Sex Offender Treatment Outcome,Actuarial Risk,and the Aging Sex Offender in Canadian Corrections

Abstract

The present study is an examination of sex offender treatment outcome in a large national cohort of Canadian Federally incarcerated sex offenders followed up an average of 11.7 years postrelease. A brief actuarial risk scale (BARS), which predicted sexual and violent recidivism, was created for the purposes of the present study to control for risk-related differences between treated and untreated offenders. In total, 732 offenders were identified as having completed (n = 625) or not attended (n = 107) a sex offender treatment program and for whom sufficient information was available to complete the scale. Controlling for risk and individual differences in follow-up time using Cox regression survival analyses and an 8-year fixed follow-up period, treated sex offenders demonstrated significantly lower rates of violent, but not sexual, recidivism. When the treated and untreated groups were stratified by risk level, significant differences were observed only among moderate or high risk offenders. Some significant group differences also emerged on indicators of recidivism severity, with treated offenders demonstrating slower times to sexual reoffense and lower scores on a quantified metric of sexual and violent recidivism severity after controlling for risk. Differences in recidivism base rates between treated and untreated offenders were also larger in magnitude for younger offenders (i.e., under age 50 at release), than for older offenders; however, interactions between age and treatment were not found. The findings are consistent with the risk principle and have possible implications regarding the dynamic nature of sexual violence risk.

Keywords

sex offender treatment outcome recidivism actuarial risk age

The assessment of sex offenders to identify high risk individuals to be targeted for treatment and risk management services has important health, public safety, criminal justice, and policy implications. Substantial advances have been made over the past two decades in research and clinical practice to more accurately assess risk and to intervene accordingly to reduce sexual violence. Part of the challenge involves not only in identifying factors that exacerbate risk but also in identifying the means by which risk can be reduced or managed.

Sex Offender Treatment Outcome: A Brief Review

Meta-analytic reviews of sex offender treatment programs have provided support for the efficacy of evidence-based treatments for reducing sexual recidivism and other violent outcomes (Gallagher, Wilson, Hirschfield, Coggeshall, & MacKenzie, 1999; Hall, 1995; Hanson, Bourgon, Helmus, & Hodgson, 2009; Hanson et al., 2002; Löesel & Schmucker, 2005). The most inclusive of these (Löesel & Schmucker, 2005) examined 69 studies (N = 22,181) and found a six percentage point reduction (a relative 37% reduction) in sexual recidivism with cognitive-behavioral, behavioral, and biomedical approaches showing the largest reductions. Most recently, Hanson et al. (2009), examined 22 studies that met minimum standards for methodological quality and found an eight percentage point difference (10.9% and 19.2%, respectively, or a relative 43% reduction) between treated offenders and untreated controls.

The common thread among effective correctional programs are the principles of risk, need, and responsivity (also known as RNR; Andrews et al., 1990; Andrews & Bonta, 2010a, 2010b). A recent comprehensive survey of sex offender treatment programs in North America found that the majority of programs tended to be cognitive behavioral and social learning in orientation, incorporate a group treatment component, and target treatment domains known to be associated with criminal risk such as sexual deviance, intimacy deficits, attitudes, and associates supportive of violence (McGrath, Cumming, Buchard, Zeoli, & Ellerby, 2010) This suggests that many of these programs are informed, at least in part, by RNR principles.

Methodological Issues in the Examination of Sex Offender Treatment Outcome

Hanson, Broom, and Stephenson (2004) noted that the quality of a meta-analytic study hinges on the quality of the studies included in it. Uncontrolled variables have the ability to confound the relationship between treatment participation and outcome. For example, Harris and Hanson (2004), in studying a large Canadian sample of sex offenders followed up for more than 15 years postrelease (N = 4,724), found that variables such as having an extrafamilial child victim, any male victim, a previous conviction for a sex offense, or being under age 50 at time of release, each served to increase base rates of sexual offense recidivism. Not surprisingly, longer follow-ups generated higher rates of recidivism as individuals spending greater periods of time in the community have greater opportunity to eventually reoffend (Harris & Hanson, 2004).

Given the aforementioned considerations in conducting sex offender treatment outcome research, an important element is the matching of treatment and comparison groups on important risk-related variables (e.g., age, prior sex offenses). Hanson et al. (2004) for instance, suggested creating an actuarial scale from available risk variables to control for risk. This could be achieved either by matching treatment and comparison groups or entering the risk score as a covariate into a regression model. A second important consideration is the necessity to control for individual differences in follow-up time. This can be done through statistical techniques that adjust for individual differences in follow-up (e.g., survival analysis) or methodological adjustments to ensure that individuals are followed up for similar periods of time, and thus have equal temporal opportunity to reoffend. An example of this strategy is the use of fixed follow-up intervals (e.g., Hanson et al., 2004; Olver, Wong, & Nicholaichuk, 2009).

A third issue relates to the nature of the outcome variable. Recidivism in its various forms, particularly sexual recidivism, has been the primary criterion of interest in studies such as this. In most cases, a single recidivism event, most commonly in the form of official criminal charges or convictions, has been used. Aside from the shortcomings inherent in using official criminal records (e.g., the problem of undetected recidivism), a binary recidivism variable tends not to capture offense severity. Multiple offenses may indicate increased severity, but not necessarily (e.g., a single sexual homicide vs. numerous incidents of exposure). The use of other indicators of offense severity or harm, such as aggregate sentence length for new offenses, may be viable. In Canada, during the historical period from which the current sample was drawn, more serious crimes tended to be penalized with longer sentences (Di Placido, Simon, Witte, Gu, & Wong, 2006; Nicholaichuk, Gordon, Gu, & Wong, 2000). Although eliminating sexual recidivism outright is an important primary objective, decreasing the seriousness of new crimes may also be a marker (albeit a less desirable one) of treatment progress. For instance, the concept of harm reduction emphasizes decreasing the frequency and severity of a maladaptive behavior (see Marlatt & Witkiewitz, 2010).

A fourth consideration is the issue of randomization. Most studies in the sex offender treatment outcome field have been nonrandomized quasi-experimental evaluations. One of the few Randomized controlled trials (RCTs), the Sex Offender Treatment Evaluation Project (SOTEP), in its most recent evaluation reported no reductions in sexual recidivism among treatment completers relative to two sets of untreated controls (Marques, Wienderanders, Day, Nelson, & van Ommeren, 2005). Despite the scientific rigor of randomization, potential ethical and pragmatic considerations stand as important challenges in conducting sex offender RCTs. (See Marshall & Marshall, 2007, 2008, and Seto et al., 2008 for a discussion of RCTs, particularly with respect to SOTEP.) A final methodological consideration concerns the nature of the treatment program under study. Hanson et al. (2009) found that sex offender programs adhering to the RNR principles generated larger reductions in recidivism and that the greater the number of principles adhered to (e.g., one vs. two vs. three), the larger the reductions.

Sex Offender Treatment Outcome, Risk Reduction, and Aging

Sex offender treatment is predicated on the assumption that risk is dynamic (i.e., it can change, for better or for worse), but interestingly, this assumption has been largely untested. Although risk has been treated as an important covariate in sex offender treatment outcome research, few studies to our knowledge have systematically examined the possible reduction in degree of violence in sex offender treatment outcome studies. It is also worth noting that most studies that have examined the predictive accuracy of dynamic risk variables have focused on single time-point estimates of risk (Douglas & Skeem, 2005) although a small but growing literature has found changes in treatment-relevant domains to be associated with reductions in sexual recidivism among treated sex offenders (Beggs & Grace, 2011, Nicholaichuk et al., 2000, Olver, Wong, Nicholaichuk, & Gordon, 2007).

A second prominent change agent involved in the reduction and management of sexual violence is aging. The extant literature has found increasing age to be associated with a decline in most, if not all, forms of antisocial activity, including sexual offending (Barbaree, Langton, & Blanchard 2007; Hanson, 2002; Harris & Hanson, 2004; Sampson & Laub, 2003; Skelton & Vess, 2008; Thornton, 2006). The mechanism involved in risk mitigation among older offenders is a topic of some debate (e.g., see explanations advanced by Hanson, 2002, and Harris & Rice, 2007). Over the past decade, Canada’s Federal correctional department, the Correctional Service of Canada (CSC), has experienced an increase in its elderly offender population. The percentage of incarcerated offenders aged 50 and above increased from 11.5% in 1997-1998 to 19% in 2009-2010, a relative increase of 65% (Corrections and Conditional Release Statistical Overview, 2004, 2010). It can be reasonably anticipated that some of these older sex offenders participated in sex offender treatment. To our knowledge, although age has been treated as a covariate in sex offender treatment outcome research, age has not been examined as a possible moderator of sex offender treatment outcomes. Put another way, do older sex offenders derive fewer benefits from sex offender treatment given that they are more likely to be lower risk, as per the risk principle (Andrews & Bonta, 2010a, 2010b)?

Previous Evaluations of CSC Sex Offender Treatment Programs

There have been several previous evaluations of the sex offender programs designed by and operated within the Correctional Service of Canada (CSC). These reviews have generally evaluated individual sex offender treatment programs delivered either through designated regional psychiatric hospitals (referred to as Regional Treatment Centres) or prison-run programs. In many (but not all) instances, the results have supported the effectiveness of these programs in reducing sexual or other forms of recidivism (e.g., Barbaree, 2005; Langton, Barbaree, Harkins, & Peacock, 2006; Looman, Abracen, & Nicholaichuk, 2000; Looman, Abracen, Serin, & Marquis, 2005; Nicholaichuk et al., 2000; Olver et al., 2009, Seto & Barbaree, 1999).

In 2000, CSC implemented the National Sex Offender Program (NaSOP; Yates, Goguen, Nicholaichuk, Williams, & Long, 2000), which entailed a revision of sex offender treatment programs to reflect developments in research and practice and was implemented service wide. Cortoni and Nunes (2007) conducted an evaluation of the NaSOP (Low and Moderate Intensity streams) on 347 sex offenders who participated in the program in institutions across Canada between 2000 and 2004 and followed up 2 years postrelease. Outcome data available for 222 treated men were compared against a control group of 137 untreated CSC sex offenders from Motiuk and Porporino (1993). Very low recidivism base rates were observed for NaSOP participants compared to untreated men with respect to sexual (1.1% vs. 4.6%), violent (2.0% vs. 17.8%), and general (6.8% vs. 36.5%) recidivism. However, it is important to note that the untreated men scored significantly higher (M = 3.91, SD = 1.88 vs. M = 2.37, SD = 1.74, respectively, d = .85) on the Static-99 (Hanson & Thornton, 1999) and thus were a higher risk sample. Cox regression survival analyses, controlling for Static-99 score and time at risk, found the NaSOP program approached significant reductions in sexual recidivism and was linked to significant reductions in violent (including sexual) and general recidivism.

Present Study

In 2004, a national evaluation of sex offender treatment outcomes was conducted for the CSC’s Evaluation Branch (Gu, Wong, & Nicholaichuk, 2004). The entire cohort of 2,401 sex offenders under CSC jurisdiction who reached their Warrant Expiry Date (WED; the end of their sentences) during the fiscal years of 1997-1998, 1998-1999, and 1999-2000 was included. Although the WED date was used as the criterion to select the sample, in fact, the majority of these individuals were released on their Statutory Release Date (SRD), occurring at two thirds of their actual sentence length. Over a 3-year follow-up, lower base rates of sexual reconvictions were reported for the 688 offenders who had completed sex offender treatment (5.4%) versus the 112 offenders who did not (17%). Treatment information was not available for the remaining individuals. The original evaluation was limited by a relatively short follow-up time and the lack of controls for important risk-related variables (e.g., offense history, age) between the treated and untreated groups. The resulting report concluded that, “Without further analyses, it is not appropriate at this point to conclude the difference in reconviction rate between the two groups is the result of participating in sex offender treatment, although the result is consistent with such an interpretation” (Gu et al., 2004, p. 10).

The present study is an examination of sex offender treatment outcome in this cohort of offenders while attempting to address some of the previous design’s limitations through utilizing updated outcome data and a longer (11.7-year) follow-up interval. Enhanced statistical and methodological controls were also employed to account for risk-related differences between treatment and comparison groups. Given that the impact of actuarial risk and increasing age on recidivism is now recognized (e.g., Hanson, 2006), it was anticipated that these variables could at least partially moderate any relationship of treatment completion to outcome. The current investigation contributes to the sex offender treatment outcome and risk assessment literature by means of examining the moderating influences of actuarial risk and age on treatment outcome.

Method

Participants

Participants included 2,401 federally incarcerated sex offenders who reached WED within the 1997-1998, 1998-1999, and 1999-2000 fiscal years. As was noted above, typically their actual date of release preceded their WED. The sample was followed up an average of 11.7 years (SD = 1.4) postrelease. Information concerning whether or not the individual had participated in sex offender treatment was available for 809 sex offenders. Of these offenders, outcome data were available for 753 sex offenders given that records could not be obtained for 50 individuals and three died prior to, or very shortly after release. Complete demographic, criminal history, and offense relevant information were available for 732 sex offenders and used to develop a Brief Actuarial Risk Scale (BARS) to compare a treated group (n = 625) and an untreated (n = 107) comparison group. Descriptive data for participant characteristics are provided in Table 1 for the treatment and comparison groups.

Table 1.

Treatment and Comparison Group Demographic, Criminal History, and Offense-Related Variables.

	Sex offender treatment (n = 625)		Comparison (n = 107)
Variable	%	n	%	n	p
Single	35.5	222	50.5	54	.003
Unrelated victim	68.5	428	79.4	85	.022
Male victim	18.2	114	15.9	17	.558
Under age 35	30.9	193	46.7	50	.000
Four or more prior sentencing dates	1.8	11	2.8	3	.466
Prior sexual conviction	29.6	185	40.2	43	.029

	M	SD	M	SD	p

Follow-up years	11.7	1.3	11.5	1.7	.119
Age at release	42.2	11.7	37.5	11.6	.001
BARS score	1.8	1.2	2.4	1.2	<.001

Note: BARS = Brief actuarial risk scale.

Treatment Program

Treated offenders attended a sex offender treatment program in a CSC institution, which was located in either a prison or a Regional Treatment Centre (i.e., a psychiatric hospital serving as a designated treatment facility for CSC inmates) in one of five geographic regions across Canada (Pacific, Prairie, Ontario, Quebec, and Atlantic regions). Although the specific location of the sex offender treatment program and specific program details were not available during the time of treatment participation, sex offender treatment programs within CSC were based on the “what works” principles of effective correctional treatment (Correctional Service of Canada, 1996). Specifically, they were guided by the CSC Standards and Guidelines for the Provision of Services to Sex Offenders (1996). Sex offender programs were grounded in the principles of risk, need, and responsivity, and comprehensively addressed a number of sex offending related criminogenic needs (e.g., deviant arousal and fantasy, attitudes and cognitive distortions, social competence). They operated primarily according to a group format and provided opportunities for individual services. The standard CSC protocol at the time supported the principle that a registered psychologist be involved in service provision. This could have taken the form of cofacilitating groups, providing individual therapy, supervising program facilitators, and conducting or quality controlling psychological assessments or some combination of the above.

Procedure

Data Collection

The data collected for the original 2004 study included variables such as victim and offender demographics, offense history, and treatment completion if available. Data collection was completed by two trained research assistants over a 1-year period using the Offender Management System (OMS), a national electronic database created by the Correctional Service of Canada (CSC), to store offender information.

Victim information

Victim information was taken from the OMS Criminal Profile Report that recorded information such as victim gender, victim age, the number of adult, teenage, and child victims, the relationship of victims to the offenders, and access to potential victims. The victim demographics which were coded included victim age, the number of adult, teenage, and child victims, and the relationship of the victim to the offender. Victim gender was recorded as “female,” “male,” “both,” or “unknown” and was dichotomized for analyses (i.e., male victim vs. no male victim). A child was defined as anyone under the age of 12, an adolescent was anyone between age of 13 and 18, and an adult victim was defined as anyone above the age of 18. Victim Relationship was determined using the extra- and intrafamilial designation. An intrafamilial relationship included close and distant family members, including adopted children, foster children, family members of a common law partner, or anyone with whom marriage is normally prohibited. An extrafamilial relationship included the friends of the offender or of his child(ren), neighbors, coworkers, complete strangers, or victims who the offender had only known a very short time before assaulting them.

Offense history

Official sexual and nonsexual offending history was obtained from the Canadian Police Information Centre (CPIC). The offender’s index offence was defined as the most recent sexual conviction recorded on his CPIC record for the sentence he was serving prior to his WED. Sexual offence history (official history) included any conviction for sexual offenses listed in the CPIC database. Prior sentencing dates were the number of separate federal sentencing dates on the record prior to the index sexual offense(s).

Marital status

Marital status was initially coded as “single,” “married/common law,” “divorced/separated,” or “widowed” and then dichotomized (i.e., never married vs. married or equivalent) for these analyses.

Treatment program data

Treatment program information was coded from Program Performance Reports, which were obtained from OMS to determine if the offender had completed any of three types of core treatment programs: Sex Offender Treatment programs, Anger Management programs, and Cognitive Skills programming.¹ Completion of a treatment program was recorded as yes-no (i.e., completed/did not complete). If no information was available regarding whether or not the offender had taken a particular program, the program status field was coded as “no information.” For the present study, sex offender treatment was the primary program of interest and on which all analyses were based. Sex offender treatment status was coded as a simple dichotomous (yes-treated, no-untreated) variable. Untreated offenders were defined as not having a documented completion of a sex offender treatment program; they may have attempted or completed other programs. Information regarding whether any of the untreated offenders had attempted but failed to complete a sex offender treatment program was not available. Treated offenders were those individuals whose treatment completion was documented in OMS.

Outcome Variables

The primary outcome variables of interest were sexual and violent recidivism. For the present study, recidivism data were coded by a trained research assistant and the first two authors from official CPIC records, updated between September 18, 2009, and November 2, 2009. All were blind to the treatment status of the participants. As per Nicholaichuk et al. (2000), the SRD was used as the default date of release, with the exception of 120 (16.4%) offenders released at warrant expiry and thus for whom their WED was used. Sexual recidivism was defined as any new conviction for an offense that was clearly sexual in nature (e.g., sexual assault, indecent assault, sexual interference). Violent recidivism was defined as any new conviction for an offense that involved any actual, attempted, or threatened physical harm to the victim, including sexual offenses. Outcome variables were coded in a binary (1-0, yes-no) manner.

Three additional indictors of harm reduction were coded. First, the time to first sexual and/or violent conviction was recorded to perform survival analyses (see “Analyses” below) as well as to provide an indication of the time to recidivism. Second, aggregate sentence lengths for new sexual and violent offenses, were computed and used as a proxy indicator regarding the degree of violence associated with the offense. This was based on the observation that more serious offenses tended to result in longer sentences. Larger number of new offenses also tended to result in longer periods of incarceration (Nicholaichuk et al., 2000). Offenders who were declared Dangerous Offenders (DO; an indeterminate sentence, Sec. 753 Canadian Criminal Code) on reconviction were given a default sentence of 14.4 years. This was based on the research by Nicholaichuk, Olver, Gu, and Takahashi (in press), which found this to be the average time spent in custody by Canadian DOs prior to their conditional release (although only four individuals out of 328 eligible offenders were released during the course of the last 20 years of the study period). Third, the Cormier-Lang system of offense severity (see Quinsey, Rice, Harris, & Cormier, 1998) was scored by the first author on all new cumulative sexual and violent convictions to generate total scores of recidivism severity. If treatment is linked to harm reduction, then it stands that treated offenders should have lower Cormier-Lang scores for sexual and violent recidivism than the comparison group.

Brief Actuarial Risk Scale

Following the recommendations of Hanson et al. (2004), to control for differences in risk level between the treatment and control groups, a Brief Actuarial Risk Scale (BARS) was created. The scale consisted of six static risk variables with demonstrated relationships to sexual recidivism. Although insufficient data had been originally coded in the original data collection to allow scoring on Hanson and Thornton’s (1999) Static-99, the items selected were informed by this tool as well as the recently revised Static-99-R (cf. Helmus, Thornton, Hanson, & Babchishin, 2012). The resulting items were prior sexual convictions, unrelated victim, male victim, young age at release, single marital status, and prior sentencing dates. All items were rescaled in a binary manner with scores of “1” indicating presence of the risk factor and “0” indicating absence of the risk factor. The Static-99 scoring rules informed the scoring of three items: any unrelated victim, male victim, and prior sentencing dates. The Static-99-R age cutoff of 35 was applied to code the “young age at release” item while a 0 was assigned to age categories of 35 and higher. The two remaining items were scored in a manner that would maximize reliability given the limited amount of detail available in the existing data set: “Prior sexual convictions” was coded in a binary manner with a score of “1” assigned to any official previous sexual offense history, and “single marital” status was coded such that any indication that the individual was single and had never been married was assigned a score of “1” (information concerning the duration of the longest live-in relationship was not available). Scores ranged from 0 to 5 with a mean of 1.89 (SD = 1.26). The predictive accuracy of the BARS was evaluated using the Receiver Operating Characteristic (ROC) Area Under the Curve (AUC) values. In the current data set, the scale significantly predicted both sexual, AUC = .71 (95% CI = [.66, .76]), and violent, AUC = .69 (95% CI = [.65, .73]), recidivism. This level of predictive accuracy is consistent in magnitude to other actuarial sex offender measures reported in meta-analytic reviews (Hanson & Morton-Bourgon, 2009).

Analyses

Four sets of analyses are described below. A fifth post hoc set of analyses was conducted to contextualize the findings from the initial four sets of analyses.² First, Cox regression survival analyses were the main procedures used to compare sexual and violent recidivism rates of treated and untreated offenders given that it can be used to control for actuarial risk and individual differences in follow-up time. In this case, survival time was defined as the time period from an offender’s release to his conviction for a new sexual or violent offense, or in the case of nonrecidivists, the time between release and the data collection date; time spent in custody for other offenses not related to the outcome variable of interest was subtracted from the total survival time. The BARS was used as a covariate in these analyses to control for individual differences in risk. In the second set of analyses, treated and untreated offenders were stratified by risk and compared using both fixed follow-up intervals and Cox regression survival analyses. When offender groups are stratified, this can result in a loss of power due to shrinking cell sizes. On the other hand, the advantage of this strategy is that it permits an examination of sexual and violent recidivism base rates at different levels of risk, thus providing a finer grained examination of possible group differences. The fixed follow-up data were analyzed by means of chi-square analyses, which were also used to examine differences in sexual and violent recidivism between groups, while odds ratios (OR) were used to examine the magnitude of possible treatment effects using the formula (recid_t/nonrecid_t)/(recid_c/nonrecid_c) (see Hanson et al., 2002). Odds ratios under 1.0 indicate lower recidivism in the treated sample relative to the comparison group, while an OR of 1.0 indicates that there is no difference between treatment and control groups. A value above 1.0 would indicate higher recidivism rates in the treated group.

The third set of analyses examined indicators of possible harm reduction. These were operationalized to provide three measures of outcome severity: time to new sexual or violent conviction, aggregate sentence length for new offenses, and Cormier-Lang offense severity score for new sexual and/or violent offenses. A factorial ANOVA of risk and treatment groups, as well as one-way ANOVAs within risk groups, were conducted. Cohen’s d was also computed as a measure of the magnitude of the difference, in standard deviation units, between treated and untreated groups. In these analyses, we also controlled for risk as a continuous variable through analysis of covariance (ANCOVA). The fourth set of analyses employed the survival analysis procedure to examine violent and sexual recidivism failure rates as a function of risk and age among treated and untreated offenders. The purpose of these analyses was to examine the extent to which possible treatment effects might be more evident among a younger, and hence higher risk, cohort of offenders, than an older and potentially lower risk group. We also examined the extent to which age at release may moderate possible treatment outcome findings. We followed with a Cox regression survival analysis to examine whether age and treatment participation were associated with reductions in sexual and violent recidivism, taking into account risk (minus the age variable) and a possible three-way interaction amongst risk, age, and treatment condition. The fifth and final set of analyses revisited the earlier national evaluations of CSC’s sex offender programs by employing a fixed 2-year follow-up as in Cortoni and Nunes (2007). We examined frequencies of sexual and violent recidivism among treated and untreated offenders during this follow-up and used sequential Cox regression survival analysis to examine treatment outcome while controlling for baseline risk.

Results

Sex Offender Treatment Outcome Controlling for Actuarial Risk

The sample of 732 treated and untreated sex offenders was followed up a mean of 11.7 years (SD = 1.4) postrelease. The base rate of sexual offense recidivism during this time period was 13.7% (n = 100) and the violent recidivism rate was 32.4% (n = 237). As reported in Table 1, the treated cohort of offenders had a significantly lower mean static actuarial (BARS) score than the untreated group and thus was a significantly lower risk group. The magnitude of the difference (0.54) was significant (p < .001; d = .29).³

Cox regression survival analysis

Cox regression survival analysis was subsequently used to examine sexual and violent recidivism failure rates for treated and untreated sex offenders while controlling for static actuarial risk and individual differences in follow-up time. The BARS score and the dichotomous treatment group variable were entered in the first step (Block 1) followed by the interaction term in the second step (Block 2). As reported in Table 2 significant differences were observed between treated and untreated offenders in terms of violent, but not sexual recidivism (Block 1). The Risk (BARS) × Treatment interaction was not significant for either outcome. The BARS significantly and independently predicted both outcomes.

Table 2.

Cox Regression Survival Analysis: Sex Offender Treatment Outcome for Sexual and Violent Recidivism Controlling for Brief Actuarial Risk Scale (BARS).

						95% CI
Variable combination	B	SE	Wald	p value	e^B	Lower	Upper
Sexual recidivism
Block 1
BARS score	.549	.083	43.97	.000	1.731	1.472	2.036
Treatment	−.357	.246	2.10	.147	.700	.432	1.134
Block 2
BARS score	.739	.210	12.33	.000	2.093	1.386	3.161
Treatment	.317	.744	.18	.670	1.373	.320	5.900
BARS × Treatment	−.227	.229	.98	.322
Violent recidivism
Block 1
BARS score	.434	.053	67.25	.000	1.544	1.391	1.712
Treatment	−.531	.158	11.32	.001	.588	.431	.801
Block 2
BARS score	.404	.121	11.11	.001	1.498	1.181	1.901
Treatment	−.629	.389	2.61	.106	.533	.249	1.144
BARS × Treatment	.037	.135	.07	.786

Note: N = 732.

Risk stratified group comparisons: Fixed follow-ups

To permit a finer grained examination of the relationship of risk and treatment completion to outcome, group comparisons were conducted at each individual score on the BARS employing a fixed 8-year follow-up given that this was the longest time-period available to retain the majority (n = 720 or 98.4%) of the participants. Given that there were few offenders in either group with BARS scores of 5 or higher, these were collapsed with offenders receiving a score of 4 into a single group. The difference in mean raw score for the 4+ risk group for treated (M = 4.16, SD = .37) and comparison (M = 4.10, SD = .31) groups was not significant, F(1, 74) = 0.43, p = .514. As reported in Table 3, significant differences were observed between the treated and untreated cohorts for sexual (10.7 vs. 20.2%, respectively, OR = .47) and violent (26.5 vs. 44.2%, respectively, OR = .45) recidivism across the total sample when risk was uncontrolled. When group differences were examined at each BARS score, significant differences in 8-year rates of sexual recidivism were observed only between the highest risk cohort (i.e., the 4+ group) of treated and untreated offenders (p = .027, OR = .30); no significant differences were observed among any of the other, lower risk cohorts. No significant differences were observed between treated and untreated sex offenders with respect to violent recidivism at any of the BARS thresholds (OR = .35 to .77).

Table 3.

Sexual and Violent Recidivism Rates of Treated and Untreated Offenders at Different Levels of Risk Over a Fixed 8-Year Follow-up.

	Sex offender treatment (n = 616)		Comparison (n = 104)
Variable	%	n	%	n	Odds ratio	p value
Sexual recidivism
BARS score
0	2.8	3/106	0.0	0/8	−	.630
1	4.2	6/142	11.8	2/17	.33	.179
2	13.1	22/168	10.7	3/28	1.26	.727
3	16.7	24/144	22.6	7/31	.69	.434
4+	19.6	11/56	45.0	9/20	.30	.027
Total sample	10.7	66/616	20.2	21/104	.47	.006
Violent recidivism
BARS score
0	9.4	10/106	12.5	1/8	.73	.777
1	12.7	18/142	29.4	5/17	.35	.064
2	33.3	56/168	39.3	11/28	.77	.539
3	36.8	53/144	51.6	16/31	.55	.126
4+	46.4	26/56	65.0	13/20	.47	.154
Total sample	26.5	163/616	44.2	46/104	.45	.000

Note: An odds ratio statistic cannot be computed for sexual recidivism outcome for actuarial scores of 0, given the denominator derived from the comparison group is 0/8 = 0. BARS = Brief actuarial risk scale.

Risk stratified group comparisons: Survival analysis

Survival analyses were subsequently conducted to examine failure rates for sexual and violent recidivism over time among broader risk categories of treated and untreated offenders. Given the small cell sizes for some of the comparisons by individual risk score, the risk scores on the BARS were collapsed together to create three risk groups of low (score 0-1), moderate (score 2-3), and high (score 4+). There were no significant differences in mean BARS scores between treated and untreated offenders in the low, F(1, 280) = 0.75, p = .388, moderate, F(1, 372) = 1.13, p = .289, or high (as reported above) risk groups. Survival curves illustrating sexual recidivism failure rates for the treated and untreated risk groups are presented in Figure 1. The high risk untreated offender group had a significantly higher and faster rate of sexual recidivism than all risk/treatment groups, including the high risk treated group, χ² (1, N = 76) = 4.01, p = .045. The treated high risk group and untreated moderate risk group did not have significantly different sexual recidivism failure rates, χ²(1, N = 116) = 1.57, p = .211. No significant differences in failure rates of sexual recidivism were observed between treated and untreated offenders for both the low and moderate risk groups. Survival analyses were subsequently repeated for violent recidivism (Figure 2). The high risk untreated group had significantly faster and higher rates of violent recidivism than all low and moderate risk groups; however, differences in violent failure from the high risk treated group approached but did not attain significance, χ²(1, N = 76), 3.24, p = .072. Moderate risk treated offenders had significantly lower rates of violent failure than moderate risk untreated offenders, χ²(1, N = 374), 5.81, p = .016. Again, no differences in violent failure were found between treated high risk and untreated moderate risk offenders, χ²(1, N = 116) = 0.114, p = .736.

Figure 1.

Survival analysis: Sexual recidivism failure rates among treated and untreated sex offenders as a function of actuarial risk level.

Figure 2.

Survival analysis: Violent recidivism failure rates among treated and untreated sex offenders as a function of actuarial risk level.

Sex Offender Treatment Outcome and Harm Reduction

The next set of analyses examined indicators of treatment outcome according to a harm reduction model. In these analyses, risk was analyzed both categorically (to facilitate interpretation) and continuously (to maximize variance). The results are reported in Table 4.

Table 4.

Comparisons between Treated and Untreated Offenders on Indictors of Harm Reduction (Time to Reconviction, Aggregate Sentence Length, and Cormier-Lang Score) as a Function of Risk.

	Sexual recidivism						Violent recidivism
	Treated		Comparison				Treated		Comparison
Criterion	M	SD	M	SD	d	p	M	SD	M	SD	d	p
Time (years) to first new conviction
Low risk	4.12	2.77	4.88	4.06	−.20	.741	4.73	2.84	5.80	3.46	−.28	.387
Moderate risk	4.02	3.02	2.56	1.14	.39	.138	3.96	2.77	3.93	2.93	.16	.963
High risk	5.77	2.93	1.30	0.95	1.83	.000	4.60	2.64	2.07	1.64	.99	.003
Overall	4.39	3.01	2.24	1.71	.63	.002	4.21	2.77	3.72	2.93	.15	.266
Aggregate sentence (years) for new convictions
Low risk	2.83	2.63	6.00	5.66	−.82	.201	1.15	1.90	2.12	3.66	−.32	.304
Moderate risk	4.17	4.81	4.61	3.94	−.07	.788	2.77	4.05	2.65	4.16	.02	.885
High risk	3.40	3.51	3.17	2.47	.07	.863	2.53	3.35	4.96	3.29	−.67	.040
Overall	3.83	4.32	4.13	3.45	−.05	.771	2.43	3.67	3.12	3.97	−.15	.239
Cormier-Lang score for new convictions
Low risk	5.63	5.99	15.00	9.99	−1.14	.085	4.14	4.11	9.57	9.41	−.79	.017
Moderate risk	7.71	6.26	13.70	11.68	−.61	.021	9.10	7.17	12.84	17.60	−.30	.068
High risk	5.38	3.77	7.11	3.33	−.48	.263	9.03	14.23	13.08	7.12	−.30	.338
Overall	6.95	5.84	11.00	9.10	−.50	.015	8.15	8.49	12.46	14.56	−.35	.007

Note: ns for treated risk groups (Sexual recidivism): Low n = 11, Moderate n = 52, High n = 16, Overall n = 79 (Violent recidivism): Low n = 35, Moderate n = 120, High n = 30, Overall n = 185; ns for untreated risk groups (Sexual recidivism): Low n = 2, Moderate n = 10, High n = 9 Overall n = 21; (Violent recidivism): Low n = 7, Moderate n = 32, High n = 13, Overall n = 52. Cohen’s d: negative values indicate the treatment group has a lower value than the comparison group, positive values indicate the treatment group has a higher value.

Time to first sexual and violent conviction

We examined potential differences between the treatment and comparison groups in time to first sexual and violent conviction according to risk level and overall. A 2 × 3 (treatment condition by risk level) factorial ANOVA demonstrated a significant main effect for treatment condition, F(1, 94) = 4.03, p = .048, but not for risk level, F(2, 94) = .55, p = .581. The Risk × Treatment interaction was significant, F(2, 94) = 3.25, p = .043. Simple main effects analyses demonstrated that high risk untreated offenders were convicted for a new sexual offense significantly faster than high risk treated offenders, F(1, 23) = 19.41, p < .001. Differences among treated and untreated moderate and low risk offenders were not significant. An ANCOVA controlling for continuous BARS score also showed a significant main effect for treatment condition, F(1, 97) = 10.24, p = .002.

These analyses were repeated for time to first violent conviction. No significant main effect was found for treatment condition, F(1, 231) = 0.90, p = .343. The main effect for risk level, F(1, 231) = 3.54, p = .031, and the Risk × Treatment interaction, F(1, 231) = 3.75, p = .025, were each significant. Simple main effects analyses demonstrated that high risk untreated sex offenders were convicted for any new violent offense significantly faster than high risk treated offenders, F(1, 41) = 10.11, p = .003. Differences among treated and untreated moderate and low risk offenders were not significant. Results of ANCOVA controlling for continuous BARS score found no significant differences by treatment condition in time to first violent conviction, F(1, 234) = 0.92, p = .339.

Aggregate sentence length for new sexual and violent offenses

We then examined potential differences between the treatment and comparison groups in aggregate sentence length, that is, the sum total of time sentenced (in years) for all new sexual or violent convictions. A 2 × 3 (treatment condition by risk level) factorial ANOVA demonstrated no significant main effect for treatment condition, F(1, 94) = 0.73, p = .394, risk level, F(2, 94) = 0.51, p = .604, or the interaction, F(2, 94) = 0.43, p = .651. Results of ANCOVA controlling for continuous BARS score found no significant differences by treatment condition in aggregate sentence length for new sexual offenses, F(1, 97) = 0.03, p = .855.

When repeated for aggregate sentence length for all violent offenses, no significant effect was found for treatment condition, F(1, 231) = 2.24, p = .136, risk level, F(2, 231) = 2.22, p = .111, or the interaction, F(2, 231) = 1.44, p = .239. Untreated high risk offenders, however, did have significantly longer aggregate sentences for all new violent offenses than treated high risk offenders, F(1, 40) = 4.52, p = .04. Results of ANCOVA controlling for continuous BARS score found no significant differences by treatment condition in aggregate sentence length for new violent convictions, F(1, 234) = 0.90, p = .344.

Cormier-Lang offense severity scores

Finally, we examined differences between treatment and comparison groups on Cormier-Lang offense severity scores coded for all new sexual and violent offenses. Cormier-Lang scores for sexual and all new violent convictions were each significantly correlated with the aggregate sentence length variables for new sexual and violent offenses (r = .72 and r = .51, ps < .001, respectively). A 2 × 3 (treatment condition by risk level) factorial ANOVA demonstrated significant main effects for treatment condition, F(1, 94) = 7.80, p = .006, and risk level, F(1, 94) = 3.36, p = .039. The interaction was not significant, F(1, 94) = 1.19, p = .308. Simple main effects analyses demonstrated moderate risk treated offenders had significantly lower Cormier-Lang scores for sexual recidivism, F(1, 60) = 5.60, p = .021. Treatment-comparison group differences in Cormier-Lang scored sexual recidivism were not significant among the high or low risk offender groups. An ANCOVA controlling for continuous BARS score found treated offenders overall to have significantly lower Cormier-Lang scored sexual recidivism, F(1, 97) = 6.39, p = .013.

When repeated for Cormier-Lang scored violent recidivism, a significant main effect was again found for treatment condition, F(1, 231) = 5.39, p = .021. No significant effects were found for risk level, F(1, 231) = 1.69, p = .186, or the interaction, F(1, 231) = 0.067, p = .935. Simple main effects analyses found the differences between the treated and comparison groups achieved significance for the low risk offenders only, F(1, 40) = 6.21, p = .017. An ANCOVA controlling for continuous BARS score also found treated offenders overall to have significantly lower Cormier-Lang scored violent recidivism, F(1, 234) = 6.18, p = .014.

Sex Offender Treatment Outcome and Age

The final set of analyses examined the relationship of sex offender age at release, risk, and treatment completion to sexual and violent recidivism. In these analyses, age at release was examined categorically using Kaplan-Meier survival analyses as well as continuously through Cox regression survival analysis.

Survival analysis

Kaplan-Meier survival analysis was conducted to examine rates of sexual and violent recidivism among four age-treatment groups: younger (< age 50) treated (n = 471), younger untreated (n = 90), older (> age 50) treated (n = 154), and older untreated (n = 17). For space considerations we have not included figures for these analyses. Compared to older offenders, younger offenders had significantly higher sexual recidivism failure rates, χ²(1, N = 732) = 7.09, p = .008, corresponding to base rates of 7.6% and 15.5%, respectively. No differences were found between older treated and untreated offenders in sexual recidivism failure rate, χ²(1, N = 171) = 0.55, p = .457. The difference between the younger treated and untreated offenders approached significance on this outcome, χ²(1, N = 561) = 3.54, p = .060. With respect to general violence, compared to older offenders, younger offenders again demonstrated significantly higher violent failure rates, χ²(1, N = 732) = 36.13, p < .001, corresponding to base rates of 12.9% and 38.3% respectively. No differences were observed between older treated and untreated offenders on this outcome, χ²(1, N = 171) = 0.01, p = .923. Younger untreated offenders had significantly higher violent failure rates than younger treated offenders, χ²(1, N = 561) = 17.86, p < .001.

Cox regression survival analysis

Cox regression survival analyses were conducted to examine the relationship of offender age, actuarial risk, and treatment completion to sexual and violent recidivism as follows: (a) effect of age, treatment completion, and their interaction; and (b) repeating these analyses incorporating the BARS score (minus the age item) and the potential interaction of age, risk, and treatment completion. The results are reported in Table 5.

Table 5.

Cox Regression Survival Analysis: Sex Offender Treatment Outcome for Sexual and Violent Recidivism Controlling for Age and Risk.

	Sexual recidivism							Violent recidivism
						95% CI							95% CI
Variable combination	B	SE	Wald	p	e^B	Lower	Upper	B	SE	Wald	p	e^B	Lower	Upper
Age and treatment
Block 1
Age at release	−.032	.010	11.20	.001	.968	.950	.987	−.058	.007	73.92	.000	.943	.931	.956
Treatment	−.405	.249	2.65	.104	.667	.410	1.086	−.466	.159	8.59	.003	.627	.459	.857
Block 2
Age at release	−.039	.022	3.02	.082	.962	.920	1.005	−.063	.015	17.30	.000	.939	.912	.967
Treatment	−.703	.896	.615	.433	.495	.085	2.868	−.661	.586	1.270	.260	.516	.164	1.629
Age × Treatment	.009	.025	.118	.731				.006	.017	.118	.731
Age, risk, and treatment
Block 1
Age at release	−.029	.010	8.52	.004	.972	.953	.991	−.057	.007	67.93	.000	.945	.932	.958
BARS score (minus age)	.544	.089	36.97	.000	1.723	1.446	2.053	.309	.060	26.88	.000	1.362	1.212	1.530
Treatment	−.329	.248	1.77	.184	.720	.443	1.169	−.426	.159	7.17	.007	.653	.479	.892
Block 2
Age at release	−.032	.012	6.975	.008	.969	.946	.992	−.058	.008	51.57	.000	.944	.929	.959
BARS score (minus age)	.478	.166	8.227	.004	1.612	1.164	2.233	.285	.111	6.57	.010	1.330	1.070	1.655
Treatment	−.501	.433	1.336	.248	.606	.259	1.417	−.479	.266	3.24	.072	.619	.368	1.043
Age × Risk × Treatment	.002	.004	.226	.635				.001	.003	.062	.803

Note: BARS = brief actuarial risk scale.

For the first set of analyses, age and treatment were entered in the first step (Block 1) followed by the interaction term in the second step (Block 2). Age at release uniquely predicted sexual and violent recidivism; however, treatment completion was uniquely associated only with reductions in violent recidivism (Block 1). The age × treatment interaction was not additive or significant for either outcome (Block 2). In the second set of analyses, age at release, BARS score (minus the age item), and treatment condition were entered in the first step (Block 1), followed by a three-way age-risk-treatment interaction term in the second step (Block 2). Age at release and risk significantly uniquely predicted sexual and violent recidivism at all stages of the analysis. Treatment completion uniquely predicted reductions in violent recidivism only, controlling for age and risk (Block 1). The interaction of continuously measured age, risk, and binary treatment completion was not significant for either outcome.

Ancillary Analyses: Previous CSC Sex Offender Program Evaluations Revisited

In light of previous national CSC evaluations of sex offender treatment, we ran a set of ancillary analyses employing a fixed 2-year follow-up consistent with Cortoni and Nunes (2007) to contextualize the above findings. It is difficult to draw direct comparisons between the two sets of studies given that risk was appraised using different actuarial tools (i.e., Static-99 vs. the shorter BARS). Controlling for actuarial risk using the same scale would make such a comparison more empirically justifiable. Without accounting for risk, marked group differences were evident between treated (n = 624) and untreated (n = 105) offenders with respect to 2-year rates for sexual (3.2% vs. 12.4%) and violent (7.9% vs. 16.2%) recidivism. Sequential Cox regression survival analyses, controlling for risk via the BARS, found treatment completion to be significantly associated with reductions in sexual (Wald [1] =11.20, e^B = .301, p = .001, 95% CI = [.149, .608]) and violent (Wald [1] = 4.78, e^B = .537, p = .029, 95% CI = [.308, .938]) recidivism over a 2-year follow-up. In our view, the pattern and magnitude of these findings are consistent with Cortoni and Nunes (2007) who reported trend level reductions in sexual (e^B = .32, 95% CI = [.09, 1.12]) and significant reductions in violent (e^B = .17, 95% CI = [.07, .38]) recidivism after controlling for risk in sequential Cox regression survival analyses.

Discussion

The present study was an outcome evaluation of sex offender treatment in a large national cohort of Canadian sex offenders followed up for more than 11 years postrelease. A brief actuarial risk scale (BARS) consisting of six static items that predicted sexual and violent recidivism was created to control for important risk-related differences between treated and untreated offenders. This seemed to be a predominantly moderate to low risk cohort of sex offenders given the relatively low mean score on the BARS and base rate of sexual recidivism (13.7%) relative to other reported base rate estimates over comparable follow-up periods in Canadian samples (cf. Harris & Hanson, 2004).

Actuarial Risk and Risk Reduction

The importance of controlling for risk was underscored by the fact that untreated offenders scored significantly higher on the BARS and thus were higher risk for sexual and violent recidivism overall. Using Cox regression survival analyses to control for risk and individual differences in follow-up time, treated offenders evinced significantly lower rates of violent, but not sexual, recidivism. When stratified by risk level, significant differences in outcome were observed only among the moderate or high risk offender groups. This finding is consistent with the risk principle (Andrews & Bonta, 2010a, 2010b). The absence of a significant risk by treatment interaction may indicate that although higher risk offenders demonstrated larger decreases in recidivism, treating low risk offenders did not necessarily result in increases in recidivism as reported elsewhere in outcome evaluations of sexual (Lovins, Lowenkamp, & Latessa, 2009) and nonsexual (Bonta, Wallace-Capretta, & Rooney, 2000) offenders.

A further set of analyses examined possible indicators of harm reduction, that is, decreases in the severity of new sexual or violent offenses as operationalized by time to new conviction, aggregate sentence length for new sexual and violent convictions, and Cormier-Lang scores for new sexual and violent convictions. Whether or not risk was controlled as a categorical or continuous covariate treatment completion was associated with significantly longer times to new sexual convictions and, among high risk offenders, any new violent convictions. Few differences were observed between treatment groups in aggregate sentence length for new sexual and violent convictions with the exception of high risk treated offenders who had shorter cumulative sentences for new violent convictions. An intriguing pattern was observed when examining the final indictor of offense severity; this was based on the Cormier-Lang scores of sexual and violent recidivism. Treatment completers had significantly lower Cormier-Lang scores for both sexual and violent recidivism, indicating their new offenses were less severe, but the magnitude of these differences was not strongly linked to risk level, particularly for sexual violence.

Taken together, the pattern of findings reported above appear to be consistent with the dynamic nature of risk, specifically (a) that risk had changed (i.e., decreased) among moderate and high risk groups both with respect to the frequency and severity of recidivism outcomes, and (b) that entirely static tools can overestimate risk among treated offender groups, particularly moderate or high risk offenders, as reductions in risk cannot be captured by means of static scores. These findings are consistent with Olver and Wong (2011) who found that an actuarially high risk cohort of sex offenders who made substantial risk-related changes in treatment had significantly lower rates of sexual recidivism than a group of treated offenders with nearly identical Static-99 scores who made few changes.

The fact that a larger and more robust treatment effect was found for general violence compared to sexual recidivism alone across most analyses is an intriguing pattern with multiple possible explanations. For one, consistent with the need principle, comprehensive sex offender programs should target multiple criminogenic needs, both sex-offender specific (e.g., deviant sexual interests, sex offender attitudes) and nonspecific (e.g., anger problems, substance abuse, unemployment). It seems reasonable that the therapeutic effects of programming could extend beyond sexual offending to other acts of interpersonal violence such as robbery or nonsexual assault. Second, this may also be a base rate issue given that the pattern of results for the sexual recidivism analyses paralleled those for general violence. Group differences were more substantial with respect to the latter criterion given that it included both sexual and nonsexual violence. A further possibility, and not inconsistent with the previous two, is that violent recidivism also serves to capture sexually motivated offending that is prosecuted under nonsexual Criminal Code categories. For instance, these might be a result of plea bargains or prosecutorial decisions to lay a charge for a nonsexual crime (e.g., assault) when the offense appeared to be sexually motivated, because there was insufficient evidence to support a conviction for a sexual crime (Rice, Harris, Lang, & Cormier, 2006). For instance, Rice et al. (2006) found that one third of nonsexual violent Criminal Code convictions, in a sample of violent sexual offenders, were sexually motivated.

The results presented here are consistent with other Canadian sex offender program evaluations (Cortoni & Nunes, 2007; Nicholaichuk et al., 2000). In their evaluation of CSC’s National Sex Offender Program, Cortoni and Nunes (2007) reported very low base rates of sexual and violent recidivism over a 2-year follow-up. It is difficult to draw direct comparisons with the present study given that the samples were rated on different risk tools and thus may vary with respect to risk; however, there is some justification for comparison as risk was controlled as in both cases discussed above. As with Cortoni and Nunes (2007), a set of ancillary analyses employing a 2-year follow-up found sex offender treatment completion to be associated with reductions in sexual and violent recidivism.

Age, Risk, and Treatment Outcome

Although age has a highly robust inverse relationship to all forms of recidivism (Hanson, 2002; Hanson & Bussière, 1998; Harris & Hanson, 2004) the reasons that such a relationship exists is a matter of debate. It appears that increasing age does have a mitigating effect on risk for sexual violence. In the present study we examined the relationship of age, actuarial risk, and treatment completion to subsequent sexual and violent recidivism. As expected, younger offenders reoffended at higher rates irrespective of treatment condition. Moreover, significant differences in outcome were observed only among younger offenders and treatment completion uniquely predicted reductions in subsequent violence after controlling for age and risk. There was, however, no evidence for a three-way interaction of age, risk, and treatment completion to outcome. Although it may be argued that younger offenders derived greater benefit from treatment than older offenders, the differences in outcome and the lack of an interaction is likely explained by the fact that the older offenders were already a lower risk group, both in terms of their actuarial risk (i.e., lower BARS score) as well as older age.

Strengths, Limitations, and Conclusions

There are some important strengths and limitations with the present study. First, the study methodology involved extracting data for a cohort of offenders shortly following their release who were subsequently followed up in real time for more than 11 years in the community. This in turn, yielded comprehensive outcome data that enabled a wide breadth of analyses. Moreover, risk assessment and scale construction procedures were applied to develop a brief actuarial tool (BARS) to control for important risk-related differences between treated and untreated groups. The analytic procedures chosen (fixed follow-up, survival analysis) also enabled us to control for differences in follow-up time to permit a more rigorous examination of group differences.

There are also some important limitations that we must acknowledge. First, we did not have access to detailed information about the content of the sex offender treatment programs, but rather, relied on the existing guidelines and best practices of CSC for sex offender treatment programming at the time (i.e., mid- to late 1990s). This was prior to the implementation of the National Sex Offender Program in Canada (Yates et al., 2000). This limitation, however, is tempered by our knowledge of the sex offender program philosophy in most CSC institutions, which, according to the 1996 Standards and Guidelines, was driven by the evidentiary base and was thus cognitive behavioral in nature.

A second, and perhaps more significant, limitation is that information regarding sex offender treatment participation status (including the specific circumstances of nonparticipation) was available for only about one third of the sample. We attribute the incompleteness of treatment information to inconsistencies that existed at the time in the manner in which treatment participation was documented and from which fields in the electronic database (OMS) this could be accessed. With the implementation of the NaSOP in late 2000, all participation (or nonparticipation) in sex offender programming became thoroughly documented and information regarding treatment status (including noncompletion) could be readily accessed from consistent fields in OMS. Given the lack of specificity of information about the individuals who definitively did not complete sex offender treatment, it is possible that at least some of these men were sex offender treatment refusers or dropouts. This group has been shown to be higher risk offenders on average (Olver, Stockdale, & Wormith, 2011).

In conclusion, the findings from the present study are consistent with the risk principle and provide at least partial support for the effectiveness of the sex offender programs operated across Canadian Corrections in the 1990s in terms of reducing sexual and violent recidivism, particularly among moderate to high risk offenders. The findings also offer implications for the dynamism of sexual violence risk. Risk has the potential to be moderated or reduced when suitable treatment and risk management services are available and appropriately delivered. An important challenge facing researchers and clinicians alike is to determine how to best integrate risk-related knowledge from multiple sources and to incorporate these data into prudent release recommendations and effective supervisory strategies to prevent future sexual violence.

Footnotes

Acknowledgements

The authors thank Richard Coupland for his assistance with data collection.

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.

Notes

References

Andrews

D. A.

Bonta

(2010a). The psychology of criminal conduct (5th ed.). New Providence, NJ: LexisNexis.

Andrews

D. A.

Bonta

(2010b). Rehabilitating criminal justice policy and practice. Psychology, Public Policy, and Law, 16, 39-55. doi:10.1037/a0018362

Andrews

D. A.

Zinger

Hoge

R. D.

Bonta

Gendreau

Cullen

F. T.

(1990). Does correctional treatment work? A clinically-relevant and psychologically informed meta-analysis. Criminology, 28, 369-404.

Barbaree

H. E.

(2005). Psychopathy, treatment behavior, and recidivism: An extended follow-up of Seto and Barbaree. Journal of Interpersonal Violence, 20, 1115-1131. doi:10.1177/0886260505278262

Barbaree

Langton

Blanchard

(2007). Predicting recidivism in sex offenders using the VRAG and SORAG: The contribution of age at release. International Journal of Forensic Mental Health, 6, 29-46. doi:10.1080/14999013.2007.10471247

Beggs

S. M.

Grace

R. C.

(2011). Treatment gains for sexual offenders against children predicts reduced recidivism: A comparative validity study. Journal of Consulting and Clinical Psychology, 79, 182-192. doi:10.1037/a0022900

Bonta

Wallace-Capretta

Rooney

(2000). A quasi-experimental evaluation of an intensive rehabilitation supervision program. Criminal Justice and Behavior, 27, 312-329. doi:10.1177/0093854800027003003

Corrections and Conditional Release Statistical Overview. (2004). Public safety and emergency preparedness Canada, Ottawa, Ontario, Canada.

Corrections and Conditional Release Statistical Overview. (2010). Public Safety Canada, Ottawa, Ontario, Canada.

10.

Correctional Service of Canada. (1996). Standards and guidelines for the provision of services to sex offenders. Ottawa, Ontario, Canada.

11.

Cortoni

Nunes

K. L.

(2007). Assessing the effectiveness of the National Sexual Offender program (Research Report R-183). Ottawa, Ontario: Correctional Service of Canada.

12.

Douglas

Skeem

(2005). Violence risk assessment: Getting specific about being dynamic. Psychology, Public Policy, and Law, 11, 347-383. doi:10.1037/1076-8971.11.3.347

13.

Di Placido

Simon

T. L.

Witte

T. D.

Wong

S. C. P.

(2006). Treatment of gang members can reduce recidivism and institutional misconduct. Law and Human Behavior, 30, 93-114. doi:10.1007/s10979-006-9003-6

14.

Gallagher

C. A.

Wilson

D. B.

Hirschfield

Coggeshall

M. B.

MacKenzie

D. L.

(1999). A quantitative review of the effects of sex offender treatment on sexual reoffending. Corrections Management Quarterly, 3, 19-29.

15.

Wong

Nicholaichuk

(2004). Reconviction rates of federal offenders after release on warrant expiry date. Ottawa, Ontario: Prepared for Performance Assurance, National Headquarters, Correctional Service of Canada.

16.

Hall

G. C.

(1995). Sex offender recidivism revised: A meta-analysis of recent treatment studies. Journal of Consulting and Clinical Psychology, 63, 802-809. doi:10.1037/0022-006X.63.5.802

17.

Hanson

R. K.

(2002). Recidivism and age: Data from 4,673 sexual offenders. Journal of Interpersonal Violence, 17, 1046-1062. doi:10.1177/088626002236659

18.

Hanson

R. K.

(2006). Does Static-99 predict recidivism among older sexual offenders? Sexual Abuse: A Journal of Research and Treatment, 18, 343-355. doi:10.1177/107906320601800403

19.

Hanson

R. K.

Bourgon

Helmus

Hodgson

(2009). The principles of effective correctional treatment also apply to sexual offenders. Criminal Justice and Behavior, 36, 865-891. doi:10.1177/0093854809338545

20.

Hanson

R. K.

Broom

Stephenson

(2004). Evaluating community sex offender treatment programs: A 12-year follow-up of 724 offenders. Canadian Journal of Behavioural Science, 36, 87-96.

21.

Hanson

R. K.

Bussière

M. T.

(1998). Predicting relapse: A meta-analysis of sexual offender recidivism studies. Journal of Consulting and Clinical Psychology, 66, 348-362. doi:10.1037/0022-006X.66.2.348

22.

Hanson

R. K.

Gordon

Harris

A. J. R.

Marques

Murphy

W. D.

Quinsey

V. L.

Seto

M. C.

(2002). First report of the Collaborative Outcome Data Project on the effectiveness of psychological treatment for sexual offenders. Sexual Abuse: A Journal of Research and Treatment, 14, 169-194. doi:10.1177/107906320201400207

23.

Hanson

R. K.

Morton-Bourgon

K. E.

(2009). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis of 118 prediction studies. Psychological Assessment, 21, 1-21. doi:10.1037/a0014421

24.

Hanson

R. K.

Thornton

(1999). STATIC-99: Improving actuarial risk assessments for sex offenders (User Report 1999-02). Ottawa, Ontario: Department of the Solicitor General of Canada.

25.

Harris

A. J. R.

Hanson

R. K.

(2004). Sex offender recidivism: A simple question (User Report 2004-03). Ottawa, Ontario: Department of the Solicitor General of Canada.

26.

Harris

G. T.

Rice

M. E.

(2007). Adjusting actuarial violence risk assessments based on aging or the passage of time. Criminal Justice and Behavior, 34, 297-313. doi:10.1177/0093854806293486

27.

Helmus

Thornton

Hanson

R. K.

Babchishin

K. M.

(2012). Improving the predictive accuracy of Static-99 and Static-2002 with older sex offenders: Revised age weights. Sexual Abuse: A Journal of Research and Treatment, 24, 64-101. doi:10.1177/ 1079063211409951

28.

Langton

C. M.

Barbaree

H. E.

Harkins

Peacock

E. J.

(2006). Sex offenders’ response to treatment and its association with recidivism as a function of psychopathy. Sexual Abuse: A Journal of Research and Treatment, 18, 99-120. doi:10.1007/s11194-006-9004-5

29.

Löesel

Schmucker

(2005). The effectiveness of treatment for sexual offenders: A comprehensive meta-analysis. Journal of Experimental Criminology, 1, 117-146. doi:10.1007/s11292-004-6466-7

30.

Looman

Abracen

Nicholaichuk

T. P.

(2000). Recidivism among treated sexual offenders and matched controls: Data from the Regional Treatment Centre (Ontario). Journal of Interpersonal Violence, 15, 279-290. doi:10.1177/088626000015003004

31.

Looman

Abracen

Serin

Marquis

(2005). Psychopathy, treatment change, and recidivism in high-risk, high-need sexual offenders. Journal of Interpersonal Violence, 20, 549-568. doi:10.1177/0886260504271583

32.

Lovins

Lowenkamp

C. T.

Latessa

E. J.

(2009). Applying the risk principle to sex offenders: Can treatment make some sex offenders worse? The Prison Journal, 89, 344-357. doi:10.1177/0032885509339509.

33.

Marlatt

G. A.

Witkiewitz

(2010). Update on harm-reduction policy and intervention research. Annual Review of Clinical Psychology, 6, 591-606. doi:10.1146/annurev.clinpsy.121208.131438

34.

Marques

J. K.

Wienderanders

Day

D. M.

Nelson

van Ommeren

(2005). Effects of a relapse prevention program on sexual recidivism: California’s Sex Offender Treatment and Evaluation Project (SOTEP). Sexual Abuse: A Journal of Research and Treatment, 17, 79-107. doi:10.1177/107906320501700108

35.

Marshall

W. L.

Marshall

L. E.

(2007). The utility of the random controlled trial for evaluating sexual offender treatment: The gold standard or an inappropriate strategy? Sexual Abuse: A Journal of Research and Treatment, 19, 175-191. doi:10.1007/s11194-007-9046-3

36.

Marshall

W. L.

Marshall

L. E.

(2008). Good practice and the valuation of treatment. A response to Seto et al. Sexual Abuse: A Journal of Research and Treatment, 20, 256-260. doi:10.1177/1079063208323839

37.

McGrath

R. J.

Cumming

G. F.

Buchard

B. L.

Zeoli

Ellerby

(2010). Current practices and emerging trends in sexual abuser management: The Safer Society 2009 North American Survey. Brandon, VT: Safer Society Press.

38.

Motiuk

L. L.

Porporino

(1993). An examination of sex offender case histories in federal corrections (Research Report No. R-30). Ottawa, Ontario: Correctional Service Canada.

39.

Nicholaichuk

Gordon

Wong

(2000). Outcome of an institutional sexual offender treatment program. A comparison between treated and matched untreated offenders. Sexual Abuse: A Journal of Research and Treatment, 12, 139-153. doi:10.1177/107906320001200205

40.

Nicholaichuk

T. P.

Olver

M. E.

Takahashi

(in press). Correctional careers of Dangerous Offenders. The Criminal Law Quarterly.

41.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2011). A meta-analysis of predictors of offender treatment attrition and its relationship to recidivism. Journal of Consulting and Clinical Psychology, 79, 6-21. doi:10.1037/a0022200

42.

Olver

M. E.

Wong

S. C. P.

(2011). A comparison of static and dynamic assessment of sex offender risk and need in a treatment context. Criminal Justice and Behavior, 38, 113-126. doi:10.1177/0093854810389534

43.

Olver

M. E.

Wong

S. C. P.

Nicholaichuk

(2009). Outcome evaluation of a high intensity inpatient sex offender treatment program. Journal of Interpersonal Violence, 24, 522-536. doi:10.1177/0886260508317196

44.

Olver

M. E.

Wong

S. C. P.

Nicholaichuk

Gordon

(2007). The validity and reliability of the Violence Risk Scale-Sexual Offender version: Assessing sex offender risk and evaluating therapeutic change. Psychological Assessment, 19, 318-329. doi:10.1037/1040-3590.19.3.318

45.

Quinsey

V. L.

Rice

M. E.

Harris

G. T.

Cormier

C. A.

(1998). Violent offenders: Appraising and managing risk. Washington, DC: American Psychological Association.

46.

Rice

M. E.

Harris

G. T.

Lang

Cormier

C. A.

(2006). Violent sex offenses: How are they best measured from official records? Law and Human Behavior, 30, 525-541. doi:10.1007/s10979-006-9022-3

47.

Sampson

R. J.

Laub

J. H.

(2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41, 555-592. doi:10.1111/j.1745-9125.2003.tb00997.x

48.

Seto

M. C.

Barbaree

H. E.

(1999). Psychopathy, treatment behavior, and sex offender recidivism. Journal of Interpersonal Violence, 14, 1235-1248.

49.

Seto

M. C.

Marques

J. K.

Harris

G. T.

Chaffin

Lalumière

M. L.

Berliner

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, 247-255. doi:10.1177/1079063208317733

50.

Skelton

Vess

(2008). Risk of sexual recidivism as a function of age and actuarial risk. Journal of Sexual Aggression, 14, 199-209. doi:10.1080/13552600802267098

51.

Thornton

(2006). Age and sexual recidivism: A variable connection. Sexual Abuse: A Journal of Research and Treatment, 18, 123-135. doi:10.1177/107906320601800202

52.

Yates

P. M.

Goguen

B. C.

Nicholaichuk

T. P.

Williams

S. M.

Long

C. A.

(2000). National Sex Offender Programs: Vol. 2. Moderate intensity. Ottawa, Ontario: Correctional Service of Canada.