Group differences in social inclusion between young adults aged 18 to 25 with serious mental illness and same-aged peers from the general community

Abstract

Aims:

Young adults with serious mental illness (SMI) are thought to be socially excluded. However, psychometric measures of social inclusion have not been employed to examine group differences relative to peers from the general community. The aim of this study was to employ such a measure to determine differences in social inclusion between young adults with SMI and peers from the general community.

Methods:

A cross-sectional hierarchical logistic regression was conducted to identify which dimensions and individual indicators from the Filia Social Inclusion Measure (F-SIM) discriminated between n = 152 young adults aged 18–25 from the general community (M = 21.36, SD = 2.16) and n = 159 young adults aged 18–25 with SMI (M = 21.13, SD = 2.21).

Results:

Group membership was accounted for by Interpersonal Connections (Nagelkerke R² = .32), Vocational & Financial Security (Nagelkerke R² = .32) and Healthy Independent Lifestyle (Nagelkerke R² = .08) dimensions of the F-SIM. Relative to young adults from the general community, those with SMI were five times less likely to feel they had friends who would call on them in a crisis, odds ratio (OR) = .19 (95%CI = .04, .53), p = .04, almost five times more likely to live with their parents, OR = 4.79 (95%CI = 1.98,11.15), p = .004, almost four times less likely to have worked/studied any time over the past 12 months, OR = .27 (95%CI = .11,.64), p < .001, and three-and-a-half times more likely to report unstable accommodation, OR = 3.58 (95%CI = 1.14, 11.15), p = .03.

Conclusion:

Young adults with SMI are socially excluded relative to peers from the general community in terms of interpersonal connections, vocational engagement, autonomy/independence and housing stability. In addition to the well-established focus on vocational engagement, interventions to improve social inclusion in this population must promote reciprocity within social relationships and healthy autonomy/independence (including stable housing).

Keywords

Social inclusion social exclusion young adult group difference mental illness

Hallmark characteristics of social inclusion, for example, high-quality interpersonal relationships, are robustly associated with better health (Holt-Lunstad, Robles, & Sbarra, 2017). There is no unanimously agreed definition of social inclusion, but it is broadly considered to comprise objective and subjective factors (Berry & Greenwood, 2018). It has been defined as a sense of belongingness/acceptance in relation to opportunities to participate in valued social roles (Gardner, Filia, Killackey, & Cotton, 2019). Social inclusion initially emerged in antithesis to the established social exclusion construct (Spandler, 2007), which may be defined as a sense of loneliness/isolation in relation to dissatisfaction with opportunities to participate in such roles. This typically involves active processes of exclusion by society, for example, enforced lack of participation (Morgan, Burns, Fitzpatrick, Pinfold, & Priebe, 2007). Social inclusion and exclusion are widely considered to be dimensional opposites (Baumgartner & Burns, 2014). They are also widely agreed to be multidimensional (Huxley, Evans, & Munroe, 2006). Commonly cited dimensions include the following: social relationships, employment/education, leisure activities, housing and accommodation, financial situation and health (Filia, Jackson, Cotton, Gardner, & Killackey, 2018; Huxley et al., 2012). The particular roles that are valued across these dimensions likely differ throughout the lifespan. Focusing on subgroups defined by developmental stage may therefore be helpful – particularly during key transitions such as young adulthood (Hayes, Gray, & Edwards, 2008).

The terms youth and young adulthood have been applied somewhat interchangeably to the transition from childhood to adulthood for people in their late-teens to mid-20s (UNESCO, 2017). It is important to examine social inclusion during this stage because it appears to be a sensitive period for related processes, for example, peer group acceptance (Blakemore & Mills, 2014). Mental illness onset typically occurs during this stage and may complicate such processes (McGorry et al., 2013). Serious mental illness (SMI) is often defined by diagnosis (e.g. schizophrenia) but may be defined as any mental illness that makes it difficult for a person to complete everyday activities without assistance (Australian Bureau of Statistics, 2015).

Young adults with SMI are considered socially excluded (Social Exclusion Unit, 2005). This is largely based on evidence from objective indicators, for example, lower vocational engagement relative to peers from the general community (Iyer et al., 2018). However, subjective indicators are under-researched in this population (Sündermann, Onwumere, Bebbington, & Kuipers, 2013), where loneliness remains poorly understood (Lim, Gleeson, Alvarez-Jimenez, & Penn, 2018). Psychometric measures of social inclusion combine objective and subjective indicators (Shepherd & Parsonage, 2011). Such measures have not been used to examine group differences in social inclusion between young adults with SMI and peers from the general community: there is a lack of research targeting young adults in the social inclusion literature (Evans-Lacko et al., 2014). This makes it difficult to adequately assess group differences between young adults from the general community and those with SMI. That is, to move beyond gross discrepancies in for example, vocational engagement toward a more granular understanding that incorporates subjective perceptions of social opportunities. It is at this granular level where individual interventions may be most effective (Australian Mental Health Outcomes Classification Network, 2016).

The primary aim of this study was to examine group differences between young adults aged 18–25 with SMI and those from the general community on a newly developed measure of social inclusion. The secondary aim was to identify particular dimensions and individual indicators of social inclusion that discriminated between groups.

Method

This cross-sectional study involved analysis of baseline data from studies exploring the psychometric properties of a social inclusion measure among 18–25 year olds with SMI (Gardner, Cotton, O’Donoghue, et al., 2019) and from the general community (Gardner, Cotton, Killackey, et al., 2019). This age range was selected for reasons explained above (i.e. it appears to be a sensitive period for social inclusion).

Setting and Sample

Clinical sample

Participants were recruited from (a) Orygen Youth Health, a public mental health service for people aged 15–25 in the northwest of Melbourne, Australia; and (b) Mind Australia, youth psychosocial rehabilitation services for people aged 16–25 across Melbourne, Australia. Both services support young adults with diverse manifestations of SMI. This study employed a transdiagnostic conceptualization (Hartmann et al., 2019), reflecting the diagnostic instability endemic of early-stage SMI (Menezes & Milovan, 2000). Diagnoses of mental illness were recorded via self-report, a method used in genome-wide studies of depression (Hyde et al., 2016). Participants were approached in consultation with clinical staff from these services. Exclusion criteria were clinician-assessed symptom severity precluding informed consent (e.g. florid psychosis) and lack of fluency in English (due to pragmatic reasons).

General community sample

Participants were living in Melbourne, Australia and were recruited from a range of settings (e.g. universities, restaurants/cafés, public housing, shopping centers). Recruitment techniques included advertisement (e.g. social media, flyers/posters) and snowballing. Exclusion criteria were lack of sufficient fluency in English, and diagnosis of SMI. As this was a general community sample (not ‘healthy controls’), a diagnosis of mental illness per se did not necessarily exclude participants. Anxiety and depression are highly prevalent among the general population of young adults (Gustavson et al., 2018). General community participants reporting a lifetime mental illness diagnosis were asked followup questions to help determine whether this ought to be considered SMI. For example, they were asked whether the illness made it difficult for them to manage everyday tasks by themselves and whether they had ever accessed psychiatric services.

Measures

The Filia Social Inclusion Measure (F-SIM) is a 72-item self-report measure of social inclusion over the past month. Item creation was based on a thematic analysis identifying indicators of social inclusion (Filia et al., 2018) and a Delphi study seeking consensus on which indicators to include in the measure (Filia, Jackson, Cotton, & Killackey, 2019). The F-SIM employs dichotomous-, ordinal-, nominal- and ratio-level items to assess objective (e.g. How many of the past 12 months did you work/study for?) and subjective (e.g. I often feel unwelcome, like I don’t belong) aspects of social inclusion. Like other social inclusion measures (Coombs, Reed, & Rosen, 2016), frequencies/percentages can be used to examine responses to individual items. This maximizes clinical utility by targeting specific areas of social inclusion (Australian Mental Health Outcomes Classification Network, 2016). Items are grouped into three dimensions: Interpersonal Connections (Cronbach’s α = .94), Vocational and Financial Security (Cronbach’s α = .87) and Healthy Independent Lifestyle (Cronbach’s α = .82). The F-SIM is one of the only measures to demonstrate reliability and validity specifically among young adults aged 18–25 with SMI (Gardner, Cotton, O’Donoghue, et al., 2019). Demographic information and self-reported lifetime mental disorder diagnosis were obtained via the F-SIM. Participants were asked whether they had ever received a diagnosis from the following DSM-5 categories: depressive disorders, anxiety disorders, schizophrenia spectrum & other psychotic disorders, bipolar and related disorders, eating disorders, obsessive-compulsive and related disorders, trauma-related disorders, substance-related and addictive disorders or personality disorders. Participants who indicated that they had been diagnosed with a mental disorder at some point in their lives were asked a series of follow-up questions (e.g. history of accessing psychiatric services and/or psychiatric hospital admissions).

Ethics

This study received ethical approval from the Melbourne Health Human Research and Ethics Committee (HREC/16/MH/325) and was endorsed by the Mind Australia Research and Evaluation Committee. The Youth Research Council at Orygen, the National Centre of Excellence in Youth Mental Health, provided consumer perspective and feedback on the study plan and design. Data were collected via online surveys. Informed consent was implied by survey submission, which was explained via the participant information form. Participants received a retail voucher for their participation.

Data analyses

Analyses were conducted using IBM^® SPSS^® Statistics Version 25. Little’s Missing Completely at Random (MCAR) test was used to assess missing data. Descriptive statistics were inspected for accuracy of input and compared between groups. Such comparisons were conducted via bivariate logistic regressions, with group membership as the outcome variable. This method was selected so as to produce the same type of test statistic (i.e. Wald) and effect size (i.e. odds ratios) reported in the primary analysis of interest outlined below.

A four-block hierarchical logistic regression was conducted to predict membership to one of two groups: young adults aged 18–25 from the general community and young adults aged 18–25 with SMI. Logistic regression was chosen because many variables were non-continuous. There should be ⩾10 cases for each predictor variable in a logistic regression model (Vittinghoff & McCulloch, 2007). With N = 311, this afforded the inclusion of ⩽31 predictor variables. To avoid over-fitting of the model, 22 predictor variables were included. Therefore, not all F-SIM items (n = 72) could be included in the analysis. A rationale for the selection of F-SIM items is provided below.

Age and gender were entered into the first block to control for the potentially confounding effect of these variables. The dimensional structure of the F-SIM has been reported in a study examining its psychometric properties among young adults aged 18–25 with SMI (Gardner, Cotton, O’Donoghue, et al., 2019). This dimensional structure was used as a rationale for the hierarchy of the final three blocks in the model. The seven best-performing items (i.e. those with the highest dimension loadings) from the Interpersonal Connections dimension were entered as predictors in Block 2 (dimension loading range = .64–.80). The seven best-performing items from the Vocational and Financial Security dimension (loading range = .48–.80) were entered as predictors in Block 3. Finally, the seven best-performing items from the Healthy Independent Lifestyle dimension (loading range = .46–.70) were entered as predictors in Block 4. Figure 1 depicts the four blocks of the hierarchical logistic regression and details the individual F-SIM items included within each block.

Figure 1.

F-SIM dimensions that comprise each block, and the individual F-SIM item indicators within each dimension, for the hierarchical logistic regression with group membership (general community = 0, serious mental illness = 1) as the outcome variable.

From the logistic regression model, we were able to determine how each block contributed to the explanation of group differences. Omnibus tests of model coefficients were examined to determine chi-square change between blocks ( $χ_{Δ}^{2}$ ). Chi-square (χ²) and Hosmer–Lemeshow tests were conducted to assess the overall goodness-of-fit of each of the steps of the model, as well as the final model. Nagelkerke R² values were generated to indicate the amount of variation in group membership explained by the cumulated blocks of variables. Classification tables were generated to examine the ability of each of the steps in the model to accurately identify group membership. The relative contribution of individual predictors in the models was tested with the Wald statistic. Odds ratios (OR) and their 95% confidence intervals [95%CI] were calculated to assess the effect size of individual predictors. Effect sizes were interpreted as small (OR = 1.5), medium, (OR = 2.0) or large (OR = 3 (Sullivan & Feinn, 2012)).

Results

Cohort characteristics

Between April 2017 and August 2018, N = 311 participants (general community n = 152; SMI n = 159) completed surveys. Table 1 describes the two samples. In the SMI group, seven people identified as transgender and seven did not identify as female, male or transgender. In the general community group, one individual identified as transgender. There were no significant differences in age or binary gender (male/female) between groups. Young adults with SMI were six times less likely to have access to the Internet at home (OR = 0.16, 95%CI [0.04, 0.74], p = .02) and two-and-a-half times less likely to have a current drivers license, (OR = 0.38, 95%CI [0.23, 0.62], p < .001). Young adults with SMI were three times less likely to receive income from work/study (OR = 0.34, 95%CI [0.22, 0.55], p < .001), six-and-a-half times more likely to receive government payments (OR = 6.54, 95%CI [3.92, 10.87], p < .001) and almost four times less likely to receive money from family members (OR = 0.27, 95%CI [0.15, 0.46], p < .001). The most commonly self-reported diagnostic categories in the SMI group were depression (53%, n = 84), anxiety (43%, n = 69) and schizophrenia spectrum disorders (30%, n = 48). Comorbidity was common: 64% (n = 103) in the SMI group reported more than one diagnosis of mental illness. In the general community group, n = 7 reported the following lifetime diagnoses: anxiety disorders (2.63%, n = 4), depressive disorders (1.32%, n = 2), eating disorders (1.32%, n = 2), trauma-related disorders (0.66%, n = 1) and obsessive-compulsive-related disorders (0.66%, n = 1). None of these seven participants reported that the illness made it difficult for them to complete everyday tasks by themselves or that they had ever accessed psychiatric services. Hence they were retained in the general community sample.

Table 1.

Demographic and other characteristics of (a) a sample of young people aged 18–25 from the general community (GC) and (b) a sample of young people aged 18–25 with serious mental illness (SMI).

Characteristic		GC (n = 152)	SMI (n = 159)	Wald test of group difference	OR [95%CI]
Gender
Female	% (n)	51.3 (78)	48.40 (77)
Male	% (n)	48.0 (73)	42.80 (68)
				0.06	0.94 [0.60, 1.49]
Transgender	% (n)	.70 (1)	4.40 (7)	–	–
I do not identify as female, male or transgender		–	4.40 (7)	–	–
Age in years	M (SD)	21.36 (2.16)	21.13 (2.21)	0.86	0.95 [0.86, 1.06]
How many dollars (AUD) per week do you personally spend on rent or mortgage payments?	M (SD)	121.32 (130.42)	96.28 (122.78)	2.99	1.00 [0.99, 1.01]
	Median	109.50	60.00
Do you have access to the Internet at home? (% Yes)	% (n)	98.7 (150)	92.50 (147)	5.50*	0.16 [0.04, 0.74]
Do you have a current drivers license (% Yes)	% (n)	77.00 (117)	56.00 (89)	14.92***	0.38 [0.23, 0.62]
Nett weekly income in dollars (AUD)	M (SD)	420.06 (322.45)	317.87 (239.23)	9.45**	0.99 [0.98, 1.00]
	Median	310.00	280.00
Where do you currently receive an income from?
Payment from work or study (e.g. scholarship) (% Yes)	% (n)	58.60 (89)	32.70 (52)	20.45***	0.34 [0.22, 0.55]
Government payment (e.g. youth allowance, newstart allowance, single parent payment, disability support pension) (% Yes)	% (n)	19.70 (30)	61.60 (98)	51.71***	6.54 [3.92, 10.87]
Spouse or other family member provides money (% Yes)	% (n)	40.10 (61)	15.10 (24)	23.04***	0.27 [0.15, 0.46]
I don’t receive an income – I have no money (% Yes)	% (n)	7.90 (12)	11.30 (18)	1.04	0.67 [0.31, 1.45]
Have you ever received a diagnosis of a mental illness from a health professional? (% Yes)	% (n)	4.60 (7)	96.20 (153)	–	–
DSM-V category:
Depressive Disorders	% (n)	1.32 (2)	52.83 (84)	–	–
Anxiety Disorders	% (n)	2.63 (4)	43.40 (69)	–	–
Schizophrenia Spectrum and Other Psychotic Disorders	% (n)	–	30.20 (48)	–	–
Personality Disorders	% (n)	–	25.79 (41)	–	–
Bipolar and Related Disorders	% (n)	–	18.87 (30)	–	–
Trauma-related Disorders	% (n)	0.66 (1)	17.61 (28)	–	–
Eating Disorders	% (n)	1.32 (2)	10.06 (16)	–	–
Obsessive-Compulsive and Related Disorders	% (n)	0.66 (1)	6.29 (10)	–	–
Substance-Related and Addictive Disorders	% (n)	–	4.40 (7)	–	–

M: mean; SD: standard deviation; n: number of participants; OR: odds ratio; 95%CI: 95% confidence interval; AUD: Australian Dollars; TAFE: Technical and Further Education; DSM-V: Diagnostic and Statistical Manual of Mental Disorders, 5th Edition.

p < .05; **p < .01; ***p < .001.

Logistic regression to predict group membership

Statistical assumptions require ⩾5 observations in 80% of cells for χ² analyses (McHugh, 2013). Hence two logistic regressions were performed: one without n = 15 participants who did not identify as male or female and one including these participants but dropping the gender variable from the analysis. Binary gender did not predict group membership. The two analyses produced near identical results (e.g. statistically significant individual predictors of group membership were exactly the same). In the spirit of inclusion, results of the logistic regression that excluded the gender variable and included all participants are reported here.

Age was entered in Block 1 of the hierarchical logistic regression, explaining <1% of variance in group membership, over and above that which was explained by the constant in Block 0. Seven items from the Interpersonal Connections dimension of the F-SIM were entered in Block 2. After controlling for age, this block explained a statistically significant additional 32% of variance in group membership, Nagelkerke R² = .32, $χ_{Δ}^{2}$ (df = 7) = 84.65, p < .001. The model was statistically significant, χ² (df = 8) = 85.51, p < .001, and demonstrated goodness-of-fit, Hosmer–Lemeshow χ² (df = 8) = 11.28, p = .19. Seven items from the Vocational and Financial Security dimension of the F-SIM were entered in Block 3. After controlling for variables included in previous blocks, this block explained a statistically significant additional 32% of variance in group membership, Nagelkerke R² = .64, $χ_{Δ}^{2}$ (df = 7) = 119.53, p < .001. The model was statistically significant, χ² (df = 15) = 205.04, p < .001, and demonstrated goodness-of-fit, Hosmer–Lemeshow χ² (df = 8) = 5.48, p = .71. Finally, seven items from the Healthy Independent Lifestyle dimension of the F-SIM were entered in Block 4. After controlling for variables included in previous blocks, this block explained a statistically significant additional 8% of variance in group membership, $χ_{Δ}^{2}$ (df = 7) = 36.37, p < .001. The final model with 22 predictor variables was statistically significant and explained 72% of variance in group membership, Nagelkerke R² = .72, χ² (df = 22) = 241.41, p < .001. It demonstrated goodness-of-fit, Hosmer–Lemeshow χ² (df = 8) = 7.20, p = .52. The final model correctly classified 90% of general community cases, 85% of SMI cases and 87% of overall cases.

Table 2 provides a summary including individual variables that made a unique statistically significant contribution to the final model. Given group membership coding (general community = 0, SMI = 1), an OR between 0 and 1 denotes reduced likelihood for the SMI group whereas an OR > 1 denotes increased likelihood for the SMI group. Young adults with SMI were five times less likely to feel they had friends who would call on them in a crisis (OR = 0.19, 95%CI [0.04, 0.52], p = .04) and almost five times more likely to live with their parents (OR = 4.79, 95%CI [1.98, 11.15], p < .001). They were almost four times less likely to have worked/studied at any time in the past 12 months (OR = 0.27, 95%CI [0.11, 0.64], p < .001) and three-and-a-half times more likely to report unstable accommodation (OR = 3.58, 95%CI [1.14, 11.15], p = .04). Young adults with SMI were more than two times less likely to be currently studying at a formal institution (OR = 0.44, 95%CI [0.30, 0.64], p < .001). For every month worked/studied over the past 12 months, SMI group membership was 33% less likely than general community group membership (OR = 0.67, 95%CI [0.56, 0.81], p < .001). For every glass/cup, small bottle or can of alcoholic beverage consumed per week, SMI group membership was 1.18 times more likely (OR = 1.18, 95%CI [1.02, 1.37], p = .02).

Table 2.

Summary of the hierarchical logistic regression with group membership of N = 311 young adults aged 18–25 years (GC = 0, SMI = 1) as the outcome variable and age entered in Block 1, F-SIM Interpersonal Connections dimension items entered in Block 2, F-SIM Vocational and Financial Security dimension items entered in Block 3 and F-SIM Healthy Independent Lifestyle dimension items entered in Block 4 as predictor variables.

Regression block	Individual predictor variables	B	SE	Wald χ² (df = 1)	OR	95% CI for OR
Block 1: Age
Nagelkerke R² =< .01Cases correctly classified = 52%	Age in years	−0.01	0.10	0.01	1.00	0.82, 1.20
Block 2: F-SIM Interpersonal Connections dimension
Nagelkerke R² = .32Cases correctly classified = 71%	How satisfied are you with your social life?	−0.31	0.35	0.82	0.73	0.37, 1.44
	How satisfied are you with your social activities?	−0.41	0.39	1.12	0.67	0.31, 1.42
	I don’t have a group of friends to do anything with socially	0.24	0.34	0.51	1.28	0.65, 2.50
	How often do you catch up with a friend for social activity?	−0.33	0.27	1.53	0.72	0.42, 1.22
	Do you have friends who would call on you in a crisis?	−1.66	0.84	3.93*	0.19	0.04, 0.53
	Do you have good friends you see or speak to regularly?	−1.07	0.62	2.98	0.34	0.10, 1.16
	Do you have friends who would call on you to discuss everyday experiences?	−0.21	0.78	0.07	0.81	0.18, 3.73
Block 3: F-SIM Vocational and Financial Security dimension
Nagelkerke R² = .64Cases correctly classified = 81%	Are you currently unemployed?	0.38	0.72	0.27	1.45	0.36, 5.90
	Are you currently working for pay?	−0.07	0.24	0.09	0.93	0.58, 1.49
	How many of the past 12 months did you work or study for?	−0.40	0.10	17.44***	0.67	0.56, 0.81
	Over the past 12 months did you work or study at any time?	−4.61	1.17	15.58***	0.27	0.11, 0.64
	Have you been enrolled in any formal studies in the past 12 months?	−0.32	0.57	0.30	0.73	0.24, 2.22
	Are you currently studying at a formal institution?	−0.82	0.19	18.80***	0.44	0.30, 0.64
	Are all household members ⩾18 years old currently employed or attending formal education?	−0.97	0.50	3.76	0.38	0.14, 1.01
Block 4: F-SIM Healthy Independent Lifestyle dimension
Nagelkerke R² = .72Cases correctly classified = 87%	Do you currently live with your parents?	1.57	0.46	11.91**	4.79	1.98, 11.15
	Are health services in your local community of a good standard?	0.06	0.43	0.02	1.06	0.46, 2.46
	I don’t have the time to do what I think I should do. I need more time in the day	0.47	0.30	2.48	1.60	0.89, 2.88
	Is your accommodation unstable (e.g. Are you worried about being evicted)?	1.27	0.58	4.74*	3.58	1.14, 11.15
	How many alcoholic beverages (e.g. glass/cup, small bottle or can) do you drink per week?	.17	.07	5.20*	1.18	1.02, 1.37
	Are you living in a location other than where you would like?	0.21	0.57	0.14	1.24	0.41, 3.70
	Do you currently use illegal substances?	1.24	0.90	1.91	3.70	0.60, 20.00

SE: standard error; OR: odds ratio; GC: General Community; SMI: Serious Mental Illness; F-SIM: Filia Social Inclusion Measure; Nagelkerke R²: estimated % of variance in group membership explained cumulatively by blocks in the logistic regression model.

p < .05; **p < .01; ***p < .001.

Discussion

To our knowledge, this is the first study to use a psychometric measure of social inclusion to examine group differences between young adults aged 18–25 with SMI and those from the general community. This contributes to the much-needed establishment of normative and clinical data regarding social inclusion among young adult populations (Evans-Lacko et al., 2014). It also deepens understanding of gross objective discrepancies (e.g. smaller social networks) by elucidating nuances of the social exclusion experienced by young adults with SMI. In so doing, these findings suggest the need for novel approaches to improving social inclusion in this population.

Summary of findings

The F-SIM dimensions of Interpersonal Connections, Vocational and Financial Security and Healthy Independent Lifestyle significantly discriminated between groups and accounted for a large amount of variation in group membership. Individual indicators suggested that young adults with SMI were socially excluded relative to peers from the general community in a number of areas. Those with SMI were significantly less likely to feel they had friends who would call on them in a crisis, to have been vocationally engaged in the past year, and to be currently studying at a formal institution. They were significantly more likely to live with their parents and to report unstable accommodation. This logistic regression model demonstrated a high level of accuracy in group classification. Alongside evidence that young people at risk of developing SMI are socially excluded (Robustelli, Newberry, Whisman, & Mittal, 2017), these findings suggest that models of social inclusion may have a part to play in the early identification of SMI.

Comparison to previous literature

Comparisons to previous literature are complicated by the lack of studies using psychometric measures of social inclusion to examine group differences between young adults from the general community and those with SMI. Nonetheless, these findings are consistent with evidence that young adults with SMI have smaller social networks (Gayer-Anderson & Morgan, 2013), are less vocationally engaged (Iyer et al., 2018), experience lower levels of independence (Breitborde, Woolverton, Frost, & Kiewel, 2014) and higher levels of housing instability (Narendorf, Cross, Santa Maria, Swank, & Bordnick, 2017) than those from the general community.

The findings support evidence that vocational engagement contributes to social inclusion for young adults with SMI (Gardner, Cotton, Allott, et al., 2019). They also suggest subtle differences in historical versus current vocational engagement and in employment-specific versus education-specific factors. Unspecified historical vocational engagement (i.e. Did you work or study at any time over the past year?) significantly discriminated between groups. The education-specific indicator of enrolment in formal studies over the past year did not discriminate between groups. This raises the question as to whether observed differences in vocational engagement over the past year were driven by employment-specific factors. Young adults with SMI may have been less likely to have worked, but not necessarily less likely to have studied, than their peers from the general community over the past year. This remains speculative, because no indicators exclusively assessing work history over the past year were included in the analysis. It is also important to distinguish enrolment in formal studies from attendance at such studies. Although young adults with SMI were not significantly less likely to be enrolled in formal education over the past year, they were significantly less likely to be currently attending an educational institution. This is consistent with evidence that young adults with SMI are less likely to be engaged in postsecondary education than peers from the general community (Roy, Rousseau, Fortier, & Mottard, 2016). It also suggests that there may be some utility in examining discrepancies between rates of enrolment and rates of attendance in educational settings for young adults with SMI. Such discrepancies may provide a useful index of the disruption that SMI can engender for affected young people.

In this study, indicators of current unemployment did not significantly discriminate between groups. This is inconsistent with evidence of a higher unemployment rate among young adults with SMI relative to peers from the general community (Ramsay, Stewart, & Compton, 2012). Methodological differences may have contributed to this inconsistency. Ramsay and colleagues included a wider age range for participants (18–40 years-of-age) and found that age was independently associated with unemployment. That is, younger age predicted higher unemployment. A relatively restricted age range in this study (i.e. 18–25) may have precluded us from detecting such an effect. Periods of unemployment, though likely detrimental to wellbeing, are commonly experienced by people in this age group (Lee et al., 2019). This may partly obscure the effects of illness-related employment difficulties for young adults with SMI, with such effects emerging from early to middle adulthood.

Clinical implications

Increased social inclusion (e.g. improved social relationships and vocational engagement) is a common treatment goal among young adults with SMI (Cotton et al., 2011; Iyer, Mangala, Anitha, Thara, & Malla, 2011; Ramsay et al., 2011). Individual Placement and Support (IPS) is an intervention that has demonstrated efficacy in terms of increasing vocational engagement for young adults with SMI (Killackey et al., 2019). However, perceptions of social exclusion in terms of interpersonal relationships are an important and under-researched treatment area in this population (Lim & Gleeson, 2014). The findings of the present study suggest that interventions to improve social inclusion should target such perceptions in addition to increasing vocational engagement. Young adults with SMI were significantly more likely to perceive that they had no friends who would call on them in a crisis, with a large effect size. The smaller social networks observed in this population likely reduce opportunities for young adults with SMI to receive interpersonal support. However, they may also reduce opportunities for young adults with SMI to demonstrate their ability to provide support to others – a crucial element of the reciprocity involved in satisfying interpersonal relationships. An implication of this finding is that interventions to improve social inclusion in this population need to address peer attitudes, beliefs and behaviors toward young adults with SMI. That is, a whole-of-community approach may be most effective (Saavedra, Pérez, Crawford, & Arias, 2017).

Relative to those from the general community, young adults with SMI were more likely to live with their parents, with a large effect size. Although it may be argued that this provides necessary social support for some individuals, there is evidence that living with parents is negatively associated with social inclusion for young adults with SMI (Gardner, Cotton, O’Donoghue, et al., 2019). Social inclusion interventions may therefore need to help empower young adults with SMI to establish an optimal level of autonomy/independence while ensuring access to support. Young adults with SMI were also more likely to report unstable accommodation, with a large effect size. This suggests a role for housing support services in any interventions aimed at promoting adaptive autonomy/independence in this population.

These findings have several implications regarding how early intervention services may best respond to the psychosocial needs of their clients. It may be helpful for clinicians to ask whether young adults with SMI feel they have opportunities to demonstrate their capacity to be a source of support to others. This may be an overlooked element of social inclusion compared to the more common enquiries about available support from others. Related to this are notions of autonomy and independence. It may be helpful for clinicians to assess how satisfied young adults with SMI are with their level of independence (particularly as it relates to living situation), while being sensitive to individual support needs. Such approaches may be combined with evidence-based vocational programs (e.g. IPS) to improve social inclusion for young adults with SMI.

Strengths and limitations

It is plausible that mental illness causes social exclusion (the ‘social drift’ hypothesis), and plausible that social exclusion causes mental illness (the ‘social causation’ hypothesis (O’Donoghue, Roche, & Lane, 2016)). A limitation of this study is that causality cannot be inferred from these cross-sectional, correlational findings. Given that the F-SIM is administered entirely via self-report, none of its items can be considered strictly objective. This is a further limitation. However, the F-SIM has been developed through consultation with experts, carers and consumers (Filia et al., 2019). Also, its psychometric properties have been examined specifically among young adults with SMI (Gardner, Cotton, O’Donoghue, et al., 2019). These are strengths of the study. Another strength is the item-level examination of social inclusion variables, which afforded a sufficiently granular level of detail so as to be clinically informative. The novelty of the finding that young adults with SMI were less likely to experience reciprocity in interpersonal relationships is another strength. It is important to note that the 95%CI around the OR for this effect was relatively wide, and its lower bound estimate was approaching zero. This highlights the need for replication of the effect, and is a potential limitation.

Future research directions

It will be helpful for future research to further examine actual and perceived reciprocity in interpersonal relationships (e.g. do members of the wider peer group around young adults with SMI note the proposed discrepancy in reciprocity?). Exploring additional methods for measuring reciprocity may help increase the precision of the proposed effect. There is no ‘gold standard’ measure of social inclusion – existing tools require ongoing development (Cordier, Milbourn, Martin, Buchanan, & Chung, 2017). This includes the F-SIM, which is one of the first measures to examine social inclusion specifically among young adults with SMI (Gardner, Cotton, O’Donoghue, et al., 2019). Although it has demonstrated reliability and validity in this population, the F-SIM requires further refinement to become clinically useful (e.g. the production of a shortened version). It would also be helpful to examine potential mediators of the relationship between group membership (general community vs SMI) and social inclusion among young adults. There is evidence that self-stigma predicts reduced social inclusion for young adults with SMI (Berry & Greenwood, 2018). Further examination of such potentially mediating variables may enable the development of more effectively targeted interventions.

Conclusion

Young adults with SMI are considered socially excluded relative to peers from the general community, but this has not been examined via psychometric measures of social inclusion. We undertook such an examination and found evidence of social exclusion in terms of interpersonal connections, vocational engagement and independence. Improving the social inclusion of young adults with SMI may involve addressing peer attitudes and maximizing independence, in addition to the more established vocational interventions.

Footnotes

Acknowledgements

This article was completed while the corresponding author was receiving a Research Training Program scholarship from the Australian Government Department of Education and Training.

Conflict of interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

ORCID iDs

Andrew Gardner

Sue Cotton

Kate Filia

References

Australian Bureau of Statistics. (2015). Mental health statistics, 2015. Retrieved from http://www.abs.gov.au/ausstats/abs@.nsf/mf/4330.0

Australian Mental Health Outcomes Classification Network. (2016). Living in the Community Questionnaire (LCQ): A measure of social participation. Retrieved from http://www.amhocn.org/special-projects/living-community-questionnaire-lcq

Baumgartner

J. N.

Burns

J. K.

(2014). Measuring social inclusion – A key outcome in global mental health. International Journal of Epidemiology, 43, 354–364. doi:10.1093/ije/dyt224

Berry

Greenwood

(2018). Direct and indirect associations between dysfunctional attitudes, self-stigma, hopefulness and social inclusion in young people experiencing psychosis. Schizophrenia Research, 193, 197–203. doi:10.1016/j.schres.2017.06.037

Blakemore

S.-J.

Mills

K. L.

(2014). Is adolescence a sensitive period for sociocultural processing? Annual Review of Psychology, 65, 187–207. doi:10.1146/annurev-psych-010213-115202

Breitborde

N. J. K.

Woolverton

Frost

R. B.

Kiewel

N. A.

(2014). Self-determination theory and first-episode psychosis: A replication. Advances in Psychiatry, 2014, Article 240125. doi:10.1155/2014/240125

Coombs

Reed

Rosen

(2016). Developing the Living in the Community Questionnaire: Reporting the social outcomes of mental health care. Asia Pacific Journal of Social Work & Development, 26, 178–189. doi:10.1080/02185385.2016.1218359

Cordier

Milbourn

Martin

Buchanan

Chung

(2017). A systematic review evaluating the psychometric properties of measures of social inclusion. PLoS ONE, 12(6), e0179109.

Cotton

S. M.

Luxmoore

Woodhead

Albiston

D. D.

Gleeson

J. F. M.

McGorry

P. D.

(2011). Group programmes in early intervention services. Early Intervention in Psychiatry, 5, 259–266. doi:10.1111/j.1751-7893.2011.00277.x

10.

Evans-Lacko

Courtin

Fiorillo

Knapp

Luciano

Park

A. L.

. . . Thornicroft

(2014). The state of the art in European research on reducing social exclusion and stigma related to mental health: A systematic mapping of the literature. European Psychiatry, 29, 381–389. doi:10.1016/j.eurpsy.2014.02.007

11.

Filia

K. M.

Jackson

H. J.

Cotton

S. M.

Gardner

Killackey

E. J.

(2018). What is social inclusion? A thematic analysis of professional opinion. Psychiatric Rehabilitation Journal, 41, 183–195. doi:10.1037/prj0000304

12.

Filia

K. M.

Jackson

H. J.

Cotton

S. M.

Killackey

(2019). Understanding what it means to be socially included for people with a lived experience of mental illness. International Journal of Social Psychiatry, 65, 413–424. doi:10.1177/0020764019852657

13.

Gardner

Cotton

Killackey

O’Donoghue

Norton

P. J.

Filia

(2019). Preliminary psychometric properties of a measure of social inclusion among young adults aged 18 to 25 with serious mental illness. Psychiatric Rehabilitation Journal. Advance online publication. doi:10.1037/prj0000382

14.

Gardner

Cotton

Killackey

Filia

(2019). Preliminary psychometric properties of a measure of social inclusion among young adults aged 18 to 25 from the general community. Manuscript under review.

15.

Gardner

Cotton

S. M.

Allott

Filia

K. M.

Hester

Killackey

(2019). Social inclusion and its interrelationships with social cognition and social functioning in first-episode psychosis. Early Intervention in Psychiatry, 13, 477–487. doi:10.1111/eip.12507

16.

Gardner

Filia

Killackey

Cotton

(2019). The social inclusion of young people with serious mental illness : A narrative review of the literature and suggested future directions. Australian & New Zealand Journal of Psychiatry, 53(1), 15–26. doi:10.1177/0004867418804065

17.

Gayer-Anderson

Morgan

(2013). Social networks, support and early psychosis: A systematic review. Epidemiology and Psychiatric Sciences, 22, 131–146. doi:10.1017/S2045796012000406

18.

Gustavson

Knudsen

A. K.

Nesvåg

Knudsen

G. P.

Vollset

S. E.

Reichborn-Kjennerud

(2018). Prevalence and stability of mental disorders among young adults: Findings from a longitudinal study. BMC Psychiatry, 18(1), Article 65. doi:10.1186/s12888-018-1647-5

19.

Hartmann

J. A.

Nelson

Spooner

Paul Amminger

Chanen

Davey

C. G.

. . . McGorry

P. D.

(2019). Broad clinical high-risk mental state (CHARMS): Methodology of a cohort study validating criteria for pluripotent risk. Early Intervention in Psychiatry, 13, 379–386. doi:10.1111/eip.12483

20.

Hayes

Gray

Edwards

(2008). Social inclusion: Origins, concepts and key themes. Australia: Social Inclusion Unit, Department of the Prime Minister and Cabinet. Retrieved from https://www.voced.edu.au/content/ngv:741

21.

Holt-Lunstad

Robles

T. F.

Sbarra

D. A.

(2017). Advancing social connection as a public health priority in the United States. American Psychologist, 72, 517–530. doi:10.1037/amp0000103

22.

Huxley

Evans

Madge

Webber

Burchardt

McDaid

Knapp

(2012). Development of a social inclusion index to capture subjective and objective life domains (Phase II): Psychometric development study. Health Technology Assessment, 16, 1–248. doi:10.3310/hta16010

23.

Huxley

Evans

Munroe

(2006). Development of a social inclusion index to capture subjective and objective domains (Phase I): Final report to the NCCRM. Birmingham, UK. Retrieved from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.469.5761&rep=rep1&type=pdf

24.

Hyde

C. L.

Nagle

M. W.

Tian

Chen

Paciga

S. A.

Wendland

J. R.

. . . Winslow

A. R.

(2016). Identification of 15 genetic loci associated with risk of major depression in individuals of European descent. Nature Genetics, 48, 1031–1036. doi:10.1038/ng.3623

25.

Iyer

Mustafa

Gariépy

Shah

Joober

Lepage

Malla

(2018). A NEET distinction: Youths not in employment, education or training follow different pathways to illness and care in psychosis. Social Psychiatry and Psychiatric Epidemiology, 53, 1401–1411. doi:10.1007/s00127-018-1565-3

26.

Iyer

S. N.

Mangala

Anitha

Thara

Malla

A. K.

(2011). An examination of patient-identified goals for treatment in a first-episode programme in Chennai, India. Early Intervention in Psychiatry, 5, 360–365. doi:10.1111/j.1751-7893.2011.00289.x

27.

Killackey

Allott

Jackson

H. J.

Scutella

Tseng

Y.-P.

Borland

. . . Cotton

S. M.

(2019). Individual placement and support for vocational recovery in first-episode psychosis: Randomised controlled trial. The British Journal of Psychiatry, 214, 76–82. doi:10.1192/bjp.2018.191

28.

Lee

J. O.

Jones

T. M.

Yoon

Hackman

D. A.

Yoo

J. P.

Kosterman

(2019). Young adult unemployment and later depression and anxiety: Does childhood neighborhood matter? Journal of Youth and Adolescence, 48, 30–42.

29.

Lim

M. H.

Gleeson

J. F.

(2014). Social connectedness across the psychosis spectrum: Current issues and future directions for interventions in loneliness. Frontiers in Psychiatry, 5, Article 154. doi:10.3389/fpsyt.2014.00154

30.

Lim

M. H.

Gleeson

J. F. M.

Alvarez-Jimenez

Penn

D. L.

(2018). Loneliness in psychosis: A systematic review. Social Psychiatry and Psychiatric Epidemiology, 53, 221–238. doi:10.1007/s00127-018-1482-5

31.

McGorry

Bates

Birchwood

Kessler

Berglund

Demler

. . . Nordentoft

(2013). Designing youth mental health services for the 21st century: Examples from Australia, Ireland and the UK. The British Journal of Psychiatry, 54(s54), s30–s35. doi:10.1192/bjp.bp.112.119214

32.

McHugh

M. L.

(2013). The chi-square test of independence. Biochemia Medica, 23, 143–149. doi:10.11613/BM.2013.018

33.

Menezes

N. M.

Milovan

(2000). First-episode psychosis: A comparative review of diagnostic evolution and predictive variables in adolescents versus adults. The Canadian Journal of Psychiatry, 45, 710–716. doi:10.1177/070674370004500803

34.

Morgan

Burns

Fitzpatrick

Pinfold

Priebe

(2007). Social exclusion and mental health: Conceptual and methodological review. The British Journal of Psychiatry, 191, 477–483.

35.

Narendorf

S. C.

Cross

M. B.

Santa Maria

Swank

P. R.

Bordnick

P. S.

(2017). Relations between mental health diagnoses, mental health treatment, and substance use in homeless youth. Drug and Alcohol Dependence, 175, 1–8.

36.

O’Donoghue

Roche

Lane

(2016). Neighbourhood level social deprivation and the risk of psychotic disorders: A systematic review. Social Psychiatry and Psychiatric Epidemiology, 51, 941–950. doi:10.1007/s00127-016-1233-4

37.

Ramsay

C. E.

Broussard

Goulding

S. M.

Cristofaro

Hall

Kaslow

N. J.

. . . Compton

M. T.

(2011). Life and treatment goals of individuals hospitalized for first-episode nonaffective psychosis. Psychiatry Research, 189, 344–348. doi:10.1016/j.psychres.2011.05.039

38.

Ramsay

C. E.

Stewart

Compton

M. T.

(2012). Unemployment among patients with newly diagnosed first-episode psychosis: Prevalence and clinical correlates in a US sample. Social Psychiatry and Psychiatric Epidemiology, 47, 797–803. doi:10.1007/s00127-011-0386-4

39.

Robustelli

B. L.

Newberry

R. E.

Whisman

M. A.

Mittal

V. A.

(2017). Social relationships in young adults at ultra high risk for psychosis. Psychiatry Research, 247, 345–351.

40.

Roy

Rousseau

Fortier

Mottard

J. P.

(2016). Postsecondary academic achievement and first-episode psychosis: A mixed-methods study. Canadian Journal of Occupational Therapy, 83, 42–52.

41.

Saavedra

Pérez

Crawford

Arias

(2018). Recovery and creative practices in people with severe mental illness: Evaluating well-being and social inclusion. Disability and Rehabilitation, 40, 905–911. doi:10.1080/09638288.2017.1278797

42.

Shepherd

Parsonage

(2011). Measuring the costs and benefits of promoting social inclusion. Mental Health and Social Inclusion, 15, 165–174. doi:10.1108/20428301111186804

43.

Social Exclusion Unit. (2005). Transitions: Young adults with complex needs. London, England: Author.

44.

Spandler

(2007). From social exclusion to inclusion? A critique of the inclusion imperative in mental health. Medical Sociology Online, 2(2), 3–16.

45.

Sullivan

G. M.

Feinn

(2012). Using effect size-or why the p value is not enough. Journal of Graduate Medical Education, 4, 279–282. doi:10.4300/JGME-D-12-00156.1

46.

Sündermann

Onwumere

Bebbington

Kuipers

(2013). Social networks and support in early psychosis: Potential mechanisms. Epidemiology and Psychiatric Sciences, 22, 147–150. doi:10.1017/S2045796012000601

47.

UNESCO. (2017). What do we mean by ‘youth’? Retrieved from https://en.unesco.org/youth

48.

Vittinghoff

McCulloch

C. E.

(2007). Relaxing the rule of ten events per variable in logistic and Cox regression. American Journal of Epidemiology, 165, 710–718. doi:10.1093/aje/kwk052