Program description
The BBBS community-based mentoring programs [
19] provide youth with a one-to-one relationship with an adult volunteer mentor. For one year (minimum), mentors spend 2-4 h (average) each week with their mentee in recreational, skill, or career-oriented activities. Mentors attend a training session prior to being matched. To determine a match, caseworkers interview qualified families and mentors to assess common interests, preferences and mentor ability to meet youth needs. Caseworkers contact families and mentors at least monthly for six months then bi-monthly until 12 months then quarterly. Caseworkers may provide advice on match-related problems or information on organizational events. Enrolled youth are assigned to a waiting list until they are paired to a mentor.
Sample Selection and Recruitment:
Twenty agencies were recruited by the national BBBS office to participate. Most (80%) were in metropolitan centers and chosen based on a long history of operation, large annual caseloads, and culturally diverse clients. To qualify for the study, families were required to be new admissions, have passed the agency’s qualifying assessment, and youth had to be 6-17 years old. One youth was randomly selected to participate in families with more than one eligible youth. Parents/guardians had to have primary parenting responsibility for the youth. Across all agencies, 1279 families meeting the study eligibility criteria were approached by BBBS caseworkers, and 997 (78%) agreed to participate and completed a baseline assessment. Detailed study methodology is reported elsewhere [
18]. The study was approved by the Centre for Addiction and Mental Health Research Ethics Board.
The initial sample for the current study consisted of 859 youth with mentoring status information at the 18 month follow-up. Three-quarters (75%) had a previous or ongoing mentoring relationship; 46% for at least 12 months [
20], the minimum period of BBBS mentor commitment expected. Information on the mentoring status of youth not completing an 18 month follow-up (
n = 212/859) (scheduling difficulties 71%, drop out 29%) was obtained from earlier or later follow-ups for the missing 18 month values. Youth not completing 18 month follow-up were compared to completers on baseline demographics, personal characteristics, and environmental factors. Results revealed non-completers were older (
OR = 1.10,
p < .01), had more recent family moves (
OR = 1.08,
p < .05), and younger parents (
OR = 1.03,
p < .05).
To assess youth chronic health problems, parents were asked if their child currently had any long-term illness or medical condition. Parents replying yes were asked to list the specific conditions. Dominant conditions included breathing difficulties/asthma (55%), allergies (29%), speech, hearing/vision problems (9%) and neurological disorders (7%). Less common (< 5%) were cancer, heart disease, liver and kidney disease, digestive problems, migraines, autoimmune disorders, skin disease, diabetes, and arthritis/rheumatism. To assess youth activity limitations, parents were asked if their child had any long-term health problems or medical conditions that prevented/limited activities at school, play or other age appropriate activities (yes/no).
Among the sample of 859 youth, 285 had one or more chronic physical health problems in the period of time between baseline and 18 month follow-up. Excluded were youth with only mental health problems, learning difficulties or autism spectrum disorder. A total of 574 youth had no chronic physical health problem. Of the 285 youth with health problems, 94 had an activity limitation associated with their problem; 191 did not. Thirty-eight youth were reported to have an activity limitation but no health problem and were removed from the analysis. The remainder consisted of 536 youth with no health problem or activity limitation. The final sub-sample consisted of 821 youth with complete information on their mentoring status at 18 months and who satisfied our criteria for inclusion in one of the three health status categories. Youth missing mentoring status at the 18 month follow-up (n = 138) (997-859) and those with an activity limitation but no physical health problem (n = 38) did not differ significantly from the final sub-sample (821) on baseline demographics, personal characteristics, and environmental factors.
Measurement.
Dependent variables
Dependent variables (youth outcomes) included youth self-reports of social anxiety, depressed mood, and peer self-esteem. Social anxiety was assessed using the Social Anxiety Scale for Children-Revised (SASC-R) [
21] which has18 anxiety-related items and four fillers. Sub-dimensions include: fear of negative peer evaluations (SAD-FNE) (8 items), social avoidance and distress with new situations or unfamiliar peers (SAD-NEW) (6 items), and generalized social avoidance and distress (SAD-G) (4 items). Five response options range from “not at all” to “all the time”. Sub-scales have good internal consistency, stability, and divergent and discriminate validity [
22]. In this study, internal consistency of sub-scales at baseline and 18 months follow-up was as follows: SAD-FNE (
α = .90, .94); SAD-NEW (
α = .76, .81); and SAD-G (
α = .68, .72).
Depressed mood was assessed using 8 items from the Center for Epidemiologic Studies Depression Scale (CES-DC) [
23]. Four response options range from “not at all” to “a lot or all the time”. There is good internal consistency and retest stability and moderate support for concurrent validity [
22]. For this study, internal consistency at baseline and 18 month follow-up was
α = .76 and
α = .83 respectively.
Peer self-esteem was measured using a 6 item abbreviated version of the peer sub-scale of the HARE Self-Esteem Scale [
24]. Five response options range from “strongly agree” to “strongly disagree” (
αs = .63 and .73 at baseline and follow-up, respectively).
Independent variable
Youth mentoring status was defined as the presence or absence of a BBBS mentoring relationship (i.e., mentored versus not mentored) between the baseline assessment and18 month follow-up. Non-mentored youth were the comparison group.
Moderator
Youth health status served as the hypothesized moderator in the SEM models (see Analytic Method section).
Covariates
Study covariates were chosen based on previous theory and research on youth mentoring [
25]. Youth characteristics were: gender (1 = boys 0 = girls), age (continuous), living arrangements (two dummy-coded categories: living with a single biological parent only and living in other arrangements vs. a reference group of both biological parents), ethnic/racial minority status reported by the parent/guardian (1 = Aboriginal/First Nations/Metis/Inuit, African, Asian, and Hispanic Canadian 0 = all others), number of siblings at home (continuous), number of family moves (past five years) (continuous), and sought help from a mental health or social service professional in the past 12 months (1 = yes 0 = no). Parent/guardian characteristics included: age (continuous) and education (1 = < high school 0 = other) and family economic deprivation (count of: parent-reported gross annual household income < 20 K, government social assistance receipt, living in government subsidized dwelling). Other parent-reported covariates included: family functioning (
α = .86), using 13-item general functioning sub-scale of McMaster Family Assessment Device, [
26] parent depression (
α = .92), using 20-item CES-D, [
27] parent social anxiety (
α = .92), using 17-item Social Phobia Inventory (
α = .92), [
28] and neighbourhood problems (
α = .86), using 6 items from revised Simcha-Fagan Neighbourhood Questionnaire [
29].
Analytic method
Structural equation modeling (SEM) multiple groups analysis in M-Plus [
30] was used to examine the relationship between mentoring status and outcomes (social anxiety, depressed mood, peer self-esteem) across health status groups. At the 18 month follow-up, each outcome was specified as a latent endogenous construct defined by three or more indicators and simultaneously regressed across the health status groups on: 1) youth mentoring status; 2) baseline scores for the same outcome; and 3) all covariates measured as observed variables. Baseline latent constructs and covariates were allowed to co-vary with youth mentoring status since youth eventually paired to a mentor may be differ from those without a mentor. Standard errors for estimated model parameters were adjusted for the nested data structure using the M-Plus Complex command. Missing data on outcomes at baseline and follow-up (2% and 26% respectively) were handled using Full Information Maximum Likelihood [
31]. Goodness of fit was evaluated using the Chi-Square statistic and three standardized indices: Comparative Fit Index (CFI), Tucker-Lewis Index (TLI) and Root Mean Square Error of Approximation (RMSEA).
To reduce skewness on the latent construct indicators, construct items were parceled [
32] into four sets for depressed mood (
α = .78 and
α = .85) and three for self-esteem (
α = .61 and
α = .72). The three SASC-R sub-scales formed the indicators of the social anxiety latent construct (
α = .75).
To establish measurement invariance on the latent construct indicators, equality constraints were initially imposed on the factor loadings and intercepts across the three health status groups. Nested model comparisons (with and without constraints) based on chi-square difference values were conducted with non-significant values taken as evidence of measurement invariance. Minimal differences in model fit between constrained and unconstrained models revealed that the chosen indicators operated in a similar fashion across health status groups, and the decision was made to use the constrained models in subsequent analyses.