Abstract
Forensic evaluators commonly conduct evaluations of sexually violent persons (SVPs) but often retrofit generic risk assessments to assess SVP legal criteria. As a 20-year follow-up to “The Risky Enterprise of Risk Assessment,” this article critically evaluates updated versions of common risk measures (Historical-Clinical-Risk Management Scale 20 (HCR-20V3), Sexually Violent Risk 20 (SVR-20 V2), Violence Risk Appraisal Guide-Revised (VRAG-R), Static-99R, and Risk for Sexual Violence Protocol (RSVP V2)) to determine their appropriateness for SVP evaluations. For more balanced appraisals, risk assessments have more recently considered protective factors (Inventory of Offender Risk, Needs, and Strengths (IORNS), Structured Assessment of Protective Factors-Sex Offenders (SAPROF-SO), and Short-Term Assessment of Risk and Treatability (START)), given their growth in forensic research. Importantly, some risk measures demonstrated substantial improvements. Empirical support for predicting violent sexual recidivism, the legal standard in many SVP jurisdictions, has yet to be reached. Most studies assess general sexual recidivism rather than sexual violence, per se. This disconnect presents a professional challenge for evaluators in determining whether previously convicted offenders pose such a risk of sexual violence, following their criminal sentence, as to warrant indefinite civil commitment. This review underscores the need for stronger alignments between SVP legal criteria and forensic assessment instruments to solidly support their empirical use in SVP determinations.
The Journal’s 2005 article, “The Risky Enterprise of Risk Assessment,” provided an early analysis of assessment methods for sexually violent predators.1 In the intervening two decades, many advances have been documented regarding risk assessment, including sexually violent recidivism.2,3 Because the term predators is stigmatizing and sometimes inaccurate, we prefer the updated description of sexually violent persons (SVPs) found in recent statutes (e.g., Adam Walsh Child Protection and Safety Act4). Despite focusing on an offender population, forensic psychiatrists and psychologists conduct SVP evaluations in the context of civil commitment proceedings.
SVP determinations have grown over the last several decades. In 2006, the total was 4,534.5 In 14 years, the estimate increased nearly 40 percent to approximately 6,300 persons civilly committed under SVP laws.6 Notably, Black men were overrepresented in the SVP population by more than double of White men (i.e., 7.72 versus 3.11 per 100,000).
Most SVP statutes provide indeterminate commitments.7 Typically, states annually review the risks of recidivism and dangerousness. The standard of proof for SVP release varies remarkably among preponderance (10%), clear and convincing (60%), and beyond a reasonable doubt (30%).7 The lengths of SVP commitments are largely unknown. A 2019 report averaged six years of SVP detainment,8 but this may increase, given the low rates (5%) of granted conditional releases.9
SVP determinations are highly consequential for those civilly committed and for community safety. The next section critically reviews components of SVP legal standards and SVP measures in the last two decades.
Legal Standard of Sexually Violent Persons
Institutionalizing violent sex offenders predates current SVP statutes. Early “sexual psychopath laws” were passed,10 such as the 1937 Michigan Statutes Annotated.11 This allowed transferring individuals deemed “sexual psychopaths” to rehabilitative treatment.
In the United States, the first modern SVP law was enacted in Washington State through the Community Protection Act of 1990.12 This law defined an SVP as “any person who has been convicted of or charged with a crime of sexual violence and who suffers from a mental abnormality or personality disorder which makes the person likely to engage in predatory acts of sexual violence if not confined in a secure facility” (Ref. 12, Sec. 19).
In the years following, several states passed their own SVP laws, including Arizona,13 California,14 and Wisconsin.15,16 Kansas17 enacted an SVP law that became the subject of the landmark Supreme Court case, Kansas v. Hendricks.18
Indefinite civil commitment of offenders after serving their sentences faced several constitutional challenges. In Kansas v. Hendricks, the Supreme Court of the United States upheld the constitutionality of Kansas’s SVP law, ruling it is a civil proceeding and avoided double jeopardy because it allowed for the release of individuals who were no longer deemed dangerous. They also upheld Kansas’s definition of “mental abnormality,” defined as “a congenital or acquired condition affecting the emotional or volitional capacity which predisposes the person to commit sexually violent offenses in a degree constituting such person a menace to the health and safety of others” (Ref. 19, Sec. (b)).
Statutory definitions of behavioral abnormality or mental disorder vary across states. DeMatteo and colleagues10 summarized these differences in definitions of mental abnormality. Some states (e.g., California, Illinois, and Wisconsin) use the term mental disorder, whereas Texas uses the term behavioral abnormality, although most definitions resemble Kansas’s mental abnormality. Texas additionally requires the assessment of psychopathy as part of a behavioral abnormality.20 Arizona defines mental disorder as “a paraphilia, personality disorder or conduct disorder or any combination of paraphilia, personality disorder and conduct disorder that predisposes a person to commit sexual acts to such a degree as to render the person a danger to the health and safety of others.”21 Pragmatically, such states provide clear parameters regarding specified disorders.
In Kansas v. Crane,22 the Court expanded on Hendricks, clarifying that SVP examinees must have “substantial difficulty” controlling their behavior. This clarification was crucial because volitional impairment is key to determining the risk of reoffending if released. Rather than requiring a complete loss of volitional control, an unattainable standard, substantial difficulty appears more feasible to conceptualize and evaluate. Nonetheless, states define and apply this standard differently.
At the federal level, the Adam Walsh Child Protection and Safety Act4 and the subsequent U.S. Supreme Court decision in United States v. Comstock23 permit civil commitment of individuals deemed sexually dangerous for reasons of public and individual safety. They upheld the standards of mental abnormality and volitional impairment, as defined by Hendricks and Crane.
Twenty states have passed SVP laws. For differences in SVP terminology across states, please see previously cited reviews.7,10 In the next two sections, we summarize the 2005 SVP analysis and examine subsequent advances.
Summary of the 2005 SVP Analysis
Rogers and Jackson1 evaluated risk-assessment tools and their appropriate use in SVP civil commitments. They applied five cumulative steps for SVP determinations based on two U.S. Supreme Court cases.18,22 These steps are reproduced here:1. Does the defendant have the requisite clinical condition (e.g., mental abnormality, mental disorder, or personality disorder), as mandated by the particular jurisdiction?
2. If yes, does the defendant have volitional impairment?
3. If yes, does the volitional impairment directly arise from the clinical condition?
4. If yes, does the defendant have the statutorily defined likelihood of sexual violence?
5. If yes, is that likelihood of sexual violence directly the result of volitional impairment? (Ref. 1, p 525)
In their review, common risk-assessment methods failed to address most steps, as seen in Table 1. The authors evaluated eight risk measures: Historical-Clinical-Risk Management Scale 20 (HCR-20V2),24 Sexually Violent Risk 20 (SVR-20),25 Static-99,26 Rapid Risk Assessment for Sex Offense Recidivism (RRASOR),27 Minnesota Sex Offender Screening Tool-Revised (MnSOST-R),28 Violence Risk Appraisal Guide (VRAG),29 Sex Offender Risk Appraisal Guide (SORAG),30 and Psychopathy Checklist-Revised (PCL-R).31 They also examined a physiological measure of arousal, the penile plethysmography (PPG).32 There were few efforts in assessing for the breadth of clinical conditions. Overall, most risk assessments failed to adequately address Step 1. Three steps were largely omitted: Step 2 (volitional impairment), Step 3 (nexus between volitional impairment and the clinical condition), and Step 5 (volitional impairment predicting sexual violence).
Fulfillment of SVP Standards by Generic Risk-Assessment Measures1
In 2005, other limitations included limited data on violent sexual recidivism. Research was scattered across types of sexual offenses (e.g., adult rapists and child molesters) and levels of aggression (e.g., noncontact versus violent sexual offenses).
Advances in SVP Measures
Psychometric advances are noteworthy over the last two decades. Improvements were documented with substantial revisions for measures for structured professional judgment: HCR-20V3,33 SVR-20 V2,34 Risk for Sexual Violence Protocol (RSVP V2),35 and Structured Assessment of Protective Factors-Sex Offenders (SAPROF-SO).36 For actuarial measures, the VRAG-R37 was updated. Finally, the Inventory of Offender Risk, Needs, and Strengths (IORNS)38 was published as a general risk assessment measure.
Four tools evaluated in the 2005 review were omitted in this article: RRASOR, SORAG, PPG, and MnSOST-R. First, the RRASOR27 was omitted because its author stated, “We no longer support the use of the RRASOR” (Dr. Karl Hanson, personal communication, January 2025). Second, the SORAG30 was not included because of the creation of the VRAG-R,37 which combined elements of the VRAG29 and SORAG. Third, the PPG32 was excluded because of its lack of standardization and inconsistent legal admissibility in the United States.39,40 Fourth, although public scoring materials for the MnSOST-R are available (e.g., Washington Association of Sheriffs and Police Chiefs), we were unable to verify their completeness or accuracy. The MnSOST-R was therefore omitted to avoid potentially outdated or inaccurate instructions.
Continuing Challenges for SVP Measures
Like most legal standards, SVP legislation and subsequent case law lack clear definitions to demarcate SVP from non-SVP determinations. According to Pfaffenroth, “Despite the constitutional importance of only committing those with ‘serious difficulty’ controlling themselves, the Supreme Court neither clarified where to draw that critical line nor specified whether existing state procedures comport with the lack of volitional control requirement” (Ref. 41, p 2245). Rogers commented, “The unintended irony is that volitional impairment lacks the empirical basis for exculpation in federal insanity cases but clearly is justified for the indefinite detention of sex offenders” (Ref. 1, p 159).
Forensic evaluators must apply general risk assessment literature to the specific requirements of SVP determinations. Outlined in Hendricks and Crane, constructs like “mental abnormality” and “volitional impairment” remain ill-defined, increasing their subjectivity in forensic practice. Although volitional impairment remains a cornerstone in criminal responsibility assessments,42,43 the retrospective nature of insanity is not aligned with prospective judgments of ongoing SVP threats. Therefore, a pivotal question remains regarding whether these measures are inappropriately retrofitted for legal standards they were not designed to address.
Step 4 is the most relevant in the current literature. For each risk assessment measure, predictive validity was summarized for sexual recidivism. Sexually violent recidivism is the focus and differs fundamentally from nonsexual violent recidivism and general recidivism. Some researchers measured sexual recidivism as any future sexual offense (contact and noncontact). Blurring classifications is inappropriate, as sexually violent recidivism is required for SVP civil commitments.
There is an important methodological question about what standards should be applied when measuring effect sizes for risk assessments. Rice and Harris44 proposed standards for receiver operating characteristic (ROC) analyses. The area under the curve (AUC) determines a risk measure’s predictive validity. At chance levels, 50 percent or .50 represents no improvement (i.e., completely hit or miss) in prediction. Using Cohen’s45 standard, Rice and Harris recommended the following standards: small (AUC < .56), medium (AUC = .64–.71), and large (AUC > .72) effect sizes. The present review avoids tools with small AUCs because of their minimal predictive validity. As acknowledged by Cohen46 and quoted in Rice and Harris,43 small was characterized “by a barely perceptible (but real) difference” (Ref. 46, p 618). Given the consequences of SVP determinations, a reasonable conclusion is that barely perceptible should be clearly insufficient.
Scholars have criticized AUC for its imprecision. Although it reflects a tool’s ability to distinguish between recidivists and nonrecidivists at the group level, it does not provide an individual’s likelihood of reoffending, required in SVP evaluations. Hart and colleagues47 noted that individual-level predictions from some risk assessments produce wide confidence intervals, making risk categories difficult to distinguish. This limitation is especially relevant in SVP evaluations, where evaluators make individual judgments.
Scholars have identified methodological shortcomings in risk assessment tool development and use, potentially limiting their utility in SVP evaluations.48,–,50 A recent systematic review found that 82 percent of studies only report AUC.50 AUC does not reflect whether predicted outcomes match real-world outcomes (i.e., calibration), and poor calibration increases misclassifications and false positives.50 Only 18 percent of studies reported sensitivity (i.e., identifying true positives) and specificity (i.e., identifying true negatives), limiting understanding of a tool’s accuracy. Fazel and colleagues49 noted several limitations. First, current risk assessments often rely on validation samples and risk factors that are outdated and not generalizable for broader forensic samples. Second, SVP determinations lack definitions of medium- to high-risk individuals. These critiques provide important context when evaluating the appropriateness and performance of risk measures.
Analyses of Risk Measures
Given advancements in risk assessment measures over the past 20 years, we evaluated each tool’s ability to address four of the five sequential steps: clinical condition, volitional impairment, volitional impairment because of a clinical condition, and likelihood of sexual violence because of volitional impairment. Step 4, the likelihood of sexual violence, is examined separately in the section “Predictive Validity.” Because these steps are conceptually interdependent, measures that failed to adequately address an earlier step were not assessed for the subsequent step. Major findings for each risk assessment measure are summarized in Table 1.
HCR-20V3
The HCR-20V3 33 is sometimes used in SVP evaluations, as literature indicates the HCR-20V2 51 appears in approximately 25 percent of evaluations.52 The HCR-20V3 33 answers mental disorders as part of Step 1. It assesses major mental disorders, personality disorders, including antisocial personality disorder (APD), and deficits in interpersonal relationships. The authors recommend an additional risk measure focused on sexual violence to gain more information.
Volitional impairment (Step 2) is addressed by recent problems linked to lack of insight (C1), violent ideation or intent (C2), and instability (C4). Regarding C1, the manual states, “Denies or greatly minimizes the interpersonal consequences of mental, personality, or substance-related disorder” (Ref. 33, p 85). It also notes social consequences and failure to understand external factors that aggravate the disorder. Other facets of HCR-20V3 are relevant to volitional impairment. C2 includes “thoughts, plans, desires, fantasies or urges to cause harm to others” (Ref. 33, p 87). This reflects volitional impairment and self-regulation, cognitively (thoughts and plans), affectively (desires), and behaviorally (urges). For major mental disorders (C3), evaluators consider “deficits in … impulse control” (Ref. 33, p 89).
When an examinee lacks awareness into the examinee’s mental disorder, the examinee may struggle to identify triggers, assess risks, and implement strategies to control sexual violence. This inability to self-regulate stems from the examinee’s mental disorder, as evaluated by C1 and C3. Thus, the HCR-20V3 33 begins to address Step 3 by linking volitional impairment because of one’s mental condition or abnormality.
PCL-R
The PCL-R31 is used in approximately 32.6 percent to 41.1 percent52,53 of SVP evaluations. It does not sufficiently address any of the five steps. The PCL-R has two factors: interpersonal and affective and social deviance. These are divided into four facets: interpersonal, affective, lifestyle, and antisocial. The PCL-R may provide information about mental abnormality by assessing for a lack of remorse or guilt and a lack of empathy (Facet 2). It also measures poor behavioral control and impulsivity (Facet 4), although these items do not assess clinical conditions (Step 1) or volitional control (Step 2).
VRAG-R
The VRAG-R37 is an updated version of the VRAG,29 used in about 10 percent of SVP evaluations.53 The VRAG-R incorporated the PCL-R into its actuarial determination of risk. The VRAG-R partially addresses only Step 1 (clinical condition) and Step 2 (volitional control). For Step 1, the VRAG-R is narrowly focused on a history of conduct disorder and antisociality, measured with PCL-R Facet 4 (i.e., juvenile delinquency and criminal versatility). These components are broad and encompass elements of mental abnormality and impulse control. Consequently, evaluators may find it challenging to interpret the VRAG-R, as it reflects general risk rather than violent sexual recidivism. It assesses history of sexual offending and focuses on hands-on offenses; however, it de-emphasizes sexual violence against children under the age of 14. Therefore, it only partially addresses Step 1 and Step 2.
SVR-20 V2
The SVR-2054 is used in about seven percent of SVP evaluations.52 The updated SVR-20 V234 addresses Step 1 (clinical condition). It evaluates mental abnormality across three domains: psychopathic personality disorder, sexual deviations, and major mental disorders. The SVR-20 V2 also includes sexual health problems and chronic sexual offending, which may contribute to the presence of a mental abnormality.
The SVR-20 V2 does not directly assess volitional impairment (Step 2), although impulsive behavior is considered in evaluating for psychopathic personality disorder. This narrow application does not determine whether individuals have general difficulty with controlling their behavior.
Static-99R
The Static-99R55 is a brief screen for risk assessment using demographic and criminal information that rarely change. One could argue that the Static-99R is not a psychological measure, as it lacks clinical variables. Despite this, it is widely used in SVP evaluations, with prevalence rates ranging from 59.2 percent52 to 70.6 percent.56 It does not answer any step.
The Static-99R categorizes examinees based upon empirical research on sexual recidivism.55 Although it discusses use with persons civilly committed under SVP statutes, it does not address its utility for SVP determinations.
IORNS
Although the IORNS38 is not commonly used in SVP evaluations, we include it because it incorporates protective factors, which are often overlooked.
The IORNS has limited validity for Steps 1 (clinical condition) and 2 (volitional impairment). For Step 1, it does not address the breadth of mental abnormalities. It focuses on the Psychopathy (PPY) subscale, assessing for prototypical psychopathy (e.g., manipulativeness and poor impulse control). Difficulty controlling anger and impulses are measured by the Aggression (AGG) and Impulsivity (IMP) subscales, respectively. Such loss of control is not necessarily connected to a mental disorder or abnormality. It does address Step 2.
SAPROF-SO
The Structured Assessment of Protective Factors (SAPROF57) developed the Sexual Offending version (SAPROF-SO36). It is designed for biological males previously convicted of a sexual offense. It measures protective factors against sexual reoffending and should be used with another sexual risk tool.36 It does not evaluate the sequential steps needed for an SVP evaluation. Although the SAPROF-SO evaluates impulsivity and sexual self-regulation, which may indicate volitional impairment, it does not establish their nexus to a mental abnormality.
We included this measure because protective factors are undervalued in sexual recidivism risk assessment. Given the indeterminate nature of civil commitment, protective factors should be strongly evaluated. SVP evaluations are typically completed at the end of the offender’s criminal sentence. Years or decades later, careful consideration of protective factors may help ensure that SVP determinations are both fair and evidence-based.
RSVP V2
The RSVP V235 is a structured professional judgment measure used to assess risk factors related to sexual violence. No published evidence of its utility in SVP evaluations exists, but its structure may inform these cases.
The RSVP V2 addresses Step 1 (clinical condition) by assessing psychopathic personality disorders, major mental disorders, and sexual deviance. Across these three areas, it assesses psychopathy, paraphilias, antisocial personality disorder, and interpersonal problems (e.g., problems with empathy, impulsivity, callousness). Sexual deviance “reflects serious problems stemming from a relatively stable abnormality in the focus of the person’s sexual appetite that motivates or incites sexual violence” (Ref. 35, p 91). These aspects can be assessed separately or together to determine whether sexual deviance reflects a mental abnormality linked to future sexual violence.
For Step 2 (volitional impairment), it examines sexually violent ideation, self-awareness, and deficits with stress or coping. The former includes thoughts, urges, plans, or impulses to commit sexual violence. Although not explicitly stated in the manual, persistent obsessive thoughts about committing sexual violence may impair inhibitory control. Inhibitory control is measured through cognitive understanding, self-awareness, appraisal of behavior, and behavioral disinhibition related to sexual violence. Cognitive denial alone may not demonstrate volitional impairment, but combined with behavioral disinhibition, it strengthens the argument for difficulty in controlling sexual violence.
The RSVP V2 offers advances in addressing Step 3 (volitional impairment arising from clinical condition). When evaluating psychopathic personality disorders, the manual states, “This risk factor reflects serious problems stemming from symptoms of psychopathic personality disorder, including …an impulsive and irresponsible behavioral style” (Ref. 35, p 93). Therefore, volitional impairment is a key part of evaluating psychopathic personality disorder. It does not, however, include aspects of volitional impairment when assessing mental disorders or sexual deviance. Part of evaluating for sexual deviance is how it could motivate future sexual behavior, although it does not include difficulty with controlling sexual behavior.
START
The Short-Term Assessment of Risk and Treatability (START)58 is a structured professional judgment measure used to assess risk of violence and other variables (e.g., self-harm, suicide, substance use). Quoting the manual, “individuals being at risk for particular kinds of violence should be further evaluated using a measure appropriate to the identified risk” (Ref. 58, p 4). We included this assessment because identifying strengths is important in SVP evaluations, despite no published documentation of its use in such evaluations. Strength-based factors allow more nuanced assessments and limit over-pathologizing individuals.
The START informs Step 1 (clinical condition) by evaluating the examinee’s mental state. This domain includes symptoms of mental disorders (e.g., delusions, hallucinations, and concentration difficulties). It offers valuable clinical data for SVP determinations. Under Kansas legal standards, the mental abnormality must be linked to volitional impairment.
The START assesses parts of Step 2, volitional impairment, including impulse control, rule adherence, and conduct. These domains reflect the ability to control behaviors following social norms. Rule adherence reflects whether a person understands, appreciates, and accepts the need to follow reasonable expectations (Ref. 58, p 64), particularly within a risk reduction plan. This reflects volitional impairment through poor cognitive and behavioral self-regulation. A high vulnerability score indicates individuals do not engage with rules (Ref. 58, p 65), suggesting difficulty regulating behavior despite clear conditions. Although helpful, this is insufficient for evaluating volitional impairment.
Predictive Validity
For Step 4, we reviewed the literature on each measure’s predictive validity. Limited data exist for the RSVP V2, SVR-20 V2, and the SAPROF-SO. These newer versions should be prioritized in future research, as they are increasingly used in forensic practice. These findings are summarized in Table 2.
Summary of Predictive Validity (AUC) by Risk Assessment and Recidivism Type
HCR-20V3
Neil et al.59 examined the predictive validity of the HCR-20 V3 in Scottish and Northern Irish inpatients with “dangerous, violent or criminal propensities” (Ref. 59, p 86). The authors measured institutional violence, including physical or sexual violence. Sexual violence was defined as any behavior or verbal comment with “sexual content” (Ref. 59, p 88). The HCR-20 V3 total score did not significantly predict sexual violence; however, the Clinical scale significantly predicted it with a large effect size (AUC = .83).
In Australian forensic inpatients, the HCR-20V3 total score, summary risk rating, and relevance ratings significantly predicted violent recidivism (AUC = .70–.77).60 Violent recidivism included new sexual offenses. O’Shea and colleagues61 found the HCR-20V3 total scores and Clinical scales significantly predicted violent, inappropriate sexual behavior (AUC = .64–.68). This is encouraging, although evidence specific to violent recidivism is limited. Some evidence suggests that the HCR-20V3 can predict aspects of sexual recidivism, including sexual violence.59
PCL-R
The PCL-R shows mixed results for predicting sexual recidivism.62 A meta-analysis of 20 studies summarized its predictive validity for sexual recidivism. Effect sizes (Cohen’s d) were transformed to AUC for easier comparison using Rice and Harris criteria.43 The PCL-R total (AUC = .61), Factor 2 (AUC = .62), and Facet 4 scores (AUC = .61) significantly predicted sexual recidivism with medium effect sizes. They analyzed potential moderators, including offender type (rapist versus child molester), but did not find significant effects for PCL-R total or Factor scores. The authors suggested forensic evaluators should focus on Factor 2 and Facet 4 scores when measuring predictive validity for sexual recidivism.
Another study found the PCL-R significantly predicted sexual recidivism (AUC = .73), but not violent sexual recidivism (AUC = .70) in extrafamilial molesters from a German sample.63
A recent systematic review evaluated 14 articles on the PCL-R’s utility for predicting sexual recidivism based on psychopathic traits.64 Their review can be summarized by the following quote: “Although there is no unanimous agreement on the results analyzed, an elevated PCL-R score …would not …appear to be significantly associated with the risk of sexual recidivism, being a poor predictor of the same” (Ref. 64, p 48). Overall, the PCL-R shows small predictive power for sexual recidivism, insufficient for Step 4.
VRAG-R
A cross-validation study65 examining the predictive validity of the VRAG, SORAG, and VRAG-R found that the VRAG-R showed significant predictive validity for sexual recidivism at both five- and 10-year follow-up periods, with an AUC of .63, approaching a medium effect size. The authors measured general sexual recidivism, not sexual violence, as required by SVP determinations.
In an Austrian sample, Gregório Hertz and colleagues66 found that the VRAG-R total score significantly predicted contact sexual recidivism for the total sample (AUC = .61) and child sexual abuse (AUC = .66). These results were echoed when evaluating VRAG-R risk bins. Notably, it did not distinguish between violent and nonviolent contact offenses.
A correctional sample67 found the VRAG-R had moderate predictive validity for violent recidivism (AUC = .66). Violent recidivism included sexual assaults, but not as a separate category. Thus, the VRAG-R can moderately predict sexual recidivism and general violent recidivism. It does not, however, address sexual violence, as required by Step 4.
SVR-20 V2
The SVR-20 V1 has been extensively evaluated, whereas the SVR-20 V2 has limited predictive validity since its 2017 release. In Hart and Boer’s chapter about the SVR-20 V1 and V2,68 only one study is cited about the V2’s predictive validity in Singapore. Tsao and Chu69 completed a retrospective study of male sexual offenders using the SVR-20 V2 to predict general sexual recidivism. Sexual recidivism was broadly defined as any new offenses that were sexual in nature, including contact and noncontact offenses. The SVR-20 V2 total score (AUC = .76) and risk category (AUC = .78) significantly predicted sexual recidivism, with large effect sizes. Like other risk measures, violent sexual recidivism was not addressed as required for Step 4.
Static-99R
Two recent Static-99R meta-analyses evaluated the predictive validity of the Static-99R for general sexual recidivism70,71 but did not address sexual violence. Results found the Static-99R demonstrated moderate predictive validity, with AUC values ranging from .65 to .66 for follow-up periods between 5 and 15 years.71
A major concern involved its effectiveness across races. AUC of .54 failed to be effective with Hispanic Americans, whereas African Americans showed an AUC of .76, indicating high predictive validity.70 Because these meta-analyses included noncontact sexual offenses, their relevance to Step 4 is compromised.
IORNS
No published study addresses the IORNS predictive validity for sexual recidivism, making it insufficient for Step 4.
SAPROF-SO
The SAPROF-SO36 only has one published study on its predictive validity,72 with a sample of 100 adult male SVPs in Wisconsin. They found that the SAPROF-SO total score had a small but significant ability (AUC = .61) to predict individuals who were found to sexually breach their conditional release in two years. This single study is not sufficient to address Step 4.
RSVP V2
Darjee and colleagues73 investigated the RSVP with sexual offenders in Scotland, with no significant relationship between RSVP summary judgments and serious sexual offending. The RSVP Past total items (AUC = .71) and RSVP Relevance total items (AUC = .72) significantly predicted any serious sexual offending (e.g., rape and sexual homicides). Only one dissertation examined the RSVP V2 specifically, finding that Recent (AUC = .81), Past (AUC = .76), and Ever Present (AUC = .79) total scores significantly predicted sexual recidivism, though sexual recidivism was broadly defined.74 Overall, there is limited evidence that the RSVP and its current version predict violent sexual recidivism; Step 4 was not addressed.
START
As a general risk measure, no studies have examined START’s predictive validity for sexual recidivism. One study75 looking at institutional aggression with sexual offenders found the Vulnerability and Strength total scores to be significantly predictive of any aggression (AUCs = .71–.73). Meanwhile, only the Vulnerability total score was significantly predictive of physical violence against others (AUC = .73). Therefore, there is insufficient evidence of predictive power for violent sexual recidivism, and it does not answer Step 4.
Discussion
Over the past 20 years, risk-assessment tools have advanced, with three measures showing marked improvements for SVP determinations. The HCR-20V3 appears to assess mental abnormality and volitional impairment more than its previous version. The SVR-20 V2 comprehensively evaluates mental abnormality and includes some aspects of volitional impairment not addressed in V1. The VRAG-R also shows improvements relevant to Steps 1 and 2.
Other risk measures were less successful. The RSVP V2 fully addresses Step 1 and offers insight into Step 2 (impulsivity and behavioral disinhibition). The PCL-R minimally addresses Step 1. The IORNS, START, and SAPROF-SO do not adequately address any of the five SVP steps, although they may provide ancillary information for Step 1. Although clinical judgment can sometimes address these gaps, it may increase subjectivity and variability among evaluators, reducing reliability and standardization.
Our systematic approach for reviewing SVP measures relied on test manuals used by practitioners in these determinations. As a limitation, we did not gain access to the STABLE-2007 manual.76 As a result, this measure was excluded despite extensive recent research.77,78
Predictive validity of risk measures in SVP contexts remains concerning. A major limitation is that sexual recidivism is often defined broadly as any sexual reconviction, without distinguishing between violent and nonviolent sex offenses or even contact and noncontact behaviors. The low base rate of sexual recidivism also complicates accurate risk predictions.79 Given the severe consequences of SVP commitment, future research should utilize precise definitions of sexual recidivism. For example, voyeuristic behaviors are concerning but different from sexual violence, which is the primary focus of SVP evaluations. Empirical literature should reflect such differences.
Most studies did not assess for calibration, making it difficult to evaluate how accurately these tools perform in practice.50 Future research should prioritize calibration analyses to improve accuracy in classification and reduce false positives, which carry significant consequences in SVP evaluations.
Some progress has been documented in developing risk measures more aligned with SVP determinations. Still, the retrofitting of psychological measures persists. The PCL-R, designed to assess psychopathy, was reportedly used in 67.4 percent of SVP evaluations to inform the risk of sexual offending.54 It has not established predictive validity for sexual violence and falls short on the other SVP steps.
We do not argue that these risk measures are entirely unhelpful. Rather, we ask forensic evaluators to consider whether these measures help answer the legal criteria required by SVP determinations. The decisive matter centers on the volitional impairment resulting in sexual violence.
We return to the main question about whether these measures are appropriate for SVP evaluations or being incorrectly retrofitted to meet legal standards they were not designed to address. Although some instruments have improved, many still lack a clear conceptual link between key legal standards of mental abnormality and volitional impairment. The HCR-20V3 stands out as the most aligned with SVP standards, having made meaningful progress in evaluating constructs relevant to legal definitions. Forensic evaluators are likely to be divided on whether its current progress is sufficient to meet the requirements for being classified as a forensic assessment instrument (FAI) for SVP evaluations. On pragmatic grounds, implementation of the HCR-20V3 is certainly far better than no standardized SVP assessments. Perhaps a middle ground would be to accept the HCR-20V3 results as an FAI when “too-close-to-call” cases are excluded. The challenge remains in the absence of established cut scores for the HCR-20V3. As a heuristic, practitioners experiencing more than transient ambivalence should make this uncertainty known in their forensic reports. This cautionary note should be highlighted when the SVP determinations are close to the crucial threshold (i.e., between moderate and low levels).
The role of the HCR-20V3 in SVP determinations will continue to evolve. More generally, the momentum for stronger alignment between legal criteria and the psychological constructs should continue unabated. Greater precision is needed to ensure that risk-assessment measures are tailored to sexual violence with rigorous psychometric standards.
Footnotes
Disclosures of financial or other potential conflicts of interest: None.
- © 2026 American Academy of Psychiatry and the Law
References
- 1.↵RogersRJacksonRL. Sexually violent predators: The risky enterprise of risk assessment. J Am Acad Psychiatry Law. 2005 Dec; 33(4):523–8
- 2.↵WertzMSchobelSSchiltzKRettenbergerM. A comparison of the predictive accuracy of structured and unstructured risk assessment methods for the prediction of recidivism in individuals convicted of sexual and violent offense. Psychol Assess. 2023 Feb; 35(2):152–64
- 3.↵SreenivasanSHoffmanACahanJ. Applying collaborative justice to sexually violent predator civil commitment. J Am Acad Psychiatry Law. 2020 Dec; 48(4):496–508
- 4.↵Adam Walsh Child Protection and Safety Act, 5 U.S.C. 553 (2006)
- 5.↵GookinK. Comparison of State Laws Authorizing Involuntary Commitment of Sexually Violent Predators. Olympia, WA: Washington State Institute for Public Policy [Internet]; 2007. Available from: https://www.wsipp.wa.gov/ReportFile/989/Wsipp_Comparison-of-State-Laws-Authorizing-Involuntary-Commitment-of-Sexually-Violent-Predators-2006-Update-Revised_Full-Report.pdf. Accessed August 25, 2026
- 6.↵HoppeTMeyerIHDe OrioS. Civil commitment of people convicted of sex offenses in the United States [Internet]; 2020. Available from: https://williamsinstitute.law.ucla.edu/publications/civil-commitment-us/. Accessed May 11, 2025
- 7.↵FelthousARKoJ. Sexually violent predator law in the United States. East Asian Arch Psychiatry. 2018 Dec; 28(4):159–73
- 8.↵California Sex Offender Management Board. Annual report [Internet]; 2019. Available from: https://casomb.org/pdf/2019_Annual_Report.pdf. Accessed May 11, 2025
- 9.↵California Department of State Hospitals. Sexually violent predator (SVP) fact sheet and civil commitment process [Internet]; 2021. Available from: https://www.dsh.ca.gov/Treatment/docs/SVP_FactSheetAndProcess.pdf. Accessed May 11, 2025
- 10.↵DeMatteoDMurphyMGallowayMKraussDA. A national survey of United States sexually violent person legislation: Policy, procedures, and practice. Int J Forensic Ment Health. 2015 Oct; 14(4):245–66
- 11.↵Mich. Comp. Laws Ann. §§ 780.501–.509 (1937)
- 12.↵Wash. Rev. Code Ann. § 71.09.020 (1990)
- 13.↵Ariz. Rev. Stat. Ann. § 36-3701 (1998)
- 14.↵Cal. Welf. & Inst. Code §§ §6600–6609.3 (1996)
- 15.↵1993 Wisconsin Act 479, Wis. Stat. ch. 980 (1994)
- 16.↵LiebR. Washington’s sexually violent predator law: Legislative history and comparisons with other states [Internet]; 2020. Available from: https://www.wsipp.wa.gov/reportfile/1244/wsipp_washingtons-sexually-violent-predator-law-legislative-history-and-comparisons-with-other-states_full-report.pdf. Accessed May 20, 2025
- 17.↵Kan. Stat. Ann. § 59-29a (1994)
- 18.↵Kansas v. Hendricks, 521 U.S. 346 (1997)
- 19.↵Kan. Stat. Ann. § 59-29a02 (1994)
- 20.↵Tex. Health & Safety Code Ann. § 841.023 (2015)
- 21.↵Ariz. Rev. Stat. Ann. § 36-3701(5) (1998)
- 22.↵Kansas v. Crane, 534 U.S. 407 (2002)
- 23.↵United States v. Comstock, 560 U.S. 126 (2010)
- 24.↵WebsterCDDouglasKSEavesD. HCR-20: Assessing Risk for Violence (Version 2). Burnaby, BC, Canada: Mental Health, Law, and Policy Institute, Simon Fraser University; 1997
- 25.↵BoerDPHartSDKroppPR. Manual for the Sexual Violence Risk-20: Professional Guidelines for Assessing Risk of Sexual Violence. Vancouver, BC, Canada: The British Columbia Institute Against Family Violence; 1997
- 26.↵HansonRKThorntonD. Static-99: Improving Actuarial Risk Assessments for Sex Offenders. Ottawa, ON, Canada: Solicitor General of Canada; 1999
- 27.↵HansonRK. The Development of a Brief Actuarial Risk Scale for Sexual Offense Recidivism. Ottawa, ON, Canada: Department of the Solicitor General of Canada; 1997
- 28.↵EppersonDLKaulJDHuotS. Minnesota Sex Offender Screening Tool-Revised (MnSOST-R). St. Paul, MN: Minnesota Department of Corrections; 2000
- 29.↵HarrisGTRiceMEQuinseyVL. Violent recidivism of mentally disordered offenders: The development of a statistical prediction instrument. Crim Just & Behav. 1993 Dec; 20(4):315–35
- 30.↵QuinseyVLHarrisGTRiceME. Violent Offenders: Appraising and Managing Risk. Washington, DC: American Psychological Association; 1998
- 31.↵HareRD. Manual for the Hare Psychopathy Checklist-Revised, Second Edition. Toronto, ON, Canada: Multi-Health Systems; 2003
- 32.↵WardTLawsRDHudsonSM. Penile plethysmography: Will we ever get it right? In WardTLawsRDHudsonSM, editors. Sexual Deviance: Issues and Controversies. Thousand Oaks, CA: SAGE Publications Inc; 2013. p. 82–102
- 33.↵DouglasKSHartSDWebsterCD. HCR-20V3: Assessing Risk of Violence – User Guide. Burnaby, BC, Canada: Mental Health, Law, and Policy Institute, Simon Fraser University; 2013
- 34.↵BoerDPHartSDKroppPR. Manual for the Sexual Violence Risk-20, Version 2. Professional Guidelines for Assessing Risk of Sexual Violence. Vancouver, British Columbia, Canada: Protect International, Risk and Safety Services Inc; 2017
- 35.↵HartSDKroppPKWattKA. RSVP-V2: Version 2 of the Risk for Sexual Violence Protocol. In Structured Professional Judgement Guidelines for Assessing and Managing Risk for Sexual Violence. Burnaby, BC, Canada: Protect International; 2022
- 36.↵WillisGMThorntonDTKelleySMde Vries RobbéM. Structured Assessment of Protective Factors against Sexual Offending (SAPROF-SO) Version 1. Scoring Manual. 2021
- 37.↵RiceMEHarrisGTLangC. Validation of and revision to the VRAG and SORAG: The Violence Risk Appraisal Guide–Revised (VRAG-R). Psychol Assess. 2013 Sep; 25(3):951–65
- 38.↵MillerHA. IORNS: Inventory of Offender Risk, Needs, and Strengths: Professional Manual. Odessa, FL: Psychological Assessment Resources; 2006
- 39.↵MarshallWL. Clinical and research limitations in the use of phallometric testing with sexual offenders. Sex Offender Treat. 2006; 1(1):1–8
- 40.↵MurphyLGottfriedEDiMarioK. Use of penile plethysmography in the court: A review of practices in Canada, the United Kingdom and the United States. Behav Sci & L. 2020 Mar; 38(2):79–99
- 41.↵PfaffenrothPC. The need for coherence: states’ civil commitment of sex offenders in the wake of Kansas v. Crane. Stan L Rev. 2003; 55(6):2229–66
- 42.
- 43.↵RogersRCavanaughJLJr. The Rogers criminal responsibility assessment scales. IMJ Ill Med J. 1981 Sep; 160(3):164–6, 169
- 44.↵RiceMEHarrisGT. Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law & Hum Behav. 2005 Oct; 29(5):615–20
- 45.↵CohenJ. A power primer. Psychological Bulletin. 1992; 112(1):155–9
- 46.↵CohenJ. Statistical Power Analysis for the Behavioral Sciences. New York, NY: Academic Press; 1969
- 47.↵HartSDMichieCCookeDJ. Precision of actuarial risk assessment instruments: Evaluating the ‘margins of error’ of group v. individual predictions of violence. Br J Psychiatry Suppl. 2007 May; 49(S49):s60–5
- 48.↵FazelSChangZFanshaweT. Prediction of violent reoffending on release from prison: Derivation and external validation of a scalable tool. Lancet Psychiatry. 2016 Jun; 3(6):535–43
- 49.↵FazelSSariaslanAFanshaweT. Towards a more evidence-based risk assessment for people in the criminal justice system: The case of OxRec in the Netherlands. Eur J Crim Pol Res. 2022; 28(3):397–406
- 50.↵OgonahMGTSeyedsalehiAWhitingDFazelS. Violence risk assessment instruments in forensic psychiatric populations: A systematic review and meta-analysis. Lancet Psychiatry. 2023 Oct; 10(10):780–9
- 51.↵WebsterCDDouglasKSEavesD. HCR-20: Assessing Risk for Violence (Version 2). Burnaby, BC: Mental Health, Law, and Policy Institute, Simon Fraser University; 1997
- 52.↵DeMatteoDEdensJFGallowayM. The role and reliability of the Psychopathy Checklist–Revised in U.S. sexually violent predator evaluations: A case law survey. Law & Hum Behav. 2014 Jun; 38(3):248–55
- 53.↵BoccacciniMTChevalierCSMurrieDCVarelaJG. Psychopathy Checklist–Revised use and reporting practices in sexually violent predator evaluations. Sex Abuse. 2017 Sep; 29(6):592–614
- 54.↵BoerDPHartSDKroppPR. Manual for the Sexual Violence Risk–20: Professional Guidelines for Assessing Risk of Sexual Violence. Vancouver, BC: The Mental Health, Law, and Policy Institute; 1997
- 55.↵PhenixAFernandezYHarrisAJ. Static-99R Coding Rules, Revised-2016. Ottawa, Ontario, Canada: Public Safety Canada; 2017
- 56.↵ChevalierCSBoccacciniMTMurrieDCVarelaJG. Static-99R reporting practices in sexually violent predator cases: Does norm selection reflect adversarial allegiance? Law & Hum Behav. 2015; 39(3):209–18
- 57.↵VogelVDde RuiterCBoumanY. SAPROF: Guidelines for the Assessment of Protective Factors for Violence Risk, English version. Utretcht, The Netherlands: Forum Educatief; 2009
- 58.↵WebsterCDMartinMLBrinkJ. Short-Term Assessment of Risk and Treatability (START). Ontario, Canada: St. Joseph’s Healthcare Hamilton; 2004
- 59.↵NeilCO’RourkeSFerreiraNFlynnL. Protective factors in violence risk assessment: Predictive validity of the SAPROF and HCR-20V3. Int J Forensic Ment Health. 2020 Jan; 19(1):84–102
- 60.↵BrooksteinDMDaffernMOgloffJRP. Predictive validity of the HCR-20V3 in a sample of Australian forensic psychiatric patients. Psychiatr Psychol Law. 2021 Jun; 28(3):325–42
- 61.↵O’SheaLEThakerDPicchioniMM. Predictive validity of the HCR-20 for violent and non-violent sexual behaviour in a secure mental health service. Crim Behav Ment Health. 2016 Dec; 26(5):366–79
- 62.↵HawesSWBoccacciniMTMurrieDC. Psychopathy and the combination of psychopathy and sexual deviance as predictors of sexual recidivism: Meta-analytic findings using the Psychopathy Checklist–Revised. Psychol Assess. 2013 Mar; 25(1):233–43
- 63.↵RettenbergerMMatthesABoerDPEherR. Prospective actuarial risk assessment: A comparison of five risk assessment instruments in different sexual offender subtypes. Int’l J Offender Therapy & Comp Criminology. 2010 Apr; 54(2):169–86
- 64.↵MorettiGFluttiEColaninoM. Recidivism risk in male adult sex offenders with psychopathic traits assessed by PCL-R: A systematic review. Med Sci & L. 2024 Jan; 64(1):41–51
- 65.↵OlverMESewallLA. Cross-validation of the discrimination and calibration properties of the VRAG-R in a treated sexual offender sample. Crim Just & Behav. 2018 Jun; 45(6):741–61
- 66.↵Gregório HertzPEherREtzlerSRettenbergerM. Cross-validation of the revised version of the Violence Risk Appraisal Guide (VRAG-R) in a sample of individuals convicted of sexual offenses. Sex Abuse. 2021 Feb 1; 33(1):63–87
- 67.↵GloverAJJChurcherFPGrayAL. A cross-validation of the Violence Risk Appraisal Guide–Revised (VRAG–R) within a correctional sample. Law & Hum Behav. 2017 Dec; 41(6):507–18
- 68.↵HartSDBoerDP. Structured professional judgment guidelines for sexual violence risk assessment. In DouglasKSOttoRK, editors. Handbook of Violence Risk Assessment, Second Edition. New York, NY: Routledge; 2020. p. 322–58
- 69.↵TsaoITChuCM. An exploratory study of recidivism risk assessment instruments for individuals convicted of sexual offenses in Singapore. Sex Abuse. 2021 Mar; 33(2):157–75
- 70.↵AhmedSLeeSCHelmusLM. Predictive accuracy of Static-99R across different racial/ethnic groups: A meta-analysis. Law & Hum Behav. 2023 Feb; 47(1):275–91
- 71.↵HelmusLMKelleySMFrazierA. Static-99R: strengths, limitations, predictive accuracy meta-analysis, and legal admissibility review. Psychol Pub Pol’y & L. 2022 Aug; 28(3):307–31
- 72.↵CarrAEKelleySMWillisGMThorntonD. The predictive validity of the SAPROF-SO for success on supervised release from a secure treatment center. Sex Abuse. 2025; 37(7):759–86
- 73.↵DarjeeRRussellKForrestL. Risk for Sexual Violence Protocol (RSVP): A real world study of the reliability, validity and utility of a structured professional judgement instrument in the assessment and management of sexual offenders in South East Scotland. Edinburgh: NHS Lothian Sex Offender Liaison Service; 2016. Available from: https://www.researchgate.net/publication/294718597_Risk_of_Sexual_Violence_Protocol_RSVP_A_real_world_study_of_the_reliability_validity_and_utility_of_a_structured_professional_judgement_instrument_in_the_assessment_and_management_of_sexual_offenders. Accessed August 15, 2026
- 74.↵VargenLM. Risk assessment of men convicted of child sexual exploitation material offences: a case-control study of predictive validity [dissertation]. Burnaby (BC): Simon Fraser University; 2026. Available from: https://summit.sfu.ca/item/40615. Accessed August 25, 2026
- 75.↵CartwrightJKDesmaraisSLHazelJ. Predictive validity of HCR-20, START, and Static-99R assessments in predicting institutional aggression among sexual offenders. Law & Hum Behav. 2018; 42(1):13–25
- 76.↵FernandezYMHarrisAJRHansonRK. STABLE-2007 coding manual: Revised 2014 [unpublished report]. Ottawa, ON: Public Safety Canada; 2014
- 77.↵HolperLCerulloEMokrosAHabermeyerE. Predictive and incremental validity of the Static-99, Static-99R, and STABLE-2007 for sexual recidivism: A diagnostic test accuracy network meta-analysis (DTA-NMA). Psychol Assess. 2024 Feb; 36(2):134–46
- 78.↵HelmusLMHansonRKMurrieDCZabarauckasCL. Field validity of Static-99R and STABLE-2007 with 4,433 men serving sentences for sexual offences in British Columbia: New findings and meta-analysis. Psychol Assess. 2021 Jul; 33(7):581–95
- 79.↵AbbottBR. Sexually violent predator risk assessments with the violence risk appraisal guide-revised: a shaky practice. Int’l J L & Psychiatry. 2017 May-June; 52:62–73
