The current research utilizes a multi-level analysis of historical, clinical, situational and neighbourhood factors to predict violence severity among persons with major mental illness. In addition, it draws on the typologies of offenders proposed by Moffitt (1993) and Hodgins and Janson (2002) to explore whether different predictors of violence severity exist for early-start, persistent offenders and late-start offenders. Finally, it compares early-start and late-start offenders with major mental illness to determine if differences exist in their criminal history, clinical presentation, motive for violence, crime-scene behaviours and neighbourhood backgrounds. A retrospective chart review of a mental health court support program in Toronto, Canada is utilized to explore the correlates of violence severity. Clinical charts and supplemental arrest records are content analyzed to extract data on arrestee/offender characteristics and on crime scene behaviours and tract-level data from the 2001 Canada Census is used to identify structural features of the neighbourhood environment of arrestees/offenders at the time of their arrest. Violence severity is measured using the Cormier-Lang System of Quantifying Criminal History (Quinsey, Harris, Rice, & Cormier, 1998). In total 1806 charts were reviewed and 245 subjects were subsequently included within the analyses. Using a variety of analytic techniques, the following results were obtained: 1) offense characteristics such as victim gender, victim-offender relationship, instrumental motive, and use of a weapon were the most robust predictors of violence severity while clinical factors such as diagnosis and comorbid clinical conditions were marginally significant predictors and historical factors such as previous violence and early-start offending were not significant predictors of violence severity; 2) context-specific measures accounted for more of the explained variation in violence severity than did individual-specific measures; 3) early-start and late-start offenders did differ with respect to history of violence, presence of a comorbid clinical condition such as a personality disorder or substance abuse and current life circumstances. Implications for theory refinement, clinical practice and program development are discussed and future avenues of research are considered.

The current research utilizes a multi-level analysis of historical, clinical, situational and neighbourhood factors to predict violence severity among persons with major mental illness. In addition, it draws on the typologies of offenders proposed by Moffitt (1993) and Hodgins and Janson (2002) to explore whether different predictors of violence severity exist for early-start, persistent offenders and late-start offenders. Finally, it compares early-start and late-start offenders with major mental illness to determine if differences exist in their criminal history, clinical presentation, motive for violence, crime-scene behaviours and neighbourhood backgrounds. A retrospective chart review of a mental health court support program in Toronto, Canada is utilized to explore the correlates of violence severity. Clinical charts and supplemental arrest records are content analyzed to extract data on arrestee/offender characteristics and on crime scene behaviours and tract-level data from the 2001 Canada Census is used to identify structural features of the neighbourhood environment of arrestees/offenders at the time of their arrest. Violence severity is measured using the Cormier-Lang System of Quantifying Criminal History (Quinsey, Harris, Rice, & Cormier, 1998). In total 1806 charts were reviewed and 245 subjects were subsequently included within the analyses. Using a variety of analytic techniques, the following results were obtained: 1) offense characteristics such as victim gender, victim-offender relationship, instrumental motive, and use of a weapon were the most robust predictors of violence severity while clinical factors such as diagnosis and comorbid clinical conditions were marginally significant predictors and historical factors such as previous violence and early-start offending were not significant predictors of violence severity; 2) context-specific measures accounted for more of the explained variation in violence severity than did individual-specific measures; 3) early-start and late-start offenders did differ with respect to history of violence, presence of a comorbid clinical condition such as a personality disorder or substance abuse and current life circumstances. Implications for theory refinement, clinical practice and program development are discussed and future avenues of research are considered.