Appendix 1 - American Psychological Association



Supplemental MaterialsDoes Reassessment of Risk Improve Predictions? A Framework and Examination of the SAVRY and YLS/CMIby J. Viljoen et al., 2016, Psychological Assessment 1Shelf-Life Hypothesis88455551435Risk AssessmentOffending Time 1Offending Time 200Risk AssessmentOffending Time 1Offending Time 2Dynamic Change Hypothesis for Two Assessments884555210185Risk Assessment Time 1 Risk Assessment Time 2OffendingWithin-Individual Change0Risk Assessment Time 1 Risk Assessment Time 2OffendingWithin-Individual ChangeDynamic Change Hypothesis for More than Two Assessments88138020320Risk Assessment at t – Person-Mean Risk Total (t1… ti)Offending at t lagged0Risk Assessment at t – Person-Mean Risk Total (t1… ti)Offending at t laggedFamiliarity Hypothesis848360186055Risk Assessment Time 1Offending Risk Assessment Time 2Offending 00Risk Assessment Time 1Offending Risk Assessment Time 2Offending Appendix 2Shelf-Life Hypothesis: Predictive Validity of Total Scores on Risk Assessment Tools Will Diminish Over TimeAuthorsSampleToolMethodologyResultsYouth StudiesMeyers & Schmidt (2008)121 male and female adolescent offenders, CanadaSAVRYRAs rated SAVRY from files. Follow-ups were at 1 and 3 years.AUC for violent recidivism was .66 for a 1-year follow-up and .77 for a 3-year follow-up. This was not significant (i.e., CIs overlapped).Olver et al. (2012)167 male and female adolescent offenders, CanadaYLS/CMI RAs rated YLS/CMI from files in most cases. Mean follow-up was 6.8 years.AUCs for YLS/CMI total score were somewhat higher for adolescent vs. adult recidivism (.74 to .77 vs. .66 to .75), but this was not significant.Ralston & Epperson (2013)636 male adolescents who had sexually offended, USAMnSOST-R, Static-99, J-SOAP-II, JRASRAs rated tools from files. Mean follow-up was 2.8 years for juvenile records, and 12-14 years for adult records.AUCs were lower for adult than adolescent sexual recidivism (.58 – .63 vs. .72 – .81). The MnSOST-R and JRAS were significantly less accurate for adult vs. juvenile sexual recidivism. Schwalbe (2007)58,052 male and female adolescent offenders, USA28 toolsConducted a meta-analysis of 28 published and unpublished studies. Average follow-up length did not moderate the predictive validity of tools; association between follow-up length and AUC values was r = -.01. Schwalbe (2008)57,938 male and female adolescent offenders, USA11 toolsConducted a meta-analysis of 19 published and unpublished studies.Studies with longer follow-up lengths had smaller effect sizes; association between follow-up length and AUC values was r = -.38 to -.39. Takahashi et al. (2013)389 male adolescent offenders, JapanYLS/CMIPsychologists rated the YLS/CMI. Examined reoffense records for 6, 12, and 18-month cumulative follow-ups. No significant differences in AUCs across time (i.e., CIs overlapped). For violent offending, AUCs for the YLS/CMI total score were .72, .80, and .80 for the 6, 12, and 18-month follow-ups. Viljoen et al. (2012)6,196 male adolescents with sexual offenses4 tools (e.g., J-SOAP-II, ERASOR)Conducted a meta-analysis of 33 published and unpublished studies. Average follow-up length (i.e., > 5 years vs. < 5 years) did not moderate the predictive validity of tools for sexual and general reoffending. Worling et al. (2012)191 male youth with sexual offenses, CanadaERASORClinicians rated the ERASOR. Some youth were followed for 3.7 years, others for 1.4 years.ERASOR Total Score showed significantly higher predictive validity for sex offense at 1.4 years vs. 3.3 years (AUC = .93 vs. .72).Meta-Analyses with Adults or Primarily AdultsBlair et al. (2008)13,653 male and female offenders/ patients, USAVRAG, SORAG, Static-99Meta-analysis of 38 published studies.Average follow-up length did not significantly moderate the validity of tools in predicting recidivism. Fazel et al. (2012)24,847 males and females (some youth)9 tools (e.g., SAVRY, HCR-20)Meta-analysis of 68 published and unpublished studies.Average follow-up length did not significantly moderate the effect sizes for predictions of violent, sexual, or any recidivism. Singh, Grann, & Fazel (2011)25,980 males and females (some youth)9 toolsMeta-analysis of 68 published and unpublished studies.Follow-up length did not significantly moderate predictive validity when measured categorically (i.e., > vs. < 2 years) or continuously.Smith et al. (2009)14,737 female offendersLSI-RMeta-analysis of 25 published and unpublished studies.Studies with shorter follow-ups (i.e., < 12 vs. >12 mo.) had larger effect sizes (r = .43 vs. .28).Yang et al. (2010)6,348 to 7,221 males and females 9 toolsMeta-analysis of 28 published and unpublished studies.Longer follow-up time was associated with larger effect size for violence, 2 = 7.73, p < .01.Note: ERASOR (Estimate of Risk of Adolescent Sexual Offense Recidivism; Worling & Curwen, 2001); HCR-20 (Historical–Clinical–Risk Management-20; Webster, Douglas, Eaves, & Hart, 1997); JRAS (Juvenile Risk Assessment; Hiscox, Witt, & Haran, 2007); J–SOAP–II (Juvenile–Sex Offender Assessment Protocol–II; Prentky & Righthand, 2003); JSORRAT-II (Juvenile Sexual Offender Recidivism Risk Assessment Tool—II; Epperson et al., 2006); LSI-R (Level of Service Inventory-Revised; Andrews & Bonta, 1995); MnSOST-R (Minnesota Sex Offender Screening Tool–Revised; Epperson et al., 2004); Static-99 (Hanson & Thornton, 2000); SORAG (Sex Offender Risk Appraisal Guide; Quinsey, Harris, Rice, & Cormier, 1998); VRAG (The Violence Risk Appraisal Guide; Harris, Rice, & Quinsey, 1993); YLS/CMI (Youth Level of Service/Case Management Inventory; Hoge & Andrews, 2002). Numerous adult studies have compared more proximal and distal assessments (e.g. Chu et al., 2013; Craig, Beech, & Browne, 2000; Dahle, 2006; Dickens & O’Shea, 2015; Snowden, Gray, Taylor, & MacCulloch, 2007; Wilson et al., 2013; Wong & Gordon, 2006). As such, our review of adult studies focused on meta-analyses. Appendix 3Dynamic Change Hypothesis: Changes in Total Scores Will Predict Likelihood of ReoffendingAuthorsSamplea ToolMethodologyResultsYouth StudiesStockdale (2008)22 male and female adolescent offenders, Canada VRS-YVRated VRS-YV from files at pre- and post-treatment. Examined new juvenile offenses.Correlations between VRS-YV Total Change scores and any and violent offenses were not significant (r = .12 and -.16, respectively).Viljoen et al. (2015)163 male youth in program for sex offending, USAJ-SOAP-II, SAVRYRAs rated tools from files. Mean treatment length was 1.13 years. Mean follow-up was 8.07 yearsDynamic change scores did not predict reoffending (r = -.03 to .08), even after controlling for pre-treatment scores and risk (r = -.02 to .08).Adult StudiesBeggs & Grace (2011)218 male adult sex offenders (incest), New Zealand VRS-SO RAs rated files. Time between assessments is unknown. Mean follow-up was 12.24 years.Dynamic Change scores predicted reduced sexual reoffending (AUC = .70), but was not significant after controlling for static and dynamic factors. Blanchard (2013)235 adult civil psychiatric patients and correctional offendersHCR-20STARTRAs rated tools from interviews and files every month over a 6-month period. Examined violence in the subsequent assessment.Increases in START Strengths predicted decreased serious violence, but Vulnerabilities were not predictive. Increases in HCR-20 Clinical and Risk scales predicted increased violence. Cohen et al. (2016)32,647 male and female adult probationers, USAPCRAProfessionals rated PCRA. Mean of 9 months between assessments. Follow-up was 12 months.Increases in PCRA total scores were related to increased rearrests (OR = 1.92 – 4.17). de Vries Robbé et al. (2015) 108 male adult forensic patients, NetherlandsHCR-20, SAPROFRAs rated tools from files. Mean length of treatment 5.56 years. Followed for a mean of 11.28 years. After controlling for historical factors, age, and follow-up length, SAPROF and HCR-20 Total Change Scores predicted reduced violent recidivism (O.R. = 1.26 and 1.15, respectively). Goodman-Delahunty & O’Brien (2014)197 male adult sex offenders, Australia VRS-SORAs rated tools from files. Mean treatment length was 23.4 months. Followed for 9.1 years.After controlling for treatment completion and pre-treatment Total Scores, Total Change did not predict sexual, violent, or general reoffending.Hanson et al. (2007)612 male adult sex offenders, Canada and USASTABLE-2000/2007ACUTE-2000/2007Parole officers rated tools at baseline, and again at 10 months and every 6 months. Offenders followed for a mean of 3.4 years.Difference scores on STABLE-2007 and ACUTE did not predict any type of reoffending. Hogan & Olver (2016) 99 adult forensic inpatients (mostly male), CanadaVRS, START, HCR-20 V3RAs coded tools for pre- and post-treatment based on file information. Examined aggression during inpatient stay.Change scores on VRS Dynamic, HCR-20 (Relevance Ratings), and START Vulnerability Total Scores significantly predicted decreased inpatient aggression, controlling for baseline scores and length of stay.Labrecque et al. (2014)828 male and female adult offenders on probation, USALSI-RProfessionals conducted LSI-R assessments. Mean of 9.5 months between assessments. One-year reoffense records were obtainedLSI-R Total Change scores predicted likelihood of rearrest (O.R. = 1.05) in a multivariate model that included demographics (age, gender, race, offense type) and initial LSI-R score.Lewis et al. (2013)150 male adult offenders in a violence treatment program, CanadaVRSResearchers coded VRS from files for pre- and post-treatment (with some clinical cases). Followed sample for a mean of 4.9 years. VRS Total Dynamic Change scores predicted violent reoffending (r = -.21 to -.23). When VRS static and pre-treatment scores were controlled, change predicted violent recidivism (eB = 0.92).Michel et al. (2013)248 male patients, Canada, Finland, Sweden, GermanyHCR-20Psychiatric researchers conducted HCR-20 assessments every six months. Followed for a two years.Changes in the Risk score predicted reduced aggression (O.R. = 2.81). Changes in the Clinical score were not significant (O.R. = 1.56). Olver, Christofferson, Grace, & Wong (2014)539 adult male sex offenders, Canada and New ZealandbVRS-SORAs rated VRS-SO from files for pre- and post-treatment. Follow-up was mean of 15.5 years post-release.VRS Total Dynamic Change scores predicted reduced sexual, violent, and general reoffending (r = -.16 to .24). It remained predictive after controlling for static and pre-treatment scores. Olver, Nicholaichuk, Kingston, & Wong (2014)676 adult male sex offenders in treatment programs, CanadaVRS-SOProfessionals rated VRS-SO pre- and post-post treatment. Program was 2 – 9 months. Followed offenders for a mean of 6.31 years. VRS Total Dynamic Change scores did not predict reduced sexual, violent, and general recidivism in zero-order correlations (r = -.04 to -.05). However, after partialling out pre-treatment scores, it reached significance (r = -.09 to -.13).Olver et al. (2007) 321 adult male sex offenders in treatment program, CanadaVRS-SORAs rated VRS-SO from files at pre- and post-treatment. Program length was 6 – 8 months. Follow-up was a mean of 10 years.Total Dynamic Change scores did not directly predict sex recidivism (r = -.09), but reached significance after controlling pre-treatment scores and follow-up length. Change predicted in high but not low risk offenders (r = -.15 vs. .01). Raynor (2007)360 male and female adult probationers, United Kingdom LSI-RLSI-R was rated in probation practice. Length of time between reassessments is unknown. The follow-up was 12 months.Offenders whose LSI-R score decreased vs. increased were significantly less likely to be reconvicted (42% vs. 67%). This pattern was found in both low and high risk offenders.Vose et al. (2013) 2,849 adults on probation and parole, United States LSI-RLSI-R completed as part of practice. Time between first and second assessment was 1 year. Mean follow-up was 2.79 years.Change scores predicted reduced reoffending when controlling for demographics, time-at-risk, and initial LSI-R scores. Change was more predictive in high vs. low risk offenders. Note: ACUTE-2007 (Hanson, Harris, Scott, & Helmus, 2007); HCR-20 (Historical–Clinical–Risk Management-20; Webster, Douglas, Eaves, & Hart, 1997); J–SOAP–II (Juvenile–Sex Offender Assessment Protocol–II; Prentky & Righthand, 2003); LSI-R (Level of Service Inventory–Revised; Andrews & Bonta, 1995); PCRA (Post Conviction Risk Assessment; Administrative Office of U.S. Courts, 2011); SAPROF (Structured Assessment of Protective Factors for Violence Risk; De Vogel, De Ruiter, Bouman, & De Vries Robbé, 2012); SAVRY (Structured Assessment of Violence Risk in Youth; Borum, Bartel, & Forth, 2006); STABLE-2007 (Hanson, Harris, Scott, & Helmus, 2007); VRS (Violence Risk Scale; Wong & Gordon, 1999-2003); VRS-SO (Violence Risk Scale – Sex Offender Version; Wong, Olver, Nicholaichuk, & Gordon, 2003); VRS-YV (Violence Risk Scale: Youth Version; Lewis, Wong, & Gordon, 2004). Some studies have also tested the relationship between change scores on specific risk factors and reoffending (e.g., Bowen, Gilchrist, & Beech, 2007; Brown et al., 2009; Greiner, Law, & Brown, 2014; Hudson, Wales, Baker, & Ward, 2002; Kroner & Yessine, 2013; Olver, Nicholaichuk, & Wong, 2013; Quinsey, Jones, Book, & Wakeling, Beech, & Freemantle, 2013), or on particular subscales of a tool. However, these were not included, as the above review focuses on change scores on the overall risk assessment tools (i.e., total change scores).a In some studies the change analyses occurred on a smaller subset of cases. We report n for this smaller subset.b This study combined samples from Beggs & Grace (2011) and Olver et al. (2007).Appendix 4Familiarity Hypothesis: Reassessments will be More Predictive than Initial Risk Total ScoresAuthorsSamplea ToolMethodologyResultsYouth StudiesViljoen et al. (2015)163 male youth in residential program for sex offending, USAJ-SOAP-II, SAVRYRAs rated tools from files for admission and discharge. Mean treatment length was 1.13 years. Mean follow-up was 8.07 yearsAUCs for pre- and post-treatment Dynamic Risk Total scores did not differ significantly (i.e., confidence intervals [CI] overlapped).Adult Studiesde Vries Robbé et al. (2015) 108 male adult forensic patients, NetherlandsHCR-20, SAPROFRAs rated tools from files for pre- and post-treatment. Mean length of treatment 5.56 years. Followed for a mean of 11.28 years. On HCR-20, post-treatment scores added incremental validity to pre-treatment scores for 1-year recidivism. On SAPROF, post-treatment scores added incrementally to pre-treatment scores for long-term and 1-year recidivism. Goodman-Delahunty & O’Brien (2014)197 male adult sex offenders in diversion program VRS-SORAs rated tools from files for pre- and post-treatment (n = 93). Mean treatment length was 23.4 months. Followed for a mean of 9.1 years.Pre- and post-treatment VRO:SO risk categories showed same pattern for predicting sexual and general recidivism, but only pre- and not post-treatment categories predicted sexual recidivism.Hanson et al. (2007); Hanson (2015)612 male adult sex offenders, Canada and USASTABLE-2000/2007ACUTE-2000/2007Probation and parole officers rated tools, and again at 10 months and then every 6 months. Offenders followed for a mean of 3.4 years.Second STABLE assessment was no more predictive than the first except for sexual recidivism. Recent and past acute ratings were comparable except recent 6-month average was slightly more predictive than past 6-month average. Lewis et al. (2013)150 male adult offenders in a violence treatment program, CanadaVRSResearchers coded VRS from files for pre- and post-treatment (with some clinical cases). Followed sample for a mean of 4.9 years. AUCs for pre- and post-treatment Total and Dynamic Total score did not differ significantly (i.e., CI’s overlapped). However, post-treatment AUCs were slightly higher.Olver, Christofferson, Grace, & Wong (2014)539 adult male sex offenders, Canada and New ZealandaVRS-SORAs rated VRS-SO from files for pre- and post-treatment. Follow-up was mean of 15.5 years post-release.AUCs for pre- and post-treatment Total and Dynamic Total score did not differ significantly (i.e., CI’s overlapped, AUCs were within .01 to .04 points).Olver, Nicholaichuk, Kingston, & Wong (2014)676 adult male sex offenders in treatment programs, CanadaVRS-SOProfessionals rated VRS-SO pre- and post-treatment. Program was 2 – 9 months. Followed offenders for a mean of 6.31 years. AUCs for pre- and post-treatment total did not differ significantly (i.e., CI’s overlapped).Olver et al. (2007) 321 adult male sex offenders in treatment program, CanadaVRS-SORAs rated VRS-SO from files at pre- and post-treatment. Program length was 6 – 8 months. Follow-up was a mean of 10 years.AUCs for pre- and post-treatment total did not differ significantly (i.e., CI’s overlapped).O’Keefe, Klebe, & Hromas (1998)172 parolees and 85 corrections residents, USALSIOffenders were assessed at release (Time 1) and after 6 months of parole (Time 2).Time 1 LSI total scores were higher than Time 2 scores for the parolees (r = .35 vs. .22) but comparable for the residents (r = .08 vs. .11). It is unknown if these differences were significant.Wilson et al. (2013)30 male patients in a forensic hospital, CanadaHCR-20, STARTRAs rated START from files at 3-month intervals over a 1-year period. Each assessment had a follow-up of 3 months.AUCs for assessment 1 vs. assessments 2-4 did not differ significantly (i.e., CI’s intervals overlapped). Note: ACUTE-2007 (Hanson et al., 2007); HCR-20 (Historical–Clinical–Risk Management-20; Webster, Douglas, Eaves, & Hart, 1997); J–SOAP–II (Juvenile–Sex Offender Assessment Protocol–II; Prentky & Righthand, 2003); LSI (Level of Service Inventory; Andrews, 1982); SAPROF (Structured Assessment of Protective Factors for Violence Risk; De Vogel, De Ruiter, Bouman, & De Vries Robbé, 2012); SAVRY (Structured Assessment of Violence Risk in Youth; Borum, Bartel, & Forth, 2006) STABLE-2007 (Hanson et al., 2007); VRS (Violence Risk Scale; Wong & Gordon, 1999-2003); VRS-SO (Violence Risk Scale – Sex Offender Version; Wong, Olver, Nicholaichuk, & Gordon, 2003); VRS-YV (Violence Risk Scale: Youth Version; Lewis, Wong, & Gordon, 2004). a This study combined samples from Beggs & Grace (2011) and Olver et al. (2007). ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download