Executive Summary

Sex Offender Characteristics, Response to Treatment,

And Correctional Release Decisions At the

Warkworth Sexual Behaviour Clinic

Howard E. Barbaree, Ph.D.1
Michael C. Seto, M.A.1 
Alexandra Maric, H.B.A.2 
Forensic Division 
Clarke Institute of Psychiatry

1 Contractors

2 Research Assistant

To obtain copies of the full report, contact:

Michael Seto, Forensic Division, Clarke Institute of Psychiatry, 250 College St., Toronto, Ontario, M5T 1R8 Phone: (416) 979-4747 x 2680 Fax: (416) 979-6965 email: setom@cs.clarke-inst.on.ca

Sex Offender Characteristics, Response to Treatment,
And Correctional Release Decisions At the
Warkworth Sexual Behaviour Clinic

 The Warkworth Sexual Behaviour Clinic (WSBC) was inaugurated in June, 1989, as a sex-offender treatment programme at Warkworth Institution, a medium-security federal penitentiary located approximately 11/2 hours' drive northeast of Toronto. Warkworth Institution is one of Canada's largest penitentiaries, housing over 600 inmates serving sentences of two or more years in length. Approximately half of these offenders were convicted of a sexual offense or a violent offense in which sexual motivation or behaviour were considered to be important. Since its inauguration, the WSBC has provided treatment for approximately 75 sex offenders per year. The current report presents an analysis of data compiled from the first 250 cases processed through the clinic. Of these 250 cases, 123 were rapists, 15 were sex-killers (men convicted of sex-related homicides), 56 were incest offenders, and 56 were extrafamilial child molesters. At the time of the data analysis, conducted during the summer and fall of 1995, 132 of these 250 offenders had been conditionally released to the community: 23 on day parole, 15 on full parole, and 94 by their statutory release date.

The treatment programme in the WSBC was intended to reduce the likelihood of recidivism by these offenders, especially violent or sexual reoffenses. The WSBC was carefully implemented as a state-of-the-art programme, informed by expert opinion and empirical data in the scientific and professional literature on sex offender treatment and treatment delivery. Accordingly, it was designed as a cognitive-behavioural programme that relied heavily on the principles of Relapse Prevention theory. The programme utilized a group therapy format as an economical means of service delivery, in order to make the most of limited resources. During the period of time covered by this report, the cost of treatment and assessment for each participant was between $6,000 and $7,000. It was designed to fit into the institutional work and job site organization, and consequently, offenders in treatment were assigned to the WSBC as a job site, and participants reported to "work" on a five day per week basis for the duration of the five month treatment programme.

A novel feature of the WSBC programme was the use of a "multifactorial assessment of sex offender risk for reoffense" (MASORR). At the time the WSBC was implemented, a comprehensive review of the scientific literature on the prediction of reoffense among sex offenders was conducted and formed the basis of the MASORR. This review indicated that four factors were predictive of sexual reoffense: (1) a history of sexual offending; (2) deviant sexual arousal; (3) a history of antisocial behaviour and other indicators of an antisocial personality (as measured by the Psychopathy Checklist- Revised); and (4) social competence (estimated from apparent intelligence level and socioeconomic status).

The MASORR is therefore based on static predictors of reoffense. It was initially implemented as a pre-treatment assessment of risk. Later, however, it was modified to incorporate two dynamic factors reflecting the man's performance while in treatment. These dynamic factors were motivation for treatment and degree of behaviour change achieved. An overall clinical impression based on the man's involvement with the WSBC was also incorporated into the post-treatment risk evaluation. These various factors were combined to form overall ratings of risk for reoffense, first at pre-treatment (static factors only) and again at post-treatment (initial risk score and consideration of the dynamic factors).

The MASORR represents what we believe to be two improvements over many previous studies of prediction of reoffense among sex offenders. First, while many previous studies have been retrospective, scoring these predictors after outcomes have already been recorded (i.e., reoffense or no reoffense), the present study is prospective in nature because predictors were scored prior to treatment and before the offender had been released into the community. Second, the MASORR was systemic in the sense that the result of the MASORR was in each case reported to decision-makers, so that outcome decisions (i.e., release or no release) were presumably influenced by the results of the MASORR. This allows us to analyse the impact of risk assessment on decision-making in the release process. At the same time, all offenders with determinate sentences are eventually released, allowing us to analyse the prediction of recidivism after the potential selection bias for conditional release has "washed out".

The following issues were addressed in this research report:

Comparisons between the offender types (rapists, sex-killers, incest offenders, and extrafamilial child molesters) showed that they did not differ in terms of the level of education they achieved or their socioeconomic status, as estimated from the most typical occupation in their work history (Blishen, Carroll, & Moore, 1987). The rapists and sex-killers were significantly younger than the two groups of child molesters, and the two groups of child molesters were more likely to have been married than the rapists. In terms of criminal history, the rapists had more prior criminal offenses, both nonviolent and violent in nature, while the incest offenders had the lowest number of prior criminal offenses. A notable finding was that the rapists were much more likely than the other types to have previously committed a violent offence.

Overall, the administration of the psychological tests did not yield very useful data. Most of the tests, even those which were presumed to have offender type-specific content, e.g., the Rape Myth Acceptance Scale for rapists, did not differentiate between the offender types, and few test scores changed over the course of treatment. In addition, psychological test scores were not good predictors of the major outcome variables. Of all the tests, the Multiphasic Sexual Inventory (MSI: Nichols & Molinder, 1984) was the most useful in its assessment of denial and minimization; subscale scores from the MSI did change following treatment targeting denial and minimization.

The results of phallometric testing indicated significant and large differences between groups: the two groups of child molesters showed higher deviant indices than either the rapists or sex killers. The group means for these indices revealed that both incest offenders and extrafamilial child molesters showed a sexual preference for children over adults; however, there was a great deal of heterogeneity within the groups (see Barbaree, Bogaert, & Seto, 1995; Barbaree & Seto, in press). The results from the phallometric testing were not predictive of any of the major outcome variables in this analysis.

We used a series of crude factor analyses to reduce the large number of quantitative variables in the database (300+) to a more reasonable and manageable number. These analyses reduced the number of historical variables pertaining to the domains of education, occupation, relationships, family history, juvenile antisocial behaviour, and adult antisocial behaviour a set of 10 historical factors: Childhood Behaviour Problems, Erratic Employment, Previous Treatment, Quality of Early Life, Separation from Family of Origin, Sexual Promiscuity, Alcohol Problems, Severity of Index Offense, Antisocial History, and Criminal History. Behaviour during treatment was rated on a number of dimensions and similarly submitted to factor analysis, resulting in three treatment process factors: Treatment Behaviour, Treatment Change, and Clinical Impression. The psychological tests included in the pre-treatment assessment were also subjected to factor analysis, producing three test factors: Overt Hostility, Covert Hostility, and Social Functioning. Most of these factors had three or four items loading onto them. For example, Childhood Behaviour Problems represented the common variance for the following set of items: Cheating in school (yes/no), stealing in school (yes/no), total number of behaviour problems on a checklist of the criteria for a psychiatric diagnosis of conduct disorder, and drinking or drug use in school (yes/no).

One hundred and ninety three (77.2%) of the 250 subjects completed the treatment programme. Treatment completion was unrelated to offender type or whether the offender had previously committed a sexual offense. None of the historical factors emerged as a significant predictor of treatment completion. Psychological test scores also did not predict treatment completion, except for the Treatment Attitudes subscale of the Multiphasic Sex Inventory. Men who reported more positive attitudes about treatment and expected to get more out of treatment were more likely to complete the treatment programme.

An analysis of overall pre-treatment risk ratings showed that offender types differed in their initial risk scores; incest offenders were assessed as significantly lower in initial risk than the other types of offenders. After combining the types, two factors were significant predictors, explaining 23.5% of the variance in pre-treatment level of risk: Antisocial History and Criminal History. Addition of the deviant sexual arousal scores did not contribute to the predictive equation. Higher risk subjects had more extensive antisocial and criminal histories. The Justification subscale of the Multiphasic Sex Inventory was also a significant predictor, but the new set of three predictors explained slightly less variance (17.0%) in the subset of subjects who completed the Multiphasic Sex Inventory. The psychological test factors did not predict initial level of risk.

In contrast to the comparison for the initial risk assessments, there was no significant difference between offender types on their overall post-treatment risk scores. The average risk scores of incest offenders did not change, while the average risk scores of the other types decreased slightly. A multiple regression analysis was then conducted to identify predictors of post-treatment risk score. The initial risk score was entered first in order to determine if the historical and treatment factors contributed something more to the prediction of post-treatment risk. The regression equation was highly significant, explaining 68.1% of the variance in post-treatment risk score. Significant predictors were initial risk score, Treatment Behaviour, Treatment Change, and Clinical Impression. Higher post-treatment risk was associated with higher initial level of risk, poorer behaviour in treatment, less positive gains over the course of treatment, and poorer overall clinical impression. In other words, post-treatment risk could be explained in terms of a large static component (reflected in the initial risk score) and dynamic factors associated with treatment (the treatment process factors). A second regression equation showed that the Rape subscale of the Multiphasic Sex Inventory was a small but significant predictor. A third regression analysis found that the psychological test factors did not predict post-treatment level of risk. Not surprisingly, change in risk score (i.e., pre-treatment risk level minus post-treatment risk level) was significantly predicted by Treatment Change and Clinical Impression.

As part of the process of reporting to the institutional case manager and the National Parole Board, the WSBC made specific recommendations for each offender's post-treatment disposition. These recommendations, referred to as the Level of Management Index, were based on the individual's post-treatment risk score and on the various case management options available for that individual (e.g., whether a graduated release to a community residential centre would be supported by case management). There was no difference between offender types in the level of management that was recommended. There were three significant predictors in the regression equation, explaining 51.4% of the variance in recommended level of management: Post-treatment risk, Previous Treatment, and Clinical Impression. In other words, a more restrictive recommendation for management was associated with higher post-treatment risk, previously being involved in treatment, and poorer overall clinical impression.

Data for 215 of the 250 offenders were obtained during a review of National Parole Board files. Of these 215 offenders, 198 (92.1%) were eligible for conditional release: 1 man died while incarcerated, 12 men were serving a life sentence and were not yet eligible for parole, and 4 men had been designated as dangerous offenders. Of the 198 eligible offenders, 132 (66.7%) were conditionally released; 3 men had passed their parole eligibility dates but had not reached their statutory release date, and the remaining 63 men were detained following their statutory release date. There was no difference between offender types in the proportion who were detained. Only one of the sex-killers had been released by his statutory release date. Being detained was significantly related to whether the offender was a first-time sexual offender and whether the offender had completed the WSBC treatment programme.

A stepwise discrimination function analysis was conducted to identify predictors of detention (i.e., not being released on day parole, full parole, or statutory release). There were 148 valid cases, with 48 (32.4%) of these individuals being detained. Post-treatment level of risk and the recommended level of management were entered in the first block. The treatment process factors were then entered in a stepwise fashion in a second block. There were two significant predictors of being detained: post-treatment level of risk and recommended level of management. In other words, offenders assessed at higher risk for reoffending in the WSBC post-treatment and given recommendations for more restrictive levels of management report were more likely to be detained. The MSI subscales and psychological test factors were not significantly related to detention.

Out of the 132 subjects who were conditionally released, a total of 42 (31.8%) men failed their conditional release for one of the following reasons: a relapse in which no official action was taken, suspension for the breach of a condition related to their relapse plan, or revocation of their conditional release. Rapists were more likely than child molesters to fail their conditional release, 40.7% vs. 25.0%, although this trend was not quite statistically significant. These failure rates are very similar to those found for a sample of 145 sex offenders, recently released on parole and with an average follow-up time of one year, who were identified through the National Sex Offender Census: 42.3% of rapists and 22.5% of child molesters (see Motiuk & Brown, 1993). It should be pointed out that the WSBC sample's official failure rate was slightly lower than the percentages given here because we also recorded relapses for which no official action was taken.

The average time at risk for failure during a conditional release was approximately 43 months, ranging from a week to 5.2 years. Survival analysis showed that 29.1% of the rapists and 14.4% of the child molesters had failed after one year of follow-up. After two years, 47.7% of the rapists and 28.2% of the child molesters had failed, and after three years, these proportions were 62.9% and 43.0% respectively. These results indicate that the rapists failed at approximately twice the rate of child molesters, but this difference in failure rate decreased in the third year of follow-up. These results should be considered tentative because only a small number of offenders in the present sample were at risk of failing their conditional release for three years, and because more than half of the cases were censored at the time of this analysis, i.e., they had not passed their warrant expiry date and were therefore still at risk. We conducted a stepwise discriminant analysis to identify predictors of conditional release failure. The resulting predictors were Antisocial History and Treatment Behaviour. In other words, highly antisocial subjects who behaved poorly in treatment were more likely to fail their conditional release.

At the time of the current data analysis, a total of 218 individuals have been released from prison. As noted, 132 of these offenders were conditionally released and the remainder left prison following the expiration of their warrants. One man died after being released and 15 other men were deported from Canada, so the follow-up group was comprised of 202 individuals. At the present time, we have identified 13 individuals who committed a new sexual offense, and an additional four subjects who committed a new violent offense but who did not commit a new sexual offense. A total of 36 individuals committed a new offense of any kind. Therefore, a total of 17 WSBC treatment participants have committed a new serious offense after being released from prison (i.e., violent or sexual reoffense), giving the WSBC a serious recidivism rate of 8.4% and a sexual recidivism rate of 6.4% after an average follow-up period of approximately 2.5 years. These rates compare favourably with the reoffense rates reported by other large treatment programmes (see reviews by Marshall, Jones, Ward, Johnston, & Barbaree, 1991; Hall, 1995). However, they were too low to conduct discriminant function or logistic regression analyses. The base rates will presumably be higher and amenable to analysis as the length of the follow-up period increases.

Parole board data were available for a comparison group of 74 offenders who had been offered treatment at the WSBC but who had refused it. Sixty five of these offenders had been released from prison, 26 at warrant expiry and the other 39 on parole. There was a statistical trend towards treatment refusers being less likely than treatment participants to be conditionally released. Slightly more than half of these offenders were rapists (52.2%), a proportion that did not differ from the WSBC sample of 250 offenders.

In terms of their histories, treatment refusers did not differ in the likelihood of experiencing emotional, physical, or sexual abuse. Treatment refusers also did not differ in the proportion who had a parent with an alcoholic, psychiatric, or criminal history. However, treatment refusers did differ from treatment acceptors by having a higher global rating of the quality of their home environment while growing up. Treatment refusers were almost twice as likely to have had problems with alcohol as a teenager, as well as with other drugs as a teenager. These differences continued to a lesser degree into adulthood for alcohol, but not for other drugs.

There was a large difference between subjects who entered treatment and those who refused in terms of the proportion who had previously participated in mental health treatment of any kind. Treatment refusers were much less likely to have previously been involved in treatment. However, the two groups did not differ in the proportion of subjects who had previously received sex offender treatment. Treatment refusers were more likely (21.2% vs. 7.2%) to have been diagnosed with antisocial personality disorder. Treatment refusers did not differ from those who accepted treatment in the number of prior violent or sexual offenses in their criminal records. They were significantly older (20.7 vs. 16.1 years) when they committed their first crime, but did not differ in their age when they committed their first sexual offense. Treatment refusers did not differ in the likelihood of using or threatening to use a weapon while committing their index offense, but they did use more force and they tended to be more instrumental in their use of force.

Fifteen of the 39 treatment refusers (38.5%) who were conditionally released failed for one of the following reasons: relapse in which no official action was taken, suspension for the breach of a condition related to their relapse plan, or revocation of parole. This proportion did not differ from the proportion of conditional release failures in the treatment sample. However, the average follow-up time for treatment refusers was approximately 30 months, ranging from 4 months to 3.3 years, a shorter period of time than the follow-up period for the treatment sample. Treatment refusers might have a higher failure rate than treatment acceptors after an equivalent follow-up period. Survival curves were plotted to compare the failure rates of the WSBC sample and the treatment refusers. The survival curves were interesting, showing that the treatment refusers consistently failed at a greater rate than the WSBC follow-up sample. For example, after one year of follow-up, 77.8% of the WSBC sample survived, compared to 61.1% of the refusers; after two years, the surviving proportions were 60.0% and 40.1%, respectively.

Two findings deserve emphasis here. First, there were sensible relationships between decisions that were made at different stages of the offender's involvement with the WSBC and case management, indicating that information about the offender was being used in a systemic way. Initial risk scores were based on historical (i.e., static) factors that reflected information obtained through file reviews and a clinical interview. Post-treatment risk scores were conservative in that they were heavily influenced by these initial risk scores; nonetheless, treatment process factors did have an influence, demonstrating that the offender's performance in treatment was taken into account in making this decision. Similarly, recommendations about level of management were informed by post-treatment risk score and overall clinical impression, and parole board decisions were related to these recommendations.

Second, the relatively prominent role of treatment process factors in predicting various outcomes suggests the potential importance of examining treatment responsivity (see Stewart & Millson, 1995). Based on an analysis of 2400 offender assessments conducted in the Ontario Region, Stewart and Millson found that offenders judged to be at high risk for reoffending were also rated as less motivated than lower risk offenders; not surprisingly, lower risk offenders who did well in treatment were the least likely to fail during their conditional release.

The present report describes the preliminary results of a research programme at the WSBC designed to evaluate a risk assessment methodology, and to evaluate the programme of treatment offered at the WSBC. The preliminary data analysis indicates that the risk assessment completed at the WSBC is predictive of decisions taken by the Parole Board, and is subsequently predictive of failure on conditional release. Though not conclusive, this preliminary evaluation of the treatment programme suggests that the programme could be effective in reducing the rate of failure on conditional release, and in reducing the rate of recidivism among these treated sex offenders.

References

Barbaree, H. E., Bogaert, A. F., & Seto, M. C. (1995). Sexual reorientation therapy: Practices and controversies. In L. Diamant & R. D. McAnulty (Eds.), The psychology of sexual orientation, behavior, and identity: A handbook (pp. 357-383). Westport, CT: Greenwood.

Barbaree, H. E., & Seto, M. C. (in press). Pedophilia: Assessment and treatment. In Laws, D. R., & O'Donohue, W. T. (Eds.), Handbook of sexual deviance: Theory and application. New York: Guilford.

Blishen, B. R., Carroll, W. K., & Moore, C. (1987). The 1981 socioeconomic index for occupations in Canada. Canadian Review of Sociology and Anthropology, 24, 465-488.

Hall, G. C. N. (1995). Sexual offender recidivism revisited: A meta-analysis of recent treatment studies. Journal of Consulting and Clinical Psychology, 63, 802-809.

Hare, R. D. (1991). Manual for the revised Psychopathy Checklist. Toronto: Multi-Health Systems.

Marshall, W. L., Jones, R., Ward, T., Johnston, P., & Barbaree, H. E. (1991). Treatment outcome with sex offenders. Clinical Psychology Review, 11, 465-485.

Motiuk, L. L., & Brown, S. L. (1993). Survival time until suspension for sex offenders on conditional release (Report No. 31). Ottawa: Correctional Service of Canada.

Nichols, H. R., & Molinder, I. (1984). Multiphasic Sex Inventory. (Available from authors at 437 Bowles Drive, Tacoma, WA, 98466).

Stewart, L., & Millson, W. A. (1995). Offender motivation for treatment as a responsivity factor. Forum on Corrections Research, 7, 5-7.