Criminal Justice

IResearchNet

Academic Writing Services

Academic Writing Services
Criminal Justice > Criminal Psychology > Criminal Risk Assessment > Risk Matrix 2000

Risk Matrix 2000




Of the various approaches used in sexual offender risk assessment (unstructured clinical judgment, actuarial risk assessment instruments [ARAIs], and structured professional judgment), ARAIs outperform all other methodologies. Of the numerous sexual offender ARAIs currently available, the Risk Matrix 2000 (RM2000) is one of the most widely used and researched and is the focus of this article.




Description

The RM2000 is probably the most frequently used second-generation ARAIs in the United Kingdom. In developing the scale, researchers created a two-dimensional risk assessment system for sex offenders referred to collectively as RM2000. The revised system has two scales, one for measuring risk of sexual recidivism—Risk Matrix 2000/ Sexual (RM2000/S)—and one for measuring risk of nonsexual violent recidivism—Risk Matrix 2000/Violent (RM2000/V) in sexual offenders.

The RM2000/S and RM2000/V scales were constructed to yield four summary risk categories: low, medium, high, and very high. The scores of both scales can be combined to give a composite risk of reconviction for sexual or nonsexual assault—Risk Matrix 2000/Combined. In developing the RM2000 scales, the authors referred to a landmark meta-analysis of sexual reoffending predictors as they felt this study offered a more complete guide to the literature, containing more precise and representative estimates of the predictive accuracy of individual factors than would be obtained from any individual study.

The RM2000/S uses a two-step system to risk assessment. Step 1 contains three risk items (number of previous sexual appearances, number of criminal appearances, and age divided into three age bands), the sum of which is translated into a risk category. Step 2 considers four aggravating risk factors (any conviction for sexual offense against a male, any conviction for a sexual offense against a stranger, any conviction for a noncontact sex offense, and single–never been married); the presence of two or four aggravating factors raises the risk category by one or two levels, respectively.

The RM2000 can only be used if the offender meets the following tests:

  1. The offender must be aged 18 years or over.
  2. The offender must have at least one conviction for a sexual offense committed when the offender was aged 16 years or over.
  3. The offender must be male.

All appearances for sexual offenses are counted when scoring RM2000/S, including ones where the offender was under the age of 16 years. The scale can also be used in cases where an individual has been charged with a sexual offense but not yet convicted.

The scale has demonstrated good interrater reliability with an absolute agreement for a single rating of 0.90 (which is considered excellent) across 50 cases. The item with the lowest reliability was the noncontact item, which still had good interrater reliability (93.2% agreement). Agreement for the risk category rating is also good (94.9% agreement).

Predictive Accuracy

Since the 2003 development study, the scale has been independently validated by several research studies using samples in Scotland, Denmark, England and Wales, Canada, and the United States. In the development study, the research group examined the predictive accuracy of the RM2000/S on a treated and an untreated sexual offender sample using the area under the curve (AUC), which is a measure of prediction accuracy ranging from 0 to 1.0; values of 0.56, 0.64, and 0.71 represent small, medium, and large prediction effects, respectively. The RM2000/S yielded good accuracy in predicting sexual reconviction (AUC = 0.77 treated sample, and AUC = 0.75 untreated sample), and the RM2000/V obtained higher validity indices (AUC > 0.78) in predicting of nonsexual violent reconviction in different long-term samples. Conducting a series of cross-validation studies with different England and Wales samples elsewhere reported varying results for the predictive accuracy of the RM2000 scales. On the one hand, the tools yielded impressive validity indices up to AUC = 0.86 for the prediction of violence including sexual recidivism by the RM2000/V. On the other hand, the predictive accuracy was hardly better than chance (AUC = 0.56).

A further validation study was completed using the RM2000 on a Scottish sample of sexual offenders compared against a sample from England and Wales. Of the 771 sexual offenders, 116 (15.0%) were reconvicted of a sexual offense at any time following their release from prison, while 83 (10.8%) were reconvicted of a sexual crime within 5 years of their prison release. This compares with a 19.6% 5-year sexual reconviction rate for the 1979 England and Wales cohort. The predictive accuracy of the RM2000/S in the Scottish sample was AUC = 0.73 compared with an AUC of 0.75 in the England and Wales 1979 sample.

In 2010, another research group examined the predictive accuracy of the RM2000 scales on a large sample of 9,824, of which 4,946 sexual offenders were followed up at 2 years and 578 followed up at 4 years. The predictive validity of the scale was explored for different subgroups of sexual offenders. The higher risk groups offended more quickly and at a higher rate than lower risk groups. The RM2000/S obtained moderate results (AUC = 0.68) in predicting sexual reconviction, whereas the RM2000/V obtained good results in predicting nonsexual violent reconviction (AUC = 0.80).

As a subsequent evaluation of nine ARAIs on a sample of 590 sexual offenders in Massachusetts followed up over a 15-year period, the RM2000/S scale performed reasonably (AUC = 0.68) in predicting sexual reconviction. Compared to the other scales in the study, the RM2000 scales were outperformed by the STATIC-99, a brief actuarial tool. Incidentally, the RM2000/V (AUC = 0.71) obtained the highest score across all nine scales in predicting violent reconviction.

In a 2013 meta-analysis, the findings were integrated from 16 unique samples (derived from 14 studies), which have examined the ability of the RM2000 to discriminate between recidivists and nonrecidivists. All three RM2000 scales provided significant predictive power for all recidivism types that were investigated (i.e., sexual, nonsexual violent, any violent, nonviolent, and any recidivism). The RM2000/S yielded the best predictive accuracy for sexual recidivism (Cohen’s d = 0.74, considered large). The RM2000/V and the Risk Matrix 2000/Combined predicted nonsexual violent recidivism and any violent recidivism with similarly large effect sizes. The effect sizes were significantly higher using samples from the United Kingdom compared to studies with samples from other countries. The RM2000/S demonstrated moderate to large predictive accuracy among sex offenders, and the RM2000/V and Risk Matrix 2000/Combined are similarly effective for both types of violent recidivism as well as nonviolent and any recidivism. However, the authors noted that the predictive accuracy was quite low in sexual offender samples preselected as high risk or high need.

More recently, in a study examining the relationship between psychometric changes in treatment and recidivism in a sample of 3,773 sex offenders, the RM2000/S was used to establish a baseline of risk. Of the 3,027 offenders who had been released for at least 2 years, 1.7% had reoffended sexually to 4.4% who were reconvicted for sexual and violent offenses combined. Due to limitations in the study data, it was not possible to score two items: intimate relationship and stranger items of RM2000/S, so they used a modified RM2000/S. Since the modified RM2000/S scoring may have resulted in an underestimate of risk, those who were scored as having any aggravating factors were raised a risk category. The authors compared the modified version against a subsample (about half of the total cases) for whom full RM2000/S scores were available and found that approximately 64% of these were in the same risk category, 9% were rated as a higher risk category, and 27% were rated as a lower risk category. This indicated that the modified scoring procedure appeared to result in a greater level of underscoring than overscoring. Using the modified version, the RM2000/S obtained moderate results (AUC = 0.64).

Applying the RM2000/S to Child Molesters and Rapists

A 2011 study followed 275 sexual offenders (n = 199 child molesters; n = 76 rapists) over a period of 9 years and 2 months as part of an evaluation of a community-based sexual offender treatment program. For the child molester sample, there is a fairly robust relationship between increased risk level and increased rate of reoffending, with the medium–high categories having a significantly higher failure rate. However, the RM2000/S did not perform well with rapists (AUC = 0.64) compared to child molesters (AUC = 0.71).

The RM2000/S fared better in a Danish sample of 160 rapists and 144 child molesters, followed up over an average of 14.8 years, obtaining moderate results for sexual (AUC = 0.71) and nonsexual violence reconviction (AUC = 0.70).

Combining Static and Dynamic Frameworks

A 2015 study investigated the incremental validity of combining the RM2000 scales with the STABLE-2007 (a dynamic actuarial risk scale) on a Canadian sample of 710 sexual offenders (mean follow-up of 7.7 years). The RM2000/S obtained moderate predictive accuracy (AUC = 0.69) for sexual recidivism compared to the RM2000/V (AUC = 0.63), which performed better in predicting nonsexual violence (AUC = 0.74). The RM2000/S had a larger incremental effect over the STABLE-2007 than vice versus.

Special Populations

Sexual Offenders With Intellectual Developmental Disabilities (IDD)

Researchers in this field have noted that sexual offenders with IDD (IQ below 69) would have been rare in the original research samples, but offenders with IQs between 70 and 80 should have been present in the original samples in reasonable numbers.

A number of studies have examined the utility of the scale in a sample of sexual offenders with IDD with moderate results. The predictive accuracy of the RM2000/S and the RM2000/V scales were compared against other scales with a mixed group of 212 violent and sexual offenders with IDD in England and Wales (mean IQ = 66), across three different levels of security. After a 12-month follow-up period, both the RM2000/V and RM2000/S AUCs did not attain significance for violent (AUC = 0.62) and sexual incidents (AUC = 0.61), respectively.

A 2011 study compared the Rapid Risk Assessment for Sexual Offense Recidivism (RRASOR), the Sexual Violence Risk-20 (a 20-dynamic item structured professional judgment sexual violence risk tool), the RM2000-V, and the Assessment of Risk Manageability for Intellectually Disabled Individuals Who Offend-Sexually-Stable and -Acute dynamic client subscales on a sample of 88 offenders: 44 mainstream and 44 sexual offenders with special needs (mean IQ = 69). The mean length in treatment was 18 months with a mean follow-up period of 109 months. While the sexual reconviction predictive validities of the RRASOR (AUC = 0.53) and the RM2000-V (AUC = 0.50) were little better than chance, the Assessment of Risk Manageability for Intellectually Disabled Individuals Who Offend-Stable (AUC = 0.60) and Assessment of Risk Manageability for Intellectually Disabled Individuals Who Offend-Acute, (AUC = 0.73) performed better. It should be noted that the RM2000/V is generally better at predicting violent reconviction than sexual reconviction.

A further study examined the predictive accuracy of the STATIC-99, RRASOR, and the RM2000/S in a sample of 27 sexual offenders with IDD in England and Wales followed up over a 76-month period. The STATIC-99 (AUC = 0.64) outperformed the RM2000/S (AUC = 0.58) and the RRASOR (AUC = 0.42). These results suggest caution when using the RM2000 scales with sexual offenders with IDD.

Internet Sexual Offenders

Offenders who have only been convicted of Internet offenses would not have been present in the original development RM2000 samples. However, in the development sample, the RM2000 was tested on a sample of adult males with a history of being convicted for Internet sex offenses and attained an AUC that is comparable to what it has in regular samples. Like most ARAIs, the RM2000/S does not discriminate among individuals whose only known criminal behavior is downloading indecent images of children as these groups typically score low on such scales and have a low sexual recidivism rate. Individuals with Internet-only offenses would meet the scoring criteria for noncontact offenses. However, a key issue is whether these offenses were a prelude to a contact offense.

Of the few studies validating the scale on Internet sexual offenders, the predictive accuracy of a modified version of the RM2000/S was examined in a sample of 1,344 Internet sexual offenders. The overall rates of sexual reoffending were very low at 2.1% at 1-year follow-up, which increased to just 3.1% at the 2-year follow-up stage. At the 2-year follow-up, 74% (23/31) of the offenders were reconvicted of Internet-related offenses, 19% (6/31) were reconvicted of non-Internet offenses. The RM2000/S obtained moderate results in predicting reoffending (AUC = 0.67).

Conclusions

A number of cross-validation and meta-analytical studies have demonstrated the psychometric properties and predictive utility of the RM2000 scales. Compared with other ARAIs, the RM2000/S demonstrates moderate to large predictive accuracy among contact sex offenders (rapists and child molesters) and Internet sexual offenders although the predictive accuracy of the scale reduces slightly when applied to sexual offenders with IDD. Nevertheless, the RM2000/S can be used with confidence in assessing future risk of sexual reconviction in sexual offenders.

References:

  1. Boer, D. P., Hart, S. D., Kropp, P. R., & Webster, C. D. (1997). Manual for the sexual violence risk-20: Professional guidelines for assessing risk of sexual violence. Vancouver, Canada: The Mental Health, Law and Policy Institute.
  2. Craig, L. A., Browne, K. D., & Beech, A. R. (2008). Assessing risk in sex offenders: A practitioner’s guide. Chichester, UK: Wiley.
  3. Hanson, R. K., & Thornton, D. (2000). Improving risk assessments for sex offenders: A comparison of three actuarial scales. Law and Human Behavior, 24, 119–136. Retrieved from https://doi.org/10.1023/A:1005482921333
  4. Thornton, D., Mann, R., Webster, S., Blud, L., Travers, R., Friendship, C., & Erikson, M. (2003). Distinguishing and combining risks for sexual and violent recidivism. In R. A. Prentky, E. S. Janus, & M. C. Seto (Eds.), Annals of the New York Academy of Sciences, Vol. 989. Sexually coercive behavior: Understanding and management (pp. 225–235). New York, NY: New York Academy of Sciences.
  5. Wakeling, H., Beech, A. R., & Freemantle, N. (2013). Investigating treatment change and its relationship to recidivism in a sample of 3773 sex offenders in the UK. Psychology, Crime and Law, 19, 233–252. Retrieved from https://doi.org/10.1080/1068316X.2011.626413