Long Term Recidivism Rates Among Individuals at High Risk to Sexually Reoffend

Preventive detention provisions in the US and Canada assume we can identify, in advance, individuals at high risk for sexual recidivism. To test this assumption, 377 adult males with a history of sexual offending were followed for 20 years using Canadian national criminal history records and Internet searches. Using previously collected information, a high risk/high need (HRHN) subgroup was identified based on an unusually high levels of criminogenic needs (n = 190, average age of 38 years; 83% White, 13% Indigenous, 4% other). A well above average subgroup of 99 individuals was then identified based on high Static-99R (6+) and Static-2002R (7+) scores. In the HRHN group, 40% reoffended sexually. STATIC HRHN norms overestimated sexual recidivism at 5 years (Static-99R, E/O = 1.44; Static-2002R, E/O = 1.72) but were well calibrated for longer follow-up periods (20 years: Static-99R, E/0 = 1.00; Static-2002R, E/O = 1.16). The overall sexual recidivism rate for the well above average subgroup was 52.1% after 20 years, and 74.3% for any violent recidivism. The highest risk individuals (top 1%) had rates in the 60%–70% range. We conclude that some individuals present a high risk for sexual recidivism, and can be identified using currently available methods.

To construct an implementation for the current study the data dictionary was reviewed and variables classified according to their potential relevance to the constructs in the above table .Where alternative descriptions were available variables which implied some persistence of the construct were chosen.Within each domain selected variables were subjected to exploratory factor analysis to group them into more homogeneous categories, better corresponding to the long-term vulnerabilities.For each domain at least two summary scores and levels of these were recoded so that they fell between 0 and 2. The summary scores for each domain were averaged to create a summary score for each domain and then these domain summary scores were summed to give a Need score running between 0 and 8.
The following table shows the within domain summary scores that resulted from this process.

Resistance to Supervision
It was not possible to find an operationalization of Sexualized Violence as a long-term vulnerability with the available variables, nor were there appropriate indicators for Dysfunctional Coping, but otherwise coverage of the domains was fairly good.This implementation has two substantive limitations.First, with the exception of the PCL-SV, none of the indicators are established measures with known reliability and validity.Secondly, with some exceptions, most ratings are heavily based on the conduct of individuals as seen through the eyes of supervising agents who saw them for a limited period of time.This latter limitation means they likely overemphasize how long-term vulnerabilities manifested themselves during recent periods of supervision.
Determining a Threshold Level of Need to Identify HRHN Individuals Three considerations informed the selection of this threshold.
Conceptually we understand "high need" individuals as having multiple intense criminogenic needs.This can be contrasted with individuals whose criminogenic needs are less intense, or more circumscribed, or those whose offending resulted from transitory, situational, factors, and who therefore do not have sustained criminogenic needs.In SRA terms this concept suggests that "high need" individuals should have serious criminogenic needs from than one domain and that if only two domains are involved then there should be serious problems in more than one criminogenic need within each domain.
For clinical practice, Thornton & Helmus (2021) have recommended using the HRHN norms when an individual's score on a dynamic instrument is equal to or exceeds the mean score typically found in HRHN samples.Hanson and Thornton (2012) reported a review of mean scores in HRHN samples of 27.24, 3.26.and 14.70 for the VRS-SO Initial Dynamic Risk scale, the SRA-FV, and the STABLE-2007.Although these means vary substantially, resulting from how ratings are scaled in the different instruments, in each case the HRHN sample mean is close to half the maximum possible score on the instrument.The maximum on the VRS-SO Initial Dynamic Risk scale is 51, for SRA-FV it is 6, and for STABLE-2007 it is 26.Thus, in each case the means in HRHN samples are close to but just a little above half the maximum possible score on the instrument.This suggests that the threshold adopted on our instrument should correspond approximately to half the maximum possible score.
The threshold adopted should be consistent with selecting a higher need group while also assigning sufficient numbers to that group to permit statistical analyses.This suggests that the threshold adopted should be close to the sample median.These considerations do not fully constrain what the threshold should be but they do provide guidance.The actual threshold adopted was a score of at least 4.Such a score can be obtained by having serious, multi-need, problems within two domains; having a serious single-need problem in all four domains; or some mixture of serious, single and multi-need, problems in three domains.It therefore is consistent with the first consideration.It corresponds to half the maximum possible score, so it is consistent with the second consideration.And it is close to the median found in the complete sample which was 4.2.It is therefore consistent with the third principle.
Importantly, both construction of the present SRA Need score and selection of the threshold were done while blind to the recidivism data.