The Juvenile Sex Offender Recidivism Risk Assessment Tool-II (JSORRAT-II) was originally developed by Douglas Epperson in collaboration with the Utah Juvenile Justice Services as an actuarial assessment of risk. Actuarial assessments aim to make risk assessments based on well-established statistical relationships (e.g., weather forecasting relies on statistical models to predict storms but cannot predict with 100% accuracy), as opposed to clinical assessments (e.g., making a judgment of risk based on information gained from the psychosocial interview). The JSORRAT-II was designed to evaluate the risk of sexual recidivism for male sex offenders whose sexual offenses occurred between the ages of 12 and 18 years. The tool provides an overall score (ranging from 0 to 21) that is comprised of scores on 12 static, or unchanging, variables: number of adjudications for sexual offenses, number of victims, the time span of the documented sexual offending, presence of court-ordered supervision while offense occurred, offense occurred in a public place, presence of grooming or deception, treatment history, physical abuse history, sexual abuse history, special education placement, discipline problems during education time periods, and adjudications for nonsexual offenses. The tool, if used properly, can provide a fairly reliable measure of sexual recidivism risk, although it is currently only validated for use in Utah, Iowa, California, and Georgia.
Scoring
A number of items on the assessment are scored as either 0 or 1, signifying the absence or presence of risk factors, respectively (e.g., presence of court-ordered supervision while offense was committed, offense committed in a public place). Other items indicate a range of scores from 0 to 2 or 0 to 3 to denote a level of severity for specific risk factors (e.g., number of different victims is scored as 0 for one victim, 1 for two victims, or 2 for three or more victims). Individuals who receive an overall score of 8 of 21 or higher are placed at a moderate-to-high risk of recidivism, those who receive a score of 4–7 are deemed at moderate risk, scores of 1–3 classify an individual at a low-to-moderate risk, and those with a score of 0 are viewed as low risk.
Prior to completing the assessment, it is crucial that evaluators review the entire file and only include information contained within the file. The assessment requires data from the offender’s officially adjudicated offense case file; uncharged offense information should not be included in the scoring of items. This is due to the fact that the items that were found to contribute to the predictive validity of the assessment were developed by using only officially adjudicated offense information, and thus, predictive validity is maintained when users follow the same standard. When information provided in the case file is insufficient, evaluators are instructed to make an estimate when possible. If the information in the file is deemed insufficient to make an estimate, the item should be coded as U (unable to score). Further, if the item is not relevant to the individual, it should be coded as NA. The unambiguous and objective nature of the items, along with the detailed scoring instructions, potentially allows for high reliability.
Psychometrics
Research has evaluated the basic psychometric properties of the JSORRAT-II including its validity and reliability as a risk assessment measure. It was originally developed through examination of possible predictor variables from a large sample of juvenile sex offenders in Utah. Statistical analyses were then utilized to identify pertinent variables most predictive of juvenile sexual recidivism. Overall, evaluations of the accuracy of the JSORRAT-II demonstrate that its effectiveness as a predictor of recidivism prior to age 18 significantly exceeds chance (Area Under the Curve, AUC = 0.89, d = 1.74). The JSORRAT-II was also subjected to cross-validation using another sample in Utah (Area Under the Curve, AUC = 0.65, d = 0.53) as well as a sample in Iowa (Area Under the Curve, AUC = 0.65, d = 0.54). Additionally, it has demonstrated high inter-rater reliability across users (ICC = 0.96 in one study and 0.89 in another). While the JSORRAT-II has demonstrated predictive utility and adequate psychometric properties, no actuarial assessments are 100% accurate; thus, the tool should be used in the context of a larger assessment procedure.
It should also be noted that in one study, when validated on a sample totally independent from which it was originally developed, the JSORRATII did not accurately predict sexual reoffending. Interestingly, it has also been found that the JSORRAT-II did not achieve the same levels of predictive accuracy for juveniles who had offended exclusively against their siblings (ROC = 0.58; 95% CI from 0.43 to 0.73), which may prompt clinicians to use caution when interpreting results with this population.
Since the initial study of the JSORRAT-II’s accuracy, subsequent psychometric studies have revealed a decrease in the predictive accuracy of the assessment tool. Researchers have found that this diminished level of accuracy for the tool cannot be explained by missing data, severity of the index sexual offense, or age at risk assessment. However, it is still unclear exactly why the predictive accuracy decreased from the initial validation study compared to the subsequent studies.
Critiques
Despite the decreased accuracy over time with repeated validation studies, a strength of the JSORRAT-II is that it utilizes an actuarial approach to prediction. Given the many heuristic errors that can erode the accuracy of clinical predictions, it follows that actuarial prediction has generally been found to have increased accuracy compared to clinical prediction. However, this is not to say that any actuarial measure is satisfactory—its actual quality depends on the demonstrated accuracy of measurement, which in turn can only be interpreted in the context of what sort of accuracy is sought in the particular prediction task. For instance, a highly accurate prediction of cancer is more important than a highly accurate prediction of acne, and false negatives are more troublesome than false positives in cancer predictions, as providing treatment to someone who does not need it is less harmful than not treating someone with the disease. However, the examination of the rates and types of errors that are acceptable in the context of sex offending prediction have been insufficient. Further, as noted earlier, this instrument has significant limits on its predictive validity. There has been too little evaluation of the kinds and amounts of error that are acceptable in the context of sex offending predictions.
It is also important to note that many of the other psychometric properties (e.g. discriminate validity), as well as the degree to which the instrument generalizes beyond the validation sample, remain unknown. Specifically, it is unclear if this instrument can be used with special populations, such as those who are developmentally delayed, non-English speakers, females, and members of minority groups. Clearly, there is a need for more studies evaluating the psychometric properties, particularly with diverse samples. Further, the JSORRAT-II is primarily used in conjunction with other measures including record reviews, other diagnostic testing, and clinical interviews. These additional assessments are important as they allow for a more inclusive consideration of dynamic factors. However, this is also problematic as those utilizing the JSORRAT-II are instructed to solely rely on the information contained in the file, and violating this may disrupt the instrument’s predictive utility. A final issue to consider regards the broken leg problem—in which a dynamic factor (e.g., a recent broken leg) disrupts a well-established statistical relationship (e.g., playing basketball every Friday). Currently, there is a need for researchers to investigate how information from the JSORRAT-II ought to be synthesized with such additional dynamic information to produce more accurate conclusions (i.e., incremental validity).
Competitors
In addition to the JSORRAT-II, two of the other most widely used risk assessments for the prediction of juvenile sexual offenses are the Juvenile Sex Offender Risk Assessment Protocol-II (J-SOAP-II) and the Estimate of Risk of Adolescent Sexual Offense Recidivism (ERASOR). All three scales differ in several important ways, including their intended purpose, development, structure, and length. Additionally, moderators such as age and type of offense committed impact the validity of each scale. For instance, the J-SOAP-II and the ERASOR have been found to have higher levels of predictive validity for juveniles who have committed only sexual offenses compared to those who also had a history of nonsexual offending.
The J-SOAP-II is the most widely employed risk assessment tool and is designed to predict risk of sexual violence and general delinquency for adolescent males aged 12–18 years with a history of sexual aggression. The scale is composed of a 28-item checklist of risk factors that are clinically relevant and empirically supported. Items on the scale are scored on a three-point system, and like the JSORRAT-II, higher scores indicate increased risk of recidivism. Despite its wide use, the J-SOAP-II has no classification related to the various scores, and thus this scale is not an actuarial assessment but rather functions as an empirically informed guide.
The ERASOR is composed of 25 items and was created based on structured professional judgments. Evaluators make their own judgments of low, moderate, or high risk based on the items, as well as other factors, such as the psychosocial interview, not included in the file that may be relevant. The ERASOR, like the J-SOAP-II, distinguishes between static or immutable risk factors and dynamic or contextual risk factors. While the ERASOR and J-SOAP-II consider contextual information, both fail to provide actuarial assessments. The JSORRAT-II, on the other hand, provides a brief actuarial measure but is not always attuned to other relevant information. Overall, the JSORRAT-II has some evidence that it provides a fairly accurate risk estimate for juvenile sex offenders, although its predictive validity has been strongest in samples from which it was validated, prompting its use only within a larger assessment context.
References:
- Cottle, C. C., Lee, R. J., & Heilbrun, K. (2001). The prediction of criminal recidivism in juveniles: A metaanalysis. Criminal Justice and Behavior, 28(3), 367–394. Retrieved from https://doi.org/10.1177/0093854801028003005
- Epperson, D. L., & Ralston, C. A. (2015). Development and validation of the Juvenile Sexual Offense Recidivism Risk Assessment Tool-II. Sexual Abuse: A Journal of Research and Treatment, 27(6), 529–558. doi:10.1177/1079063213514452
- Epperson, D. L., Ralston, C. A., Fowers, D., DeWitt, J., & Gore, K. S. (2006). Actuarial risk assessment with juveniles who offend sexually: Development of the Juvenile Sexual Offense Recidivism Risk Assessment Tool-II (JSORRAT-II). In D. Prescott (Ed.), Risk assessment of youth who have sexually abused: Theory, controversy, and emerging strategies. Oklahoma City, OK: Woods N Barnes.
- Klima, T., & Lieb, R. (2008). Risk assessment instruments to predict recidivism of sex offenders: Practices in Washington State. Olympia, WA: Washington State Institute for Public Policy.
- Ralston, C. A. (2008). Validation of the Juvenile Sexual Offense Recidivism Risk Assessment Tool-II. Retrospective Theses and Dissertations.
- Viljoen, J. L., Scalora, M., Cuadra, L., Bader, S., Chávez, V., Ullman, D., & Lawrence, L. (2008). Assessing risk for violence in adolescents who have sexually offended: A comparison of the J-SOAP-II, J-SORRAT-II, and SAVRY. Criminal Justice and Behavior, 35, 5–23. Retrieved from https://doi.org/10.1177/0093854807307521