Gender Not a Significant Moderator for Predicting Violence in Insanity Acquittees

Debbie Green, Farleigh Dickinson University
Melanie Shneider, Farleigh Dickinson University
Hali Griswold, Farleigh Dickinson University
Brian Belfi, Kirby Forensic Psychiatric Center
Michelle Herrera, Farleigh Dickinson University
Ashley DeBlasi, Blue Ridge Behavioral Healthcare

The current retrospective risk assessment study evaluated the use of the Historical-Clinical-Risk Management-20 Version 3 (HCR-20V3) in a sample of 100 male and 24 female insanity acquittees, comparing the presence of risk factors and its validity in assessing violence in a state forensic hospital across males and females. Over an average of 15.5 months, 44.4% of the total sample engaged in any act of violence within the institution, highlighting the need for effective risk assessment and management tools. An equivalent proportion of males and females engaged in violence (42.0% of males and 54.2% of females). Results indicated higher interrater reliability on scoring risk factors among males as compared to females, calling for future research into the role of item indicators across genders and possible differences in interpretations of scoring guidelines. Results indicated that females exhibited similar numerical ratings of presence risk factors as males, although they were rated higher on Past Problems in Relationships and Traumatic Experiences. The relationships between scale scores and violence were higher among males than females. However, gender was not a significant moderator in logistic regression analyses predicting likelihood of violence. Additional research is necessary to evaluate how ratings of risk factors are translated into summary risk estimates, opinions about dangerousness, and treatment interventions, with a particular focus on how such decision-making is influenced by the gender of examinees.

Gender, HCR-20, risk assessment, institutional aggression

“Although rarely explicitly stated, the risk assessment literature presumes that tools should work equally well for offenders and patients, regardless of gender. However, some researchers suggest that a gendered perspective of violence risk is more appropriate (Garcia-Mansilla, Rosenfeld, & Nicholls, 2009). This perspective proposes that the presence and salience of risk factors and, therefore, risk prediction tools are likely to vary as a function o fgender. Indeed, research suggests that the manifestation of violence is moderated by gender (Kellermann & Mercy, 1992; Strand & Belfrage, 2001). Further, although certain risk factors may be shared among men and women, some may be more prevalent in one gender over the other or more influential with regard to risk in one gender (Krakowski & Czobor, 2014; McKeown, 2010; Sadeh, Javdani, Finy, & Verona, 2011)” (p. 48).

“In a more recent review of 15 studies examining the predictive validity of 12 risk assessment tools among female samples of offenders, only four measures demonstrated adequate predictive efficacy, and none of the instruments did so with a moderate to large degree of accuracy (Geraghty & Woodhams, 2015). One of the risk assessment tools that demonstrated small predictive accuracy with females was the HCR-20” (p. 49).

However, “the predictive validity of the HCR-20 in female samples has yielded mixed findings, in part due to sample characteristics and the context of violence studied” (p. 50).

“The HCR-20 was designed to inform release decisions. Further, evaluations of insanity acquittees involve highly contextualized risk assessment, prediction, and management (Douglas, 2014). Therefore, previous research suggests that the HCR-20V3 would be appropriate for use with institutionalized insanity acquittees, although there is not yet published research within this population and setting” (p. 50).

“The current study evaluated the HCR-20V3 in a sample of female and male forensic patients adjudicated Not Guilty by Reason of Mental Disease or Defect (more commonly referred to as Not Guilty by Reason of Insanity; NGRI), comparing the presence of risk factors among females and males and assessing the relation of risk factors to institutional violence” (p. 51).

Participants consisted of NGRI acquittees who were admitted to a state forensic hospital and found to have a dangerous mental disorder during the initial evaluation. Graduate students used the HCR-20V3 to code archival forensic reports for each participant. “The HCR-20V3 consists of 20 risk factors, including 10 historical, five current clinical, and five anticipated future risk factors” (p. 53). No changes were made to the measure for the study. Additionally, “risk management items were rated based on consideration of placement in a civil psychiatric setting, consistent with the step-down process that occurs with NGRI patients” (p. 52).

“The START outcome scale (SOS) was used to record incidents of violence. In the current study, violence was defined consistently with that of the HCR-20V3” (p. 53).

Results

Interrater reliability for the scoring of the HCR-20V3 tended to be higher in ratings of male patients as compared to female patients. “Females scored higher than males in Problems in Relationships, Problems with Traumatic Experience on the HCR-20V3…Men, however were rated higher than women on Violent Attitudes, and Recent Problems with Violent Ideation and the Clinical scales. Additionally, scores on the Historical and Clinical scales, and Total score were significantly associated with outcome for males” (p. 56).

Gender, and the interaction between the Clinical Scale and gender emerged as significant. Odds of violence were higher in females than males. There were common risk factors found in females and males. “Females were less likely to be arrested, more likely than males to be married, suggesting stability and social support. However, the higher presence ratings for H3 (Problems with Relationships) in women as compared to men indicate that such relationships may be characterized by chaos, aggression, and victimization. The diagnoses of patients significantly differed by gender; the rate of psychotic disorders in males exceeded that of females and, conversely, the rate of mood disorders was higher in females than males…Males were more likely to be diagnosed with Antisocial Personality Disorder, whereas the most common personality disorder diagnosis among females was Borderline Personality Disorder” (p. 59).

“Multivariate logistic regression analyses indicated that the model comprised of the Historical, Clinical, and Risk Management scales predicted classification of any violence” (p. 58).

Although there were some significant differences in the association between gender and item ratings for each scale, “results of the current study support the utility of using the HCR-20V3 to predict institutional violence in females” (P. 60). Any differences between gender and item and scale association should be addressed in treatment. Further, the HCR-20V3 Clinical Scale was useful in predicting future violence, and gender did not moderate the results.

“Future research would also benefit from investigating the contribution of individual indicators to the coding of risk factors. Further, study of how evaluators use risk information derived from the HCR-20V3 to inform opinions about the ultimate issue in evaluations of NGRI acquittees is warranted. Specifically, does an individual present a future risk of danger such that confinement is necessary? Furthermore, research is needed to determine how the HCR-20V3 is used in practice to inform management strategies (Douglas et al., 2014) and how such approaches may differ by gender. For example, a qualitative study comparing treatment plans (either through direct review of such plans or through interviews of clinicians) implemented prior to and after the incorporation of the HCR-20V3 into a clinical setting may help to highlight the effect of this risk assessment procedure” (p. 61).

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