Assessing Risk for Aggression in a Forensic Psychiatric Hospital Using the Level of Service Inventory-Revised: Screening Version

Abstract

The assessment of risk for aggression has become a central concern of many mental health services. Some structured schemes for guiding risk assessment, developed primarily to assess risk in psychiatric patients discharged from hospital and offenders being considered for release from prison, have recently been trialled in the inpatient setting. Against the background of some prior research demonstrating the potential benefits of structured assessment schemes, and some concerns about the practicality of these schemes for assessing risk in inpatients, the current research was initiated to assess the effectiveness of the Level of Service Inventory-Revised: Screening Version (LSI-R: SV) in the prediction of aggression within a forensic psychiatric hospital. Results showed that only a weak association between the LSI-R: SV total score and inpatient aggression existed. This is probably because the LSI-R: SV addresses more static variables and does not assess the clinical characteristics and recent hostility that is associated with aggression in psychiatric patients. The predictive validity of the LSI-R: SV in this context is therefore limited. There might be more effective schemes, some of which were designed specifically for the purpose of assessing inpatient aggression that should be used in preference to the LSI-R: SV.

Factors associated with seclusion in a statewide forensic psychiatric service in Australia over a 2‐year period

“…The LSI-R: SV has previously been shown to be only modestly accurate at predicting psychiatric inpatient aggression and/or violence (Daffern et al 2005). The findings presented here, that the LSI-R: SV has moderately good predictive accuracy in relation to predicting seclusion, are therefore noteworthy and in line with previous research in correctional settings and our experimental hypothesis.…”

“…The LSI‐R: SV has previously been shown to be only modestly accurate at predicting psychiatric inpatient aggression and/or violence (Daffern et al. 2005).…”

“…For example, a study appraising the accuracy of five commonly-used risk assessment instruments with incarcerated (non-mentally disordered) offenders found the LSI-R to be moderately predictive of minor (area under the curve (AUC) = 0.609) and major (AUC = 0.669) institutional misconduct (Kroner & Mills 2001). More specific to mental health services, a study that assessed the predictive accuracy of the short screening version of the LSI-R (the LSI-R: SV) in an Australian forensic psychiatric inpatient population suggested a weaker association with inpatient violence, with an AUC of 0.59 (Daffern et al 2005). Daffern et al (2005) further suggested that this less than favourable level of predictive accuracy might be attributable to the high base rates of certain of the LSI-R: SV items and aggressive behaviour in forensic psychiatry services.…”

“…More specific to mental health services, a study that assessed the predictive accuracy of the short screening version of the LSI-R (the LSI-R: SV) in an Australian forensic psychiatric inpatient population suggested a weaker association with inpatient violence, with an AUC of 0.59 (Daffern et al 2005). Daffern et al (2005) further suggested that this less than favourable level of predictive accuracy might be attributable to the high base rates of certain of the LSI-R: SV items and aggressive behaviour in forensic psychiatry services. Despite this, they and other authors still favour its continued use in inpatient settings, as well as upon release where it has shown considerable promise (Ferguson et al in press).…”

“…2005). Daffern et al. (2005) further suggested that this less than favourable level of predictive accuracy might be attributable to the high base rates of certain of the LSI‐R: SV items and aggressive behaviour in forensic psychiatry services.…”