Entertainment & Humor

Predicting the Likelihood of Future Sexual Recidivism: Pilot Study Findings From a California Sex Offender Risk Project and Cross-Validation of the Static99

Description
Predicting the Likelihood of Future Sexual Recidivism: Pilot Study Findings From a California Sex Offender Risk Project and Cross-Validation of the Static99
Published
of 15
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Transcript
  Predicting the Likelihood of FutureSexual Recidivism: Pilot Study FindingsFrom a California Sex Offender Risk Project and Cross-Validationof the Static-99 Shoba Sreenivasan, PhD, Thomas Garrick, MD, Randall Norris, PhD,Sarah Cusworth-Walker, PhD, Linda E. Weinberger, PhD, Garrett Essres, PhD,Susan Turner, PhD, and Terry Fain, MA, MS Pilot findings on 137 California sex offenders followed up over 10 years after release from custody (excluding casesin which legal jurisdiction expired) are presented. The sexual recidivism rate, very likely inflated by sampleselection, was 31 percent at five years and 40 percent at 10 years. Cumulatively, markers of sexual deviance(multiple victim types) and criminality (prior parole violations and prison terms) led to improved prediction of sexual recidivism (receiver operating characteristic [ROC]    .71,  r     . 46) than singly (multiple victim types:ROC  .60,  r   . 31; prior parole violations and prison terms: ROC  .66,  r   . 37). Long-term Static-99 statisticalpredictive accuracy for sexual recidivism was lower in our sample (ROC  .62,  r   . 24) than the values presentedin the developmental norms. Sexual recidivism rates were higher in our study for Static-99 scores of 2 and 3 thanin the developmental sample, and lower for scores of 4 and 6. Given failures to replicate developmental norms,the Static-99 method of ranking sexual recidivism risk warrants caution when applied to individual offenders.  J Am Acad Psychiatry Law 35:454–68, 2007 Since the advent of the sexually violent predator/sexually dangerous person (SVP/SDP) statutes, 1 sev-eral sex offender risk actuarial instruments have beendeveloped. For the purposes of SVP/SDP commit-ment, most states require at least a qualitative assess-ment of risk level. In some states, such as Washing-ton, a quantification of sexual recidivism risk above50 percent is required. 1 The four most commonly usedriskscalesaretheSexOffenderRiskAssessmentGuide (SORAG), the Minnesota Sex OffenderScreening Tool-Revised (MnSOST-R), the RapidRisk Assessment of Sexual Offense Recidivism(RRASOR), and the Static-99. 2–7  All four sex offender risk assessment methodsdemonstrate statistically moderate correlations withsexual recidivism. 8,9  A moderate statistical accuracy  warrants the caution of over- or underestimation of risk when group-based actuarial rates are applied toan individual. Moreover, the tools are developedgenerally on composite samples from different sitesand cohorts because of the difficulty in accessing thecomplete prison files of a single, large cohort group.Consequently, many researchers have amalgamatedexisting data sets from different sites to obtain ade-quatesamplesizes. 9,10 Thispractice,however,occursat the cost of introducing variability. There are few  Dr. Sreenivasan and Dr. Weinberger are Clinical Professors of Psychi-atryandtheBehavioralSciences,KeckSchoolofMedicine,University of Southern California, Los Angeles, CA. Dr. Garrick is Professor of Psychiatry, Geffen School of Medicine, University of California, Los Angeles, CA. Dr. Norris is Chief Psychologist, California Institutionfor Men, California Department of Corrections and Rehabilitation(CDCR), Chino, CA. Dr. Cusworth-Walker is a Research Associate,UniversityofWashingtonSchoolofMedicine,Seattle,WA.Dr.Essresis Senior Psychologist, California Men’s Colony, CDCR, San LuisObispo, CA. Dr. Turner is Professor of Criminology, Law, and Soci-ety,UniversityofCalifornia,Irvine,CA.Mr.FainisProjectAssociate,Public Safety and Justice, RAND Corporation, Santa Monica, CA. Address correspondence to: Shoba Sreenivasan, PhD, GLA-VA Med-ical Center, 11301 Wilshire Blvd., Mental Health Administration,Mail code 691-B116AC, Los Angeles, CA 90073. E-mail: shoba.sreenivasan@med.va.gov  454  The Journal of the American Academy of Psychiatry and the LawR E G U L A R A R T I C L E  empirical investigations as to the applicability of ac-tuarial tools to ethnically diverse groups that differfrom the normative samples on which the tools werebased. In addition, there remains a lack of cross-validation in many of the U.S. jurisdictions, such asCalifornia, 11  where actuarial tools are used routinely in civil commitment risk assessments. While proponents of actuarial risk assessment ar-guethatrisktoolsaresuperiortoclinicaljudgmentinpredictive ability, 10 actuarial instruments have thepotential for misuse in applied risk assessments if theobtained risk percentages for a specific score are rep-resented as predictive of a specific individual’s com-mitting a future sexual offense. While beyond thescope of this report, it should be noted that use of actuarial risk assessments in SVP/SDP evaluationscontinuestofosterdebateamongforensicresearchersand clinicians. 5,12,13 The purpose of this article is to present pilot find-ings from a California sex-offender risk assessmentproject. Two primary areas are targeted: the statisti-cal identification of predictive risk markers as poten-tial variables for a risk assessment tool, and whetherthe most commonly used actuarial tool, the Static-99,offersasufficientlyreliableandaccuratemodelof risk for sexual reoffense in a racially and culturally diverse prison sample. Overview of Static-99 and Its Limitations The Static-99 was developed by Hanson andThornton 9 as an amalgamation of the RRASOR and the Structured Anchored Clinical Judgment-Minimum (SACJ-Min). 14 The risk factors for theRRASOR were derived from a factor analysis of seven follow-up studies and one replication sam-ple. 3,4 Recidivism for the RRASOR was definedprimarily as reconviction for a sexual offense. TheRRASOR items include prior sexual offenses (ex-cluding the last sexual offense, called the index offense), age at release, victim’s gender, and rela-tionship to the victim. The Static-99 includes allthe RRASOR items and adds the SACJ-Min itemsof sexual offense against a stranger, noncontactsexual offense, cohabitation status, nonsexual as-sault, and number of sentencing events greaterthan four.In a comparison of the two scales, the Static-99had a higher statistical association with sexual recid-ivism than did the RRASOR. 9 This prompted Han-son and Thornton 9 to recommend the use of theStatic-99 over the RRASOR. Of note, Hanson andMorton-Bourgon’s 8 meta-analysis found only a smallassociationwithsexoffenderrecidivismforsev-eral of the RRASOR and Static-99 variables (non-contact sexual offense, prior criminal history/history of nonsexual crimes). As with the other actuarial in-struments, the overall scores for both the Static-99and RRASOR demonstrated a statistically moderatepredictive accuracy in detecting a tendency towardsexual recidivism. 8,9 Both scales weight prior sexualoffenses heavily over other factors because of theirrobust association with sexual recidivism. 4,8 The Static-99 was not developed on a single co-hort of released sex offenders, but consisted of amal-gamating data collected previously from differentsites: two Canadian secure psychiatric facilities, oneCanadian prison, and one United Kingdom prison. While the sample size of 1,228 allowed for sufficientstatistical power, the data did not represent one co-hort group (i.e., the offenders were not all releasedduring the same period from the same facility orsame type of facility). In addition, some of the pre-dictor variables for the Static-99 risk scale were miss-ing in the developmental sample. For example, theInstitut Philippe Pinel sample did not have informa-tion about the two predictor variables, stranger vic-tims and noncontact offenses; the Millbrook sample was missing information on conviction for noncon-tact sexual offenses; and the Oakridge sample haddata for relationship to victim only for the most se-rious offense, counted any male child victims as op-posed to male victims regardless of age, and recordedonly the most serious last sexual offense. When data  were missing, statistical procedures were used to esti-matevalues,aprocedurethatislessthanideal.Thus,theStatic-99reliedonlessthanoptimalsamplingtoobtainsufficient sample sizes. This merely underscores thestate of the art in sexual recidivism risk assessment andthe difficulty in obtaining large samples with suffi-ciently broad data from a single cohort.Currently,theStatic-99hasthemostcross-valida-tionstudiesofanyoftheactuarialtools,markingthisratingmethodasthemostresearchedofthereviewedactuarial measures. 15–19 These studies and reportsdemonstrated that the Static-99 has moderate statis-tical association with sexual recidivism risk in Cana-dian, U.K., and select U.S. samples (Vermont,Texas). 15,17,20–25 TheadvantagesoftheStatic-99arethat it offers a quick method of rating risk, and thenormative data from the srcinal sample have good Sreenivasan, Garrick, Norris,  et al. 455 Volume 35, Number 4, 2007  inter-rater reliability. 26 However, risk percentagesappear to vary dependent on the base rate of sexualrecidivism in the sample studied. Doren 27 examinedthe correspondence of the developmental risk per-centages for Static-99 scores in seven studies for a five-yearfollow-upperiod.Hefoundthattheunder-lying sexual recidivism base rate of the sample af-fected the risk percentage associated with a specificStatic-99 score, which led to differences from thatderivedfromthedevelopmentalsample.Forlow-risk scores (in the 1 to 2 range), when the sample had a high underlying base rate of sexual recidivism, there werehigherriskpercentagesassociatedwiththescorethan in the developmental sample. High-risk scoresrevealed lower than expected risk percentages whenthe underlying base rate was low, but remained sim-ilar to the developmental sample when the base rates were high. The Doren data failed to replicate theHanson and Thornton 9 developmental norms. Defining Sexual Recidivism The definition of sexual recidivism varies acrossactuarial schemes. Some studies have used criminalconvictions, reflecting a conservative strategy basedonadjudicatedoffensesforwhichaguiltyverdictwasfound and the individual sanctioned. 10 However, asiswidelyacknowledged,suchadefinitioncanunder-estimate the true rate of recidivism, as it is based onboth apprehensions and punished offenses. 9,28,29 In an attempt to determine the full extent of sex offending, the U.S. Department of Justice (DOJ)compared statistics from U.S. law enforcement re-ports of sexual offense arrests to data from the Na-tional Crime Victimization Survey report. 28 Itshould be stated that these data reflect estimates of the baseline frequency of acts. The U.S. DOJ docu-ment noted that in 1995, individuals aged 12 yearsandolderreportedtotheNationalCrimeVictimiza-tion Survey that they experienced 260,300 incidentsof attempted or completed rape. By contrast, thenumberofsuchcrimesactuallyreportedtothepolicein 1995 was 97,460. Thus, only 37 percent of thesexual crimes reported to the National Crime Vic-timization Survey came to the attention of law en-forcement,leavingahighnumber(63%)ofundetec-ted offenses. Moreover, even among those crimesreported to the police, only one-half resulted in theidentificationandarrestofaperpetrator(i.e.,48,730of the reported 97,460 sexual assaults). The Bureauof Justice Statistics findings for the years 1992through 2000 mirrored these findings (i.e., 63 per-cent of completed rapes, 65 percent of attemptedrapes, and 74 percent of completed and attemptedsexual assaults against females were not reported tothe police). 30 Therefore, a comprehensive outcome definitionfor sexual recidivism should include arrests, convic-tions, and parole/probation or in-custody sexual vi-olations to address the underestimation of risk cre-atedbymorelimiteddefinitionsofsexualrecidivism,such as convicted offenses. Nonetheless, even thisbroad method represents observed rates. The truerate of sexual recidivism would include the unob-served and/or unreported sexual assaults. 31 Sample Generalizability The Static-99 is based on Canadian and U.K. de-velopmental and cross-validation samples. 9 Thesample was described as predominantly Caucasian.Consequently, it may have limited applicability toracially diverse U.S. prison samples. Sample limita-tions could reduce the efficacy and even the applica-bility of an actuarial tool, a concept articulated inevidence-based medicine. 32 Laws, enforcementmethods, judicial procedures, sanctions, and com-munity monitoring differ across countries as well asacrossU.S.jurisdictions.Suchdifferencescontributeto the variation in base rates of detected sexual recid-ivism. In relation to these sampling problems, thesexual recidivism percentages given in the develop-mental study for the Static-99 have not been corrob-orated in cross-validation studies. 27 Identifying Sex Offender Risk Predictor Variables In our study, the predictive markers for sexual re-cidivismwerederivedfromtheexistingresearchbase, with the most promising variables selected for inclu-sion in the analyses. The Hanson and Bussiere 4 meta-analysis examining 23,393 sex offenders by us-ing 61 data sets and 165 predictor variables repre-sentsalandmarkcontributiontotheidentificationof risk predictors. The authors identified several factorsthatcorrelatedwithsexualreoffending.Thesefactorsincluded past sexual offenses, male victims, strangervictimsviewedasproxyvariablesforsexualdeviance,and general criminality factors such as past non-sexual violent offenses, antisocial personality, andpsychopathy. California Sex Offender Risk Project 456   The Journal of the American Academy of Psychiatry and the Law   A second, updated meta-analysis by Hanson andMorton-Bourgon, 8 including 95 studies and 31,000sex offenders, confirmed the results of the 1998meta-analysis,butalsoaddednewpredictors,suchasthe ability to comply with conditions of supervision.Recently,Roberts et al  . 33 conductedaprincipalcom-ponents analysis on 10 actuarial items from the Stat-ic-99 and the Risk Matrix 2000 (a risk assessmentinstrument used in the United Kingdom). They identified three factors associated with sexual recidi-vism: general criminality, sexual deviance, and de-tachment.Thegeneralcriminalityfactorconsistedof ahistoryofpriorviolentandnonviolentoffenses;thesexual deviancy factor consisted of prior sexual of-fenses, noncontact sexual offenses, and male victims;and the detachment factor consisted of attacking a stranger and never being married. Barbaree  et al  . 34 conducted a similar statistical analysis on a sample of 311 sex offenders from a medium security Canadianfederalpenitentiary.Theyexamined38uniqueitemstaken from five actuarial instruments (violence risk appraisal guide [VRAG], SORAG, RRASOR, Stat-ic-99, and MnSOST-R). Six principal factors werefound to be associated with sexual recidivism, en-compassed by antisocial behavior and sexual devi-ance, including detached predatory behavior such asselecting strangers as victims or offending in a publicplace).Barbaree et al. 34 foundthatviolentrecidivismis predicted by antisocial factors, while sexual recid-ivism is predicted by the factors associated with sex-ual deviance (such as persistence of sexual offending,child sexual abuse). These studies suggest that mark-ers of sexual deviance and criminality may be predic-tive singly or cumulatively of risk of sexual recidi-vism. Aggression during the sexual offense may be a predictor, 7 although the recent Hanson and Mor-ton-Bourgon 8 meta-analysis did not support thismarker. Rationale for Presenting Pilot Data California-based cross-validation of the Static-99and the identification of groups of predictive risk variables have the advantage of including the demo-graphics specific to a racially and culturally diverseU.S. prison population. Given the intense effortneeded to collect even pilot data, these findings arepresented as preliminary in a long-term, ongoing study. Materials and Methods The project was reviewed and approved by theCalifornia Department of Corrections and Rehabil-itation (CDCR) Research Board (November 15,2001,andJanuary30,2004).Safetyprocedureswereinstituted to assure the confidentiality of any infor-mation gathered from the archival data review. Pilot Sample Files  A total of 5,898 sexual offenders, both active andinactive, were identified by the CDCR as having been released from prison between January 1, 1989,and December 31, 1990. Active files represented 29percent( n  1,709),andinactivefilesrepresented71percent ( n    4,189) of the total sample pool of 5,898. Of the active files, ( n    1,790), 137 (8%) were selected to serve as the pilot sample. Active files were those of offenders remaining un-der supervision (i.e., in custody or on parole) as re-flected by a CDCR list generated in June 2002. In-active files were cases in which the CDCR’s jurisdiction over the individual had expired some-time in the period between the person’s release in1989/1990 and the generation of the June 2002 list.These files were selected for the pilot study becausethey contained full criminal histories (i.e., state andfederal criminal records, police reports, parole re-ports, and prison rules violations) and demographicinformation. Inactive files would have had a circum-scribed data set, given the thinning of materials forstorage, and criminal histories would have been lim-ited to state criminal records at follow-up. This sam-ple selection of active files probably created a biastoward inflating the rate of sexual recidivism, as theindividuals in the sample had reoffended in somemanner after their release from prison custody in1989/1990 to June 2002, when the sample pool list was generated.Most active prison files were located in five custo-dial locations, three of which were selected for thepilot study. These three sites were chosen, as they representeddifferentlevelsofsecurity(medium-low,medium, and high). Fourteen (10%) files were se-lected from the medium- to low-security prison, 27(20%)filesfromthemedium-securityprison,and11(8%)filesfromthehigh-securityprison.Inaddition,one of two parole sites in the state (the one with thelargestnumberofparolees)wasselectedasthesourcefor inclusion in the pilot study ( n  85, 62%). All Sreenivasan, Garrick, Norris,  et al. 457  Volume 35, Number 4, 2007  active files (prison and parole) that were available atthe sites on the date of the file review were included. We acknowledge that this is not a random sample.This method of file selection was used because wehad no control over logistical matters, such as staff availabilityatcustodialsitestopullspecificidentifiedfiles;theinabilitytolocatecertainfilesatasite,giventhe fluidity of transfers between institutions and pa-role; and time and resource limitations in the avail-ability of trained file reviewers. A 1989/1990 release date assured that all in thepilot sample had had at least one period of commu-nityplacement.Allindividualsinthestudyhadcom-mitted a sexual offense sometime in their criminalhistories. In some instances, the sexual offense wasthe controlling offense for the 1989/1990 release. Inother cases, the sexual offense was a prior offense, withthe1989/1990controllingoffensebeinganon-sexual offense. Definition of Sexual Offense Sexual offenses were defined as arrests, convic-tions, parole violations, or prison rule violations in-curred for criminal sexual behavior. The sexual of-fenses included those involving force and violence orsubstantial sexual conduct, such as either the offenseitself or an attempted offense of rape with force; rape with threat of future retaliation; rape or penetrationofgenitaloranalopeningsbyforeignobjects;rapeinconcert by force or violence; spousal rape with threatof future retaliation; sodomy; oral copulation; allpenal code sections of lewd acts on a child under 14,16, or 18.Noncontact offenses such as exhibitionism, voy-eurism, or annoying/molesting a child were also in-cluded as sexual offenses. In addition, an offense wascoded as a sexual offense for charges of mayhem,battery, or murder when the file indicated a clearsexual component to the crime that was not filedseparately, or if filed, the individual was convictedonly of the nonsexual offense. Sexual offenses ex-cluded as either initial or recidivist sexual offenses were solicitation/prostitution, pimping, consensualsexualencountersincustodyresultinginprisonsanc-tions, and charges of indecent exposure in custody that would not meet the legal criteria for exhibition-ism in the community. Definition of Sexual Recidivism Sexual recidivism was defined as sexual reoffend-ing that occurred after the 1989/1990 release andduring an approximate 10  year follow-up period.The definition of sexual recidivism was any sexualbehavior following release that resulted in sanctions,suchasarrest,conviction,paroleviolation,probationviolation, or in-prison rules violations that wouldmeet the definition of a sexual crime (excluding pimping and prostitution). Procedures In addition to four of the authors, forensic psy-chologists and psychiatrists familiar with prisonrecords and the research protocol conducted the ar-chivalreview.Tenfileswerecodedindependentlyby two reviewers to determine inter-rater reliability of recording information. In all cases, the same infor-mation was recorded, suggesting that the items onthe protocol data sheet were easily coded.  Measurements Files were coded for demographics; specifics of each sexual offense, including victim type, relation-ship, and type and nature of violence; institutionalbehavior; parole behavior; medical and psychiatrictreatment; drug and alcohol abuse history; develop-mental variables, such as school functioning; gang membership; IQ scores; reading ability; and juvenileand adult criminal history. A weighted scale devisedby Quinsey   et al  . 10 and based on the Canadian crim-inal code was used to arrive at scores for all violentacts involved in the first sexual offense, to give anoverall violence score. A similar approach was usedforlessviolentornonviolentbehaviorthataccompa-nied the first sexual offense. Static-99 Ratings The Static-99 was scored at the time of the1989/90 release using only that information avail-able at the release date. The Static-99 was scored foreach case by one investigator trained in the coding rules 35  who had extensive experience using the Stat-ic-99 in sexually violent predator evaluations. More-over, the Static-99 was scored in a jurisdiction whereevaluator Static-99 rater reliabilities had been calcu-lated. 26 Hanson 26 found evaluator rater reliabilitiesto have 0.91 average item percent agreement, 0.80average item   , and 0.97 intraclass correlation fortotal scores. Other studies have also demonstratedhigh inter-rater reliability of the Static-99 (over0.90). 15,35 The tool has 10 elements that addressstatic or nonchanging factors: prior sexual offenses(excludingindexoffense);priorsentencingdates(ex- California Sex Offender Risk Project 458  The Journal of the American Academy of Psychiatry and the Law
Search
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x