Study 1
DuBois and colleagues (2018) used a quasi-experimental design to evaluate the Great Life Mentoring (GLM) program. The study sample consisted of 66 youths in the GLM program and 66 youths in the comparison group. This sample was drawn from a larger pool of 91 youths who participated in the GLM program (and who also received traditional mental health services) at any point over a period of approximately 15 years and a randomly selected comparison group of 400 youths who received mental health care (but did not participate in the GLM program) over the same period. All participants in the study received outpatient mental health services from Columbia River Mental Health Services, in Washington State. The groups were matched on a range of variables, including demographic characteristics (e.g., race/ethnicity, age, gender, family income); risk factors (e.g., involvement in the juvenile justice system, issues with substance abuse); and history of diagnosis, psychiatric hospitalization, and overall assessment of functioning on the Children’s Global Assessment of Functioning Scale.
In both groups, the average age was about 10 years, and 50 percent of the group was female. GLM group participants were white (81.5 percent), African American (9.3 percent), Hispanic (1.9 percent), and other (7.4 percent). Comparison group participants were white (76 percent), African American (8 percent), Hispanic (4 percent), and other (12 percent). Slightly less than half of both groups were of very low-income status, and the most common DSM-IV Axis I diagnosis was depressive disorder. Approximately two thirds of mentors of youth in the GLM program were female. All female youths were matched with female mentors, and some male youths were also matched with female mentors. The majority of mentors (85.8 percent) had an associate’s degree or higher, and approximately one third had a background in a helping profession or role (e.g., teaching, child care).
Participants’ general functioning was assessed using the Children’s Global Assessment Scale (CGAS), in which a clinician rates the youth’s most impaired level of general functioning for a 1-month period on a continuum of health/illness, with scores ranging from 100 to 1 (Shaffer et al.1983). Scores were categorized into deciles, with summary descriptions ranging from “Doing Very Well” (100–91) to “Extremely Impaired – so impaired that constant supervision is required for safety” (10–1). CGAS scores were obtained at intake and at varying intervals (not determined by the researchers) thereafter. Reasons for ending mental health treatment were obtained from discharge status codes in agency records. Reasons included planned ending, unplanned and client-initiated ending, or other (e.g., youth moved away from area, was transferred to another facility).
To evaluate effects of the GLM program on CGAS scores, a linear mixed model analysis was conducted. In this analysis, CGAS scores were a repeated measure predicted by group (i.e., GLM or non-GLM), time in days since intake, time in days since start of the mentoring relationship, interactions between group and each of the time variables, and selected covariates (gender, family income status, age, primary Axis I diagnosis, presence/absence of secondary Axis I diagnosis). The intercept, time-since-intake, and time-since-start of the mentoring relationship were modeled as random within-subject factors. For non-GLM youth, time-since-start of the mentoring relationship was coded relative to the start date of the mentoring relationship for the GLM youth. On average, each youth had approximately five CGAS scores. The interaction of time-since-start of mentoring relationship with group was of primary interest as it represented the extent to which the linear rate of change in CGAS changed differentially for GLM and non-GLM youths at the point in time when GLM youths were paired with their mentors. Differences between GLM and non-GLM youths on model-predicted CGAS scores were examined for up to 2 years following the initiation of mentoring. Effects of the program on reasons for ending mental health treatment were evaluated using a chi-square test of association. No subgroup analysis was conducted.
