The Effects of TF-CBT and Wraparound Services on Disadvantaged Youth: Experimental Evidence

Youth Advocate Programs and Children's Home + Aid will receive referrals from Chicago Public Schools, the Cook County Juvenile Probation Department, and other agency and community partners that work with youth who are involved in the juvenile justice system and have disengaged from school. Youth who are randomly selected from an eligible population and agree to participate in the intervention will be assigned an advocate who will interact weekly with the youth in one-on-one and in family meetings and recreational groups. Additionally, the youth will participate in group or individual therapy that uses SPARCS (Structured Psychotherapy for Adolescents Responding to Chronic Stress), an evidence-based and trauma-focused model. The program is designed specifically for youth who are actively gang-affiliated, on juvenile probation, have been found guilty of weapons offenses, are seriously disruptive in school through chronic truancy, serious misconduct or frequent suspensions, or have been direct victims of or witnesses to traumatic violence. The program will be delivered in two sessions, each lasting five months and enrolling 110 youth at a time, with the goal of reducing criminal and violent behavior and improving academic achievement by helping youth stabilize and successfully participate in school, community, and family life. Because there is little definitive evidence on the effectiveness of intensive services programs for high-risk youth, identifying programs that are successful with this population is a key policy priority for the Mayor’s Office and other key stakeholders in Chicago.

2015-10-09

2016-07-31

The two year evaluation of the Children’s Home + Aid and Youth Advocate Programs trauma-informed youth advocate collaboration will seek to address the following questions:

Education Outcomes
1. Does the TF-CBT/Wraparound model increase connection to school for treatment youth?
2. Does the TF-CBT/Wraparound model reduce in-school disciplinary infractions for treatment youth?
3. Does the TF-CBT/Wraparound model improve GPA among treatment youth? Test scores?

Crime Outcomes
1. Does the TF-CBT/Wraparound model reduce out of school arrests for treatment youth? Non-violent vs. violent crimes? Victimization vs. perpetrator?

"Connection to school" will be measured by attendance, activity involvement (such as extracurricular activities), and the CPS "on track" indicator

Randomized Controlled Trial:
For each program cohort of 110, the youth will be randomly selected from a list of at least 220 eligible youth. The Crime Lab will compare the treatment group outcomes to the control group outcomes, looking specifically at arrests and victimization, connection to school or employment, and academic performance. Given the structure of the referrals and the background of our youth, we will stratify (or block) the sample to improve our treatment effect estimates. Randomization guarantees that our treatment and control groups will be similar in expectation, but stratifying helps ensure this actually occurs in practice with our sample of youth.

More detail on RCT:
The referral sources will provide the Crime Lab with a list of names, from which the Crime Lab will randomly select half to receive the program. The Crime Lab will share the list of selected youth with CHA and YAP, who will then contact and recruit the youth. This method ensures that no youth will be contacted with a potential offer of services that is then rescinded. There is a high need for these services, but because the program is intensive and expensive, CHA and YAP only have the capacity to provide the 5-month program to 110 youth at a time. This study design ensures a fair allocation of these limited services and does not restrict the control group from getting other services for which they would normally be eligible.

We will stratify based on referral source. Stratifying according to some variables will improve the precision of our estimates to the extent that the referral source and gang affiliation of a youth helps explain variation in our treatment (Cox and Reid, 2000). This is possible given that the level of risk in engaging in violent behavior or disconnecting from school will differ by referral source.


Social Network Analysis:
If the CHA/YAP intervention is successful at improving outcomes for high risk youth, we hypothesize that this will not only improve the lives of the youth that are directly touched by the program, but also have a ripple effect in their communities. We are adding a network analysis component to the planned evaluation in order to capture the diffusion or “spillover” of the intervention. Mapping entire social networks is labor intensive, but the “friendship paradox” can be used to quickly identify social links. Using this strategy, we will add two surveys to our planned evaluation. First, before service provision begins, youth in the study will be asked a series of network name-generating questions to identify their peers and associates. We will also use this survey to collect baseline data on youth exposure to violence and psychological health. At the end of the program, we will launch a second survey to gather data on outcomes for youth in the study as well as a randomly selected subset of members of the study youth’s previously identified social network. The outcome data collected through this survey would supplement data on academic and criminal behavior collected from the Chicago Public Schools and the Chicago Police Department.

Machine Learning:
Beyond the important policy question of the effectiveness of intensive support services, we believe this study could provide broader insight to the field in two key areas. First, little is known about how to effectively target programs to youth that are most likely to be perpetrators or victims of violence and those who are most likely to benefit from a high intensity intervention. By collecting information about how program referral agencies prioritize youth for services and examining which youth are most responsive to interventions, it is our hope that this study will provide guidance on how best to target high intensity interventions.

Randomization done by computer

Eligible youth were randomized at the individual level and stratified by referral source.

No

880 youth

880 youth

440 treatment youth, 440 control youth