support (Bigby & Wiesel, 2019). This is paralleled by Contact Theory, which proposed that simple, casual contact is not enough to engender positive outcomes between people, and that certain conditions of contact (such as equal status or interdependence) must be satisfied to promote inclusion (Allport, 1954; Novak & Rogan, 2010). Bigby et al. (2018) concluded that a focus on micro-level practices with individuals, such as active mentoring or active support, may be successfully combined with approaches at a group or community level.

It is clear from our findings and the previous literature that older people with ID need enhanced opportunities for social inclusion and support to develop and maintain friendships if they are to overcome the challenges identified in this paper. The multidimensional nature of inclusion and participation suggests that an individualised or personalised approach is key to the success of such support strategies.