Study Design and Participants
A randomized controlled trial was conducted from March 2006 through June 2009. Newly homeless youth were recruited from community-based organizations (e.g., presentations at shelters or schools) and from direct recruitment (e.g., flyers, advertisements) in Los Angeles and San Bernardino counties. Eligibility criteria included having been away from home for at least two nights in the past six months, not being away for more than six months, and having the potential to return home. Additional eligibility criteria included no current abuse or neglect, no active psychosis, or no current substance intoxication; these additional criteria could not be determined prior to obtaining informed consent and were screened for in the baseline assessment.
Flyers and recruitment scripts indicated eligibility criteria for the study. Adolescents expressed interest in the study by calling phone numbers provided on recruitment materials or by approaching recruiters. Attempts were then made to obtain written informed consent from both the adolescent and parent to participate.*
The Institutional Review Board of the University of California, Los Angeles, approved the protocol for this study.
After both the adolescent and parent assented/consented, they completed baseline assessments. Eligible families were then randomly assigned to one of two treatment conditions, intervention or control. Families in the control condition received the “standard care” that they were receiving from the agencies that referred them. If they were not actively receiving any type of services, the families were given appropriate referrals based on their needs. Families in the intervention condition received the STRIVE (Support To Reunite, Involve and Value Each Other) intervention.
Separate teams were involved in the recruitment/assessment and intervention aspects of the study. Once a family provided consent and both baseline assessments, the recruitment/assessment team referred the family to the intervention team, which used a computerized coin toss to assign families to study condition, and provided the intervention to families randomized to receive it. The recruitment/assessment team continued to follow all families through June 2009, blinded as to study arm.
Following outreach efforts, 442 adolescents initially indicated interest in the study. Among these, 151 provided full consent from parent and child, completed both the parent and adolescent baseline assessments, and were randomized. Sixty-eight were assigned to receive the intervention; 83 were in the control condition. Parents included mothers/female guardians (76%), fathers/male guardians (10%), grandparents (6%) and other adults (7%).
The STRIVE intervention consisted of five sessions administered to the youth and parent(s) together by a trained facilitator at a site selected by the family (usually their home, where youth could be found despite runaway episodes). Five sessions were determined as the minimum number of sessions required for behavioral change and the maximum number of sessions anticipated to be replicable (24
). The session content was based upon cognitive-behavioral theories, designed to improve families’ problem-solving and conflict resolution skills (17
). Since the intervention was specifically designed for newly homeless youth, running away from home was framed as an ineffective attempt to resolve family conflicts. The conceptual frame underlying the intervention highlights the importance of establishing a positive family climate, improving family functioning by routinely resolving conflicts in a mutually beneficial manner, learning how to recognize and effectively manage feelings, increasing positive affirmations, learning and practicing problem solving skills, and providing role clarity (29
). As can be seen in , the sessions are based on a set of very highly interactive, semi-structured tasks involving repeated feedback and practice. New skills are learned based on the preceding sessions, and the skills learned in one session are constantly reinforced in the subsequent sessions. For example, nurturing a positive family climate is reinforced in all five sessions. The intervention includes the following tools: tokens to strengthen desired behaviors, a feeling thermometer to teach emotional regulation, a “think-feel-do” problem solving model to operationalize and tackle problems, role playing for safely practicing new skills, and reframing to conceptualize problems and solutions in a non-blaming manner.
Content of the STRIVE Intervention Sessions
An intervention manual was created to ensure fidelity. Intervention sessions usually lasted between 1½ and 2 hours, and were conducted once weekly. All sessions were audio-recorded and 20% were monitored for quality assurance. A 13-item rating instrument was developed to assess fidelity to session content and goals; overall 98% of tasks were implemented with fidelity. Attendance was high: 76% of families completed all 5 sessions; 6% attended 3–4 sessions; 16% attended 1–2 sessions; and only one family did not attend any sessions. In addition, satisfaction was assessed using the Working Alliance Inventory (WAI). The WAI is a valid and reliable instrument with scores ranging from 12 to 84 (33
). The median adolescent score was 78.0 (SD =12.5, mean = 72.0) and the median parent was 81.0 (SD = 10.0, mean = 76.4), indicating that satisfaction was very high.
A highly trained assessment team diverse in ethnicity and gender conducted computerized interviews. Audio computer-assisted self-interviewing (ACASI) was used for sensitive measures. Youth in the study completed assessments at baseline, and 3, 6 and 12 months after the baseline. All youth were paid for completing each assessment (e.g., $30, $35, and $40). Of the 151 randomized adolescents, 77% (116) completed at least one follow-up; 71% (107) completed a 3-month; 58% (87) completed a 6-month; and 46% (69) completed a 12-month assessment. On average, the 3-month assessment occurred 16 days after the last intervention session, but 21 cases had the 3-month assessment before the last intervention session.
Demographic data (age, gender, sexual orientation, race/ethnicity, and birthplace) were collected at baseline, as well as information about running away. Lifetime history of sexual- and drug-related HIV risk behaviors was also assessed at baseline, as was history of delinquent behavior.
Measures of recent risky behavior were collected at each assessment. “Recent” behaviors were defined as in the past 3 months. Sex was defined as vaginal or anal intercourse and measures of sexual behavior included: a dummy indicator for having sex, an indicator for having unprotected sex (without using a condom), number of sexual partners, and number of times had sex. Measures of recent risky behavior also included substance use variables, again reflecting behavior in the 3 months prior to each assessment (indicators for using alcohol, marijuana, and hard drugs; number of times used alcohol, marijuana, and hard drugs).
Delinquent behaviors were measured by a count of conduct problems engaged in during the prior 3 months, from a list of 13 possible behaviors, identified in the DSM-IV as being symptomatic of conduct disorder (34
). The list of behaviors is provided in ; examples include starting fights or destroying property. Runaway behavior was not included, as it was an inclusion criterion for the study.
Observed Characteristics of STRIVE Participants at Baseline, by Intervention Status*
Baseline characteristics of participants assigned to the intervention were compared to those of participants assigned to the control condition. Measures compared included demographics, runaway history, and baseline values of outcome measures. Those completing at least one follow-up assessment were compared to those lost to follow-up. For both sets of comparisons, t-tests and chi square tests were performed for continuous and categorical measures, respectively.
The impact of the STRIVE intervention on adolescents’ risk-taking and delinquent behaviors was evaluated by intent-to-treat random-intercept regression models. Outcome variables were the measures of recent risky behavior and delinquent behaviors. In these models, a random intercept was included for each youth, to account for repeated measures taken from the same individual. Model predictors include the intercept, gender, a time measure for each assessment, an intervention status indicator, and an interaction between time and intervention status. The interaction term captures the impact of the intervention over time (i.e., the relative change across assessments for those in the intervention condition as compared to control condition). To test for non-linear time trends in the outcome measures, the models were re-estimated with the addition of squared terms for the time variables. If the squared terms were significant, they were retained in the model; if non-significant, the more parsimonious model was used. We present results from significance tests of the time-intervention interaction terms, as well as observed and model-fitted mean values at each time point for youth in both arms of the study.