This study examines the effect of substance abuse and mental health services on behavioral and academic outcomes for Medicaid enrolled youth served by the Child and Family Support Team Initiative (CFST).  CFST places 100 nurse-social worker teams into 103 schools in 21 school districts identified as having disproportionate numbers of high risk youth, most of whom are likely eligible for Medicaid services.  During the 2006-2007 and 2007-2008 school years about 650 students were referred to the CFST program for suspected substance abuse.  CFST operates at the school level through a form of family group conferencing whereby the family and appropriate supports develop a coordinated strengths-based plan.  This planning coordinates the multiple services students may need, including substance abuse and mental health treatment as appropriate.

Through the existing evaluation, the Center for Child and Family Policy (CCFP) has merged education (NC-ERDC), juvenile justice and social service records for students throughout the state.  The education data include information on third through twelfth graders academic outcomes, disciplinary referrals and attendance.  Juvenile justice data include information arrests and length of stay in detention facilities.  Social services data include information suspected child abuse investigations, placement into foster care and length of time in foster care.

This study aims to examine whether students with substance use issues that are served by CFST are more likely to receive treatment.

This study aligns well with the Center for the Study of Adolescent Risk and Resilience (C-StARR)’s overriding goal to translate evidence from basic-science research on regulatory processes into novel research projects that prevent substance use and other conduct problems. It will improve our understanding of multiple neural pathways through which genetic and environmental variables bias the manifestation of substance use and antisocial behaviors in young adulthood. This study also promotes a novel integration of genetics, neuroimaging, and developmental psychopathology, which may provide new evidence at the interface of these disciplines. Furthermore, it contributes to the training of young prevention scholars by employing early career researchers such as Anne-Marie Iselin (a research scientist working with Ken Dodge) and Justin Carré (a postdoctoral associate with Ahmad Hariri) as junior investigators. Together we will publish manuscripts in the fall of 2011. This study will provide necessary feasibility data and preliminary results to support an R01 grant application to NIDA that  would assess Fast Track participants from other groups (i.e., females) and geographic sites (e.g., Seattle), as well as longer-term outcomes (e.g., adult substance dependence).

Several important steps have occurred to date.  First, the identifying information from the Department of Medical Assistance (DMA) on youth aged 8 through 21 was obtained from 2000-2009.  These data were linked to the North Carolina Education Research Data Center data.  A 67% match rate was obtained.  It is important to note that this match rate may underestimate the true percentage of youth who are students that actually were matched.  For example, we asked the DMA to include youth through age 21 because youth in special education may remain in secondary schools until this age.  However, youth typically leave school by the age of 18.  Also, students who are in private schools are expected not to match to the NC-ERDC.  This may be a small percentage of youth in the Medicaid population.  Second, in March 2010, the DMA sent information on services received by youth who matched in the Medicaid and education data.  This file contains information on claims submitted for drug treatment facilities, outpatient care, inpatient care, health check screening,  to name a few.  One challenge with working with this data set is that it is 18 gigabytes large and needs to be stored in an encrypted file on the server maintained by Duke.  This can cause logistic problems such as ensuring we have enough physical space to work with the data and the internet connection must be fast—otherwise it does slow progress considerably.  In addition, programs run slowly on a large data set so we are working to ensure programs are written efficiently—while differences in computational speed may be negligible on some data sets, care with programming is essential on this large dataset.  IRB approval for this study is covered by four existing approved protocols. The education data harmonized across years and datasets so as to reconcile differences in datasets prior to data analysis.

The next steps are as follows:  1) Currently the data are stored as multiple claims on the same day.  We need to collapse the data to create records by “episodes of care”, 2) Variables need to be created that translate the ICD-9 codes into the disease/health problems that we are interested in studying that include substance use and mental health conditions, and 3) The collapsed records need to be linked to data from the CFST case management system to determine the sample size to work with.

For researchers interested in working in this area and in collaboration on data analyses and manuscript preparation please contact the principal investigator Beth Gifford at Beth.gifford@duke.edu.