Abstract
Forensic experts are increasingly asked to consult in sexually violent predator (SVP) determinations. The substantive criteria for SVP standards vary substantially across jurisdictions, but typically include complex judgments regarding volitional impairment and predictive statements focused specifically on sexual violence. A common but questionable practice is the retrofitting of generic risk-assessment measures to address SVP criteria. The marked deficiencies of these measures in addressing the relevant questions, coupled with their methodological limitations, are noted. SVP determinations demand rigorous evaluations of relevant factors that are buttressed by empirically validated methods.
A mantle of expertise has been thrust on forensic psychiatrists and psychologists during the past 15 years through enactment of sexually violent predator (SVP) legislation. The statutes assume that forensic practitioners have the expert knowledge and competence to (1) address the ambiguity and complexity implicit in the SVP substantive criteria and (2) render accurate cross-situational predictions. Jackson et al.1 conducted the first systematic study of SVP expertise by using a sophisticated factorial design. While very experienced forensic psychologists showed some expertise in predicting general violence, the authors found that these clinicians were no more accurate than chance in predicting sexual violence. In this research, three major methods of conducting SVP evaluations were systematically evaluated: (1) general clinical judgment (i.e., the use of customary interview and history data), (2) structured clinical judgment (i.e., the review of standardized clinical data on likely risk factors), and actuarial (i.e., the mathematical quantification of risk based on scale scores). Of note, no individual method of SVP evaluation offered an appreciable advantage in predicting sexual violence.
The expertise of forensic practitioners to conduct SVP consultations is a matter of considerable importance, given the increase in SVP laws and subsequent cases. First enacted in 1990 with Washington’s Community Protection Act,2 SVP commitments are now used in at least 15 states, with recent legislative initiatives likely to increase this number.3 As of 2001, Fitch and Hammen4 found that more than 1,200 persons had already been committed as “SVP detainees.” We refer to this population as “SVP detainees,” because the primary purpose of their commitments is incapacitation rather than treatment.5
Many forensic practitioners embracing the SVP mantle of expertise have attempted to retrofit currently generic risk measures in their efforts to evaluate the specific requirements of SVP standards. While pragmatic, this retrofitting did not take into account critical differences between specific SVP determinations and general risk assessments. In this article, we examine the conceptual and empirical problems in conducting SVP evaluations, with a focus on risk assessment tools. To provide the necessary framework, we begin with a brief review of SVP substantive criteria, highlighting the statutory differences and Supreme Court decisions.
SVP Substantive Criteria
Rogers and Shuman6 have outlined the substantive criteria required for SVP determinations. Most SVP statutes share four substantive criteria: (1) past sexually harmful conduct, (2) a current clinical condition, (3) a substantial risk of future sexual violence, and (4) a causal relationship between the mental abnormality and the potential sexual harm. As outlined in the subsequent paragraphs, important differences are found in the definitions of each SVP criterion.
The requisite mental condition for SVP determinations varies substantially from jurisdiction to jurisdiction. As summarized by Lieb,7 the mental condition can be classified in four categories: (1) mental disorder (Arizona, California, Illinois, and Wisconsin), (2) mental abnormality (Iowa, Kansas, Missouri, and Texas), (3) mental abnormality or personality disorder (Florida, Massachusetts, New Jersey, South Carolina, Virginia, and Washington), and (4) idiosyncratic definitions (Minnesota and North Dakota). Mental disorders and personality disorders typically refer to DSM-IV diagnoses, which can be reliably established through the use of structured interviews.8,9 In contrast, the term mental abnormality is ambiguous and lacks any diagnostic precision. On this point, Zonana et al.10 were sharply critical of legislative attempts “to invent mental or emotional categories” (Ref. 10, p 143) for SVP commitments.
The Supreme Court in Kansas v. Hendricks11 upheld the constitutionality of SVP commitments, rejecting arguments based on diagnostic nomenclature (e.g., the imprecision of mental abnormality). In particular, the Court leaves to the states “the task of defining terms of a medical nature that have legal significance” (Ref. 11, p 359). However, the decision affirmed that the mental condition must give rise to “volitional impairment rendering them dangerous beyond their control” (emphasis added; Ref. 11, p 358). This volitional impairment was addressed further in Kansas v. Crane.12 While rejecting the necessity of a complete loss of volitional control, the Court established the following standard: “there must be proof of serious difficulty in controlling behavior” (Ref. 12, p 413). Dangerousness alone is insufficient. The Court further specified that this mental condition with its loss of volitional abilities must be distinguished “from the dangerous but typical recidivist” (Ref. 12, p 413).
The substantial risk of violence varies across jurisdictions. According to Lieb,7 most states require either sexual violence (Arizona, California, Florida, Illinois, Missouri, New Jersey, South Carolina, Virginia, and Wisconsin) or predatory sexual violence (Iowa, Kansas, Texas, and Washington). Surprisingly, two states require harmful (Minnesota) or predatory (North Dakota) conduct, but do not specify sexual violence. Contrary to the expressed intent of SVP commitment, Massachusetts apparently permits any sexual offense to satisfy commitment requirements. The level of risk6 for this sexual conduct also varies from “likely” to “substantially probable” and “more likely than not.”
Proper application of the SVP statutes requires an understanding of the nexus between the clinical condition (e.g., mental disorder or abnormality) and ability to control behavior within a legal definition. Legally, the clinical condition requires a substantial impairment in volitional control.13 The simple co-occurrence of a clinical condition and lack of volitional control is insufficient. The crucial question is whether the clinical condition causes the lack of control. This nexus cannot be inferred from diagnoses. DSM-IV14 clearly states that a clinical diagnosis “is not sufficient to establish the existence for legal purposes of a mental disorder, mental disability, mental disease or mental defect” (Ref. 14, p xxiii). Of particular relevance to SVP volitional impairment, DSM-IV further specifies that “having the diagnosis in itself does not demonstrate that a particular individual is (or was) unable to control his or her behavior at a particular time” (Ref. 14, p xxiii).
The Supreme Court in Hendricks11 and Crane12 provided guidelines for the required impairment of volition. However, the Court in Crane purposefully avoided operationalizing the construct by suggesting that “safeguards of human liberty in the area of mental illness and the law are not always best enforced through precise bright-line rules” (Ref. 12, p 413). Crane clarified that the statutes do not require a total lack of control, only a “serious difficulty” in controlling behavior (Ref. 12, p 413). In the absence of bright-line criteria, clinicians are faced with the challenging task of not only operationalizing volitionality, but also assessing it with reliable and valid methods. In earlier work on insanity evaluations, investigators15 have noted the problems in measuring volitional impairment. From a dimensional perspective, the clinician must be able to differentiate accurately serious difficulty from some or moderate difficulty.
The loss of volitional impairment directly arising from a clinical condition is pivotal to SVP statutes.16 According to the Supreme Court, it is entirely this loss of volitionality that separates civil commitment from criminal punishment. As articulated in Hendricks,11 the mental disorder requirement “adequately distinguishes Hendricks from other dangerous persons who are perhaps more properly dealt with exclusively through criminal proceedings” (Ref. 11, p 360).
In summary, the Supreme Court has upheld the constitutionality of SVP statutes and clarified the importance of volitional impairment. Forensic practitioners must take into account specific factors in SVP consultations. These are outlined as five sequential inquiries:
Does the defendant have the requisite clinical condition (e.g., mental abnormality, mental disorder, or personality disorder), as mandated by the particular jurisdiction?
If yes, does the defendant have volitional impairment?
If yes, does the volitional impairment directly arise from the clinical condition?
If yes, does the defendant have the statutorily defined likelihood of sexual violence?
If yes, is that likelihood of sexual violence directly the result of volitional impairment?
Forensic practitioners face multiple challenges in addressing these sequential steps. They must develop reliable methods to assess the definitional ambiguities posed by “mental abnormality.” They must also face squarely the long-standing controversies in the assessment of volitional impairment. In the context of insanity, the American Bar Association17 has argued forcefully for the elimination of volitional impairment because “there is still no accurate scientific basis for measuring one’s capacity for self control or for calibrating the impairment of such capacity” (Ref. 17, pp 4–5). Passage of the Insanity Defense Reform Act, strongly supported by the American Psychiatric Association and other professional associations, eradicated the volitional prong from federal insanity cases.6 As observed by Rogers,18 “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. 18, p 159).
Assessment methods must take into account the explicit requirements of SVP standards. Of particular concern is whether the widespread practice of retrofitting of general risk assessment measures to SVP determinations is warranted. The next section addresses this concern in light of conceptual and psychometric issues.
Retrofitting Generic Risk Assessment Measures
Several recent reviews19–21 emphasize the strengths and potential applicability of generic risk assessment measures for SVP determinations. However, each review fails to address whether these generic measures effectively evaluate SVP substantive criteria. As a fundamental principle of scale development, validity is never generic. According to the official standards adopted by the American Psychological Association,22 “each recommended use or interpretation requires validation” (Ref. 22, p 18). Any facile extrapolation from generic risk to SVP determinations is inconsistent with professional standards and is probably unethical.23
Generic risk assessment measures used in SVP determinations can be grouped as either structured clinical judgment or actuarial measures. For structured clinical judgment, two common measures are Historical-Clinical-Risk (HCR)-2024 and the Sexual Violence Risk (SVR)-20.25 An array of actuarial measures are available, including the Static-99,26 the Rapid Risk Assessment of Sexual Offense (RRASOR),27 the Minnesota Sex Offender Screening Tool-Revised (MnSost-R),28 the Violence Risk Appraisal Guide (VRAG),29 and the Sex Offense Risk Appraisal Guide (SORAG).30 Two additional measures are sometimes used as risk measures, although they are intended to assess psychopathy (Psychopathy Checklist-Revised; PCL-R31) and sexual arousal profiles (penile plethysmography; PPG32).
Key findings for risk assessment measures are summarized in Table 1. The first step of SVP determinations is a thorough examination of the requisite clinical condition. For mental disorders, the coverage provided by risk measures range from nonexistent to superficial. When present, forensic clinicians are typically asked about the presence of any Axis I or Axis II disorders. With the exception of schizophrenia on the VRAG, no effort is made to evaluate specific Axis I disorders and their predictive ability for sexual recidivism. Beyond disorders, mental abnormalities are likely to incorporate syndromes, salient symptoms, and other evidence of impaired functioning. Considering only syndromes and salient symptoms, we estimate the risk measures address less than one percent of the relevant clinical issues essential to Step 1. In summary, risk assessment measures fall spectacularly short in their assessment of mental disorders, personality disorders, and mental abnormalities.
Fulfillment of SVP Standards by Generic Risk-Assessment Measures
Three steps involve the detailed evaluation of volitional impairment (Step 2), its causes (Step 3), and its relationship to sexual recidivism (Step 5). As summarized in Table 1, volitional impairment is systematically ignored by all risk assessment measures. Consequently, the other two steps were also neglected. Without the systematic assessment of volitional impairment, generic risk measures cannot address SVP determinations.
The prediction of sexual violence (Step 4) is crucial to SVP determinations. Unfortunately, most research with sex offenders has focused either on general violence or nonspecific sexual recidivism. Predicting general violence among sex offenders is simply not the same as predicting sexual violence. For example, in a cross-validation of the VRAG with sexual offenders, Rice et al.33 noted that the VRAG demonstrated predictive validity for violent, but not sexual recidivism. They dismissed this crucial issue asserting that the prediction of “future sex offending may be of theoretical interest” (Ref. 33, p 239). Within the context of SVP commitments, this assertion is patently inaccurate. Clinicians, adopting this perspective, unwittingly fail to address the pivotal element of SVP determinations.
Studies that define sexual recidivism as any type of sex offense, also fail to address the SVP standards. With the possible exception of one state (Massachusetts), these predictions are irrelevant to SVP determinations. In most jurisdictions, forensic experts are required to predict sexual violence or predatory sexual violence. Such predictions are simply not possible from data on general violence or nonspecific sexual recidivism. As a specific example, the SVR-20 claims to assess sexual violence; however, its classification is overly broad and includes nonviolent offenses (obscene letters and voyeurism). Such leaps of faith from general to sexual violence are unsubstantiated.
The current research of actuarial measures is highly reductionistic, in collapsing most sex offenders into a single category. This profound disregard for the heterogeneity of sexual offenders may lead to serious errors in prediction. Even the most basic typologies (e.g., rapists and child molesters) are neglected. For example, child molesters are often motivated by sexual aspects of offending, limit their criminal acts to sexual offending, and experience negative affect preceding their offenses.34,35 In contrast, rapists are often motivated by anger and commit nonsexual offenses.35 Lumping together all paraphilias and sex offenses confounds any attempt at meaningful interpretation. Unquestionably, more focused methods are needed that take into account both clinical conditions (e.g., paraphilias) and offense types.36
In conclusion, important advances in forensic assessment have been achieved in the past decade.37 Specialized methods evidence increased sophistication in assessing legal standards, operationalizing the relevant constructs, and testing the predicted relationships.6 In stark contrast, current efforts to retrofit generic risk assessment methods fall far short. SVP determinations require that the predictions be based on (1) specified clinical conditions, which cause (2) serious volitional impairment resulting in (3) sexual violence. Table 1 underscores these fundamental problems with generic risk measures. As always, the mantle of expertise must be earned through painstaking analysis and research. SVP determinations deserve and demand rigorously validated measures.
- American Academy of Psychiatry and the Law
