Some are unclear whether risk assessment instruments, specifically dynamic risk instruments, have demonstrated utility in the risk estimation, treatment recommendations, and monitoring change over time in men at risk for or under sentence of Indeterminate Detention (ID) for sexual offenses. We compare two datasets, the first consisting of individuals representing a routine sample of persons convicted of a sexual offense and the second of men representative of a high risk/needs sample. These two distinct samples (n = 442, mean Static-99R score = 2.4; n = 168, mean Static-99R score 4.5) were then also scored on the Stable-2007. For both groups this scoring occurred in an institutional setting. The Stable-2007 predicted sexual recidivism in Sample 1 independently and in conjunction with the Static-99R. In the high-risk sample the results were the same. In both samples a compound outcome variable (Sexual + Violent reoffense) was also calculated with the Stable-2007 predicting the compound outcome variable in Sample 1 but not Sample 2. This is interesting in that it suggests that the Stable-2007 assesses constructs specific to sexual re-offense in higher risk offenders and not general traits of violence or common anti-social behaviour. Limitations and directions for further research are discussed.
The Static-99R and Stable-2007 are widely used actuarial instruments in the domain of risk assessment for men who have committed sexual offences. They have been found to be valid indicators of risk in countries around the world, in a variety of settings.
While the instruments are widely used in incarcerated populations, the Stable-2007 was normed on a community sample, and there has been concern raised regarding the validity of this instrument for men who have served lengthy prison sentences.
The Static-99R and the Stable-2007 were scored on two independent samples of men incarcerated for having committed sexual offences. One sample (n = 442, mean Static-99R score = 2.4) was a Routine sample, while the second (n = 168, mean Static-99R score 4.5) was comprised of High Risk/Needs offenders. Results showed that in both samples the Static-99R added incrementally to the predictive validity of the Static-99R for the prediction of sexual recidivism. For serious (i.e., sexual + violent) recidivism the Stable-2007 was only useful in the Routine sample.
The results indicate that the Static-99R and the Stable-2007 can be used in the prediction of sexual, but not serious reoffence for men who have been incarcerated for lengthy periods of time.
The Stable-2007 can be validly scored on incarcerated samples. The predictive validity of the Stable-2007 scored on incarcerated samples is the same as found in the developmental sample. The Stable-2007 adds incremental value to the prediction obtained with the Static-99R alone. Length of sentence does not appear to be related to the predictive validity of the Stable-2007.
The assessment of risk for sexual recidivism serves an important function when considering public safety. Efforts to increase the accuracy of prediction through the use of static factors appears to have reached a ceiling, as most extant measures are approximately equal in effectiveness and efforts to improve on these predictive values have not been successful (
Over the past decade efforts have increasingly turned to dynamic measures (
One commonly used measure of dynamic risk is the Stable-2007 (
The Dynamic Supervision Project (DSP) was launched as a more comprehensive, prospective study, led by
The results of the above analyses led to changes made to improve incremental validity and scoring in the development of the Stable-2007. First, the removal of all attitude items since they yielded the lowest relationship with recidivism (AUC = .47 to .54). Second, the scoring criteria for the following three items were refined: (a) deviant sexual interests (b) lovers/intimate partners (c) emotional identification with children. Lastly, the total score criteria were simplified. As a result, the Stable-2007 provided incremental validity to the prediction of sexual recidivism (β = .059,
In a second study
A recent meta-analysis (
One issue which has not been sufficiently addressed in the extant research is the extent to which the Stable-2007 is valid with offenders who have served lengthy sentences, given that the instrument was developed in a population of men under supervision is the community. This issue is an important one, given that in some jurisdictions (e.g., the United States), offenders may serve lo ng periods of incarceration prior to release, thus it is important to determine whether the items included in risk assessment tools are relevant for such offenders.
Thus, the purpose of this research was to examine the predictive validity in two samples of incarcerated persons convicted of a sexual offense. One sample consisted of 442 men assessed and/or treated in the Ontario Region of the Correctional Service of Canada while the second sample consisted of 168 offenders assessed and/or treated at the Regional Treatment Centre (Ontario) prior to 1992.
The Stable-2000 (
The Stable-2007 (
For those individuals for whom the Stable-2000 was scored the necessary information to convert the score to the Stable-2007 was gathered from file information. While interrater reliability was not calculated specifically for this study, since it is a retrospective analysis of data gathered for evaluation purposes, previous research conducted at the Millhaven Assessment Unit (
The Static-99R (
Sample 1 consisted of 442 persons convicted of sexual offenses assessed and/or treated in the Ontario Region of the Correctional Service of Canada. The sample consisted of two groups: For 376 of the men, the Static-99/99R and the Stable-2000/2007 were scored as part of specialized sexual offense assessment completed within three to five months of their entry to the Correctional Service of Canada at the Millhaven Assessment Unit. Of these men, 247 went on to complete a sexual offense treatment program during their sentence while 43 refused treatment, 22 were discharged from treatment prior to completion (typically for failure to comply with program rules), and for 24 there was no evidence that they were offered treatment prior to release. Data concerning the treatment status for the 40 remaining individuals was not available. Information used to score the instruments included police reports and court documents related to their trial/sentencing and when available pre-sentence reports, psychological/psychiatric assessments completed prior to sentencing and any documents available for those who had previous sentences.
The remaining 66 individuals in Sample 1 were assessed as part of the pre-treatment assessment for a sexual offense treatment program (RTCSOTP;
While approximately 15% of the sample was derived from a High Intensity Sexual Offender Treatment Program, most the men included in this analysis were consecutive admissions to the Federal Prison system in Ontario, thus the sample can be considered to represent a routine sample of men convicted of sexual offenses (
Sample 2 was derived from a group of 506 men who committed sexual offences treated or assessed at the RTC(O) prior to 1992. Of these 334 were released and available for follow-up. For this sample, the Static-99R and Stable-2007 were scored from assessment/pretreatment file information. There was sufficient information to score the Static-99R on 326 cases, and for the Stable- 2007 for 168. Therefore, there were 168 cases with complete information.
Recidivism data for both samples were collected from official criminal records maintained by the Royal Canadian Mounted Police (RCMP). The Fingerprint Service (FPS) sheets (official Canadian criminal history) for each offender were obtained electronically and new convictions were coded according to the Cormier–Lang system (
For each sample we conducted a series of analyses, starting with calculating ROC AUCs for the Static-99R and Stable-2007 to examine the ability of the instruments to discriminate recidivists from non-recidivists in the current sample. This was followed by Cox Regression analysis to evaluate the extent to which the Stable-2007 discriminates recidivists from non-recidivists over time, first on its own and then together this the Static-99R, to determine incremental validity. Finally, we examined calibration with normative data via E/O indices.
The average age at release for Sample 1 was 43.3 (
In terms of recidivism, overall, 7.3% of the sample was detected to have sexually recidivated over an average 5.8 (
Examining the relationship of the Static-99R and Stable-2007 to recidivism using the AUC statistic indicated that both measures were significantly related to sexual and sexual + violent recidivism. For this analysis a fixed follow-up of 5 years was used. The AUC for the Static-99R for sexual recidivism was .69, 95% CI [.54, .83] and for the combined sexual + violent recidivism outcome AUC = .72, 95% CI [.64, .79]. For the Stable-2007, the corresponding values for recidivism type were AUCs = .77, 95% CI [.65, .90] and .67, 95% CI [.58, .77].
Regression model | Wald | Exp( |
95% CI Exp( |
||||
---|---|---|---|---|---|---|---|
Sexual Recidivism | |||||||
Step 1 | |||||||
Stable-2007 | .16 | .04 | 19.47 | 1 | < .001 | 1.17 | 1.09 – 1.25 |
Step 2 | |||||||
Static-99R | .18 | .06 | 8.16 | 1 | .004 | 1.20 | 1.06 – 1.36 |
Stable-2007 | .10 | .04 | 6.85 | 1 | .009 | 1.11 | 1.03 – 1.20 |
Sexual + Violent Recidivism | |||||||
Step 1 | |||||||
Stable-2007 | .12 | .02 | 27.58 | 1 | < .001 | 1.13 | 1.08 – 1.18 |
Step 2 | |||||||
Static-99R | .17 | .04 | 18.39 | 1 | .001 | 1.19 | 1.10 – 1.29 |
Stable-2007 | .07 | .03 | 7.41 | 1 | .006 | 1.07 | 1.02 – 1.13 |
For sexual + violent recidivism once again the Stable-2007 is a significant predictor of recidivism χ2(1) = 28.89,
Priority Category | Observed |
Expected |
E/O Index | 95% CI | |
---|---|---|---|---|---|
Below Average | 18 | 1 (5.6) | 0.95 (5.3) | 0.95 | 0.13 – 6.77 |
Average | 140 | 5 (3.6) | 10.5 (7.5) | 2.10 | 0.87 – 5.05 |
Above Average | 54 | 4 (7.4) | 7.3 (13.6) | 1.84 | 0.69 – 4.89 |
Well-Above Average | 51 | 14 (27.5) | 13.7 (26.8) | 0.98 | 0.58 – 1.65 |
Total | 263 | 24 (9.1) | 28.4 (10.8) | 1.18 | 0.79 – 1.76 |
The average age at release for Sample 2 was 35.5 (
In terms of recidivism, overall, 45.5% of the sample was detected to have sexually recidivated over an average 14.2-year follow-up (
Examining the relationship of the Static-99R and Stable-2007 to recidivism using the AUC statistic for a fixed 20-year follow-up indicated that both measures were significantly related to sexual and sexual + violent recidivism. The AUC for the Static-99R for sexual recidivism was .67, 95 CI% [.59, .75] and for the combined sexual + violent recidivism outcome it was .68, 95% CI [.59, .77]. For the Stable-2007, the corresponding AUCs = .68, 95% CI [.60, .77] and .65, 95% CI [.55, .74].
Regression model | Wald | Exp( |
95% CI Exp( |
||||
---|---|---|---|---|---|---|---|
Sexual Recidivism | |||||||
Step 1 | |||||||
Stable-2007 | .15 | .03 | 19.88 | 1 | < .001 | 1.16 | 1.09 – 1.24 |
Step 2 | |||||||
Static-99R | .20 | .08 | 6.30 | 1 | .012 | 1.20 | 1.04 – 1.42 |
Stable-2007 | .10 | .04 | 6.88 | 1 | .009 | 1.11 | 1.02 – 1.19 |
Sexual + Violent Recidivism | |||||||
Step 1 | |||||||
Stable-2007 | .05 | .03 | 3.45 | 1 | .063 | 1.05 | 0.99 – 1.10 |
Step 2 | |||||||
Static-99R | .11 | .06 | 3.31 | 1 | .069 | 1.11 | 0.99 – 1.26 |
Stable-2007 | .02 | .03 | 0.31 | 1 | .57 | 1.02 | 0.96 – 1.08 |
For sexual + violent recidivism once again the Stable-2007 by itself approaches significance as a predictor of recidivism χ2(1) = 3.46,
Priority Category | Observed |
Expected |
E/O Index | 95% CI | |
---|---|---|---|---|---|
Average | 23 | 3 (13.0) | 1.7 (7.5) | 0.58 | 0.19, 1.78 |
Above Average | 33 | 4 (12.1) | 4.5 (13.6) | 1.13 | 0.42, 2.99 |
Well-Above Average | 107 | 34 (31.8) | 28.7 (26.8) | 0.84 | 0.60, 1.18 |
Total | 163 | 41 (25.2) | 17.6 (10.8) | 0.43 | 0.32, 0.58 |
Examining the 95% confidence intervals for the sexual recidivism E/O analysis in
An interesting comparison is presented here – one sample being an average group of persons convicted of a sexual offense versus a high-risk comparison group, a group sent for assessment or treatment at a Regional Treatment Centre (RTC). In Canada, a federal RTC is generally a multi-level security institution, (low or medium security through high security and special units) operating within an institution as a legally acknowledged hospital generally offering psychiatric treatment and assessment that assists the Correctional Service of Canada by providing special services to people transferred from their home institution for assessment and treatment of a mental disorder or a physical condition that limits their ability to benefit from the general correctional regime. This is quite a high risk sample as noted by the initial number of men who populated this cohort (
Secondly, as shown in
The results of this study show that Stable-2007 scores reliably predict sexual recidivism for both a sample of incarcerated average risk men with sexual offenses and a high-risk sample. Further useful information is provided in the analysis of Sample 2 when the Stable-2007 predicts for person convicted of a sexual offense but this predictive ability fades when the compound variable of sexual + violent recidivism is examined as seen in
The current research indicates that although the Stable-2007 was developed and normed on a community-based sample, it has the ability to predict recidivism in an incarcerated sample. The findings also reinforce the importance of utilizing both static and dynamic risk measures to estimate future risk. Although static measures of risk provide generally robust predictions of future risk, the addition of a measure of dynamic risk, such as the Stable-2007, will enhance the accuracy of such estimations. These results indicated that those who conduct SVP assessment who have been resultant to adopt the Stable-2007 in their practice have no reason for concern.
The principle limitation to this data is the age of Sample 2, all entering the system prior to 1992. While some believe it is important to have the longest possible follow-up, these efforts are confounded by change in correctional climate and practice, including loss of data partially due to the transition from paper-based to digital records. This is compounded by historical record keeping practices that would “cull” inactive paper records after a given period of inactivity due to space and personnel limitations. It must be kept in mind that of the 506 men who were assessed or treated, complete data was only available for 168, approximately one-third of the identified sample. As seen in other work within mental health institutions (
In addition, institutional treatment in the 1990’s would have focused almost totally on developing a personal crime cycle as part of a relapse prevention approach. Both are based upon avoidance techniques that have not been seen to be as promising as they were once presumed to be. While it has to be quickly added that current, more approach goal oriented, techniques may prove in time to be more effective – this only goes to underscore the point that research of this nature is presenting a picture from a particular timeframe that has variable relevance to current practices.
Research efforts in this area would greatly benefit from multi-jurisdictional research agreements allowing follow-up of DO and SVP such that the small number of men released in each jurisdiction could be compiled into composite datasets, followed for fixed time intervals, and then reliably re-assessed as to recidivism status. The large problem here is that this process would take not only researcher sophistication and energy (neither of these factors are lacking) but it would take some political will to allow for inter-jurisdictional cooperation. Given the current climate it is unfortunate that the conditions necessary for advanced research cooperation on this topic are highly unlikely. This being said, larger jurisdictions could easily implement one of the standard dynamic risk assessment instruments throughout their jurisdiction and develop and publish in-house norms and guidelines that would test, replicate, and ultimately prove the utility of dynamic assessment in high-risk high-needs samples.
The data used in this paper will be made available to others. Please contact the first author.
The authors have no funding to report.
The authors have declared that no competing interests exist.
The authors have no additional (i.e., non-financial) support to report.