The Journal of Behavioral Health Services & Research

, Volume 37, Issue 3, pp 307–321

Service System Involvement and Delinquent Offending at System of Care Entry

Authors

    • Research Triangle Institute
  • Dannia Southerland
    • Duke University Medical Center
  • Sarah A. Mustillo
    • Purdue University
  • Barbara J. Burns
    • Duke University Medical Center
Article

DOI: 10.1007/s11414-009-9179-x

Cite this article as:
Stambaugh, L.F., Southerland, D., Mustillo, S.A. et al. J Behav Health Serv Res (2010) 37: 307. doi:10.1007/s11414-009-9179-x
  • 62 Views

Abstract

This study examines correlates of delinquent offending among 2,554 youths presenting to community-based treatment in Center for Mental Health Services-funded systems of care. Variables in five key domains, including demographics, family risk, child clinical risk, school, and service system involvement, were examined in relation to severity of offending at treatment entry for early/middle (11–15 years) versus late (16–18 years) adolescents. Significant correlates of offending severity were identified in all domains except family risk for the 11 to 15 year olds and in demographics and child clinical risk for the 16 to 18 year olds. Service system involvement was a unique correlate of delinquency in the younger group. Findings add to other studies showing that correlates of offending appear to differ across stages of adolescence; specifically, school and service system involvement may be less important for older adolescents than for younger adolescents. Service system involvement should be included in longitudinal studies of risk for adolescent offending.

Keywords

disruptive behavior disordersadolescentrisk profilesjuvenile offending

Introduction

Youth at risk for behavior disorders are also at risk for early substance use, school failure, and increasing violence and arrest throughout adolescence.1,2 Behavior disorders are characterized by actions that, in many instances, constitute delinquency, or criminal acts committed by juveniles. Specifically, behavior disorders are marked by defiance of authority, failure to follow rules, aggression that threatens harm toward others, property damage, lying, and stealing. There is evidence that such a pattern of delinquent acts in adolescents with or at risk for behavior disorders can have severe functional ramifications in early adulthood including homelessness, unemployment, and the perpetuation of criminal careers.3,4

The developmental psychopathology literature includes many longitudinal and epidemiological findings on risk for adolescent offending and behavioral disorders. Risks have been established in many domains, including the family, peers, school, neighborhood, and the individual child.5,6 Although it is generally accepted that risks operate cumulatively (more risk factors = higher levels of severity), the relative influence of different risk factors on severity of offending is not well researched.2 Furthermore, the relevance of known correlates of risk across different stages of child and adolescent development is similarly underresearched. Finally, involvement in the child service system as a corollary of risk is severely understudied.

The child service system serves a critical function for youth with behavior problems. These youth often exhibit mental health comorbidity; thus, their need for services is high. Across the major child serving sectors (Child Welfare, Education, Juvenile Justice, specialty and nonspecialty Mental Health), behavior disorders are in fact the most common reason for child mental health referrals in the USA.7 Youth with behavior problems who use services in these sectors often receive many services from multiple providers prior to adulthood.810 Despite this heavy use of services and somewhat controversial findings suggesting that some more restrictive services (for example, group homes) exacerbate behavior problems, involvement with the child service system is not well studied in relation to risk. The primary purpose of the current study is to examine service system involvement as a correlate of delinquent behaviors, in the face of other well-established correlates.

Established Correlates of Delinquency

Some demographic variables have been consistently associated with severity of behavior disorders in children and adolescents. For example, male gender and age are positively linked with risk for behavioral disorders in youth.3,6,1112 Clinical variables, such as history of physical or sexual abuse, substance abuse, and depression are also related to behavior problems, both prospectively and concurrently.1315 Associated variables within the family include parent history of antisocial behavior or criminal involvement1 as well as parenting practices and family cohesion.16 Family living situation has been linked more generally to mental health risk for children, adolescents, and transition-age youth, though not specifically to behavior disorders.1718

The school domain also includes several known correlates of risk for behavior disorders. School engagement is an often-studied risk and protective variable for development of behavior disorders and other child mental health disorders.1920 Poor academic performance and school dropout are known correlates of juvenile delinquency. In addition to these well-studied domains, other risk factors include association with negative peers21 and lack of neighborhood resources.22

Domains of Risk Across Key Adolescent Stages

Correlates of risk for behavior problems have been shown to vary according to stage of adolescence. Some of the known risk factors for early adolescents include male gender, low family socioeconomic status, association with deviant peers, comorbid mental health disorders, and prior exposure to violence or abuse.6 Risk factors for late adolescents further include young age at first arrest, dropping out of school, unemployment, homelessness, substance abuse, and low self-esteem.3,2325 Considering both the biological and psychosocial changes that occur from early to late adolescence, some risk domains may have more impact on younger versus older adolescents or vice versa. For example, the family domain may become less salient as adolescents slowly gain autonomy and start to spend more time with their peers than with their parents.26,27

One study has analyzed the relative contribution of different domains of risk for early versus late adolescents in a population-based (nonclinical) sample. Sameroff and colleagues examined family, peer, school, and neighborhood risk factors across key developmental transitions—into middle school and out of high school.28 Using both cross-sectional and longitudinal data, they found that each risk domain, with the exception of neighborhood setting, was significantly associated with conduct problems when entered into a simultaneous regression model. They further found that the school domain appeared to have more relevance for early adolescents than for late adolescents. The current study provides an opportunity to further examine age-related differences in correlates of offending in youth presenting for mental health treatment.

Service System Involvement as a Correlate of Delinquency

Past and/or current child service system involvement may also be associated with delinquency severity. Youth with behavior disorders are highly represented in systems of care,29 Child Welfare,30 Education,31 and Juvenile Justice.32 Research also suggests that youth with behavior problems are at higher risk for restrictive and out-of-home placements than their peers with emotional disorders.33,34 Number of movements between various out-of-home placements has been shown in recent literature to predict negative mental health outcomes in youth.35,36 Partially, due to the use of restrictive services for this population, the costs of treating youth with behavioral disorders are the highest of all child emotional and behavioral disorders.37

Discouragingly, youth with the most severe behavior disorders who are at highest risk for placement in restrictive settings appear to be particularly resistant to treatment.38,39 This likely underlies observations that these youth often enter a revolving door of treatment throughout adolescence while they are eligible for services in the child system. This high use of services could occur for several reasons. For example, youth with severe behavior problems may have trouble finding effective treatment and thus may need to continue trying new treatments. On the other hand, a somewhat controversial line of research makes the argument that youth with severe behavior problems who receive treatment in group-based settings experience iatrogenic effects through exposure to delinquent peers, making their symptoms worse and perpetuating the need for services.40 Although concerns about peer contagion are often cited in critiques of group treatments for troubled youths, the empirical evidence remains equivocal.41 Regardless of the exact nature of the relationship between the development of behavior problems over time and the use of restrictive service placements, what is clear is that the child service system often plays a large role in the lives of youth who exhibit delinquent behavior. However, whether the service system domain correlates with severity of delinquent offending above and beyond other known correlates such as those mentioned above is unknown.

Study Aims

Using data from a large multisite study of youth presenting to community-based mental health treatment, the current analyses aimed to describe (a) the severity of offending among adolescents at risk for behavior disorders at key developmental stages—early/middle (11–15 years old) and late (16–18 years old) adolescence—and (b) correlates of offending severity at treatment entry for these two age groups. Although the sample could have been split into three age groups representing early, middle, and late adolescence, one goal of the study was to highlight differences in late adolescence when the influence of various ecological factors (e.g., family, peers, school) is known to change. The sample was divided between early/middle and late adolescence by designating age 16 as the beginning of late adolescence. This is consistent with numerous other studies of adolescent normative and abnormal development.42,43 Likewise, setting the delineation between middle and late adolescence at age 16 also corresponds with several key milestones of the transition between childhood and adulthood. Specifically, youth become eligible for full-time employment in most states, Individualized Education Plans (IEP) begin to incorporate transition planning and schools are required to invite youth to their IEP meetings, non-IEP students begin precollege coursework and postgraduation planning, and youth become eligible for their driver’s license.

Correlates of risk included in the study were categorized into the following five domains: demographics, child clinical history and current functioning, family, school, and prior service system involvement. The school domain was particularly important given the findings of Sameroff et al. that school-related risk factors discriminated between early versus late adolescents.28 The study hypothesized that child service system involvement would be positively associated with severity of offending over and above other known domains of risk and that this would be confirmed for both early/middle and late adolescents.

Methods

Study procedures

Analyses were conducted using data from the National Evaluation of the Comprehensive Community Mental Health Services for Children and Their Families Program, funded by the Center for Mental Health Services (CMHS) to establish and evaluate systems of care throughout the USA for children with emotional and behavioral disorders and their families.44 Data were derived from baseline assessments conducted at system of care entry on 2,554 youths who enrolled in the program from 1999 through 2003. This included 22 sites covering both urban and rural areas over a wide geographical range. Each CMHS study site was required to obtain approval from their local institutional review board for their procedures to enroll families and to protect participants’ confidentiality. All families who were eligible for treatment were invited to participate in the National Evaluation.

Sample selection

Adolescents were included in the current study if (a) they were between ages 11 and 18 at study entry and (b) their T score on the Externalizing Scale of the Child Behavior Checklist (CBCL)45 fell in the borderline or clinical range at the baseline assessment. The Externalizing Scale includes two subscales—Delinquent Behavior and Aggressive Behavior. These scales have been shown to differentiate children with a Diagnostic and Statistical Manual of Mental Disorders-III disruptive behavior disorder from children with other diagnoses. This ensured that the study was focused on an at-risk sample.

Participants were excluded if they had comorbid autism (n = 43) or mental retardation (n = 60), as indicated on their intake report, or if they had incomplete data for computing severity of offending (n = 504). Results from chi-square tests and t tests showed that those excluded due to missing severity data did not differ from those with complete data with respect to gender, age, ethnicity, or household income. Those retained had significantly lower CBCL externalizing T scores at baseline (t = 2.73; df = 3,157; p = 0.006). This was not considered a limitation, however, because (a) the difference in mean score for the two groups was small (74 versus 75), (b) the mean score for both groups was well into the clinical range indicating enough severity to pursue the questions of the study, and (c) the range of scores in the retained group (from 64 to 97, standard deviation (SD) = 6.7) included enough variation to obtain valid results. Nevertheless, these data indicate the most severe cases may have been underrepresented in the current sample.

Variables and measures

Severity of offending

Severity of delinquent offending was computed as the sum of items on the Delinquency Survey, which asks youth about their commission of delinquent acts over the last 6 months. The scale includes 19 items ranging in severity from nonviolent acts (e.g., being rowdy in a public place) to violent acts that would be considered felonies in adult populations (e.g., stabbing someone with a knife). All items on the Delinquency Survey were rated by youth on a scale of 0, 1, or ≥2, in terms of the number of times committed in the last 6 months. As such, responses were coded as 0, 1, or 2. Total scores were then calculated as the sum of all item scores.

Preliminary descriptives and psychometrics were performed to get a sense of the types of offenses committed in the study sample. First, internal consistency reliability was computed and was revealed to be high (α = 0.97). Second, an exploratory factor analysis was performed using principal axis factoring, applying direct oblimin rotation. Five factors were initially extracted, accounting for 72% of the total scale variance (Kaiser–Meyer–Olkin = 0.85). Subsequent models were run extracting three and four factors. Although eigenvalues increased, none of these models explained more than half of the total scale variance. Because the structure matrix from the initially extracted five-factor model was highly interpretable, this solution was used to group items into three categories: violent offenses (seven items, all involving threat of bodily harm), stealing and vandalism (ten items, all involving threat to property), and minor infractions (two items: being rowdy in public and traffic violations). Two additional items, forced sex and having sex in exchange for money or goods, did not load on any of the retained factors. As such, these items were removed from the scale for all study analyses. Less than 5% of the sample endorsed each item. Thus, the final scale score used for the study analyses captured three overall types of offending: minor infractions, stealing and vandalism, and violent offenses.1

Correlates of offending severity

Predictor variables were grouped into five conceptual domains. These variables and the instruments used to measure them are described below. Descriptives on each variable for the two age groups are presented in Table 1. These descriptives are based on raw scores for each variable.
Table 1

Study variable descriptives across groups

 

11–15 years (n = 2,093)

16–18 years (n = 461)

Descriptives

M or %

SD

M or %

SD

Demographics

 Male gender

63%

61%

 White versus other race*

61%

54%

 Parent has > high school diploma

41%

45%

 Family income ≤ $20K**

44%

50%

Family risk

 Parent clinical risk

2.90

(1.72)

2.84

(1.69)

 Living with 2 parents

36%

38%

Child clinical risk

 Child clinical risk***

1.79

(1.48)

2.42

(1.61)

 Child functioning (current)

124.93

(40.53)

128.91

(44.43)

School risk

 Academic performance in last 6 months

3.23

(1.54)

3.37

(1.60)

 School sanctions in last 6 months**

1.43

(0.99)

1.24

(1.07)

Service system involvement

 Total services used*

5.70

(0.93)

5.80

(1.04)

 Total # placements***

1.52

(0.69)

1.71

(0.72)

Delinquency severity***

2.82

(4.28)

4.10

(5.23)

*p < 0.05; **p < 0.01; ***p < 0.001

Demographics

Four demographic variables were included in the model: gender, minority status, parent education, and household income. These data were collected from caregivers using the Descriptive Information Questionnaire (DIQ), a 39-item structured interview administered at baseline to all families participating in the National Evaluation.

Family risk

This domain included two variables, both measured by the DIQ: parent clinical risk and whether the youth was living with two parents. The first variable was computed as the sum of six possible risk factors reported by the child’s caregiver: history of domestic violence, history of mental illness in the biological family, biological parent ever convicted of a crime or placed in a psychiatric hospital, history of substance abuse in the biological family, or treatment for substance abuse in the biological family. Each item under parent clinical risk was rated dichotomously (yes/no) by the caregiver. This variable, computed as described, has been predictive in other studies using National Evaluation data.46

Presence of two parents in the household was included to capture an important element of social capital in the youth’s environment, which has previously been identified as a protective factor for at-risk children.47 Caregivers were asked at baseline to give the household configuration for the child. The variable was then computed as 1 = two parents/caregivers living in the household or 0 = one parent/caregiver living in the household.

Child clinical risk

This domain also included two variables: child clinical risk and current child functioning. Child clinical risk was computed as the sum of seven possible risk factors reported on the DIQ: history of physical abuse, history of sexual abuse, prior attempt to run away, prior suicide attempt, history of substance use, prior sexually abusive behavior, and prior psychiatric hospitalization. Caregivers rated each of these items as either yes or no. The variable, child clinical risk, was then computed as the sum of items endorsed by the caregiver. Like the family risk variable, child risk as computed here has been predictive in other studies using National Evaluation data.46,48

Current child functioning was assessed using the Child and Adolescent Functional Assessment Scale (CAFAS),49 a caregiver-report measure of the degree to which a child’s mental disorder impacts his/her everyday life, covering the past 3 months. Eight domains are assessed: school/work, home, community, behavior toward others, moods/emotions, self-harmful behavior, substance use, and thinking. The CAFAS has demonstrated adequate internal consistency reliability (0.63–0.68), high interrater reliability (0.92), convergent validity with CBCL total scores (0.42–0.49), and Child Assessment Schedule total scores (0.52–0.56).50,51

School risk

Two domains of school risk were measured: educational performance and disruptive behavior at school. These variables were computed using items from the Education Questionnaire (EQ) which includes 21 parent-report items that assess the child’s school attendance and performance over the previous 6 months. In addition, three items ask about sanctions incurred by the child at school. For the current study, school performance was assessed using one item from the EQ, on which parents rated the youth’s school performance ranging from 1 (failing all or most classes) to 6 (grade average of A). The second variable (behavior at school) was coded from three items which capture the degree of sanctions imposed on the child in the last 6 months: no sanctions = 0; detention = 1; suspension = 2; expulsion = 3.

Service system involvement

This domain included two variables. Services in the last 6 months, collected from caregivers on the DIQ, was computed as the sum of the following services: outpatient mental health treatment, school-based services (in relation to an emotional or behavioral problem), day treatment, residential treatment or inpatient psychiatric hospitalization, and alcohol-substance abuse therapy. Contact with juvenile justice was not included in this variable because of its likely high correlation with delinquency. Number of out-of-home placements was computed using the Restrictiveness of Living Environments and Placement Stability Scale (ROLES-R).52 The ROLES collects information on change in living environments covering 27 possible placement settings. For the current study, these 27 settings were grouped into the following categories: foster home, therapeutic foster home, group home or emergency shelter, residential treatment, hospital, and jail or detention. These two variables (total services and total number of placements), together, cover a wide range of service sectors as well as both restrictive and nonrestrictive service types. In addition, the inclusion of the ROLES gave an indication of the level of stability in each youth’s living situation.

Analytic approach

Following descriptive analyses of offending severity across age groups, a negative binomial model was run, entering predictor variables in blocks by the five study domains: demographics, family risk, child clinical risk, school risk, and service system involvement. All predictor variables were standardized and normally scaled. All variables were imputed with single imputation. Mean substitution was used for variables with less than 5% missing, and single regression imputation was used for variables with 5–13% missing. Only one variable (race) had missing values above 10%. Imputation models fit well and explained a substantial portion of the variance in all cases. The models assume that data are missing at random, rather than missing completely at random.

Results

Sample description

Table 1 shows the demographic characteristics of the study sample and descriptive statistics for all study variables, presented by age group. These data are all based on raw scores. Some group differences emerged. Youth in the older group (16–18) were more likely to be minority race and to live in households with family income below $20,000. They had higher clinical risk scores and used more services in the past 6 months than the younger age group (11–15). The finding that they endorsed more clinical risk factors was perhaps due to their age, i.e., they had more opportunity, by virtue of their age, to have experienced these risks.

These descriptives suggest that 16 to 18 year olds were from more socially disadvantaged households and had experienced more clinical risk during their lives. They also had higher rates of recent experience with the child service system. In contrast, 11 to 15 year olds had received more sanctions at school in the past 6 months. This was not due to lack of school enrollment in the older group. Frequency statistics revealed that 83% of youth in the older group were enrolled in school, compared with 94% of youth in the younger group. These percentages are large enough to detect significant effects. The level of family risk as measured in the study was not different for the two age groups.

Further analyses were conducted to examine the types of offending reported by adolescent study participants. As shown in Table 1, the mean severity score for the younger group was 2.82 (ranging from 0–32; SD = 4.27). For the older group, the mean severity score was 4.10 (ranging from 0–28; SD = 5.23). This means that 11 to 15 year olds committed approximately three delinquent acts in the past 6 months, compared with approximately four delinquent acts among 16 to 18 year olds. The difference in mean scores was significant between the two age groups (t = −4.91, p < 0.001). Figure 1 depicts the rates of offending within category for both age groups. For both age groups, most offenses fell in the category of stealing and vandalism. As shown, more youth in the older group reported offenses in the last 6 months across all three categories of offending. These differences were significant: minor infractions (t = −2.51, p < 0.01), stealing and vandalism (t = −5.57, p < 0.001), and violent offenses (t = −2.03, p < 0.05). The greatest contrast in severity of offending between groups was for stealing and vandalism, with 65% of 16–18 year olds reporting this type of offense compared to 50% of the 11–15 year olds.
https://static-content.springer.com/image/art%3A10.1007%2Fs11414-009-9179-x/MediaObjects/11414_2009_9179_Fig1_HTML.gif
Figure 1

Percent of youth reporting one or more offenses within each offending domain (raw scores)

Bivariate correlations between predictor variables and offending severity

Table 2 shows the bivariate correlations between all study variables. Delinquency scores were positively correlated with male gender and minority race and negatively correlated with household income. Living with two parents (versus living with one parent) was negatively correlated with delinquency, while family risk was positively correlated with delinquency. Both child clinical risk and CAFAS scores were positively correlated with delinquency, and these two variables were significantly related as would be expected. Delinquency was negatively related to school performance and positively related to school behavior, again as expected. Finally, both services variables were positively correlated with delinquency. Overall, significant correlations were identified between many independent variables, and each of these fell in the expected direction. These intercorrelations, nevertheless, implied that a goodness of fit statistic should be generated for the binomial model to address any collinearity.
Table 2

Bivariate correlations between study variables

 

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(1) Delinquency

1.00

            

(2) Female gender

−0.07*

1.00

           

(3) Minority race

0.07*

0.01

1.00

          

(4) Caregiver education

−0.02

0.02

−0.11*

1.00

         

(5) Household income

0.06*

−0.01

−0.19*

0.32*

1.00

        

(6) Living with 2 parents

−0.06*

−0.04

−0.20*

0.02

0.29*

1.00

       

(7) Family risk

0.05*

0.03

−0.09*

0.05*

−0.06*

−0.08*

1.00

      

(8) CAFAS total score

0.20*

−0.02

−0.03

0.02

0.04*

−0.03

0.10*

1.00

     

(9) Child clinical risk

0.16*

0.18*

−0.07*

0.11*

0.11*

−0.06*

0.23*

0.24*

1.00

    

(10) School performance

−0.12*

0.04*

0.06*

0.06*

0.02

−0.02

0.02

−0.12

0.03

1.00

   

(11) School sanctions

0.13*

−0.10*

0.09*

−0.04*

0.00

0.01

−0.01

0.16*

−0.04

−0.20*

1.00

  

(12) Total services

0.06*

−0.06*

−0.05*

0.10*

0.11*

−0.01

0.09*

0.18*

0.26*

0.09*

0.01

1.00

 

(13) Total placements

0.18*

0.07*

0.03

0.03

0.07*

−0.07*

0.05*

0.24*

0.24*

−0.02

−0.03

0.14*

1.00

*p < 0.05

Results from regression: correlates of offending severity

The negative binomial model is similar to the Poisson model, but is used in cases of overdispersed count data (e.g., when the variance exceeds the mean).53,54 To confirm that the negative binomial model was indeed the best fitting model, both deviance statistics from a Poisson model and the alpha parameters from the negative binomial models were examined. Exponentiated coefficients (incidence risk ratios, IRR) can be interpreted in terms of factor change in the expected count of delinquent offenses.

Results from the negative binomial model are shown in Table 3. Variables were entered in blocks by domain. Cox–Snell pseudo R2 is given to show the change in R2 when each block was entered. Because the model was log-linked, an IRR was generated for each variable. In Table 3, coefficients can be interpreted as percent change in delinquency scores associated with one-unit change in the independent variable. For example, under child clinical risk for the 11- to 15-year-old group, one point change in the child clinical risk variable was associated with a 13% increase in delinquency score. For total number of placements, each change in placement was associated with a nine percent increase in delinquency score.
Table 3

Results from binomial analysis of predictors of offending severity

 

11–15 years (n = 2,093)

16–18 years (n = 461)

IRR

SE

R2

IRR

SE

R2

Demographics

  

0.01

  

0.03

 Female gender

0.80**

0.06

 

0.61***

0.08

 

 Nonwhite race

1.24**

0.09

 

0.36*

0.17

 

 Parent education

1.00**

0.02

 

1.01

0.04

 

 Household income

1.05**

0.02

 

1.02

0.03

 

Family risk

  

0.02

  

0.04

 Parent clinical risk

1.03

0.02

 

1.01

0.04

 

 Living with 2 parents

0.90

0.07

 

0.82

0.11

 

Child clinical risk

  

0.07

  

0.09

 Child clinical risk

1.13***

0.01

 

1.15**

0.05

 

 Child functioning (current)

1.04***

0.03

 

1.04**

0.02

 

School risk

  

0.09

  

0.10

 Academic performance

0.93***

0.02

 

1.00

0.04

 

 School sanctions

1.21***

0.04

 

1.05

0.06

 

Service system involvement

  

0.10

  

0.10

 Total services used

1.00

0.04

 

0.93

0.06

 

 Total # placements

1.09***

0.02

 

1.05

0.04

 

IRR incident risk ratio, R2 Cox–Snell pseudo R2

*p < 0.05; **p < 0.01; ***p < 0.001

In the 11- to 15-year-old group, 2,093 observations were included. Results show that model fit improved with the addition of each block. Significant correlates of offending were identified in the following domains: demographics, child clinical risk, school risk, and service system involvement. Male gender, minority race, parent education, and household income were each positively associated with delinquency severity. Both child clinical risk and child functioning were highly correlated with delinquency, as were academic performance and the level of sanctions imposed by the child’s school in the last 6 months. Finally, total number of placements in the last 6 months was positively associated with delinquency severity. Follow-up analysis examined discrete change which calculates the amount of change in the dependent variable for prespecified changes in the independent variables. The expected count of delinquency increased by 0.2 for each change in placement. Going from the minimum to the maximum number of placements was associated with an increase of 3.2 points on the delinquency survey.

In the 16- to 18-year-old group, 461 observations were included. The total R2 increased with the addition of each block. Significant correlates were identified in the following domains: demographics and child clinical risk. Being female was associated with a 39% decrease in delinquency score. Contrary to findings in the younger group, minority race was negatively correlated with delinquency. Being minority race was associated with a 64% decrease in delinquency score. Both child clinical risk and child functioning were positively associated with delinquency scores. No significant correlates emerged under family risk, school risk, or service system involvement.

Discussion

This study examined correlates of delinquent offending for early/middle (11 to 15 years old) and late (16 to 18 years old) adolescents, in a sample of youth presenting for community-based mental health services. Descriptive results suggest that stealing and vandalism may be the most common offenses committed by adolescents at risk for behavior disorders. Regression results revealed two main findings: (1) Family risk, school risk, and service system involvement were correlated with delinquency in early/middle adolescents but not in late adolescents and (2) service system involvement was a unique correlate of offending in early/middle adolescents.

Before the results of the predictive models are discussed, it is worth noting the cross-sectional descriptive results of the study which provide new information on patterns of delinquent behavior for early/middle versus late adolescents. These findings suggest that stealing and vandalism may be the most common delinquent acts committed by youth at risk for behavior disorders. These were the most common offenses reported by both age groups. Both self-reported minor infractions and violent acts were slightly more common in the older age group.

Results from the binomial models suggest that correlates of offending severity likely differ across stages of adolescence. Generally, associations existed in more of the measured domains for younger adolescents. There are several possible implications, both methodologically and substantively. Methodologically, the study was cross-sectional which does not rule out cohort effects to account for differences between the two groups. Similarly, unmeasured differences between the two age groups may have contributed to the differences found. A longitudinal design is the only way to study true developmental differences, and future research in this area should capitalize on such designs whenever possible.

Substantively, the results suggest that (a) there may be more avenues for effective intervention in younger adolescents than in older adolescents or (b) the domains for effective intervention may differ conceptually (not just quantitatively) for older versus younger adolescents. The findings from the current study fit with other studies showing that risk domains differ across stages of adolescent development. Sameroff et al. found that the school domain, in particular, was an area of risk for early adolescents but not for late adolescents.28 The present results provide descriptive support to these findings in a treatment-seeking sample. For younger adolescents, academic performance and behavioral sanctions at school were associated with delinquency severity. However, neither school variable was associated with delinquency in late adolescents. This is in line with research showing that juvenile offenders are often disengaged with school, not only academically but also socially.55 For older adolescents, the current findings support Sameroff et al. in suggesting that the school domain is not a unique correlate of offending as adolescents begin transitioning to adulthood. As suggested above, however, these findings should be replicated in a longitudinal sample.

The study hypothesis was partially confirmed. Service system involvement was associated with offending severity, over and above other measured domains, for early/middle adolescents at system of care entry. Specifically, experiencing multiple out-of-home placements was positively related to high levels of offending. A correlation was not found, however, in 16 to 18 year olds. The link between number of out-of-home placements and severity has been found in other studies of youth with a range of mental health problems.35,36 These prior studies tend to lump children and adolescents into age groups much broader than those of the current study, making it difficult to detect age-related differences in this relationship. The direction of the link between restrictive services and delinquency is unclear, i.e., whether severity precedes subsequent placements or vice versa. It stands to reason in the current study that those with more severe problems may have received more attention from the service system, again given the specific clinical focus of the sample. In particular, younger adolescents with recent multiple out-of-home placements demonstrated the highest levels of offending severity. The need for research on the effects of restrictive and/or unstable placements on child mental health is sorely needed.56 Specifically, research is needed to clarify the role of out-of-home placements in the continuum of care for child mental health, especially for adolescents at risk for juvenile offending, given the likelihood of being placed in juvenile detention.

Surprisingly, the family risk domain was not significantly predictive of delinquency severity in either age group. This contrasts with prior literature positively linking family resources to youth behavioral adjustment.18,57 The lack of association between parent clinical history and youth offending severity was also surprising in light of research linking maternal mental illness (primarily depression) to child behavior problems.58 While other studies reporting this link have measured parent and child clinical functioning concurrently, the current study involved retrospective reports of cumulative events in the parent’s life, including but not limited to any past episode of mental illness. It is likely that any relation between parent mental illness and child behavior problems would be more pronounced when measured more proximally. Another possible explanation for the lack of findings was the fact that many adolescents in the study may not have been living with their biological parents, and as shown in the data, many experienced multiple living placements within the last 6 months.

Child clinical risk was highly predictive of offending severity in both age groups. Current functioning likely served as a proxy for offending, given the fact that child behavioral diagnoses are largely based on observance of delinquent behaviors. The child’s clinical history was also predictive. Current clinical functioning is almost always intimately related to the child’s history, both at the individual level (substance abuse, psychiatric hospitalization) and the interpersonal level (having been physically or sexually abused).

The primary strengths of the study were the large sample size, diverse sample, and inclusion of variables that cover many important domains in a child’s life. The study was limited by the lack of peer variables in the National Evaluation. Peers have been consistently linked to risk for adolescents and thus represent an important domain to study in relation to delinquency. Nonetheless, the findings are informative descriptively and by demonstrating the utility of studying multiple conceptual domains in relation to risk, in order to detect which domains may be more salient than others. Any analysis of youth risk is limited to the aspects of a child’s life that are immediately measurable.

Another aspect of the study that may be seen as a limitation relates to the level at which variables were measured. Many of the independent variables were based on parent retrospective report. This is a limitation in two respects: (1) Many youth in the study changed living placements during the time period that variables were measured, implying that caregivers may have had trouble reporting on child-level variables, and (2) even under the best circumstances (intact nuclear family), retrospective report is not as reliable as current report. In mental health research, however, use of retrospective report for adult symptom onset and use of services is not uncommon.59

No variables measured interpersonal interactions such as parenting practices, and all information were provided by caregivers, some retrospectively. The total variance explained in the binomial model was small (about 10%), perhaps reflecting this. Still, the findings make a significant contribution by demonstrating service system involvement as a unique correlate of delinquent offending in early and middle adolescents. Finally, no direct statistical comparisons were made between the coefficients of the two regression models. Thus, when findings are compared across the two age groups, these are not necessarily statistically significant differences.

Implications for Behavioral Health

This study reconfirms the need to address the intersection of juvenile offending and mental health needs of youth with externalizing behavior disorders at both the individual clinical and system level. While promising treatment strategies are evolving,60,61 how a system of care will effectively address the needs of this particularly at-risk population of youth remains an open question. There are two central tenets that must be adhered to in developing clinical or system level strategies for this at-risk population. The first is that the services must be predicated on an ecological approach to addressing the multiple domains of risk and protective factors of offending youth with mental health conditions.62 The second is that treatment should be evidence based.63 At present, it remains unclear whether service system involvement is a precursor or a consequence of delinquency. Additional, longitudinal research is needed to untangle this relationship. A next step is to examine the interplay between longitudinal patterns of offending and longitudinal patterns of multisector service use, in the context of known risk factors for delinquency.

The current findings show that service system involvement is a critical part of what some adolescents (namely those at early to middle stages) bring to the table when they show up for community-based treatment. Moreover, the more delinquent acts they have committed, the more services they have likely received. How systems of care figure this in to individualized service plans may be particularly challenging for these youths. Creative plans which include evidence-based practices along with auxiliary support services based in the community may be beneficial.

Footnotes
1

All study models were run separately using a weighted version of the Delinquency Survey score as the dependent variable. In this version, delinquency items were weighted by severity using the results of the factor analysis to determine weights. The results of the model did not differ from the results when the model was run with the sum Delinquency Survey score (not weighted). As such, the more simplified version of the analysis using total sum scores is presented here in order to preserve clarity in presentation of analyses and results.

 

Acknowledgments

This work was supported by subcontract # 35126-4S-549 from Macro International, Inc. to Duke University.

Copyright information

© National Council for Community Behavioral Healthcare 2009