Risk Assessments Utilizing Dynamic Scores Inform the Management and Reduction of Inpatient Aggression

Neil R. Hogan, University of Saskatchewan
Mark E. Olver, University of Saskatchewan

The present study examined risk for inpatient aggression, including treatment-related changes in risk, using a battery of 5 forensic instruments. The relative contributions of different types of risk factors to the assessment of risk for inpatient outcomes were also assessed. The Historical-Clinical-Risk Management-20V3, Short-Term Assessment of Risk and Treatability, Violence Risk Scale, Violence Risk Appraisal Guide–Revised, and Psychopathy Checklist–Revised were rated from archival information sources on a sample of 99 adult forensic inpatients from a Canadian psychiatric hospital. Pretreatment and posttreatment ratings were obtained on all dynamic study measures; associations between risk and change ratings with inpatient aggression were examined. Significant pretreatment-posttreatment differences were found on the HCR-20V3, START, and VRS; pretreatment scores on these measures each demonstrated predictive accuracy for inpatient aggression (AUC .68 to .76) whereas the PCL-R and VRAG-R did not. HCR-20V3, VRS, and START dynamic scores demonstrated incremental predictive validity for inpatient aggression to varying degrees after controlling for static risk factors. Dynamic change scores from these 3 measures also demonstrated incremental concurrent associations with reductions in inpatient aggression after controlling for baseline risk. Several instruments demonstrated predictive validity for inpatient aggression and clinical/dynamic risk and change scores had unique associations with this outcome. The present findings suggest that risk assessments using the HCR-20V3, START, and VRS may inform the management and reduction of inpatient aggression, as well as assessments of dynamic risk more generally.

dynamic risk factors, institutional aggression, risk assessment

“Forensic hospitals face unique issues, functioning as part of both criminal justice and mental health systems…many individuals are admitted for violent index offenses committed in the community, and as a result, these institutions may house high risk violent offenders alongside vulnerable populations, including the acutely mentally ill. As an indication of the extent of the problem, researchers have observed base rates of violence ranging from 34% to 74% over one year follow-ups. Clearly, these programs require both effective violence-reduction interventions and accurate risk assessment practices” (p.233).

“A meta-analysis of the predictors of forensic inpatient violence found that the Historical-Clinical-Risk Management-20 (HCR-20) and the Hare Psychopathy Checklist measures (PCL-R and PCL:SV) produced the strongest results (mean r values of .33 and .26, respectively). Subsequent studies have supported the predictive validity of these and other tools for inpatient violence, such as the Short-Term Assessment of Risk and Treatability (START), across various follow-up periods. Nonetheless, the research base remains scant” (p.234)

“Rigorous evaluations of dynamic risk factors, that is, the extent to which changes over two or more time points are associated with changes in an outcome such as recidivism, are rare. This represents an area of need and significant interest in the research literature; these kinds of evaluations can identify causal dynamic risk factors that are relevant to effective risk management and treatment decisions… The research is still limited, but studies of community recidivism have demonstrated that changes on tools like the Violence Risk Scale (VRS) and the HCR-20 may provide incremental predictive validity controlling for baseline risk” (p.234).

“This study was designed to evaluate whether existing instruments, including the recently updated HCR-20^V3 and Revised Violence Risk Appraisal Guide (VRAG-R) are applicable to the assessment of risk for inpatient aggression. Instruments were selected for the following reasons: to build on the results of a previous meta-analytic study (i.e., the HCR-20 scheme and PCL-R) and to investigate other existing static (i.e., the VRAG-R) and dynamic violence risk instruments (i.e., VRS and START) of interest to the program in which the study took place. Assessing the generalizability of tools validated and used for other purposes was also an explicit aim of the study” (p.234)

Participants consisted of 99 adult forensic psychiatric inpatients admitted to a maximum security psychiatric hospital in Western Canada. The majority of the sample were white males with a mean age of 36.7 years. Half of the participants (52%) were found NCR [Not Criminally Responsible because of a Mental Disorder] and one quarter (24%) were convicted offenders admitted for mental health services. Schizophrenia/psychotic disorders and substance abuse were the most common diagnoses (61% and 70%, respectively). Most of the participants had a previous criminal charge (68%) or committed a violent index offense (70%), while more than half of participants had a documented history of institutional aggression (54%). Treatment services were offered in the psychiatric hospital as a part of a larger forensic mental health program that conducts assessments and offers treatment to those who have come in contact with the criminal justice system. “Treatment services were consistent with a general biopsychosocial model, including medications, counseling, psychoeducational groups (e.g., anger management), and institutional and community-based community reintegration programming (e.g., supervised employment/recreational activities)” (p.235).

In addition to using the HCR-20^v3, PCL-R, START, VRAG-R, and VRS, the Staff Observation Aggression Scale-Revised (SOAS-R) was also rated by trained coders who used archival data. Once pretreatment scores were completed, no later than 45 days after participant admission, coders used the participant’s clinical file to compile a posttreatment score on the HCR-20^v3, VRS, and START. The SOAS-R was then computed by scanning each participant’s file for evidence of inpatient aggression during the index admission. Only standard clinical notes and reports were used to operationally identify inpatient aggression; thus, not all elements of the SOAS-R were fulfilled.

Results

“The results of ROC analyses lend some empirical support to the argument that structured risk instruments designed to capture dynamic/clinical variables are particularly well suited to the assessment of risk for inpatient aggression. Measures designed to capture these types of variables, such as the HCR-20^V3 Clinical scale, produced the largest predictive effects, whereas measures that were not designed to capture these variables, such as the HCR-20^V3 Historical scale and the VRAG-R, produced smaller effects” (p. 238).

“Other tools also performed well. The START is theoretically well-suited to the assessment of risk for inpatient aggression and the current results provided empirical support for both the Vulnerability and Strength total scores, consistent with extant research. The VRS also demonstrated predictive validity for inpatient aggression, with the Dynamic scores demonstrating greater prediction than the Static scores, consistent with the limited previous research using the VRS to assess risk for inpatient outcomes. By contrast, more static measures fared more poorly in the prediction of aggression. The VRAG-R did not significantly predict inpatient aggression which stands in contrast to similar past research with its predecessor. PCL-R total scores were similarly not significantly predictive of inpatient aggression, although this may have been influenced by statistical power limitations, given that the observed AUC of .63 is highly consistent with past research and typically would attain significance with sufficient n” (p. 240).

“When it came down to detecting dynamic treatment-related changes in risk during forensic psychiatric admissions, pretreatment–posttreatment analyses demonstrated that the dynamic study instruments appeared to be capable of detecting changes in risk. In particular, scales capturing more clinical variables (e.g., mental health symptomatology) demonstrated the greatest change effects, consistent with the nature of the setting (a psychiatric hospital)” (p. 239).

“In this study, a number of new and more established instruments received support for their predictive validity with regard to aggression among forensic psychiatric patients. The results indicated that risk factors that have been theoretically identified as clinical/dynamic variables were incrementally predictive of inpatient aggression, beyond static factors. Additionally, support for the dynamic nature of the risk measures was provided through an examination of concurrent associations with inpatient aggression. Although the present findings require replication and extension to other patient samples and settings, these preliminary results may inform forensic mental health professionals in the selection of risk instruments based on the particular goals and intended applications of individual assessments. For professionals, the consideration of dynamic risk factors is indicated, with an eye toward improved forensic mental health services and concordant reductions in the risks and harms associated with inpatient aggression” (p.242).

“The results of the HCR-20^V3 analyses also provided some support for the addition of Relevance scores to this instrument. Indeed, although they were not compared directly, the superior performance of HCR-20^V3 Relevance scores over Presence scores may nonetheless reflect a potentially important distinction, between reliable, clinically and practically relevant changes, and transitory fluctuations or less meaningful variability in risk factors” (p.240).

“This investigation addressed an understudied behavioral issue with serious consequences for patients and professionals. By including consecutive admissions as participants, the sample can be considered highly representative of the population served by the program, and likely also of similar institutions” (p.241).

As always, please join the discussion below if you have thoughts or comments to add!