How do youth with emotional and substance use problems fare in the juvenile justice system? Alison Evans Cuellar, PhD Mailman School of Public Health Columbia University New York, New York Pinka Chatterji, PhD Center for Multicultural Mental Health Research Cambridge Health Alliance – Harvard Medical School Somerville, MA Objective To use a national sample of adolescents to investigate the association between emotional and substance use problems during adolescence and involvement with the juvenile justice system Focus on three types of involvement: Arrests before age 18 Convictions in juvenile court before age 18 (among those arrested before age 18) Detainment before age 18 (among those convicted before age 18) Motivation Ample evidence that youth with emotional and substance use disorders are over-represented in the juvenile justice system Can be explained by: Youth with emotional and substance use problems are more likely to commit crimes or more likely to commit more serious crimes Youth with emotional and substance use problems are more likely to be sanctioned conditional on committing a particular crime Public policy concern that the justice system is biased against youth with these problems – but there have been few systematic empirical tests of this claim Background Adults with mental disorders are more likely to be arrested and serve longer prison sentences than adults without mental illness (Teplin, 1984; Ditton, 1999) High prevalence of emotional disorders (other than conduct disorder) among youth in the juvenile justice system relative to youth in the community In large, random sample of arrested and detained youth in Cook County, IL (Teplin et al, 2002): 61% of males and 70% of females met DSM diagnostic criteria for any mental disorder other than conduct disorder 17% of males and 21% of females met criteria for ADHD 19% of males and 28% of females met criteria for any affective disorder 51% of males and 47% of females met criteria for substance use disorder Contribution of this Study Previous work generally is based on small and specialized samples – typically, detained or incarcerated youth – which represent a relatively small sub-set of all youth involved in the justice system Past studies generally do not control for demographic characteristics and do not have a community control group Use of the National Longitudinal Study of Adolescent Health (Add Health) builds on work based on more specialized and smaller samples Control for many potential correlates of emotional/substance use problems and juvenile justice involvement Longitudinal data Capture youth with sub-threshold disorders and earlier or more limited involvement with the juvenile justice system Methods J = b0 + b1 MH + b2 X + u + e J: Juvenile justice outcome MH: Mental health measure X: Observed personal characteristics u: Unobserved personal characteristics Primarily interested in estimating b1 Standard estimation methods such as OLS can lead to biased estimates: if a problem of reverse causality exists (e.g. juvenile justice outcome affects mental health) – although problem is limited by measurement of MH at two points in time during adolescence if unobserved characteristics (u) exist that influence both mental health and juvenile justice outcomes (e.g. u is correlated with J and MH) Because of data limitations, we cannot address both of these issues directly. We address the problem of unobserved characteristics by using a rich data source to proxy u to the extent possible. Outcome Measures All youth: Among those with 1+ arrest before age 18: Whether or not youth was arrested prior to age 18 Number of arrests prior to age 18 Whether or not youth was convicted in juvenile court prior to age 18 Among youth with 1+ conviction before age 18: Whether or not youth sentenced to a detention center (as opposed to probation or other punishment) Data The National Longitudinal Study of Adolescent Health (Add Health) – nationally representative sample of adolescents designed to study causes of health-related behaviors and young adult outcomes Waves 1, 2 and 3: Wave 1, In-Home Sample: 20,745 students in grades 7-12 in 1995 Wave 2: About 15,000 students from Wave 1 In-Home sample interviewed about one year later, in 1996 Wave 3: More than 15,000 young adults aged 18-26 interviewed in 2001-2002 Main analysis sample includes students interviewed in all three waves with information on all dependent and independent variables of interest Measures of Emotional and Substance Use Problems Elevated ADHD symptoms: Chronic elevated depressive symptoms: Binary indicator of whether or not youth’s score on the Retrospective Attention Deficit Hyperactivity Scale was at the 75th percentile or higher at Wave 3 assessment Binary indicator of whether or not youth’s CES-D score was at the 75th percentile or higher in both Waves 1 and 2 Chronic drinking: Binary indicator of whether youth in both Waves 1 and 2 reports number of alcoholic drinks in the past 12 months at the 75th percentile or higher Other Independent Variables Exogenous set: Age, race/ethnicity, gender, family structure, and proxies for family income (maternal education and welfare receipt) All measured at Wave 1 Potentially endogenous set: Delinquent and violent acts in past year Grade point average and school absences with no excuse Measured in Waves 1 and 2 Dummy variables indicating whether youth’s delinquency scale score was medium, medium high or high relative to the sample Dummy variables indicating violent behavior, theft, and drug sales Measured at Wave 1 Average GPA in Math and English Highest quartile of school absences Level of parental involvement in adolescent’s life at Wave 1 At least one parent actively engaged in activities with youth in past four weeks PRELIMINARY RESULTS Table 1: Means Full sample (N = 9,434) Male 47% Age 16.06 (1.59) African-American 22% Latino 15% Asian 8% Native American 3% Mother is high school dropout 16% Mother is welfare recipient 9% GPA in Math and English 2.21 (0.84) Table 2: Outcome Variables by Emotional/Substance Use Problems Status ADHD No ADHD Depression No Depression Arrest before age 18 7.1% 3.3%*** 5.3% 3.9%*** Number of arrests .20 .07*** .14 .09** Conviction (among those arrested) .47 .41 .47 .42 Detention (among those convicted) .19 .24 .25 .19 Table 2 cont’ Chronic Drinker Not chronic drinker Arrest before age 18 9.9% 3.8%*** Number of arrests .27 .09*** Conviction (among those arrested) .42 .44 Detention (among those convicted) .24 .21 Table 3: Effect of Emotional/Substance Use Problems on Arrests Before Age 18 Arrested before age 18 Number of arrests before age 18 Logistic Model Odds Ratio (p-value) Poisson Model Coefficient (p-value) ADHD symptoms 1.33 (0.010) n = 9,434 0.354 (0.000) n = 9,497 Chronic depressive symptoms 1.22 (0.098) n = 9,587 0.168 (0.014) n = 9,653 Chronic drinking 1.30 (0.104) n = 9,630 0.146 (0.098) n = 9,697 Table 4: Effect of Emotional/Substance Use Problems on Convictions and Detention Before Age 18 Among those arrested Among those convicted before age 18, before age 18, convicted before age 18 sentenced to detention before age 18 Logistic Model Odds Ratio (p-value) ADHD symptoms 1.00 (0.995) n = 413 0.910 (0.844) n = 172 Chronic depressive symptoms 0.998 (0.927) n = 422 1.34 (0.582) n = 178 Chronic drinking 1.03 (0.934) n = 423 0.957 (0.946) n = 178 Summary of Results Elevated ADHD symptoms, chronic depressive symptoms, and chronic drinking during adolescence are associated with statistically significant and meaningful increases in the probability of arrest, and number of arrests before age 18 Models control for most likely correlates of emotional/substance use problems and juvenile justice outcomes No apparent effects on the probability of conviction and a sentence of detention in juvenile court – however, these analyses may be under-powered Limitations Self-reported, retrospective data Cannot confirm juvenile justice data with administrative information Emotional and alcohol problem information is not collected with a diagnostic instrument – cutoffs are somewhat arbitrary Possibility of unobserved factors that are correlated with both emotional/substance use problems and juvenile justice involvement Work in progress Propensity score matching methods Run arrest models by gender Control for neighborhood characteristics Test robustness of results to alternative thresholds for ADHD, depression and alcohol use Consider effects of marijuana use, and effects of youth in treatment for emotional and substance use problems Conclusions Using a large sample of adolescents, preliminary findings show that youth with elevated symptoms of ADHD, chronic depressive symptoms and chronic drinking do fare worse in the juvenile justice system in terms of the probability of being arrested Identify a group of youth whose health needs are not being appropriately recognized and who face a greater likelihood of being punished possibly as a result of their disorder Implications for outreach and education programs in the justice system, as well as the delivery system of public mental health services