BackgroundIndividuals with severe mental illness (SMI) may be considered violent, but in most cases they are victims. Substance abuse, a clinical history of violence and gender are among the main risk factors for aggressive behavior. To date, evidence on the mechanisms involved in violent behavior in SMI is still scarce and controversial. From a neuropsychological point of view, the deficits in executive functions, may be a key element in increasing propensity to violence.MethodsSample: 50 offender SMI patients recruited at the Residence for the Execution of Security Measures (REMS) of Castiglione d/S (case group) and 35 non-offender SMI patients recruited at the Department of Mental Health and Addictive Disorders of Brescia (control group). Match 1: 1 by gender, education and main diagnosis. An interview was given to collect the anamnestic data. All patients were assessed with clinical (CGI-S, PANSS-EC), neurocognitive (TMT, SCWT, BACS), social cognitive (FEIT), psychosocial functioning / wellness (MCH-SF, PSP), impulsiveness (BIS-11, IGT) and aggressive behaviour risk assessment (MOAS, HCR-20, PCL-R) measures.ResultsThe two groups appeared to be significantly different as far as the total scores for both HCR-20 and PCL-R, with REMS patients have significant higher values. HCR-20 and PCL-R showed a significant predictive value independent of each other, in determining membership of the case group or control group. In particular, each additional point in the HCR-20 increased the probability of belonging to the case group of 1.8, while each additional point at the PCL-R increased this probability by 1.3. Significant differences were observed regarding the CGI-S and the PANSS-EC scores, where REMS patients had, at the same diagnosis, more severe psychopathology. PANSS-EC, BACS token test and SCWT errors showed a significant independent predictive value in HCR-20 score. In particular, for each additional point at the PANSS-EC, the HCR-20 increased 1 point, while this score is reduced by 0.16 for each additional point to the BACS token test and 0.42 for each additional point to the SCWT errors. Considering the cognitive profile, poorer performances were found in the offenders patients in the TMT-A and the BACS token exercise, indicating greater deficits in both visuomotor speed and attention. Regarding school performance (a possible proxy of cognitive reserve), the two groups appeared significantly different for the number of failures (p<0.012), which is higher among REMS patients. Friend relationships were more associated with PCL-R factor 2 (p<0.05). HCR-20 showed a strong association with both PCL-R factors and the total score (p<0.001), as well as CGI-S (p<0.005) and PSP (subscale A) (p<0.001). FEIT (happiness), friend relationships, PSP (subscale A), MHC-SF (social and psychological subscales) were independently associated with the PCL-R total score. The difference regarding the use of hashish lifetime, which is greater in the group of cases, appeared significant.DiscussionResults of the present study highlight that offenders patients with SMI have higher levels of clinical, cognitive and friends relationships severity, but better social cognition skills and a higher degree of perceived well-being. The greatest number of failures in REMS patients could be due to a greater impairment of cognitive function from an early stage of the disease. Even if the majority of crimes is carried out against the family members, family relationships do not present differences between the two groups. Although preliminary, these results could help clinicians to better understand offenders patients with SMI and to identify more homogeneous subgroups of patients, in order to plan more tailored care pathways.