A Multiyear Follow-Up Study Examining the Effectiveness of a Cognitive Behavioral Group Therapy Program on the Recidivism of Juveniles on Probation

Abstract

The present study evaluated the long-term effectiveness of a cognitive behavioral group therapy program titled Community Opportunity Growth. This study monitored juvenile delinquents’ recidivism across a 7-year time period, with the average length to follow-up being 39 months. It was hypothesized that program graduates (N = 178) would have a significantly lower recidivism rate than a control group (program nonstarters; N = 66) and program dropouts (whose predisposing factors may have influenced their program participation; N = 150). Analyses controlled for sex, ethnicity, age, prior petitions, highest class of prior petition, and months to follow-up. Results show a general trend indicating the long-term effectiveness of the program as graduates had a lower incidence of petitions at follow-up compared with dropouts and fewer petitions compared with the other two groups.

Keywords

juvenile delinquency cognitive behavioral therapy treatment effectiveness recidivism group psychotherapy

While overall juvenile arrests have declined by 24% from 2001 to 2010 (Federal Bureau of Investigation [FBI], 2010), juvenile crime continues to be a national concern (Snyder, 2008; Stahl, 2008). In 2010, juveniles accounted for 13% of all arrests made in the United States (FBI, 2010). In addition, juveniles were the perpetrators of 13% of all violent crime arrests, 23% of all property crime arrests, 42% of all arson arrests, and 24% of all robbery arrests in that same year in the United States (FBI, 2010). Equally distressing is the fact that more than half of the juveniles tried in the United States in 2004 were younger than 16 (Stahl, 2008). The prognosis of these youths is often poor, as Snyder and Sickmund (2006) found that the juveniles who commit a crime are more likely to recommit as an adult offender than those who had not offended before age 17.

Probation programs are an important component of the juvenile court system. Of the juveniles who are adjudicated, 62% are placed on court-ordered probation, which is the most widely used form of consequence in the juvenile justice system (Livsey, 2006). In 2002, 1.6 million juvenile cases were handled in U.S. courts (Livsey, 2006), and juvenile probation placement showed a significant increase between 1985 and 2005 (Puzzanchera & Sickmund, 2008). Given that a large number of juvenile offenders will be placed on probation, it is critically important that researchers identify any program that may reduce recidivism among these offenders. However, there continues to be a relative lack of research on this topic.

Juvenile Delinquents and Mental Health Treatment

In 2009, there were an estimated 2.11 million juvenile arrests (Puzzanchera & Adams, 2009). A large portion of these youths likely had serious mental health needs (Lyons, Griffin, Quintenz, Jenuwine, & Shasha, 2003). Many adolescents who have been arrested display disruptive behavior or aggression that is often linked to cognitive distortions and also a lack of appropriate social skills (Wasserman & Miller, 1998). This often leads to a lack of empathy for others as well. Specific problems include deficits in their ability to read certain social cues and the belief that aggressive behavior is a norm in civilized society (Wasserman & Miller, 1998). These cognitive and social skill deficits in juvenile offenders may contribute to a propensity to continue to commit crimes. Robertson, Grimes, and Rogers (2001) found, for example, that within 18 months, youth whose cognitive distortions were corrected through a cognitive-behavioral program required less future intervention by the criminal justice system than those receiving either traditional probation or intensive supervision and monitoring. A study by the Office of Juvenile Justice and Delinquency Prevention (OJJDP) revealed that a large majority of serious and violent juvenile offenders showed early signs of delinquency many years before the first arrest (Bilchick, 1999). Information was derived from interviews with both the mother and the offender. Although these juveniles were arrested for their first offense at an average age of 14, they showed emerging signs of delinquency as early as the age of 7, on average. The same study showed that a large majority of the delinquents had serious problems, such as substance abuse and mental health issues. Because the justice system is often the first agency that the juvenile comes into contact with, it is crucial to develop a better understanding of mental health issues and empirically supported psychological treatments that may be available to these young offenders.

Although programs such as shock incarceration or boot camps have some popularity with the public because their rationale is often intuitive and easy to understand, these programs have consistently been found to be ineffective or harmful to program participants, as they often increase recidivism rates (Sherman et al., 1997). However, other programs that contain specific and empirically supported therapeutic approaches have been found to reduce recidivism rates among adolescent offenders (Lipsey, Wilson, & Cothern, 2000). One such program is cognitive-behavioral therapy (CBT), which is a form of therapy that provides juvenile offenders with skills to examine their distorted beliefs that are maintaining their criminal tendencies as well as offering behavioral coping skills. According to Dowden and Andrews (2000), highly structured programs that incorporate cognitive behavioral techniques and social learning approaches are associated with a greater reduction in overall criminal activity. The previously cited OJJDP report found that the most effective programs when dealing with delinquents were those programs that concentrated on building interpersonal skills and incorporated cognitive-behavioral techniques (Bilchick, 1999).

One reason why CBT is so effective with this population is that juvenile offenders often engage in patterns of cognitive distortions that arise when they are faced with stressful situations (Lopez & Emmer, 2002; Wasserman & Miller, 1998). Examples of these distortions emerged in a qualitative study conducted by Lopez and Emmer (2002) that assessed how male offenders interpret their own violent criminal behavior. Lopez and Emmer conducted interviews with offenders that required the offenders to recall the time when they committed a particular crime. During the interviews, the offenders revealed that they strongly believed that their actions were justified. The results also indicated that two prevailing thought processes were often present when males under the age of 18 committed violent acts. The first thought is that the offender must protect his family and close friends ensuring that these individuals are safe and respected. The second is that that the offender must defend his own well-being or distinctiveness as a man by behaving violently. These results suggest that juvenile offenders may benefit from developing better cognitive and problem solving skills.

Evidence of CBT Treatment Effectiveness

Although the existing research on the effectiveness of CBT with juvenile offenders shows initial promise, there is a critical need for more research on this therapy as well as other, similarly promising psychological treatments. An important meta-analysis on this topic was conducted by Landenberger and Lipsey (2005). In this meta-analytic review of 58 previous studies, 17 of which were with juvenile offenders and 41 with adult offenders, the authors sought to understand the effects of CBT and what offender and program characteristics might moderate the effectiveness of CBT. Their initial findings underscored the effectiveness of CBT: Program completers had a 25% lower recidivism rate compared with those in a control group. When the authors focused on only those studies that implemented a “best practices” CBT program, effectiveness increased significantly, the program completers showed a 50% decrease in recidivism compared with those in a control group. Although the authors (Landenberger & Lipsey, 2005) had initially sought to identify potential moderators of CBT effectiveness, results indicated that in general, few characteristics of CBT programs and study designs made a difference in treatment effectiveness. For example, type of population (juvenile vs. adult), specific CBT program implemented, treatment setting, and training and background of providers all proved to have no influence on effectiveness. The authors did find that offenders’ risk level, program quality, and certain treatment characteristics did impact treatment effectiveness.

A more recent study of CBT effectiveness using a sample of mostly adult offenders in the community (Hollin et al., 2008) revealed similar results to those of Landenberger and Lipsey (2005). Specifically, these authors (Hollin et al., 2008) divided participants into four different groups: program completers, dropouts, nonstarters, and a comparison group. Results indicated that compared with the comparison group, the odds of reconviction were 0.61 for program completers. Strengths of this particular study are the size of the sample (N = 4,935) and the rigorous study design that allowed for the inclusion of a number of control variables including age, measured risk of reconviction, number of previous convictions, and severity of offense.

Although the literature on the effectiveness of CBT has been building in recent years, most of this research has concentrated on adults, and research on juveniles has almost exclusively focused on juveniles who were in a controlled setting such as a treatment center or incarceration. Therefore, there has been little to no research examining CBT effectiveness with juveniles where the program is led by probation personnel. As probation is a very large component of the juvenile justice system (Livsey, 2006; Puzzanchera & Sickmund, 2008; Stahl, 2008), research on this topic is critically needed.

Purpose of the Present Study and Hypothesis

The purpose of the present study was to estimate the long-term effectiveness of a CBT program titled Community Opportunity Growth (COG). This study monitored juvenile delinquents’ recidivism across a 7-year period, with a 39-month average length to follow-up. This study seeks to fill a gap in the research literature on the effectiveness of cognitive behavioral therapy for juvenile delinquents administered by a probation department. It was hypothesized that program graduates would have a significantly lower recidivism rate compared with either youth who dropped out of the program (dropout group) or youth who did not start the program (control group).

Method

Participants

Participants in the present study were juveniles placed on probation within a single county that is located in southwestern Illinois near the Missouri/Illinois border. The county has a unique demographic makeup of rural, urban, and suburban communities and according to the 2000 census has a juvenile population of nearly 62,000. Archival data were gathered from 394 youth who were either on probation or under court supervision in the county between 2000 and 2007. During that time, these youth were referred to the COG program, which is described in a later section. Of these, 178 youth graduated from the COG program successfully (graduate group), while 150 youth began the program but dropped out and did not graduate successfully (dropout group). As dropouts are a somewhat problematic choice for a control group due to possible preexisting differences, an alternative control group was available. The control group of 66 youths included those who were referred to the COG program by their probation officer but either (a) the referral was made shortly after the deadline to begin the program and the youth did not begin the program at any time in the future, or (b) the referral was made in the county’s computerized system but due to human error, the program facilitator never received the referral. This control group utilized, although not ideal in some ways and not conforming procedurally to random assignment, a number of properties that allow it to be considered a true control group. Specifically, these youth were referred at or very near the same time as the dropout and graduated groups, and were referred using the same procedure and criteria. Thus, although random assignment prior to treatment would have been ideal, we consider the present control group to be at least adequate given the prevalent constraints that exist when doing research in a natural community context such as this.

Descriptive statistics for juveniles who graduated, dropped out, or were in the control group (nonstarters) are presented in Table 1. A chi-square analysis indicated that the groups did not differ on the variables of sex or ethnicity. Analysis of variance (ANOVA) also indicated that the groups did not differ on the variables of age, number of petitions prior to group assignment, number of months from group assignment to follow-up, or average highest class petition prior to group assignment. Again, this lack of preexisting differences confirms to some extent the lack of selection bias in the control group.

Table 1.

Descriptive Statistics for the Graduated, Dropout, and Control Groups.

	Graduated	Dropout	Control
	M (SD)	M (SD)	M (SD)
Female (%)	21.30	24.70	24.20
Caucasian (%)	60.70	57.30	54.50
Age	15.27 (1.21)	15.15 (2.23)	15.33 (1.35)
Prior petitions	1.97 (1.32)	1.93 (1.31)	1.94 (1.29)
Highest class prior petition	5.61 (1.91)	5.93 (1.79)	5.94 (1.70)
Months to follow-up	38.08 (23.11)	39.22 (25.06)	39.53 (32.95)
Petitions at follow-up (%)	31.60	52.70	40.90
Number of follow-up petitions	0.59 (1.13)	1.21 (1.60)	0.94 (1.51)
N	177^a	150	66

The number (N) of graduates varies from 176 to 178 due to occasional missing data.

Description of the COG Program

The COG program is a 16-week group therapy program that occurred for approximately 1.5 hr per week and included both males and females in gender-specific groups. The beginning and end of the group coincided with the public school calendar, with fall groups beginning in September and ending before the winter break and spring groups beginning in January and ending before the summer break. The COG program was led by the same facilitator for the entire window of time encompassed by this study. The COG program is loosely organized, though not manualized, and was created by county probation staff and the first author, who serves as a consultant to the department. The program was built on the theoretical assumptions and clinical techniques described in Beck (1995). Regarding the organization of the program, the first session includes a description of the group’s purpose, group-building activities, and exploration of the negative consequences of criminal behavior. The second session includes a description of the cognitive model (Beck, 1995). The third and subsequent sessions focus on irrational thinking experienced by program participants, behavioral and emotional consequences of such thinking, and the legal repercussions of criminal behavior. The facilitator of the program relied primarily on Socratic questioning as the primary clinical technique, while significant peer discussion helped participants identify irrational thinking and produce rational replacement thoughts. It is important to note that juveniles in the COG program had a variety of primary probation officers who were responsible for their supervision while on probation. However, the facilitator of the program was on special assignment as director of the COG program and did not have responsibility for supervising the probation of any of the juveniles.

Data Collection Procedures

For the 7-year window chosen (from fall of 2000 to the end of 2007), archival data were gathered on all youth who were referred to the COG program. These data were gathered from the county’s computerized database that stores information regarding the youth, their criminal history, and other data relevant to their case. For the purposes of this study, recidivism was defined as a petition to the court related to a new criminal charge. Thus, recidivism did not include petitions to the court for technical violations (e.g., truancy) or when the youth was arrested on a new criminal charge but no petition to the court was made due to lack of evidence, and so on. In addition, the actual adjudication of the youth as guilty was not required to be considered as recidivism for this study. Related to this, a youth might often incur several petitions to the court for a particular set of crimes. For example, a youth who stole a car and was eventually arrested might incur petitions for automobile theft, reckless driving, speeding, and resisting a peace officer. Rather than considering these as four separate petitions, the date for all petitions was examined and any number of petitions filed on the same date was counted as only one petition, with the most serious charge noted. Thus, in the previous example, the youth’s auto theft would be counted as one petition, and a Class 3 felony (for automobile theft). This procedure was enacted to reduce variability in the data due to changes over time in how the State’s attorney might choose to file petitions for lesser charges related to a single incident.

For all three groups, the creation of an index date was required to define what would be considered prior petitions as opposed to petitions at follow-up. For the graduated group, the index date was the date of graduation from the COG program, so that the follow-up period was defined after this index date, and prior petitions were those that were filed prior to this index date. For the control group, the index date was considered to be the date the COG program began for which they had been referred but did not begin. For the dropout group, the index date was considered to be the date at which their COG program began. Follow-up data were gathered at the beginning of 2008.

Design

Given the retrospective nature of the study, a quasi-experimental design was used, and differences between the three groups on several variables were controlled for. These control variables included sex, ethnicity, age, prior petitions, highest class of prior petition, and months to follow-up. The independent variable was group as we compared youth who graduated from the program with those who dropped out of the program or those who were referred but never started the program. For a similar design, see Hollin et al. (2008).

Results

To examine the effectiveness of the COG program, we conducted analyses with two dependent variables: a dichotomous variable representing whether at least one petition was or was not generated at follow-up and a continuous variable representing number of petitions generated at follow-up. The continuous dependent variable may be more sensitive to change in response to treatment, but the dichotomous variable may also be important when there is interest in factors that may eliminate petitions completely.

The first analysis was a hierarchical logistic regression with the dichotomous dependent variable (i.e., whether or not at least one petition was generated at follow-up). Approximately 41% of the control group generated petitions at follow-up, whereas 53% of the dropout group and 32% of the graduated group generated such petitions. In the first block, the following control variables were entered: sex (1 = male, 0 = female), ethnicity (1 = Caucasian, 0 = non-Caucasian [95% of this group was African American]), age, prior petitions, highest class of prior petition, and months to follow-up. This model produced a good model fit, as indicated by a nonsignificant Hosmer and Lemeshow goodness-of-fit test, χ²(8) = 4.73, p = .79, and respectable R² values (Cox and Snell R² = .16; Nagelkerke R² = .21). This model was significantly better than a constant-only model (which included only the intercept), Δχ²(6) = 66.15, p < .001. In the second block, the crucial group variable was added (generating two dummy-coded variables with the control group as the reference group). This model accounted for significantly more variance, Δχ²(2) = 16.69, p < .001, and the percentage of correct classifications increased from 67.50% (for the Block 1 model) to 70.80% (for the Block 2 model). For the second block, model χ²(8) = 82.84, p < .001, Hosmer and Lemeshow goodness of fit χ²(8) = 3.69, p = .88, Cox and Snell R² = .19; Nagelkerke R² = .26. The coefficients for each of the predictors included in the Block 2 model are presented in Table 2.

Table 2.

Petitions at Follow-Up Logistic Regression With Control Variables and Group as Predictors.

	B (SE)	95% CI for Exp (B)
	B (SE)	Wald statistic	Lower	Exp (B)	Upper
Constant	−0.95 (1.33)	0.51	—	0.39	—
Sex	0.90 (0.29)	9.35*	1.38	2.45	4.36
Ethnicity	−0.38 (0.24)	2.52	0.43	0.69	1.09
Age	−0.08 (0.08)	0.99	0.79	0.92	1.08
No. of prior petitions	0.40 (0.10)	16.42*	1.23	1.49	1.81
Highest class prior petitions	−0.08 (0.07)	1.50	0.81	0.92	1.05
Months to follow-up	0.02 (0.01)	26.85*	1.01	1.02	1.03
Group	—	16.13	—	—	—
Group (1)	−0.47 (0.33)	1.94	0.33	0.63	1.21
Group (2)	0.57 (0.34)	2.82†	0.91	1.76	3.40

Note. CI = confidence interval. Model χ² (8, N = 391) = 82.84, p < .001. Cox and Snell R² = .19; Nagelkerke R² = .26. Group (1) = Graduate (1) vs. Control (0). Group (2) = Dropout (1) vs. Control (0).

p < .05. †p < .10.

Although the difference between the graduated and control groups was not significant (p = .16) in this logistic regression, it is worth noting that the odds ratio of 0.63 suggests that the odds of incurring a petition at follow-up were lower for the graduated group (32% of whom experienced a petition at follow-up) than for the control group (41% of whom experienced a petition at follow-up). For the sake of making a complete presentation of the data, we note that the difference between the dropout and control groups approached statistical significance (p = .09). However, we acknowledge that a plausible explanation for this difference is that juveniles in the dropout group were especially resistant to treatment before the treatment started. The associated odds ratio (1.76) suggests that the odds of experiencing a petition at follow-up was higher for the dropout group (53% of whom experienced a petition at follow-up) than for the control group.

Because of the possibility that juveniles in the dropout group may differ from those in the control group in several ways, we do not consider comparisons with the dropout group to provide strong tests of the effectiveness of the COG program. However, to make a complete presentation of differences between the groups, we did compare the graduated and dropout groups in a hierarchical logistic regression. In the first block, the same control variables used earlier were entered: sex (1 = male, 0 = female), ethnicity (1 = Caucasian, 0 = non-Caucasian), age, prior petitions, highest class of prior petition, and months to follow-up. This model produced a good model fit, as indicated by a nonsignificant Hosmer and Lemeshow goodness-of-fit test, χ²(8) = 4.62, p = .80, and good R² values (Cox and Snell R² = .13; Nagelkerke R² = .18). This model was significantly better than a constant-only model (which included only the intercept), Δχ²(6) = 46.50, p < .001. In the second block, the crucial Graduate versus Dropout variable was added. This model accounted for significantly more variance, Δχ²(1) = 15.63, p < .001, and the percentage of correct classifications increased from 66.50% (for the Block 1 model) to 71.10% (for the Block 2 model). For the second block (comparing the graduate and dropout groups), model χ² (7) = 62.13, p < .001, Hosmer and Lemeshow goodness of fit χ² (8) = 4.62, p = .80, Cox and Snell R² = .17; Nagelkerke R² = .24. The coefficients for each of the predictors included in the Block 2 model are presented in Table 3.

Table 3.

Petitions at Follow-Up Logistic Regression With Control Variables and Graduated vs. Dropout Variable as Predictors.

	B (SE)	95% CI for Exp (B)
	B (SE)	Wald	Lower	Exp (B)	Upper
Constant	−0.92 (1.45)	0.41	—	0.40	—
Sex	0.80 (0.32)	6.34*	1.20	2.23	4.17
Ethnicity	−0.51 (0.26)	3.85*	0.36	0.60	1.00
Age	−0.11 (0.09)	1.40	0.75	0.90	1.07
Number of prior petitions	0.38 (0.11)	12.66*	1.19	1.47	1.81
Highest class prior petitions	−0.03 (0.07)	0.19	0.84	0.97	1.12
Months to follow-up	0.02 (0.01)	14.32*	1.01	1.02	1.03
Graduated vs. Dropout	0.99 (0.25)	15.13*	1.63	2.69	4.42

Note. CI = confidence interval. Model χ² (7, N = 325) = 62.13, p < .001. Cox and Snell R² = .17; Nagelkerke R² = .24.

p < .05.

In this logistic regression, the difference between the graduated and dropout groups was significant (p < .001). The odds ratio of 2.69 suggests that the odds of incurring a petition at follow-up were higher for the dropout group (53% of whom experienced a petition at follow-up) than the graduated group (32% of whom experienced a petition at follow-up). Again, it is unknown to what extent this difference can be attributed to the effectiveness of the COG program or the possibility that the juveniles in the dropout group were more treatment resistant before entering treatment.

We next examined the effectiveness of the COG program by conducting a hierarchical multiple regression with number of petitions (a continuous variable) at follow-up as the dependent variable. The mean number of petitions at follow-up for the control, dropout, and graduated groups was 0.94, 1.21 and 0.59, respectively. In the first block, the same control variables used in the logistic regression analyses were entered. This control-variable-only model accounted for a significant portion of the variance in number of petitions at follow-up, F(6, 384) = 13.04, p < .001, R² = .17. In the second block, two dummy-coded group variables were added (Graduated vs. Control, and Dropout vs. Control) and this model accounted for significantly more variance, ΔF(2, 382) = 9.65, p < .001, ΔR² = .04. For the second-block model, F(8, 382) = 12.63, p < .001, R² = .21. Coefficients for the second-block model are presented in Table 4.

Table 4.

Number of Petitions at Follow-Up Multiple Regression With Control and Group Variables as Predictors.

	B (SE)	β
Constant	0.55 (0.67)	—
Sex	0.70 (0.16)	0.21*
Ethnicity	−0.15 (0.13)	−0.05
Age	−0.05 (0.04)	−0.06
Number of prior petitions	0.20 (0.05)	0.18*
Highest class prior petitions	−0.04 (0.04)	−0.05
Months to follow-up	0.02 (0.003)	0.27*
Group (1)	−0.37 (0.19)	−0.13*
Group (2)	0.26 (0.19)	0.09

Note. Model F(8, 382) = 12.63, p < .001, R² = .21. Group (1) = Graduate (1) vs. Control (0). Group (2) = Dropout (1) vs. Control (0).

p < .05.

In this model focused on number of petitions at follow-up, the difference between the graduated and control groups was significant, p < .05, part r = −.09. As is evident in the penultimate row of Table 1, graduates (M = 0.59, SD = 1.13) experienced fewer petitions at follow-up than youth in the control group (M = 0.94, SD = 1.51). This result, when compared with the results of the logistic regression analysis, suggests that the COG program may have had a stronger effect on the number of petitions at follow-up than on whether any petition at follow-up was generated. The results in both analyses were in the same direction, but the difference between the graduated and control groups was only significant when the number of petitions at follow-up was compared.

In the same multiple regression analysis, the difference between the dropout (M = 1.21, SD = 1.60) and control groups (M = 0.94, SD = 1.51) was not significant (p = .17, part r = .06), even though, as shown in Table 1, dropouts experienced more petitions at follow-up than any other group. When compared with the comparable result that emerged in the logistic regression analysis, this nonsignificant difference suggests that dropouts and juveniles in the control group did not substantially differ in the number of petitions at follow-up, but did differ somewhat (p = .09) in whether they generated any petitions at follow-up.

In a final analysis, we compared the graduated and dropout groups on number of petitions at follow-up. Although comparisons with the dropout group may not provide strong tests of the effectiveness of the COG program (because juveniles who dropped out may be more treatment resistant), it is appropriate to acknowledge how these groups differ. In the first block of a hierarchical multiple regression, the control variables (the same used in previous analyses) accounted for a significant portion of the variance in number of petitions at follow-up, F(6, 318) = 9.90, p < .001, R² = .16. In the second block, a Graduated versus Dropout variable was added. This second-block model accounted for significantly more variance, ΔF(1, 317) = 18.90, p < .001, ΔR² = .05. For the second-block model, F(7, 317) = 18.64, p < .001, R² = .21. Coefficients for the second-block model are presented in Table 5.

Table 5.

Number of Petitions at Follow-Up Multiple Regression With Control and Graduated vs. Dropout Variables as Predictors.

	B (SE)	β
Constant	1.11 (0.68)	—
Sex	0.65 (0.17)	0.20*
Ethnicity	−0.23 (0.15)	−0.08
Age	−0.07 (0.04)	−0.09†
Number of prior petitions	0.22 (.06)	0.20*
Highest class prior petitions	−0.03 (0.04)	−0.03
Months to follow-up	0.01 (0.003)	0.22*
Graduated vs. Dropout	−0.62 (0.14)	−0.22*

Note. Model F(7, 317) = 18.64, p < .001, R² = .21.

p < .05. †p < .10.

In this second-block model focused on number of petitions at follow-up, the difference between the graduated (M = 0.59, SD = 1.13) and dropout (M = 1.21, SD = 1.60) groups was significant, p < .001, part r = −.21. This result, when compared with results in the logistic regression analysis reported earlier, which also compared the graduated and dropout groups, suggests that the graduated and dropout groups differed in both the number of petitions at follow-up and whether any petition at follow-up was generated.

Discussion

Although there is a growing literature on the effectiveness of CBT with offender populations, there are relatively few studies that have focused on the effectiveness of CBT in juvenile populations (Landenberger & Lipsey, 2005). In addition, there are even fewer studies on this topic using community-based treatment providers such as probation departments. Given the large number of youth who have contact with probation departments each year (Livsey, 2006), these departments have the potential to play a critical role in the application of evidence-based interventions such as CBT.

Results of the present study provided some signs of the long-term effectiveness of a CBT group therapy program in reducing recidivism in juveniles. In our study, we examined two indicators of effectiveness, which were whether program graduates had any reoffenses (a dichotomous variable) and also the number of offenses. When recidivism was dichotomously coded, these results indicated that although the difference between the graduated and control groups was not statistically significant (p = .16), the odds ratio for the graduate group of 0.63 does indicate that there was a lower rate of petitions in the graduate group compared with the control group. However, these results were more robust when considering recidivism as the actual number of petitions. In this analysis, the graduated with control group comparison was significant (p < .05), with the graduated group experiencing about half as many petitions (0.59) as the control group (1.21; part r = −.21). We also found that the graduate group had significantly fewer petitions and were less likely to have a petition than the dropout group, although this difference does not necessarily attest to the effectiveness of the therapy program (see Hollin et al., 2008).

Results of the present study are similar to other studies on this topic. For example, Lipsey, Wilson, and Cothern (2000) conducted a meta-analysis and review of 200 studies of program effectiveness with institutionalized and noninstitutionalized juveniles. The authors found that the 200 programs reviewed reported recidivism reductions ranging from negligible to a 40% reduction, with the average being a 12% reduction when comparing program participants with a control group. When examining simple recidivism rate reductions, the present study found that 32% of the graduate group had reoffended compared with 41% of the control group, which is a 22% reduction in recidivism and a favorable comparison with the results found by Lipsey et al. (2000).

Although the present results point to the potential effectiveness of a group CBT program with juveniles probationees, one should also acknowledge the potential conflict that may occur when probation officers are called to both supervise as well as treat juveniles. The therapeutic treatment process relies on building rapport, which can be negatively affected when the client feels overly judged, admonished, or given strict rules to follow, such as may be the case with most probationee/probationer relationships. This potential role conflict was avoided as all participants were supervised by various probation officers who were not the facilitator of the COG program. However, if a program similar to COG were to be administered within other probation departments, potential role conflict would need to be considered.

While juvenile crime continues to be a significant problem in society, treatment that is provided to these offenders is often not evidence based and is thus relatively ineffective (Lipsey et al., 2000). The consequences of using relatively ineffective treatments that fail to reduce recidivism are not only related to public safety, but also have financial considerations. Left without effective treatment, many of these offenders will continue to offend throughout their adolescent and adult years, costing taxpayers millions of dollars to prosecute and incarcerate in the future. In addition, although there are a number of effective treatments for juvenile offenders (for a review, see Kimonis & Frick, 2010), some of the most effective treatments (e.g., multisystemic therapy) are typically provided by clinicians rather than probation personnel. However, research shows that there are significant financial benefits to providing effective treatment to youths in the juvenile justice system. For example, Robertson et al. (2001) conducted a short-run cost analysis that focused on juveniles (N = 293) who were adjudicated and placed within a cognitive-behavioral group therapy program. Program participants were compared with a control group that followed the typical probation or parole guidelines. It was found that for nearly every dollar that was spent on the cognitive-behavioral treatment program, close to two dollars was saved by the courts (Robertson et al., 2001).

Although the results of the present study fill a critical gap in the literature in terms of understanding the long-term effects of CBT therapy with a juvenile probationer population, limitations of the study should be noted. To begin, we acknowledge the limitations that exist when relying on archival data. For example, although a control group was obtained (youth referred but who did not start the program), the study did not involve random assignment. However, many possible confounding variables were considered in the analyses, thus minimizing the effect of any possible differences between the graduated and control groups. This study was also able to obtain a nonstarter control group, which has been noted by others as a favorable alternative when the design does not allow for random assignment (Hollin et al., 2008). Another limitation includes the fact that the sample size, and especially the low percentage of females in each group, did not allow for separate analyses by sex of the participant. It would be of interest to examine whether the COG graduates’ long-term recidivism rate differed for males and females in future studies. In addition, as previously mentioned, although group differences were found when comparing the dropout group with the graduate group, it is unknown whether such differences existed due to the effectiveness of the COG program or other preexisting variables such as treatment resistance that were not measured. Finally, this study drew participants from a single county in the Midwest and also relied on a single facilitator. Future research should seek to gather data from other counties and probation departments and also other similarly trained facilitators using the COG program to estimate the generalizability of these results.

Available research on the effectiveness of CBT programs with juvenile probationers is scarce. The present study stands as one of the longest follow-up studies of a CBT treatment program with juvenile offenders, with the average time to follow-up at 39 months. CBT is empirically supported as being effective in reducing recidivism in the juvenile population and is also more cost-effective. In the current times of financial deficits by both federal and state agencies, it is crucial that the juvenile courts begin to implement programs that are both effective and cost efficient in the long term while discontinuing programs that may have popular support (e.g., boot camps or shock incarceration) but are known to be ineffective and even harmful to youth and society.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Beck

(1995). Cognitive therapy: Basics and beyond. New York, NY: Guilford Press.

Bilchick

(1999). OJJDP research: Making a difference for juveniles. Washington, DC: Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice.

Dowden

Andrews

D. A.

(2000). Effective correctional treatment and violent reoffending: A meta-analysis. Canadian Journal of Criminology, 42, 449-467.

Federal Bureau of Investigation. (2010). Crime in the United States: Ten year arrest trends, 2010. Retrieved from http://www.fbi.gov/about-us/cjis/ucr/crime-in-the-u.s/2010/crime-in-the-u.s.-2010/tables/10tbl32.xls

Hollin

C. R.

McGuire

Hounsome

Hatcher

Bilby

Palmer

(2008). Cognitive skills behavior programs for offenders in the community: A reconviction analysis. Criminal Justice and Behavior, 35, 269-283.

Kimonis

E. R.

Frick

P. J.

(2010). Oppositional defiant disorder and conduct disorder grown up. Journal of Developmental & Behavioral Pediatrics, 31, 244-254. doi:10.1097/DBP.0b013e3181d3d320

Landenberger

N. A.

Lipsey

M. W.

(2005). The positive effects of cognitive-behavioral programs for offenders: A meta-analysis of factors associated with effective treatment. Journal of Experimental Criminology, 1, 451-476. doi:10.1007/s11292-005-3541-7

Lipsey

M. W.

Wilson

D. B.

Cothern

(2000). Effective intervention for serious juvenile offenders. Washington, DC: Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice.

Livsey

(2006). Juvenile delinquency probation caseload, 1985–2002. Washington, DC: Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice.

10.

Lopez

V. A.

Emmer

E. T.

(2002). Influences of beliefs and values on male adolescents’ decision to commit violent offenses. Psychology of Men & Masculinity, 3, 28-40. doi:10.1037/1524-9220.3.1.28

11.

Lyons

J. S.

Griffin

Quintenz

Jenuwine

Shasha

(2003). Clinical and forensic outcomes from the Illinois mental health juvenile justice initiative. Psychiatric Services, 54, 1629-1634. doi:10.1176/appi.ps.54.12.1629

12.

Puzzanchera

Adams

(2009). Juvenile arrests 2009. Washington, DC: Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice.

13.

Puzzanchera

Sickmund

(2008). Juvenile court statistics, 2005. Washington, DC: Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice.

14.

Robertson

A. A.

Grimes

P. W.

Rogers

K. E.

(2001). A short-run cost-benefit analysis of community-based interventions for juvenile offenders. Crime & Delinquency, 47, 265-284. doi:10.1177/0011128701047002006

15.

Sherman

L. W.

Gottfredson

MacKenzie

D. L.

Eck

Reuter

Bushway

(1997). Preventing crime: What works, what doesn’t, what’s promising. A report to the United States Congress. College Park: Department of Criminology and Criminal Justice, University of Maryland.

16.

Snyder

H. N.

(2008). Juvenile arrest, 2006. Washington, DC: Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice.

17.

Snyder

H. N.

Sickmund

(2006). Juvenile offenders and victims: 2006 national report. Washington, DC: Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice.

18.

Stahl

A. L.

(2008). Delinquency cases in juvenile courts, 2004. Washington, DC: Office of Juvenile Justice and Delinquency Prevention, U.S. Department of Justice.

19.

Wasserman

G. A.

Miller

L. S.

(1998). The prevention of serious and violent juvenile offending. In Loeber

Farrington

D. P.

(Eds.), Serious and violent juvenile offenders (pp. 197-247). Thousand Oaks, CA: Sage.