Assessing the Effectiveness of a Specialized,Field-Based Treatment Program for Youth Who Have Committed Sexual Offenses in an Australian Jurisdiction

Abstract

This study provides an evaluation of recidivism outcomes for a specialized, field-based treatment program for youth who perpetrate sexual offenses in an Australian jurisdiction. Using survival analyses, recidivism outcomes for the treatment group (n = 200), who were followed for an average of 5.07 years (SD = 3.13), were contrasted with a sample of sexually offending youth who were either referred but not accepted or not referred to the program (n = 295). Rates of sexual recidivism were low and not significantly different between the groups (9.5% for treated and 10.8% for untreated). Unadjusted Cox regression results indicated that the treated group were less likely to violently recidivate compared to the untreated group (HR = 1.41, 95% CI [1.01, 1.96]), but this effect became nonsignificant when controlling for offense history covariates (HR = 1.22, 95% CI [0.87, 1.72]). Both groups exhibited high rates of nonsexual offending during the follow-up period, and treatment factors including clinician-rated success, were found to be associated with a lower frequency of reoffending after treatment. Findings highlight important considerations for both practice and research. First, findings suggest the need for specialized programs to ensure factors associated with general recidivism are also addressed in treatment; second, findings reinforce potential utility for clinician-rated and structured assessments to inform treatment planning and outcomes. Finally, the findings raise the importance of appropriate comparison groups when designing evaluation studies, to accurately inform policy and practice.

Keywords

sexual offending treatment juveniles reoffending evaluation

Accumulated evidence suggests there are a range of treatment programs for individuals who commit sex offenses that reduce rates of recidivism, particularly those targeted toward moderate- to high-risk individuals (Schmucker & Lösel, 2015). However, more detailed outcome evaluations focused on youth who commit sexual offenses (YSOs) are needed. Although some treatment programs show “promise” in terms of their effectiveness for reducing recidivism by YSOs, the evidence base is far from complete (Kettrey & Lipsey, 2018; Reitzel & Carbonell, 2006; Schmucker & Lösel, 2015). The aim of the current study is to contribute toward advancing this evidence base by providing an evaluation of a specialized field-based treatment program for YSOs in an Australian jurisdiction.

Treatment Effectiveness for Youth Who Commit Sexual Offenses

Despite specialized youth sexual offending programs being embedded in youth justice systems internationally and in Australia, there is mixed evidence regarding the effectiveness of these programs in reducing rates of sexual recidivism (Fortune & Lambie, 2006; Kettrey & Lipsey, 2018; Reitzel & Carbonell, 2006; Walker et al., 2004). Several meta-analytic studies have examined the collective effectiveness of juvenile-targeted programs, with evidence to support both neutral (e.g., Hanson et al., 2002) and positive (e.g., Reitzel & Carbonell, 2006) effects of treatment on recidivism. Reitzel and Carbonell (2006) found that adolescents receiving specialized treatment for sexual offenses had significantly lower rates of sexual recidivism (7.37% average recidivism) compared to those who did not receive treatment (18.93% average recidivism), representing a relative reduction in sexual recidivism of 61.07%. However, the studies included in this meta-analysis varied in methodological quality, introducing systematic bias into the findings. Kettrey and Lipsey (2018) conducted a meta-analysis using more stringent inclusion criteria for studies of relatively high methodological quality (e.g., use of random assignment to conditions or matching of individuals across conditions). Only eight studies with high methodological quality were identified. The mean effect size for sexual recidivism favored treatment, but was not significant, and the mean effect size for general recidivism was significant and favored treatment. Kettrey and Lipsey (2018) argued that a confident conclusion that specialized treatment is effective in reducing recidivism is not supported by the best available evidence.

Current Study

This study reports an evaluation of treatment effectiveness for the Griffith Youth Forensic Service (GYFS), a specialized forensic assessment, treatment, and consultation service for court-adjudicated YSOs (aged 10–17 years at the time of their offense) operating in Queensland, Australia. Details of the program are provided in the Supplemental Material. The primary aim of the study is to compare reoffending outcomes between those who underwent GYFS treatment and a comparison group who received treatment-as-usual (TAU). Based on previous findings,¹ it was hypothesized that: (1) there would be a low rate (i.e., 7.08% ± 4.64%) of sexual recidivism; and (2) high rates of nonsexual recidivism (i.e., around 50%, but varying across violent and nonviolent recidivism) across the treatment and control group. To test the effectiveness of the treatment program, it was also hypothesized that: (3) the GYFS treatment group would have a lower sexual recidivism rate than the comparison group over the study period. Finally, given findings from other studies indicating that sexual offense-specific treatment can be effective for reducing rates of nonsexual recidivism (e.g., Tyler et al., 2021), it was hypothesized that: (4) the GYFS treatment group would have lower rates of violent and nonviolent recidivism compared to the comparison group. The secondary aim of the study was to explore possible treatment-related effects (e.g., time in treatment, success of treatment) on the frequency of reoffending during the follow-up period.

Method

Sample and Procedures

Treatment Group

The treatment group for the current study were drawn from a pool of 397 adolescent males found guilty of a sexual offense committed between the ages of 12 and 17 years and who were referred and accepted to GYFS for assessment between January 2001 and November 2012. From this total, 110 youth were excluded who were assessed but not accepted for treatment. Further exclusions included 37 youth where offense information was not available, two youth who had moved interstate/overseas and thus no longer resided in the jurisdiction, eight who were referred after the date of data extraction, four females, and 36 youth who had not completed treatment by the date of recidivism data extraction. This resulted in 200 male youth who completed treatment by the date of data extraction.

Characteristics of treatment recorded included: the duration in months from treatment start to end; individual session hours; family session hours; professional consultation hours; and number of missed appointments. In addition, all clients were evaluated according to whether treatment was considered successful (n = 160) or unsuccessful (n = 40). This classification was made by treating clinicians and based on clinical discretion and assessment in relation to the progress and engagement of clients throughout their contact with GYFS, as well as reference to structured decision-making tools summarizing risk factors associated with sexual reoffending (e.g., the Juvenile Sex Offender Assessment Protocol-II).² These assessments were conducted by the clinicians at the cessation of treatment and documented in a clinical administrative database as part of normal practice.

Comparison Group

Recidivism outcomes for the GYFS treatment youth were compared to a group of adolescents who were found guilty in a Queensland Court of sexual offenses committed between the ages of 12 and 17 years within the same time period as the treatment group. The group consisted of youth who had sexually offended and had either been referred to GYFS but not accepted for treatment (n = 96) or not referred to GYFS (n = 199) to form a comparison group (n = 295). Details of interventions and services that were provided to youth in the comparison group were not available. In the absence of receiving specialized treatment for sexual offending, TAU was assumed for this group, which typically includes the provision of manualized CBT modules for general offending and vocational services/supports delivered by caseworkers within youth justice service centers. Further details on comparison group context are provided in the Supplemental Materials.

Measures

Recidivism

Information about the offending histories of participants in treatment and comparison groups was obtained from the Queensland Police Service (QPS) database. QPS data reflect offenses reported to police in the State of Queensland, Australia, that resulted in police action, namely arrest. The QPS database includes information relating to the date, type, and detail of offenses, which were coded into three broad categories including: sexual offenses; violent offenses; and nonviolent offenses. Sexual offenses included sexual assault, rape, attempted rape, child pornography offenses, willful exposure, bestiality, and indecent exposure offenses. Violent offenses included actual or threatened violence against the person with an element of intent, such as assault, abduction, robbery, extortion, going armed to cause fear, break and enter dwelling with violence or threats, and homicide. Nonviolent offenses encompassed all remaining offenses such as dangerous driving, arson, theft, fraud, drug offenses, property damage, break and enter, and public order offenses. Within the nonviolent offenses category, traffic and vehicle regulatory, and community-based order breach offenses were excluded.

Follow-Up Period

Recidivism information was extracted from the QPS offending database on the 21st November 2012, which represented the end of recidivism follow-up for the treatment and comparison group. Where multiple instances of reoffending occurred, the first instance was taken as the date of recidivism. For the GYFS treated group, follow-up time was defined as the period between the date of treatment end (i.e., typically corresponding to court order end date) and the date of data extraction. For the comparison group, follow-up time was defined as the period between court order end date and the date of data extraction. Offending that occurred during treatment delivery or order administration was not considered when measuring recidivism.

Control Variables

Although there was extensive information available about the risk and need factors for the treatment group, the same information was not available for individuals in the TAU group to control for systematic differences. The variables consistently available across the groups that have been linked to recidivism outcomes and were used to control for differences included age at treatment/order start, Indigenous status³ and the number of violent and sexual offenses prior to treatment/order start.

Analytic Strategy

First, group differences in offense history variables were explored using Kruskal-Wallis tests (H) given some of the data were not normally distributed. Second, chi-square tests were conducted to examine group differences in the proportion of individuals who reoffended for each offense category and Kruskal-Wallis tests were conducted to examine group differences in the volume of reoffending for each offense category after treatment/order completion. Third, survival analyses were conducted to compare the rate of reoffending over time for the groups. Kaplan-Meier survival curves unadjusted for covariates were produced to visualize and compare the timing of reoffending for the treatment and comparison groups across offense categories. Cox regression models were used to estimate hazard ratios and quantify the effect of group membership and other covariates on reoffending outcomes. Proportionality of hazards was examined by plotting Schoenfeld residuals for covariates against time, and the global chi-square test of weighted residuals. Fourth, a series of zero-inflated Poisson (ZIP) regression models were estimated for the treated group only to examine the impact of offense history and treatment-related factors on the frequency of reoffending during the follow-up period.

All analyses were conducted using R statistical software (version 4.1.1; R Core Team, 2021), with survival models generated using the survival package (version 3.2-3; Therneau, 2020), and ZIP regression models estimated using the pscl package (version 1.5.5; Zeileis et al., 2008). Ethics approvals for the current project were obtained as part of a larger Australian Research Council Discovery grant (DP110102126) through the host university (CCJ/07/11/HREC).

Results

Group Differences in Offending History and Treatment Details

Offense history and treatment information for the treatment and comparison groups are displayed in Table 1. There were no differences between groups in terms of the age at first offense, but there was a small but statistically significant difference in the age at treatment/order commencement, with the GYFS group older on average compared to the TAU group at age of treatment/order commencement. There was a significant difference between groups in the sentence length for the index offense. The GYFS group had a longer sentence length compared to the TAU group, and this likely reflected the seriousness of their offending. The TAU group was slightly older at the time of data extraction, which corresponded with a significantly longer follow-up period compared to the GYFS group.

Table 1.

Offense History and Treatment Information.

Variable	GYFS treated youth (n = 200)	TAU youth (n = 295)	Kruskal-Wallis^†
Variable	Mean (SD)	Mean (SD)	H
Age at first offense	14.53 (1.81)^a	14.76 (1.73)^b	1.54
Age at treatment or order start	16.01 (1.48)	15.55 (1.34)	10.19**
Age at data extraction	22.20 (3.30)	23.42 (3.50)	13.80***
Index offense sentence length (months)	24.69 (13.19)^c	14.93 (11.74)^d	83.67***
Follow-up months (treatment/order end to data extraction)	60.77 (37.63)	79.53 (40.84)	21.92***
Offenses before treatment or order start date
Total	11.96 (17.32)	12.18 (18.97)	0.28
Sexual	1.90 (1.57)	1.52 (1.14)	10.78**
Violent	0.94 (1.85)	0.92 (1.86)	<0.01
Nonviolent	9.13 (16.12)	9.74 (17.78)	0.02
Offenses during treatment or order
Total	2.52 (5.29)	3.66 (7.77)	1.79
Sexual	0.04 (0.24)	0.06 (0.36)	0.62
Violent	0.25 (0.73)	0.26 (0.83)	0.11
Nonviolent	2.23 (4.93)	3.34 (7.43)	2.87
Offenses after treatment or order end date
Total	8.69 (17.01)	13.00 (20.56)	9.02**
Sexual	0.19 (0.70)	0.14 (0.44)	0.15
Violent	0.69 (1.61)	1.23 (2.45)	9.05**
Nonviolent	7.82 (15.94)	11.63 (19.21)	7.76**
Treatment
Duration (months)	13.70 (8.70)^e	—	—
Individual sessions (hr)	17.24 (16.97)	—	—
Family sessions (hr)	4.59 (7.26)	—	—
Professional consultation (hr)	13.16 (19.62)	—	—
Missed treatment appointments	2.97 (3.47)^f	—	—
Classified successful n (%)	160 (80.0%)
Classified unsuccessful n (%)	40 (20.0%)
Indigenous Australian n (%)	61 (30.5%)	107 (36.3%)
Remote location n (%)	21 (10.5%)	22 (7.5%)

Note. Notes regarding reduced sample sizes due to missing data. GYFS = Griffith Youth Forensic Service; TAU = treatment-as-usual.

n = 161.

n = 280.

n = 199.

n = 292.

n = 197.

n = 135.

p < .05. **p < .01. ***p < .001.

†

Series of Kruskal-Wallis H tests comparing differences between GYFS treated and nontreated groups for offense history variables.

There was a small but significant difference between groups for the number of prior sexual offenses before treatment/order commencement, with the GYFS group on average having more sexual offenses compared to the TAU group. There were no differences between the groups on the number of offenses detected during treatment or TAU.

Proportion and Volume of Reoffending

The rate of sexual reoffending was low and equivalent between the GYFS (9.5%) and TAU (10.8%) groups. The rate of nonsexual/nonviolent reoffending was moderate to high, ranging from 65.0% for the GYFS group and 68.5% for the TAU group. Overall, most individuals in both groups (71.0% GYFS and 75.3% TAU) accrued at least one reoffense during the follow-up period, typically for a nonsexual/nonviolent offense.

Chi-square analyses were conducted to examine group differences in the proportion of individuals to reoffend after treatment/order completion for each offense category. There were no significant differences between the GYFS and TAU groups for the proportion of individuals who reoffended across offense categories: any offending, χ²(1, n = 495) = 0.90, p = .34, φ_c = .04; sexual offending, χ²(1, n = 495) = 0.11, p = .74, φ_c = .01; and other offending, χ²(1, n = 495) = 0.50, p = .48, φ_c = .03. A significantly greater proportion of TAU (38.3%) individuals were arrested for a violent re-offense compared to GYFS individuals (25.5%), χ²(1, n = 495) = 8.25, p = .004, φ_c = .13. Considering the volume of reoffending (see Table 1), the TAU group on average had significantly more violent, nonviolent and total offenses after treatment/order completion compared to the GYFS group. However, no differences in the number of sexual offenses was found.

Nature and Timing of Reoffending

First, Kaplan-Meier survival curves were estimated to visualize the probability of reoffending over time, with estimated median months to reoffend documented in Table 2. Kaplan-Meier survival functions for the offense categories were estimated for the GYFS group contrasted with the TAU group, with survival curves displayed in Figure 1. Log-rank tests were used to examine differences in survival functions between groups. Second, two Cox proportional hazards models were estimated for each offense category. The first model included only group membership (i.e., GYFS vs. TAU), and the second model included group and the covariates of age at treatment/order start, Indigenous status and number of sexual and violent offenses prior to treatment/order start as factors potentially impacting reoffending rates. Table 2 displays Cox model hazard ratios for reoffending associated with group membership and Table 3 displays Cox regression results for reoffending by group membership and covariates.

Table 2.

Proportion of Individuals to Reoffend and Survival Model Hazard Ratio Estimates for Time to Reoffending by Offense Category and Group.

Reoffense type	Number of reoffenders (% within group)	Median months to reoffend [95% CI]^a	Model 1^b	Model 2^c
Reoffense type	Number of reoffenders (% within group)	Median months to reoffend [95% CI]^a	HR [95% CI]	HR [95% CI]
Any reoffense
GYFS	142 (71.0%)	13.31 [10.52, 19.82]	—	—
TAU	222 (75.3%)	10.95 [8.78, 16.47]	1.04 [0.84, 1.28]	0.96 [0.78, 1.19]
Sexual reoffense
GYFS	19 (9.5%)	na	—	—
TAU	32 (10.8%)	na	0.93 [0.53, 1.65]	0.89 [0.49, 1.60]
Violent reoffense
GYFS	51 (25.5%)	na	—	—
TAU	113 (38.3%)	na	1.41* [1.01, 1.96]	1.22 [0.87, 1.72]
Nonviolent reoffense
GYFS	130 (65.0%)	20.58 [14.70, 26.99]	—	—
TAU	202 (68.5%)	17.42 [13.18, 24.95]	1.03 [0.82, 1.28]	0.94 [0.75, 1.18]

Note. Significance tests are calculated with GYFS treated as reference group. GYFS = Griffith Youth Forensic Service; TAU = treatment-as-usual.

Median months to reoffend derived from Kaplan-Meier estimates. Where na is documented, median survival time cannot be calculated if less than half of the group has reoffended.

Cox proportional hazard model with group as the only factor. Treated group used as reference for calculating hazard ratios.

Cox proportional hazards model with group, Indigenous status, age at first offense, and total number of sexual and violent offenses pre-treatment/order as factors. Treated group used as reference for calculating hazard ratios.

p < .05.

Figure 1.

Kaplan-Meier survival functions of reoffending by offense type for treatment and nontreatment groups.

Table 3.

Cox Proportional Hazards Regression Results for Reoffending by Group and Controlling for Covariates.

Reoffense type and variable	β	SE	HR	Lower	Upper	Global Wald test
(Model 1) any reoffending						59.31***
TAU	−.04	0.11	0.96	0.78	1.19
Indigenous status	.76	0.11	2.14	1.71	2.67
Age at start	−.09	0.04	0.92	0.85	0.99
Prior sexual and violent offenses	.04	0.02	1.04	1.00	1.08
(Model 2) sexual reoffending						10.90*
TAU	−.12	0.89	0.89	0.49	1.60
Indigenous status	.73*	2.08	2.08	1.18	3.65
Age at start	−.01	0.99	0.99	0.81	2.21
Prior sexual and violent offenses	.07	1.07	1.07	0.98	1.17
(Model 3) violent reoffending						74.68***
TAU	.20	0.17	1.22	0.87	1.72
Indigenous status	1.09***	0.16	2.99	2.18	4.09
Age at start	−.16**	0.06	0.85	0.76	0.95
Prior sexual and violent offenses	.09***	0.03	1.09	1.04	1.15
(Model 4) nonviolent reoffending						74.57***
TAU	−.06	0.11	0.94	0.75	1.18
Indigenous status	.83***	0.12	2.30	1.83	2.89
Age at start	−.15***	0.04	0.86	0.80	0.94
Prior sexual and violent offenses	.06**	0.02	1.06	1.02	1.10

Note. Cox proportional hazards regression results for reoffending by group and controlling for covariates. TAU = treatment-as-usual.

p < .05. **p < .01. ***p < .001. GYFS treated group as reference in examining group differences.

Sexual Reoffending

There was no significant difference between groups in Kaplan-Meier estimates for sexual reoffending, χ²(1, n = 495) = 0.06, p = .81; with the groups exhibiting similar survival curves characterized by a low rate of sexual reoffending sustained over time (Figure 1). Cox regression models confirmed that group membership was not significantly associated with sexual reoffending both when considered alone and with covariate adjustment (Table 2). Indigenous status was the only covariate associated with an increased hazard of sexual reoffending (Table 3, model 2).

Violent Reoffending

There was a significant difference between groups in Kaplan-Meier survival functions for violent recidivism, χ²(1, n = 495) = 4.11, p = .04 (Figure 1), where the TAU group violently reoffended at a higher rate over time compared to the GYFS group. However, Cox regressions demonstrated that the hazards of violent reoffending were not significantly different between groups after adjustment of covariates. All covariates were significantly associated with the hazard of violent reoffending (Table 3, model 3), with an increased hazard of reoffending for individuals who were Indigenous, had a higher number of prior sexual and violent offenses, or were younger at treatment/order start.

Nonviolent Reoffending

There was no significant difference between groups in Kaplan-Meier survival functions for nonviolent offenses, χ²(1, n = 495) = 0.06, p = .81 (Figure 1), with the median time in months to reoffend ranging from 17.42 (95% CI [13.18, 24.95]) for the TAU group to 20.58 (95% CI [14.70, 26.99]) for the GYFS group (Table 2). Cox regression results demonstrated that all covariates were significantly associated with nonviolent reoffending (Table 3 model 4), with an increased hazard of reoffending for Indigenous individuals, individuals younger at treatment/order start, and individuals with a higher number of prior sexual and violent offenses.

Any Reoffending

There were no differences between the GYFS and TAU groups for Kaplan-Meier survival estimates for any reoffense, χ²(1, n = 495) = 0.12, p = .73 (Figure 1), with the median time in months to reoffend ranging from 10.95 (95% CI [8.78, 16.47]) for the TAU group to 13.31 (95% CI [10.52, 19.82]) for the GYFS group (Table 2). Cox regression models indicated that all covariates were significantly associated with reoffending hazards; Indigenous status, prior sexual and violent offenses, and younger age at treatment/order start were associated with increased hazards of any reoffending (Table 3, model 1).

Predictors of Reoffending Frequency

ZIP regression models were estimated for the GYFS group to examine the association between historical offense and treatment factors on the frequency of reoffending for each offense category after treatment completion (Table 4). Follow-up time was included as a predictor in these models to control for variation in time at risk.

Table 4.

ZIP Regression Models for History and Treatment Variables Predicting Frequency of Reoffending.

Variable	Count component			Zero/logit component
Variable	β	SE	z	β	SE	z
(Model 1) total reoffending
Intercept	6.46***	0.35	18.22	−6.36*	3.06	−2.08
Follow-up time	.01***	0.00	11.49	−.02**	0.01	−3.04
Indigenous status	.71***	0.07	9.53	−1.87**	0.69	−2.68
Age at start	−.29***	0.02	−14.56	.33*	0.17	1.97
Prior sexual and violent offenses	.01	0.01	1.17	−.15	0.11	−1.31
Treatment success	−.27***	0.07	−4.18	2.20*	0.90	2.44
Treatment duration (months)	−.01	0.00	−1.60	−.10*	0.04	2.23
Individual sessions (hr)	.00	0.00	0.48	−.01	0.02	−0.58
Family sessions (hr)	−.01	0.01	−1.26	.05	0.04	1.34
Professional consultation (hr)	−.01*	0.00	−2.34	−.04	0.03	−1.60
Missed treatment appointments	−.05***	0.01	−5.22	−.11	0.08	−1.46
Log likelihood = −842.05 (df = 22)
(Model 2) sexual reoffending
Intercept	−20.37***	5.64	−3.62	−1,065.15	1,862.74	−0.57
Follow-up time	.04**	0.01	2.90	.94	2.55	0.37
Indigenous status	−3.73***	1.02	−3.67	−427.97	1,104.05	−0.39
Age at start	.92**	0.30	3.09	54.66	88.66	0.62
Prior sexual and violent offenses	.30**	0.11	2.61	17.17	27.86	0.62
Treatment success	1.68**	0.63	2.65	207.65	320.37	0.65
Treatment duration (months)	−.08	0.06	−1.19	−17.98	28.27	−0.64
Individual sessions (hr)	.02	0.04	0.60	3.01	8.85	0.34
Family sessions (hr)	.08	0.09	0.92	3.28	20.16	0.16
Professional consultation (hr)	.17**	0.06	3.09	21.87	40.33	0.54
Missed treatment appointments	.18*	0.08	2.27	11.11	19.58	0.57
Log likelihood = −32.20 (df = 22)
(Model 3) violent reoffending
Intercept	.96	1.52	0.63	−11.09*	5.27	−2.10
Follow-up time	.03***	0.01	4.13	.02	0.02	1.03
Indigenous status	−.50	0.32	−1.54	−2.75	1.42	−1.94
Age at start	−.16	0.09	−1.76	.56*	0.26	2.14
Prior sexual and violent offenses	.08*	0.04	2.15	−.16	0.16	−0.99
Treatment success	.32	0.27	1.19	2.27*	1.11	2.05
Treatment duration (months)	−.04	0.03	−1.36	.01	0.07	0.15
Individual sessions (hr)	.02	0.02	1.15	.01	0.04	0.20
Family sessions (hr)	−.10**	0.03	−3.09	−.15	0.13	−1.08
Professional consultation (hr)	.03*	0.02	2.13	.07	0.05	1.30
Missed treatment appointments	−.01	0.04	−0.36	−.09	0.12	−0.80
Log likelihood = −112.85 (df = 22)
(Model 3) nonviolent reoffending
Intercept	6.55***	0.37	17.59	−6.18*	3.08	−2.01
Follow-up time	.01***	0.00	10.77	−.02**	0.01	−2.99
Indigenous status	.72***	0.08	9.19	−1.85**	0.70	−2.65
Age at start	−.30***	0.02	−14.26	.33	0.17	1.90
Prior sexual and violent offenses	.00	0.01	0.48	−.15	0.12	−1.26
Treatment success	−.24***	0.07	−3.52	2.20*	0.92	2.38
Treatment duration (months)	−.01	0.00	−1.79	.10*	0.04	2.23
Individual sessions (hr)	.00	0.00	−0.17	−.01	0.02	−0.63
Family sessions (hr)	−.01	0.01	−1.33	.05	0.04	1.33
Professional consultation (hr)	−.01*	0.00	−2.02	−.04	0.03	−1.62
Missed treatment appointments	−.06***	0.01	−5.24	−.12	0.08	−1.47
Log likelihood = −822.44 (df = 22)

p < .05. **p < .01. ***p < .001.

A higher frequency of total reoffending (count component) was significantly associated with a longer follow-up time, Indigenous status, younger at treatment start, unsuccessful treatment, fewer professional consultation hours, and fewer missed appointments. The odds of no reoffending (zero/logit component) were significantly increased for individuals with shorter follow-up times, non-Indigenous status, older at treatment start, successful treatment, and longer treatment duration.

For the GYFS group, only 19 (9.5%) individuals committed a sexual reoffense during the follow-up time, with 10 of these individuals only having a single sexual reoffending incident, which should be considered in interpreting the following results. A higher frequency of sexual reoffending (count component) was significantly associated with a longer follow-up time, non-Indigenous status, older age at treatment start, more prior sexual and violent offenses, being assessed as successfully completing treatment, more professional consultation hours, and more missed appointments. No significant associations were evident for the zero/logit component of the sexual reoffending model.

A higher frequency of violent reoffending was significantly associated with a longer follow-up time, more prior sexual and violent offenses, fewer family session hours, and more professional consultation hours. The odds of no reoffending were significantly increased for individuals who were older at treatment start and were assessed as successfully completing treatment.

Finally, a higher frequency of nonviolent reoffending was significantly associated with a longer follow-up time, Indigenous status, younger age at treatment start, unsuccessful treatment, fewer professional consultation hours, and fewer missed appointments. The odds of no reoffending were significantly increased for shorter follow-up times, non-Indigenous status, individuals assessed as successfully completing treatment, and a longer treatment duration.

Discussion

The aim of the study was to compare reoffending outcomes for YSOs who had participated in a specialized, field-based treatment program to reduce sexual recidivism, compared to TAU. Consistent with existing evidence (Caldwell, 2016; Schmucker & Lösel, 2015) and study hypotheses 1 and 2 respectively, there were low rates of sexual recidivism for the treatment (9.5%) and TAU comparison (10.8%) groups, and high rates of nonsexual recidivism for both groups. The high rate of general recidivism for YSOs suggests that treatment programs for sexually abusive youth must do more to target interventions that reduce rates of nonsexual reoffending post-treatment. Continued nonsexual offending by adolescents not only results in significant socio-economic costs but may also impact on the capacity of programs to effectively engage youth and promote meaningful change and further lock youth into persistent offending trajectories.

Only a small proportion of individuals committed a further sexual offense within the first 6 months following treatment/order completion. This is consistent with other evidence indicating that recidivism generally occurs during the developmental period of adolescence and in the first few years after sexual offending is detected (Caldwell, 2016; Worling et al., 2010). The third hypothesis of the study was not supported; sexual reoffending rates were equivalent between the treatment and TAU groups, including after controlling for prior offense history characteristics. We were unable to control for, or match, groups on index sexual offense characteristics, given these features are considered heavily in pre-intervention assessments of risk and therefore the selection and separation of the groups. As a result, the treatment group was systematically biased to include those individuals with more serious and/or persistent sexual offenses. It is possible that the absence of a difference in recidivism rates may reflect this systematic bias between the groups stemming from selection effects. It is also possible that the finding of equivalence in sexual recidivism rates between a high and relatively low risk group represents a possible treatment effect. However, significant caution is warranted in drawing this conclusion due to group matching limitations.

There was partial support for hypothesis 4, with unadjusted Kaplan-Meier and Cox regression survival results indicating that the treatment group had a lower rate of violent reoffending over time. However, this effect became not significant when controlling for covariates (i.e., Indigenous status, prior sexual and nonsexual offenses, and age at treatment/order start). This finding highlights a potential generalizable effect of the treatment program in reducing rates of violent reoffending by YSOs, which would be consistent with treatment outcomes of other specialized programs (e.g., Aebi et al., 2022; Worling et al., 2010). This suggests that there may be common features shared across treatment approaches that operate to reduce recidivism in general.

Historical offense-related factors continued to be significantly associated with reoffending outcomes in the presence of treatment factors. Higher frequencies of historical sexual and violent offenses before treatment were associated with higher rates of both sexual and violent reoffending, suggesting that individuals with more entrenched offending problems represent a greater challenge to treatment. Age at treatment start was consistently associated with reoffending frequency, but the direction of the relationship differed across reoffense type. For total and nonviolent reoffending, a younger age at start was associated with higher rates of reoffending, being consistent with findings that an early onset of offending is associated with more chronic/persistent offending pathways (DeLisi & Piquero, 2011). However, an older age at treatment was associated with a higher rate of sexual reoffending, which may represent a small group of older adolescents who continue to sexually offend into adulthood. These individuals may be distinct from the predominant group of youth whose sexual offending remains limited to adolescence.

Although higher frequencies of total and nonviolent/nonsexual reoffending were evident for Indigenous youth, this was not the case for sexual and violent reoffending. This provides some evidence that the treatment model characterizing the GYFS program may be a promising approach to reduce the disparity in Indigenous and non-Indigenous reoffending outcomes (Allard et al., 2015).

Clinician-rated treatment success was the most consistent treatment-related factor associated with reoffending outcomes. A higher volume of total and nonviolent/nonsexual reoffending after treatment was significantly associated with an individual being classified as unsuccessfully completing treatment. Further, the odds of non-reoffending for total, violent and nonviolent/nonsexual offenses were increased significantly for individuals classified as successfully completing treatment. However, a higher frequency of sexual reoffending was significantly associated with being classified as a treatment success, which could possibly reflect impression management of clinicians by the small number who sexually reoffended (e.g., minimizing perception of risk).

Implications for Evaluation and Clinical Practice

The results clearly highlight how the selection and matching of comparison groups can impact the findings of a study. The inability to construct a comparison group of equivalent risk level to the treatment group may have obscured potential treatment effects, emphasizing a major deficiency of incidental assignment designs in conducting evaluation research (Hanson et al., 2002). Future evaluations of “real world” programs like GYFS would benefit from more rigorous research designs, which could include random assignment to alternative treatments or treatment components (e.g., social skills training, relapse prevention planning), or controlling dosage levels across groups (Seto et al., 2008).

As experienced in the delivery of the current treatment program, adherence to the risk principle (Bonta & Andrews, 2007) in practice for delivering treatment presents significant clinical challenges, with high-risk clients often being the most difficult to engage in interventions. This highlights the importance of developing methods to both identify the high-risk subgroup at risk of poor treatment outcomes and facilitate their continued engagement in interventions. Greater effort and research are needed in developing and testing techniques and methods that facilitate the engagement of high-risk individuals and increase their chances of successfully completing and benefiting from treatment. This directly relates to the responsivity principle of effective treatment, where interventions should be designed explicitly to the learning styles, cognitive capacities, motivation, personality, and cultural background of clients (Bonta & Andrews, 2007).

Limitations

The most significant challenge experienced in the current evaluation was the identification of a suitable comparison group, including the inability to randomly assign individuals to treatment and control conditions. Despite the study groups being comparable on some key demographic and historical offense variables, the groups may have varied on some unmeasured variables that were associated with later recidivism outcomes. This impacts the internal validity of the findings (i.e., the ability to definitively conclude that treatment resulted in lower rates of recidivism), but alternatively may have strengthened the external validity of the study, with participants more accurately representing those found in typical clinical practice. Future studies may benefit from using comparison groups comprised of individuals receiving TAU or existing interventions, whereby new treatments may best be compared to current interventions with known positive effects as a benchmark. Randomization may then be used in such designs to improve the internal validity of findings without compromising ethical standards of withholding potentially effective treatments from clients.

Offenses reported to police were used as the estimate of recidivism, and this was likely to have been more sensitive to recidivism outcomes compared to convictions. This could have resulted in an overestimation of recidivism for some groups and introduced potential bias to the findings. Further, the measurement of reoffending did not take the harm associated with offenses into account, which is an important direction for future research. Measurement of harm associated with reoffending would allow for more detailed analysis of treatment effects, where analysis of offense volume may not capture potential changes in the nature of offending after intervention (e.g., volume of offending may not change, but the seriousness/harmfulness of offenses may decrease). Further, this evaluation adopted a time-to-first-event analytical approach (i.e., survival analysis), which does not capture the complete picture of reoffending. Future research would benefit from analytical approaches that can model the longitudinal course of offending (e.g., recurrent event analyses).

Conclusion

The results of this evaluation are consistent with evidence demonstrating equivocal results on the effectiveness of specialized treatment for YSOs to reduce sexual recidivism (Kettrey & Lipsey, 2018). The finding of equivalent sexual recidivism rates between a high-risk treatment and relatively low-risk comparison group illustrates of the drawbacks associated with incidental allocation evaluation designs. High rates of nonsexual recidivism emphasize the need for treatment programs to address more general reoffending that may lock youth into longer-term antisocial trajectories. Results provided preliminary evidence that the program may be effective for reducing violent recidivism among youth who commit sexual offenses. There remains a clear need for research to move toward being more specific in determining what works for whom, and under what circumstances (e.g., treatment setting, timing of intervention). A major challenge for clinical practice and evaluation will be the identification of the specific elements of treatment programs that are most strongly associated with effectiveness (e.g., reduction in recidivism), efficiency (e.g., dose of treatment), and the mechanisms through which these elements exert their effects on the lives of youth in conflict with the law.

Supplemental Material

sj-docx-1-ijo-10.1177_0306624X231219216 – Supplemental material for Assessing the Effectiveness of a Specialized, Field-Based Treatment Program for Youth Who Have Committed Sexual Offenses in an Australian Jurisdiction

Supplemental material, sj-docx-1-ijo-10.1177_0306624X231219216 for Assessing the Effectiveness of a Specialized, Field-Based Treatment Program for Youth Who Have Committed Sexual Offenses in an Australian Jurisdiction by James M. Ogilvie, Nadine McKillop, Jesse Cale, Troy Allard, John Rynne and Stephen Smallbone in International Journal of Offender Therapy and Comparative Criminology

Supplemental Material

sj-docx-2-ijo-10.1177_0306624X231219216 – Supplemental material for Assessing the Effectiveness of a Specialized, Field-Based Treatment Program for Youth Who Have Committed Sexual Offenses in an Australian Jurisdiction

Supplemental material, sj-docx-2-ijo-10.1177_0306624X231219216 for Assessing the Effectiveness of a Specialized, Field-Based Treatment Program for Youth Who Have Committed Sexual Offenses in an Australian Jurisdiction by James M. Ogilvie, Nadine McKillop, Jesse Cale, Troy Allard, John Rynne and Stephen Smallbone in International Journal of Offender Therapy and Comparative Criminology

Footnotes

Acknowledgements

The authors gratefully acknowledge the assistance of the Queensland Police Service and the Department of Children, Youth Justice and Multicultural Affairs. The views expressed are not necessarily those of the Queensland Police Service or the Department of Children, Youth Justice and Multicultural Affairs or the Australian Research Council and any errors of omission or commission are the responsibility of the authors. The authors take responsibility for the integrity of the data, the accuracy of the data analyses, and have made every effort to avoid inflating statistically significant results.

Author Contributions

SS, NM, and TA conceived the research idea and obtained funding to complete the data collection. JO prepared and analyzed the data, interpreted results, and developed the initial manuscript draft with guidance from SS. NM, TA, JC, and JR contributed to revisions of the manuscript. All authors read and approved the final manuscript.

Availability of data and materials

The data for the study are held by the Griffith Youth Forensic Service (GYFS) at Griffith University. Due to privacy, clinical sensitivity, and ethical considerations, the data cannot be shared without direct approval from GYFS University and the QPS. Any researcher interested in accessing the data can contact GYFS (gyfs@griffith.edu.au).

Consent for publication

Not applicable.

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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Australian Government through the Australian Research Council Discovery Program (Project Number DP110102126).

Ethical standards

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation, and with the Helsinki Declaration of 1975, as revised in 2008.

Ethics approval and consent to participate

The study has received ethical approval from Griffith University Human Research Ethics Committee (GU Ref No.: CCJ/07/11/HREC). Informed written consent from participants was waived due to the use of secondary deidentified data.

ORCID iDs

James M. Ogilvie

Nadine McKillop

Jesse Cale

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Aebi

Krause

Barra

Vogt

Vertone

Manetsch

Imbach

Endrass

Rossegger

Schmeck

Bessler

(2022). What kind of therapy works with juveniles who have sexually offended? A randomized-controlled trial of two versions of a specialized cognitive behavioral outpatient treatment program. Sexual Abuse, 34(8), 973–1002. https://doi.org/10.1177/10790632211070804

Allard

Rayment-McHugh

Adams

Smallbone

McKillop

(2015). Responding to youth sexual offending: A field-based practice model that “closes the gap” on sexual recidivism among Indigenous and non-Indigenous males. Journal of Sexual Aggression, 22(1), 82–94. https://doi.org/10.1080/13552600.2014.1003107

Behrendt

Cunneen

Libesman

Watson

(2019). Aboriginal and torres strait islander legal relations. Oxford University Press.

Bonta

Andrews

D. A.

(2007). Risk-need-responsivity model for offender assessment and rehabilitation. Rehabilitation, 6(1), 1–22.

Caldwell

M. F.

(2016). Quantifying the decline in juvenile sexual recidivism rates. Psychology, Public Policy, and Law, 22(4), 414–426. https://doi.org/10.1037/law0000094

DeLisi

Piquero

A. R.

(2011). New frontiers in criminal careers research, 2000–2011: A state-of-the-art review. Journal of Criminal Justice, 39(4), 289–301. https://doi.org/10.1016/j.jcrimjus.2011.05.001

Fortune

C. A.

Lambie

(2006). Sexually abusive youth: A review of recidivism studies and methodological issues for future research. Clinical Psychology Review, 26(8), 1078–1095. https://doi.org/10.1016/j.cpr.2005.12.007

Hanson

R. K.

Gordon

Harris

A. J. R.

Marques

J. K.

Murphy

Quinsey

V. L.

Seto

M. C.

(2002). First report of the collaborative outcome data project on the effectiveness of psychological treatment for sex offenders. Sexual Abuse: A Journal of Research and Treatment, 14(2), 169–194. https://doi.org/10.1177/107906320201400207

Kettrey

H. H.

Lipsey

M. W.

(2018). The effects of specialized treatment on the recidivism of juvenile sex offenders: A systematic review and meta-analysis. Journal of Experimental Criminology, 14(3), 361–387. https://doi.org/10.1007/s11292-018-9329-3

10.

R Core Team. (2021). R: A language and environment for statistical computing (version 4.1.1) [computer software]. R Foundation for Statistical Computing. http://www.R-project.org/

11.

Reitzel

L. R.

Carbonell

J. L.

(2006). The effectiveness of sexual offender treatment for juveniles as measured by recidivism: A meta-analysis. Sexual Abuse: A Journal of Research and Treatment, 18(4), 401–421. https://doi.org/10.1007/s11194-006-9031-2

12.

Schmucker

Lösel

(2015). The effects of sexual offender treatment on recidivism: An international meta-analysis of sound quality evaluations. Journal of Experimental Criminology, 11(4), 597–630. https://doi.org/10.1007/s11292-015-9241-z

13.

Seto

M. C.

Marques

J. K.

Harris

G. T.

Chaffin

Lalumière

M. L.

Miner

M. H.

Berliner

Rice

M. E.

Lieb

Quinsey

V. L.

(2008). Good science and progress in sex offender treatment are intertwined: A response to Marshall and Marshall (2007). Sexual Abuse: A Journal of Research and Treatment, 20(3), 247–255. https://doi.org/10.1177/1079063208317733

14.

Therneau

(2020). A package for survival analysis in R (R package version 3.2-3) [computer software]. https://CRAN.R-project.org/package=survival

15.

Tyler

Gannon

T. A.

Olver

M. E.

(2021). Does treatment for sexual offending work? Current Psychiatry Reports, 23(8), 51. https://doi.org/10.1007/s11920-021-01259-3

16.

Walker

D. F.

McGovern

S. K.

Poey

E. L.

Otis

K. E.

(2004). Treatment effectiveness for male adolescent sexual offenders: A meta-analysis and review. Journal of Child Sexual Abuse, 13(3–4), 281–293. https://doi.org/10.1300/J070v13n03_14

17.

Worling

J. R.

Litteljohn

Bookalam

(2010). 20-year prospective follow-up study of specialized treatment for adolescents who offended sexually. Behavioral Sciences & the Law, 28(1), 46–57. https://doi.org/10.1002/bsl.912

18.

Zeileis

Kleiber

Jackman

(2008). Regression models for count data in R. Journal of Statistical Software, 27(8), 1–25. https://doi.org/10.18637/jss.v027.i08

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB