The main finding of this study is that psychoses, mood disorders and personality disorders have different patterns of violent experiences when combined with age and gender.
Overall, in our sample, people with personality disorders showed to experience the highest levels of violence, and in particular reported significant differences compared to mood disorders in adulthood; young age is more implicated than the adult age in both perpetration and victimization and females are more victimized than males, both in childhood and adulthood, whereas males engaged more often in violent behaviors than females in early life. When we explored the episodes of violence victimization and perpetration that occurred in childhood and adulthood, a distinct diagnosis-age pattern did emerge for the three diagnostic entities, with mood disorders showing a strong victimization pattern, personality disorders a strong perpetration pattern and psychoses a less defined pattern. For people affected by a mood disorder, being victimized in childhood is positively correlated to being victimized in adulthood. Moreover, victimization was positively correlated with perpetration in adults with mood disorders, based on a medium “victimization-perpetration pattern”. This finding is in line with the evidence that victimized subjects are more likely to engage in violent events [7, 12]. On the contrary, in personality disorders, engaging in violent behavior in childhood is positively and strongly correlated to violent acts perpetrated in adulthood. People with psychoses showed weaker patterns of victimization (from childhood to adulthood) and victimization-perpetration (both childhood and adulthood) than other patients (see Figure 1).
Our findings on the diagnosis-gender pattern show that male subjects suffering from a personality disorder had higher scores than the other patients. We could claim that in these subjects a cluster of risk factors interact with each other (i.e. gender, marital status, diagnosis, substance misuse) increasing exponentially the risk of violence.
According to a more holistic perspective, our data relative to the gender-diagnosis interaction allows us to do some comments. First, the absence of the interaction in childhood was expected: a psychiatric condition might not have developed or be pervasive yet at that time. Second, gender and diagnosis had an interaction effect only on expression of violence in adulthood, that was the only subscale not to show a gender difference, although showed differences among diagnostic groups. This finding could mean that a specific SMI in a specific gender might have a more pervasive impact on the perpetration of violence, not on victimization. This data should be interpreted carefully given its inherent methodological shortcomings.
Males got higher scores in used violence subscales than females with the same diagnosis and in the same period of life; on the contrary, females got the highest scores on victimization subscales. This pattern did not repeat for “used violence as an adult” and “victim of violence in childhood”: in the first case, females with mood disorders reported greater expression of violence than males; in the second case, female victims of violence in childhood with a mood disorder got higher scores than females with psychoses. Moreover, males with psychoses reported higher levels of victimization in childhood then females. Although the results are not statistically significant in most cases, they and their graphic representations provide stimulating cues. For example, mood disorders showed to have a trend of a lower involvement in episodes of violence, in particular males; males and females with personality disorders showed a trend of greater involvement in the expression of and in the exposure to violence, respectively (see Figure 2).
Our findings regarding specifically the three risk factors taken into account (i.e. age, gender and diagnosis) are globally consistent with literature. We found that people with a personality disorder got the highest scores in KIVS subscales of perpetration and victimization, both in childhood and adulthood, followed by people with psychoses and then with mood disorders. A recent large American population-based study confirms that personality disorders show higher odds for violence , compared to other DSM 5 psychiatric disorders. Cluster B and paranoid personality disorders are considered mostly linked to violent offending and aggression [41, 42]. They are also more frequently prone to suicidal behaviors and criminal arrest . Personality disorder diagnosis increases the probability of violence perpetration, likely linked to their intrinsic impulsiveness, substance abuse and bio-psychological mechanisms [8, 44]. As regards major psychoses, only a modest relation with violence has been found in several large population-based studies ; regarding mood disorders, some evidence suggests an increased risk of violence in particular those with bipolar disorder .
Our results regarding the correlation between age and violence are consistent with other researches [2, 6, 12, 15, 17, 19]. Using the KIVS, which distinguishes exposure to and expression of violence in childhood and adulthood, we demonstrated the strong, inverse correlation between age and both perpetration and victimization. Further studies are necessary to understand if the young are actually more violent than older people, or they merely tend to remember more often or disclose more freely violent episodes.
With regard to gender, the results concerning the victimization pattern are in line with the literature: females are the main victims of IV [7, 9, 12, 22, 25–27, 29]. When broken down by diagnosis, the exposure to violence from childhood to adulthood, was greater in females just for mood disorders and not for psychoses or personality disorders. This finding is somewhat in line with that studies which more frequently found an association between victimization and diagnosis of mood disorders in females [6, 9, 27]. As regards perpetration of violence, we confirm that males are more likely than females to bully the peers and act aggressively in childhood [19, 31, 35, 47, 48]. On the other hand, we did not find a difference in expression of violence as an adult between genders, contrary to previous studies. This data could be due to several reasons, as the small sample or the different methodology. The largest epidemiological studies on the issue found that males commit acts of violence at greater rates than females [6, 14, 15, 49]; just the National Institute of Mental Health Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE), which investigated violent outcomes in schizophrenia patients as part of a large multisite randomized clinical trial, found an association between female, not male gender, and minor violence . A paucity of other studies came to different conclusions: Gillies & Brien  stated in their review that there is no clear pattern as to whether male or female psychiatric patients are more likely to act violently, due to inconsistent results from literature; Hamberger  conducted a “gender analysis” about intimate partner violence and stated that males and females are equal in terms of “frequency”, but not in terms of severity or initiative, of violent behaviors; lastly, Desmarais and colleagues , by pooling from five studies of adults with mental illnesses from across the United States, found instead that women reported significantly higher rates of violence perpetration than men, but commented that this finding might reflect, for women in the community, increased opportunity of being violent or the more likelihood to disclose violence-related experiences.
Overall, these findings might suggest that people with personality disorders are at greater risk for perpetuating cycles of violence perpetration, whereas people with mood disorders are more predisposed to be a victim across the life span. In another perspective, we could argue that being a victim of violence in early life plays a role in developing a mood disorder and predisposes to revictimization. Regards to schizophrenic spectrum, results are more ambiguous, so no comment can be made regarding any specific pathway or pattern of violence.
Although somehow in line with previous literature [5, 6, 9, 41, 44] future research comparing diagnoses, gender and age impact, including larger size samples and using more sophisticated methodological approaches are warranted.
The cross-sectional nature of the study design is a limitation. A study aiming to research the cycle of violence must be non-experimental, thus having to overcome multiple methodological obstacles in order to reach useful results . The observational approach we used, however, makes impossible to determine any causal association between the variables of interest. As a matter of fact, we could not affirm that suffering from a SMI represents a condition favoring experiences of violence.
Additionally, this study failed to include any control or norm groups. Furthermore, the sample is relatively small. The most of participants belonged to the same ethnicity and socioeconomic class, so the chance that interethnic or social differences played a meaningful role was limited.
Our study solely based on KIVS, whose ratings was not compared with other measurements, including for example self-report questionnaires. Higher rates in self-reports of both victimization and perpetration of violence have been reported in comparison to clinical interview , so, in this study, participants’ ratings could depend on their openness to disclose their experiences to others, not on the truth of the facts. Moreover, we did not perform an interrater reliability analysis of this clinician-administered interview, so we cannot exclude inhomogeneity in the ratings given by various clinicians. Finally, previous researches based on KIVS failed to consider issues of ethnic diversity since the instrument was administered just to Sweden and Italian populations; therefore, the findings reported in this study cannot be generalized.
Furthermore, we conducted the research in an inpatients psychiatric unit; therefore, the findings reported in this study cannot be generalized to other clinical or outpatient samples likely affected by less severe disorders. Circumstances related to hospital admission may have rekindled memories of past experiences of violence, above all in early life, and may have influenced the patients’ report.