A classification tree approach to the development of actuarial violence risk assessment tools

Examining the utility of the PCL:SV as a screening measure using competing factor models of psychopathy

The correspondence between the Hare Psychopathy Checklist: Screening Version (PCL:SV; S. D. Hart, D. N. Cox, & R. D. Hare, 1995) and the Hare Psychopathy Checklist-Revised (PCL-R; R. D. Hare, 1991, 2003) was examined in forensic (N = 175) and correctional (N = 188) samples. Intermeasure correlations for Total scores (.95 forensic, .94 correctional) and the original 2-factor, D. J. Cooke and C. Michie's (2001) 3-factor, and R. D. Hare's (2003) 4-facet models (range = .87-.95) were high. Area under the curve values for the PCL:SV were .98 in both samples (cutoff = PCL-R Total score of 25). The PCL:SV performed well as a screen, maximizing false positive relative to false negative errors. Close correlations for prediction of violent recidivism in the correctional sample were obtained for the PCL-R (.42) and PCL:SV (.37). Results indicate the robust relation between the measures is maintained whether they are completed on the basis of file review only or file plus interview and whether the same or different raters score the measures.

Predicting inpatient violence using an extended version of the Brøset-Violence-Checklist: instrument development and clinical application

BACKGROUND

Patient aggression is a common problem in acute psychiatric wards and calls for preventive measures. The timely use of preventive measures presupposes a preceded risk assessment. The Norwegian Brøset-Violence-Checklist (BVC) is one of the few instruments suited for short-time prediction of violence of psychiatric inpatients in routine care. Aims of our study were to improve the accuracy of the short-term prediction of violence in acute inpatient settings by combining the Brøset-Violence-Checklist (BVC) with an overall subjective clinical risk-assessment and to test the application of the combined measure in daily practice.

METHOD

We conducted a prospective cohort study with two samples of newly admitted psychiatric patients for instrument development (219 patients) and clinical application (300 patients). Risk of physical attacks was assessed by combining the 6-item BVC and a 6-point score derived from a Visual Analog Scale. Incidents were registered with the Staff Observation of Aggression Scale-Revised SOAS-R. Test accuracy was described as the area under the receiver operating characteristic curve (AUCROC).

RESULTS

The AUCROC of the new VAS-complemented BVC-version (BVC-VAS) was 0.95 in and 0.89 in the derivation and validation study respectively.

CONCLUSION

The BVC-VAS is an easy to use and accurate instrument for systematic short-term prediction of violent attacks in acute psychiatric wards. The inclusion of the VAS-derived data did not change the accuracy of the original BVC.

The classification of violence risk

The Classification of Violence Risk (COVR) is an interactive software program designed to estimate the risk that a person hospitalized for mental disorder will be violent to others. The software leads the evaluator through a chart review and a brief interview with the patient. At the end of this interview, the software generates a report that contains a statistically valid estimate of the patient's violence risk-ranging from a 1% to a 76% likelihood of violence-including the confidence interval for that estimate, and a list of the risk factors that the program took into account to produce the estimate. In this article, the development of the COVR software is described and several issues that arise in its administration are discussed.

Reliability and Validity Evaluation of the Psychopathy Checklist: Screening Version (PCL:SV) in Swedish correctional and forensic psychiatric samples

This study evaluated the structural reliability, construct-related validity, and cultural validity generalization of the Hare Psychopathy Checklist: Screening Version (PCL:SV) in a sample of more than 560 male and female Swedish forensic psychiatric treatment patients, forensic evaluation patients, and criminal offenders. Structural reliability was excellent for most indices. PCL:SV scores were higher for males than females for total and Part 1 scores (interpersonal/affective features) but not for Part 2 (behavioral features). With some exceptions, PCL:SV scores were meaningfully related to aggression to others, a measure of risk for violence, substance use problems, personality disorder (positive), and psychosis (negative). Correlations between PCL:SV and aggression were larger for females than males, although the difference was smaller when personality disorder was held constant. The structural reliability and pattern of validity coefficients were comparable in these Swedish samples to other non-North American samples. Implications for the cross-cultural manifestation and correlates of psychopathy are discussed.

Family representative payeeship and violence risk in severe mental illness

Although representative payeeship is prevalent among people with mental illness and shows promise to positively influence clinically relevant outcomes, research also suggests this legal mechanism could be implemented in ways that are problematic. The current study examined whether family representative payeeship was associated with elevated risk of family violence perpetrated by persons with severe mental illness (SMI). Data were collected every 4 months for 1 year in structured interviews with N = 245 persons with SMI who received disability benefits. Multivariate analyses showed that substance abuse, history of violence, frequency of family contact, and family representative payeeship were associated with elevated odds of family violence. Analyses also showed family contact and family representative payeeship had a cumulative effect on increasing the predicted probability of family violence (controlling for covariates such as violence history and substance abuse). The data shed light on the potential for family representative payeeship to be associated with increased risk of interpersonal conflict and violence in SMI.

A comparison of statistical models in predicting violence in psychotic illness

BACKGROUND

The application of statistical modeling techniques, including classification and regression trees, in the prediction of violence has increasingly received attention.

METHODS

The predictive performance of logistic regression and classification tree methods in predicting violence was explored in a sample of patients with psychotic illness.

RESULTS

Of 2 logistic regression models, the forward stepwise method produced a simpler model than the full model, but the latter performed better. The performance of the classification tree appeared to be high before cross-validation, but reduced when cross-validated. The standard logistic model was the most robust model. A simplified tree with extra weight given to violent cases was a reasonable competitor and was simple to apply.

CONCLUSION

Although classification trees can be suitable for routine clinical practice, because of the simplicity of their decision-making processes, their robustness and therefore clinical utility was problematic in this sample. Further research is required to compare such models in large prospective epidemiologic studies of other psychiatric populations.

Assessing violence risk in stalking cases: a regression tree approach

Advances in the field of risk assessment have highlighted the importance of developing and validating models for problematic or unique subgroups of individuals. Stalking offenders represent one such subgroup, where fears of and potential for violence are well-known and have important implications for safety management. The present study applies a Classification and Regression Tree (CART) approach to a sample of stalking offenders in order to help further the process of identifying and understanding risk assessment strategies. Data from 204 stalking offenders referred for psychiatric evaluation to a publicly-funded clinic were used to develop and assess putative risk factors. A series of nested models were used to generate tree algorithms predicting violence in this sample of offenders. Both simplified and more extensive models generated high levels of predictive accuracy that were roughly comparable to logistic regression models but much more straightforward to apply in clinical practice. Jack-knifed cross-validation analyses demonstrated considerable shrinkage in the CART, although the models were still comparable to many other actuarial risk assessment instruments. Logistic regression models were much more resilient to cross-validation, with relatively modest loss in predictive power.

Structural models of psychopathy

Psychopathy traditionally is defined by a cluster of inferred personality traits and socially deviant behaviors. The accepted standard for the reliable and valid assessment of psychopathy is the Psychopathy Checklist-Revised (PCL-R). Because of its importance in basic and applied research, and in the mental health and criminal justice systems, the PCL-R has been subjected to intense scrutiny by researchers and clinicians. In this article we discuss issues surrounding its structural properties and those of its derivatives. Using factor analysis, item response theory, and multidimensional scaling, we propose that the PCL-R and its derivatives are underpinned by at least four correlated factors: Interpersonal, Affective, Lifestyle, and Antisocial. We argue that attempts to characterize antisocial behaviors as merely "downstream" manifestations of more central traits are inconsistent with the structural properties of the PCL-R and with evidence that the development of traits and actions are interactive and reciprocal. We also report new evidence that psychopathy and its factors are dimensional in nature, perhaps extreme variants of normal personality traits and behaviors.

Twenty years of progress in violence risk assessment

Violence risk assessment has advanced considerably in the last 20 years. In the 1980s, leading professionals questioned the very possibility of valid violence risk assessments; now, many of the major risk factors have been identified, and professional debate focuses on how best to combine these risk factors into meaningful evaluations. An important contributor to this advance in knowledge has been the rise of meta-analytic reviews. Through quantitative summaries, the cumulative findings of small, potentially insignificant studies have provided important answers to questions concerning the effective assessment and treatment of violent offenders.

The state of contemporary risk assessment research

The focus on assessing dangerousness in routine psychiatric practice developed when relatively little was known about factors related to violence, and the accuracy of predicting violence was distinctly below chance. Since the 1990s, however, significant research attention has been directed toward factors related to violence and mental illness, as well as toward factors related to the accuracy of risk assessment techniques. Sociodemographic and environmental variables have been identified as significant predictors of violence, as has the presence of substance abuse. However, the data on specific mental health variables are somewhat mixed. Many studies point to a modest increased risk of violence associated with major mental illness and psychosis, whereas other noteworthy studies have failed to confirm such findings. Studies of the accuracy of risk assessments indicate that both actuarial and clinical methodologies perform better than chance, although the former achieve greater statistical accuracy. Despite ongoing controversies, risk management strategies that encompass the strengths and limitations of our present knowledge are available to clinicians.

Applying a forensic actuarial assessment (the Violence Risk Appraisal Guide) to nonforensic patients

The actuarial Violence Risk Appraisal Guide (VRAG) was developed for male offenders where it has shown excellent replicability in many new forensic samples using officially recorded outcomes. Clinicians also make decisions, however, about the risk of interpersonal violence posed by nonforensic psychiatric patients of both sexes. Could an actuarial risk assessment developed for male forensic populations be used for a broader clientele? We modified the VRAG to permit evaluation using data from the MacArthur Violence Risk Assessment Study that included nonforensic male and female patients and primarily self-reported violence. The modified VRAG yielded a large effect size in the prediction of dichotomous postdischarge severe violence over 20 and 50 weeks. Accuracy of VRAG predictions was unrelated to sex. The results provide evidence about the robustness of comprehensive actuarial risk assessments and the generality of the personal factors that underlie violent behavior.

Do the print media "hype" genetic research? A comparison of newspaper stories and peer-reviewed research papers

BACKGROUND

The public gets most of its information about genetic research from the media. It has been suggested that media representations may involve exaggeration, called "genohype." To examine the accuracy and nature of media coverage of genetic research, we reviewed the reporting of single-gene discoveries and associated technologies in major daily newspapers in Canada, the United States, Great Britain and Australia.

METHODS

We used neutral search terms to identify articles about gene discoveries and associated technologies hosted on the Dow Jones Interactive and Canadian NewsDisk databases from January 1995 to June 2001. We compared the contents, claims and conclusions of the scientific journal article with those of the associated newspaper article. Coders subjectively assigned the newspaper articles to 1 of 3 categories: moderately to highly exaggerated claims, slightly exaggerated claims or no exaggerated claims. We used classification tree software to identify the variables that contributed to the assignment of each newspaper article to 1 of the 3 categories: attention structure (positioning in the newspaper and length of the article), authorship, research topic, source of information other than the scientific paper, type and likelihood of risks and benefits, discussion of controversy, valuation tone (positive or negative), framing (e.g., description of research, celebration of progress, report of economic prospects or ethical perspective), technical accuracy (either omissions or errors that changed the description of the methods or interpretation of the results) and use of metaphors.

RESULTS

We examined 627 newspaper articles reporting on 111 papers published in 24 scientific and medical journals. Only 11% of the newspaper articles were categorized as having moderately to highly exaggerated claims; the majority were categorized as having no claims (63%) or slightly exaggerated claims (26%). The classification analysis ranked the reporting of risks as the most important variable in determining the categorization of newspaper articles. Only 15% of the newspaper articles and 5% of the scientific journal articles discussed costs or risks, whereas 97% of the newspaper articles and 98% of the scientific journal articles discussed the likelihood of benefits of the research.

INTERPRETATION

Our data suggest that the majority of newspaper articles accurately convey the results of and reflect the claims made in scientific journal articles. Our study also highlights an overemphasis on benefits and under-representation of risks in both scientific and newspaper articles. The cause and nature of this trend is uncertain.

Prediction of violence and self-harm in mentally disordered offenders: a prospective study of the efficacy of HCR-20, PCL-R, and psychiatric symptomatology

The efficacy of the Historical, Clinical, and Risk Management Scales (HCR-20; C. D. Webster, D. Eaves, K. S. Douglas, & A. Wintrup, 1995), Psychopathy Checklist-Revised (PCL-R; R. D. Hare, 1991), Beck Hopelessness Scale (BHS; A. T. Beck, A. Weissman, D. Lester, & L. Trexler, 1974), and Brief Psychiatric Rating Scale (BPRS) to predict violence and self-harm in 34 institutionalized mentally disordered offenders was assessed. Both the HCR-20 and BPRS were strong predictors of violence whereas the PCL-R had moderate predictive ability. BHS was the only variable predictive of self-harm. Although risk assessment measures were successful at predicting in-patient violence, a clinical measure of mental state was at least as effective in these mentally disordered offenders.

The emotional experience of intercourse and sexually transmitted diseases: a decision-tree analysis

BACKGROUND

Epidemiologic data document high risks for many sexually transmitted diseases (STDs) among US adolescents and young adults.

GOAL

This case-control study used decision trees to investigate the relationship between STD incidence and emotional reactions to intercourse.

STUDY DESIGN

For this study, 188 adolescents and young adults (mean age, 24.9 years [SD = 8.2]) at a regional public STD clinic completed a behavioral and psychological questionnaire and underwent a workup for STD.

RESULTS

The prevalence of STD in this group was 44.8%. Decision-tree analysis identified emotional reactions to intercourse that were associated with STD diagnosis for some patients: feeling good about oneself after sex half the time or less (OR = 3.21; 95% CI = 1.73-5.95), feeling comfortable during sex half the time or less (OR = 2.17; 95% CI = 1.07-4.40), and feeling angry after sex (OR = 1.90; 95% CI = 0.91-3.99). Findings of a logistic regression model of emotional reactions to intercourse were significant (chi-square = 24.6; df = 8; P < 0.002), but adding behavioral variables did not improve prediction.

CONCLUSIONS

For some of these young adults at the time of life when they are at highest risk of STD, emotional factors have higher odds ratios for STD diagnosis than the traditionally assessed behavioral variables. This underscores the need for interventions targeted to specific subgroups and for readily available mental health services.

Symbolic, relational, and ideological signifiers of bias-motivated offenders: toward a strategy of assessment

Developmental, ideological, and behavioral characteristics of 58 convicted hate crime offenders were examined. Ratings on the Psychopathy Checklist: Screening Version, HCR-20, and Bias Motivation Profile (BMP) were made via record review. Offense characteristics were rated on the Cornell Aggression Index and Cormier-Lang Crime Index. Results indicated that offenders with higher BMP scores engaged in more instrumental (i.e., premeditated) aggression and targeted racial-ethnic minority victims. Significant within-group variation in the prominence of offender bias motivation on the BMP was observed.

Applicability of traditional and revised models of psychopathy to the Psychopathy Checklist: screening version

Recently, psychopathy has become virtually synonymous with the Psychopathy Checklist (PCL) measures. However, the "gold standard" 2-factor model that underlines these measures has been questioned for its uncertain empirical support and emphasis on antisocial behavior that is not specific to psychopathic personality deviation. This study (N = 870 civil psychiatric patients) compares the fit of the traditional 2-factor model with that of a revised 3-factor model of psychopathy. The revised model better describes the structure of the Screening Version of the PCL (PCL:SV) than the traditional model. Although the revised model's exclusion of some items that assess antisocial behavior reduces the PCL:SV's power in predicting patient violence, this model arguably assesses psychopathy in a more specific, theoretically coherent fashion that may reduce misapplications of the construct. Implications for future research are discussed.

The role of theory in the assessment of sex offenders

No generally accepted theory of sex offender behavior exists at this point. As clinical experience and research findings interact within an evolving theoretical framework, the picture of what is important, as well as what is possible to know about a sex offender, is rapidly changing. It is vital that mental health professionals, the legal system, social service agencies, and other consumers and providers of sex offender assessments be aware of both what is possible to learn about an offender as well as the limitations on that knowledge and its application. Prominent theories that influence the assessment of offenders in North America and Europe are presented and examined. Uses and limitations of each theory are discussed. Several trends emerge which could influence the future of sex offender assessment, treatment, management, and policy.

The antecedents of aggressive behavior among men with schizophrenia: a prospective investigation of patients in community treatment

The association between psychotic symptoms and violence is unclear, due in part to methodological features of investigations that have examined this question, and in part to the fact that the association likely differs by disorder and treatment conditions. Using data from The Comparative Study of the Prevention of Crime and Violence by Mentally Ill Persons, we examined 128 men with schizophrenia or schizoaffective disorder discharged from general and forensic psychiatric hospitals in Canada, Finland, Germany, and Sweden. The association between symptoms and aggressive behavior was studied during two 6 month periods when the patients lived in the community. Severe positive and negative symptoms of psychosis, depression, and anxiety were measured at the beginning of each of the 6 month periods. In addition, at the beginning of the second 6 month period changes in symptoms in the previous period were indexed. Aggressive behavior was measured in each 6 month period by reports from patients and from collaterals. During the first 6 months post-discharge, after controlling for the presence of antisocial personality disorder or PCL score and past diagnoses of alcohol/drug abuse/dependence, the presence of a severe positive symptom significantly increased the risk of aggressive behavior. During the second 6 month period, after controlling for antisocial personality disorder or PCL score and self-reported alcohol/drug use, the presence of a severe positive symptom, a TCO symptom, and an increase in TCO symptoms significantly increased the risk of aggressive behavior. Neither depot medications nor obligatory community treatment reduced the risk of aggressive behavior after controlling for the presence of a severe positive symptom and/or TCO symptoms. These findings suggest that, among men with schizophrenia being treated in the community, the presence of severe psychotic symptoms and the development of TCO symptoms are antecedents of aggressive behavior.

Risk assessment in offenders with intellectual disability: the evidence base

A review of the current literature on risk assessment and management in offenders with intellectual disability (ID) revealed little direct evidence for the specific population. Theoretical models and non-ID populations have been abstracted and adapted, but not validated, for those with ID. The varying conceptual frameworks of risk, and its assessment and management, must be considered in context. Difficulties remain with the consideration of offences versus offence-like behaviour, offender versus those with similar needs, and indeed, what is regarded as 'intellectual disability'. Mainstream forensic assessment has moved towards a more dynamic appreciation of risk and risk management, as opposed to risk elimination. This development is more in line with the normalization principles of 'risk-taking' in ID. Consideration is given to future research and development priorities.

Legal issues in maximum security institutions for people with mental illness: liberty, security, and administrative discretion

This article explores four legal issues relevant to the provision of care in secure hospitals. These include the current status of right to treatment litigation; the potential impact of the Americans with Disabilities Act; new developments in laws governing restraint and seclusion; and the need for uniform institutional policies on risk assessment. These issues illustrate the potential conflicts between individual autonomy and institutional control that have been at the heart of mental health law for three decades. The article suggests that because of the diminishing oversight provided by the federal judiciary, institutional custodians have a particular obligation to ensure that individual rights are not overwhelmed by concerns with security.

Validity of the MMPI-2 basic and Harris-Lingoes subscales in a forensic sample

The present study explored the MMPI-2 correlates of criminal history in a sample of prison inmates. Although Ma was the only basic scale to demonstrate a significant positive association to crime history, three Harris-Lingoes subscales also bore significant associations to this criterion. The Ma1 (amorality), Pd2 (authority problems), and Pd5 (self-alienation) subscales demonstrated positive associations with crime history. Hierarchical multiple-regression analyses were used to examine the incremental validity of the D, Pd, and Ma Harris-Lingoes subscales when added to their parent scales in the prediction of crime history. Only the Pd Harris-Lingoes subscales demonstrated incremental validity in the prediction of this criterion. The value of including the Ma and Pd Harris-Lingoes subscales in criminal risk assessment is discussed.

Psychopathy and ethnicity: structural, item, and test generalizability of the Psychopathy Checklist--Revised (PCL-R) in Caucasian and African American Participants

The Psychopathy Checklist--Revised (PCL-R) is an important measure in both applied and research settings. Evidence for its validity is mostly derived from male Caucasian participants. PCL-R ratings of 359 Caucasian and 356 African American participants were compared using confirmatory factor analysis (CFA) and item response theory (IRT) analyses. Previous research has indicated that 13 items of the PCL-R can be described by a 3-factor hierarchical model. This model was replicated in this sample. No cross-group difference in factor structure could be found using CFA; the structure of psychopathy is the same in both groups. IRT methods indicated significant but small differences in the performance of 5 of the 20 PCL-R items. No significant differential test functioning was found, indicating that the item differences canceled each other out. It is concluded that the PCL-R can be used, in an unbiased way, with African American participants.

Clinical training and the duty to protect

Given that the duty to protect is now a well established clinical and legal expectation, training in professional psychology should assist students in developing conceptual models for violence risk assessment and management. This report presents a training model incorporating recent advancements in risk assessment (such as the assessment of psychopathy), the need for basic legal understanding, and knowledge of specific risk management strategies. Additionally, academic and internship training programs in Michigan were surveyed by telephone about current training patterns. Preliminary results indicated program strengths in general awareness and training in risk factors. However, the data also suggested a need to formalize a faculty role in risk education, improve the training of students in relevant legal information, and establish coherent conceptual models of dealing with potential patient violence. It is argued that improving training of future professionals in this way will improve clinical practice, reduce legal liability, and improve public safety.

The duty to warn and protect--impact on practice

OBJECTIVE

To discuss the concepts of the duty to protect and the associated threat to confidentiality and their impact on practice for Canadian psychiatrists.

METHOD

We review these concepts and provide a synthesis of legal cases impacting psychiatric practice.

CONCLUSION

The onus is on the psychiatrist to make him or herself aware of the current state of the legal obligation with respect to duty to protect. The evolving concept of duty to protect has and will continue to have significant impact on the practice of psychiatry.

Developing a clinically useful actuarial tool for assessing violence risk

BACKGROUND

A new actuarial method for violence risk assessment--the Iterative Classification Tree (ICT)--has become available. It has a high degree of accuracy but can be time and resource intensive to administer.

AIMS

To increase the clinical utility of the ICT method by restricting the risk factors used to generate the actuarial tool to those commonly available in hospital records or capable of being routinely assessed in clinical practice.

METHOD

A total of 939 male and female civil psychiatric patients between 18 and 40 years old were assessed on 106 risk factors in the hospital and monitored for violence to others during the first 20 weeks after discharge.

RESULTS

The ICT classified 72.6% of the sample as either low risk (less than half of the sample's base rate of violence) or high risk (more than twice the sample's base rate of violence).

CONCLUSIONS

A clinically useful actuarial method exists to assist in violence risk assessment.