Structured professional judgment of violence risk in forensic clinical practice: A prospective study into the predictive validity of the Dutch HCR-20

Abstract

In this prospective study, the Dutch version of the HCR-20 (an instrument assessing risk factors for violence in the past, present and future) was coded independently by three rater groups (researchers, treatment supervisors and group leaders) for 127 male mentally disordered offenders admitted to a forensic psychiatric hospital. During case conferences, the three raters discussed their ratings and reached consensus on their ratings and final risk judgment. HCR-20 ratings were related to incidents of physical violence during treatment. Overall, the predictive validity of the HCR-20 was good. We found no differences between researchers and treatment supervisors in predictive accuracy. Group leaders performed worse compared to the other two rater groups. The consensus rating was the best predictor. Implications for structured violence risk assessment in clinical practice are discussed.

Keywords:

• Risk assessment
• violence
• HCR-20
• predictive validity

The authors thank all clinicians and researchers who participated in this study. Special thanks go to Cécile Vandeputte-van de Vijver who functioned as workshop trainer together with the first author and participated as a researcher in the study.

Notes

1. See Douglas, Cox, and Webster (1999) and Otto (2000) for a more detailed overview of risk assessment approaches.

2. This sample comprised 53 men and seven women. In the present study, the women were excluded from the analyses (see Procedure).

3. In Dutch forensic psychiatry, cluster B personality disorders are the most prevalent (see Hildebrand & de Ruiter, 2004; de Ruiter & Greeven, 2000).

4. Axis I diagnoses were lifetime clinical diagnoses based on consensus between four raters (see Hildebrand & de Ruiter, 2004), Axis II disorders were diagnosed with the Structured Interview for DSM-IV Personality (SIDP-IV; Pfohl, Blum, & Zimmerman, 1995).

5. This is not completely true, Historical items can change in an unfavorable direction. For instance, the score on item 10 increases when a patient violates the rules by escaping from the secure hospital.

6. Results of the SVR-20 are not included in this study but can be expected within 1 or 2 years.

7. The HCR-20 was developed based on research in predominantly male samples.

8. Five of seven patients that were assessed for the context Out had an incident of physical violence inside the hospital. Four of these patients were in the transmural phase. Although these patients lived outside the hospital, they frequently visited the hospital, for instance, to attend work or psychotherapy.

9. It should be noted that ROC analyses are less appropriate to apply with dichotomous or trichotomous variables. Still, we believed it was important to examine the predictive accuracy per item, and AUC values are easy to understand and provide comparison values with other similar studies.