The Minnesota Sex Offender Screening Tool-Revised (MnSOST-R) is a 16-item, “actuarial” risk assessment tool initially developed for the Minnesota Department of Corrections (MDOC) to provide empirically based estimates of risk for sexual recidivism for incarcerated male sex offenders. Indices of reliability have ranged from .76 to .86 across four studies and have generally been .80 or greater. The MnSOST-R has been successfully validated in Minnesota; North Dakota; and Ontario, Canada; however, it failed to predict significantly better than chance in an Arizona study.
MnSOST-R Description and Development
The MnSOST-R was developed in response to a 1991 MDOC report calling for a more formal and uniform process to identify predatory and violent sex offenders. An actuarial approach was used in an attempt to bring greater accuracy and utility to sex offender risk assessments, enabling the MDOC to more effectively use limited resources.
The development sample for the MnSOST-R included 256 incarcerated sex offenders in Minnesota who were released primarily in 1988 or 1990. This sample excluded only those offenders whose offenses consisted exclusively of “fondling” offenses against family members that did not involve vaginal or anal penetration of a child aged 13 or younger or the rape of an older family member. Research on an earlier version of the instrument indicated that this group was substantially different from other sex offenders and generally presented fewer concerns regarding release decisions (e.g., level of supervision, level of community notification, and potential referral for civil commitment).
Sex offenders were sampled from each relevant Minnesota correctional facility, and the offenders were 32-42 years, on average, with a range from 17 to 70 years. Sixty-six percent of the sample was White, 24% was Black, 5% was Hispanic, 4% was Native American, and 2% were from other ethnic groups. There was some oversampling of sexual recidivists in the development sample to provide more stability in any observed relationships between sexual recidivism and potential predictor variables.
Potential predictors were drawn from research on an earlier version of the tool and from an updated review of the literature. Only variables based on information routinely available in correctional records were considered as predictors to ensure that the resulting tool could be scored for the majority of sex offenders based on a file review. Sexual recidivism, the criterion variable, was defined as a formal charge for a new sex offense within 6 years of release from prison.
Empirically based item selection and scoring procedures identified 16 items as the optimal predictors of sexual recidivism, including 12 historical/static items and 4 institutional/dynamic variables. The 12 historical variables included the number of convictions for sex offenses, length of sex offending history, commission of a sex offense while under court supervision, commission of a sex offense in a public place, use or threat of force in any sex offense, perpetration of multiple sex acts in a single event contact, offending against victims from multiple age groups, offending against a 13- to 15-year-old victim with more than a 5-year age difference between the offender and the victim, victimization of a stranger, persistent pattern of adolescent antisocial behavior, recent pattern of substantial substance abuse, and recent employment history. The four institutional variables included discipline history, chemical dependency treatment recommendations and outcomes, sex offender treatment recommendations and outcomes, and age of the offender at the time of release.
Total MnSOST-R scores were significantly predictive of sexual recidivism in the development sample, as reflected by an area under the receiver operator characteristics (ROC) curve of .77 (95% confidence interval [CI] of .71 to .83). Total MnSOST-R scores were equally predictive of sexual recidivism in the development sample for rapists (ROC = .79) and molesters (ROC = .75) and for minorities (ROC = .75) and nonminorities (ROC = .77).
Reliability studies have yielded positive results across a variety of settings with varying degrees of training. A Minnesota study involving a minimal 2-hour training session for 10 participants, who then scored the same 11 cases by the end of the day, produced intra-class correlation coefficients (ICCs) of .80 for relative agreement and .76 for absolute agreement. A Florida study involving an optimal 1.5-day training session for 27 participants, who then scored the same 10 cases over the next 3 months, yielded ICCs of .87 for relative agreement and .86 for absolute agreement. Two Canadian studies produced interrater reliability coefficients of .80 and .83.
The MnSOST-R was validated in Minnesota with an exhaustive sample of 220 sex offenders released from prison in 1992 who met the same inclusion criteria used in the development study. This sample was very similar demographically to the development sample, and sexual recidivism was defined in the same way. Total MnSOST-R scores were significantly predictive of sexual recidivism in this sample (ROC=.73, 95% CI of .65 to .82).
Two validation studies were conducted in North Dakota. The first sample included 182 incarcerated sex offenders with an average time at risk of 8 years, and the second sample included 271 probated sex offenders with an average time at risk of 10 years. No sex offenders were excluded in either of these samples. The MnSOST-R significantly predicted sexual recidivism in the incarceration sample (ROC = .76, 95% CI of .66 to .85) and in the probation sample (ROC = .75, 95% CI of .63 to .88).
Howard Barbaree, Calvin Langton, and their associates conducted a validation study of the MnSOST-R with two Canadian samples. Because the second, bigger sample largely subsumed the first sample, the results of the second study are summarized here. That sample of 354 sex offenders who were at risk for an average of 5.9 years yielded a significant ROC = .70 for the MnSOST-R (95% CI of .62 to .77).
Darci Bartosh and her colleagues conducted a validation of several risk assessment tools, including the MnSOST-R, with a sample of 186 sex offenders in the state of Arizona who were at risk for approximately 5 years. The resulting ROC = .58 missed the threshold for statistical significance in this study, though it was only slightly lower than the ROC values for the other instruments assessed.
The North Dakota, Canadian, and Arizona studies also assessed the Rapid Risk Assessment for Sexual Offense Recidivism (RRASOR) and the STATIC-99, and none of these studies identified significant differences between the tools. In fact, the scores were clustered fairly tightly within each of these studies. The respective ROC values for the MnSOST-R, STATIC-99, and RRASOR were, respectively, .76, .75, and .73 in the North Dakota prison study; .75, .78, and .77 in the North Dakota probation study; .70, .64, and .68 in the Canadian study; and .58, .64, and .63 in the Arizona study.