2 Violence and Persons with Psychiatric Disorders: The Influencef oSocial Networks and Social Support It [living at home] was awful. I mean my mother was always drunk. She was screaming and yelling. She never verbally or physically abused me. I mean, I beat my mother up, and that's why people get scared of me. But I didn't know what else to do, and I was so angry at her. I mean, when she got sober I told her one time, I said, 'I wish to God you'd get drunk.' And now going to Al-Anon I understand that you're so addicted to the chaos that it's easier to live with. [I told my therapist] I don't like it. It's calm at home. I go home. There's nobody there bothering me...I don't miss the beating (from current boyfriend). I miss it being all mixed up and stirred up. Introduction This excerpt from our interviews with psychiatric patients illustrates graphically that there are complex interpersonal and intra-familial dynamics to consider along with symptoms if we want to understand violence in their lives. This paper explores the usefulness of thinking about violence and mental illness contextually, as an interpersonal issue, rather than as a primarily neurological or clinical concern.1 In the view taken here, violence and the risk for violence are best assessed by investigating what kinds of people in what kinds of situations, in what kinds of social networks, with what qualities of social relations, and at what phase of their lives and illnesses are likely to engage in dangerous behaviors towards whom.2-4 Social support characteristics and social network members constitute risk factors for violence by creating the opportunity or need for violence, or serve to prevent or decrease the opportunity and need for defensive, threatening, or assaultive behaviors. Many individuals with enduring and serious psychiatric disorders live with or depend substantially upon their families, primarily parents and siblings.5 Child sexual abuse and physical violence in families in general are unfortunately widespread.6 Yet seldom have family violence researchers focused on the extent, dynamics, or nature of violence in families where someone is seriously mentally ill. Equally rarely have researchers who study violence among mentally ill persons considered this as a type of family or domestic violence. As a result, we lack empirical studies of violence among individuals with major psychiatric disorders that include relatives and friends as the interpersonal context within which violence occurs, is identified, and responded to.7,8 The findings reported here are from an exploratory study of social network and social support factors related to violence among a cohort of 169 seriously mentally ill persons described in detail elsewhere.9,10 The main questions were: 1) What were the clinical, demographic, and social network/social support characteristics of patients who committed violent acts or threats; 2) Towards whom was the violence directed; and, 3) What was the nature and quality of the relationships between the patients and their most signficant others, particularly those who were targets of violent acts or threats? The investigation is one of very few that focuses on contextual factors related to violence outside of hospitals, with extensive data from both patients and relatives, and that uses multivariate anaysis to examine the risk for violent acts and threats separately. Face to face interviews were conducted with respondents at 6 month intervals over an 18 month period, along with retrospective reviews of hospital charts and court records, and a one- time interview with their self-identified most significant other. Social Networks, Social Support, and Violence Among Persons with Severe, Persistent Mental Illness Social networks and social supports represent the number and types of relationships in which an individual participates, and thereby delineate the potential risks or opportunities for violence. A particularly dense network composed primarily of relatives (everybody knows everybody else) may provide the individual with few unaffiliated outlets for seeking an ally in a conflict. Or, a low density, loosely connected network may 'burn out' or be brittle, leaving the individual isolated and abandoned, decompensating and becoming dangerous to self or others unnoticed. Beels11 observes that persons with schizophrenia have a 'deficit of initiative,' failing to pursue or maintain relations with individuals other than those to whom they have easy access ('dues free' relationships) like family. Other social network researchers report the majority of relationships to be with casual acquaintances or unrelated adults.12,13 Most individuals with severe persistent mental illnesses never marry. Even more rarely are they parents. For example, 86% (130) of our sample had no partner at baseline, and 61% (92) had never married. Seventy-one percent (111) had no children. This is an important consideration since in the general population, spouses far exceed any other type of relative as the target of violence.6,15 Lacking mates, individuals with severe persistent mental illness may be more likely to direct violence towards parents or siblings, who may be the source of their most intimate relationship. Social support includes the affective or emotional and instrumental or helping qualities of relationships. People with major mental illnesses are characterized as having relationships that are less intimate, reciprocal, symmetrical, and durable than those of the general population,16 due either to the pragmatic contingencies of life as a mentally ill person17, or to the interpersonal styles associated with their clinical conditions. Nelson et al13 have challenged this view, reporting that there was considerable reciprocity in interactions with others in the social network among their sample, and that more supportive transactions were reported with friends than with family or mental health professionals. Persons with enduring psychiatric disorders may be at increased risk for violence because of socioeconomic factors18, and how, where, and with whom they live17, and not solely because of their psychiatric disorders. The combination of having a major mental illness and living in meager, stressful circumstances may be much more predictive of social networks, relationships and risk for violence than any clinical factor alone. Violence, Families, and Mental Illness The factors that researchers have found to be related consistently to family violence are: 1) the cycle or intergenerational transmission of violence; 2) low socioeconomic status; 3) social and structural stress; 4) social isolation and low community embeddedness; 5) low self concept; and, 6) personality problems and psychopathology.15 These characteristics pertain to many individuals with major mental illnesses and their families, making these families at high risk for violence. In the literature on family violence, the areas most relevant to the topics explored here are adolescent-to-parent violence19 and elder abuse.20-22 In both situations, mothers are the most likely targets of violence, and most of the perpetrators are their spouses or sons. However, in the area of severe violence toward mothers, there was little difference in the rates between daughters and sons as perpetrators. Gelles and Cornell19 stress that adolescent-to-parent violence is often accompanied by violence among siblings. They conclude that "the only important consistency in the analysis of violence toward parents and other forms of family violence is that the presence in a household of one form of violence is related to the occurrence of other types of violence." (165) With regard to elder abuse, or violence directed at family members over the age of 65, Finkelhor and Pillemer22 note that the abuser is often in a dependent position vis a vis the abused parent. While such arrangements are widespread among the target population6, there is scant published research that reports the age of the parent towards whom violence has been directed. Straznickas et al 23 found that younger patients who lived at home were more likely to attack their parents than anyone else, but did not report the age of these parents. In a recent large scale study of families that belong to the National Alliance for the Mentally Ill (NAMI), Steinwachs, Kasper, and Skinner24 identified single parents, especially single female parents as "vulnerable caregivers," reporting that over one-third of the single parents in the sample altered their behavior to avoid upsetting their diagnosed relative i.e., feared violence from their relative. The vast majority (86%) of the single parents were female. A few previously published investigations touch on the direct and indirect, and causal and preventive influences social networks and social support on violence by persons with mental illnesses in community settings.7,8,23,25-28 Family and household composition and climate, employment status, relations with friends, and past substance abuse, violence, and disorder by and around the patient were included as variables. Chaotic, violent family environments where there was alcohol or substance use, a history of and ongoing conflict among family members, and a controlling atmosphere were associated with violence by the patients in at least one of these studies. Parents, especially mothers, were the relatives most likely to live with patients, the primary caretakers or instrumental supporters, and not surprisingly, the most frequent initiators of commitment petitions for dangerous behavior.29 It is clear from these studies that a family tradition of physical violence is associated with violence from the patient research subjects. While this atmosphere may be predictive of violence, it reveals little about the role of psychiatric disorders in the equation, as the same findings hold for the general population. Intrafamilial hostility is one of the three domains that comprises expressed emotion, and higher levels of expressed emotion are associated with elevated frequency and rapidity of relapse, i.e., rehospitalization, among persons with schizophrenia.30- 32 Since rehospitalization is the measure of relapse, it would be helpful to know how many of those admissions were precipitated by violent incidents within the family. Expressed emotion is a promising approach to the study of violence in context because the hostility construct is contextual and intra- familial, based on extensive observation of interactions among the relatives. Data and Methods Extensive demographic, clinical, social network and social support, and social functioing data were collected prospectively via interview from a cohort of 169 persons with major psychiatric disorders. We conducted text analysis of interview transcripts, and bivariate and multivariate analysis of the other data. Clinical information came from hospital charts (diagnosis), a self-report symptom scale (PERI 33-35), and interviewer rating of the BPRS checklist. 36 Social networks and support were elicited with a structured protocol developed for the study. For this analysis of violent threats and acts committed during an 18 month period, the sample size varies because of missing data for some individuals. The sample is described in Table 1. The median age of the cohort was 28.6 years, 53% were female, and slightly under 30% were African-American. The mean number of prior psychiatric hospitalizations was 2.78, but over 40% of the sample reported no prior mental health center visits. Diagnostically, 39.5% had schizophrenias, 33.1% affective disorders, 17.2% personality disorders, and 10.2% had other psychotic diagnoses. More than 2/3 of the sample did not commit a violent threat or act during the study, but over 40% reported engaging in some type of violence in the past. Evidence of the occurrence of violent threats and acts came from hospital charts and commitment petitions, self report during interviews with patients and their significant others, and review of court records in the counties where patients resided. The interview was not designed to elicit specific information about domestic or other violence towards others, but there were questions about hitting others and fighting in the PERI Anti-Social History scale,33 and the topic came up spontaneously during in-depth, semi-structured interviews. A violent act was defined as an arrest/criminal charge and adjudication for assault and battery, manslaughter, or murder, or a danger to others (DTO) commitment which specified that the respondent hit, hit with an object or weapon, sexually assaulted, or threatened with an object or weapon another person, or a confirmed report of other violent acts by the respondent. There were 39 acts of violence committed by 23 persons. An object or weapon was used in 41% (16 acts), while the remainder involved hitting (19 acts) and 4 sexual assaults (no rapes). A threat was defined as an arrest/criminal charge and adjudication for communicating threats/threatening conduct, or a DTO commitment specifying vague ideation or verbal threat towards another person, or a confirmed report of other threatening behaviors. There were 75 threats of violence towards persons made by 52 individuals. Over half, 53.3%, involved threats made to or about specific people, while the remainder were vague ideation regarding harm to others. Using these definitions, 157 study participants were assigned to one of three groups based on their behavior during the 18 month study period: 1) those who engaged in no reported violent behavior towards a person, N=101/64.3% of the sample; 2) those who engaged in making violent threats only, N=33/21.0%; and 3) those who engaged in violent acts, N=23/14.6%. Among the third group, there were two unduplicated types of individuals: 1) people who committed a violent act only, N=4; and 2) people who both threatened violence and committed a violent act, N=19. Overall, 56 persons in the sample, or about one-third (35.6%), engaged in threats and/or acts directed at other persons during the 18-month period. Nearly twice as many threats were made than acts were committed. Five (3.2%) of these persons had criminal charges for violent incidents, only one of which was for a violent act (assault and battery). Findings Network Composition, Size, and Targets We found a heavy concentration of network members to be relatives, with half (78) of the sample reporting networks that were composed of over 2/3 relatives. In contrast, 15% (23) of the sample listed no friends in their networks. The mean percent of friends comprising networks was 28.2%, compared to relatives at 62.7%. (Table 1) Many persons with enduring mental illnesses are unemployed and financially dependent upon their families of origin.5,24 Over 57% of all instrumental helpers indentified by respondents were relatives. A small group, 14% (21) of the sample said that they got all of their instrumental support from relatives. Individuals in the study who threatened or behaved violently were significantly more likely (p.= .042) to be financially dependent on their families than people who were not violent. The mean network size of 11.86 persons in our sample is comparable to that reported recently by others.13,14 Network size and composition varied by race, sex, diagnosis, and marital status. (Table 2) People with schizophrenia had smaller, more densely kin- based networks than any other diagnostic group, and were more likely to live with relatives other than spouses. The number and proportion of targets who were relatives, and the type of relatives who were targets varies from the general population. Over half of the targets were relatives, 78% were known to the respondent, and over 1/3 had been listed on their social networks. Among related targets, mothers were the largest category (28%), and fathers and children the smallest (7% and 9%). (Figure 1) These patterns reflect the predominance of relatives in the social networks of the sample, the large proportion of respondents who lived with their mothers, and the comparative absence of fathers and children from those households. Forty-two percent (70) of the households at baseline included parents, 67 of which included a mother, but only 36 included a father. One of the important ways that violence among this cohort differed from that among the overall population was the predominance of parents as targets. (Figure 1) In particular, the pattern was distinct with regard to the type of relative, i.e., mothers. Child, albeit adult child, to parent violence seemed to be a pattern unique to this group. Not coincidentally, mothers also played a dominant role in the social networks of a substantial portion of the sample, and were, not surprisingly, the most frequent network member to initiate DTO commitment complaints. Co-resident partners and spouses, though uncommon, were proportionately at risk of being a target nearly twice that of mothers. (Figure 1) Thirty-eight percent (8) of co-resident partners or spouses were targets compared to 19.7% (13) of co-resident mothers. This pattern coincides with that reported by Straus and Gelles,6 who note that violence committed by women is concentrated within their nuclear families. However, the pattern here differs because wife to husband violence is how women equal or surpass men in the population at large, and only 2 male spouses or partners were targets among the cohort. We suspect that mothers (and other female relatives) are substituting for husbands as the related targets of violence by the women we studied. As in the population at large, male to female violence predominated among this group. Men were much more likely to both threaten and act with violence toward women, particularly female relatives, than toward men. Sixty percent (12) of the violent acts were male to female, and only 10% were male to male. Male to male threats were similarly infrequent (8% of threats), especially within families. Men and women threatened relatives and non-relatives in equal proportions, but when they engaged in violent acts, women were more likely than men to direct violent acts toward relatives. That is, 75% (6) of female acts of violence were directed towards relatives, compared to 56% (14) among the men. Perceptions of Threat within the Social Network To investigate the quality of relations with the most signficiant others of the respondents, we used the Structural Analysis of Social Behavior (SASB).37-340 The SASB requires the respondent to choose another person, and rate both their own behavior with that individual and the other's behavior in relation to themselves. Over half (N=85/54%) of the sample rated their mothers, with an additional 11% (17) rating parents as a unit. Spouses were rated by 29 espondents, or 17% of the sample. Interviews with significant others produced 42 matched sets of ratings, including sixteen individuals who became targets of violence. The SASB is based on a two axis model of relationships: a control--autonomy dimension, and an affiliation--attack dimension. We used a shortened verison of the scale with 36 items which the respondents rated from 0 to 100 with 50 marking the border between false and true. The scale yields four Attack coefficients which have been rigorously established psychometrically. Attack 1 coefficients are ratings of the significant other in transitive action toward the respondent (he/she acts). Attack 2 coefficients are the respondents' ratings of themselves in intransitive relation to the other (I am). Attack 3 describes the others' intransitive response (he/she is), and attack 4 represents the respondents' transitive actions towards the other (I act). The language of the scale is primarily metaphorical with regard to violence, referring to emotional aggression, fear, and rage, rather than to explicit physical attack. One item states that the other "Murders, kills, destroys and leaves me as a useless heap," while another describes the respondent as, "Boiling over with rage and/or fear, I try to escape, flee, or hide from them." Results from the SASB were associated with violence in various ways. Table 3 summarizes baseline SASB Attack coefficients for those who were not violent, for those who acted or threatened, and for the targets of violence who were rated by respondents. Scores closer to zero indicate more hostility and disaffiliation, while those closer to -1 indicate friendly, affiliative sentiments. The mean Attack 1(- .57) and Attack 2 (-.57) coefficients for the entire sample exceed the norms reported for the scale, even when the respondent and others rate themselves at their worst.41 Those who were violent scored well above the norms for Attack 1 and Attack 2, particularly those who engaged in violent acts and those who rated individuals who were targets of violence. One pattern is striking and consistent in the results. Those respondents who were violent rated their "others" as more attacking and menacing than the rest of the sample, but did not describe themselves as more hostile, defensive, or fearful than the non- violent group. They felt threatened and attacked, but did not perceive themselves to be more threatening or hostile than individuals who did not behave violently. The 56 persons in the cohort who were violent described their significant others on the Attack 1 coefficient as more hostile (-.502 versus -.610), but rated themselves as more affiliative and less fearful or defensive than the non-violent group on the Attack 2 coefficient (- .633 versus -.534). They described themselves as more friendly and less hostile in demeanor than the rest of the sample, but saw their significant others as substantially more attacking. Following the same pattern, those who engaged in violent acts rated their significant others as even more menacing (Attack 1=-.458) than either those who threatened others (Attack 1=-.533), and those who were not violent (Attack 1=-.610). These differences within the violent group, and between them and the whole sample, did not reach statistical significance, except when eventual targets of violence were rated. The 16 targets of violence who were rated by respondents were seen as markedly more attacking than the significant others of the remainder of the sample. The respondent's ratings of themselves in response were more angry and wary than those of the violent group as a whole, but not significantly moreso. Further analysis of SASB findings focused on mothers and respondents who rated each other. (Table 4) The mothers of respondents who had been violent rated their children (Attack 3, he/she is) and themselves (Attack 2, I am) as more intransitively hostile than the mothers of respondents who were not violent. Both of these differences are significant at p < .05. Though not statistically significant, the respondents had the same pattern. They rated themselves and their mothers as more attacking in a defensive mode than respondents who had not been violent. Everyone feels threatened and hostile, but no one is accused directly of threatening or attacking. A second measure of perceived threat from others derives from the PERI Perception of Hostility (PH) scale.33 Of the five PH items, one that asked whether the respondent "felt that people were picking fights with you" had by far the strongest correlation with engaging in a violent threat or act. Other items concerning perceptions of people talking behind one's back, staring at, avoiding, or cheating the respondent had weaker correlations. The SASB and PERI findings suggest strongly that perceived threat and hostility from significant others are linked to violence by the respondents. The respondents who were violent described focal relationships in which they experienced significant emotional danger and damage, while perceiving themselves to be less angry, defensive, or offensive in response than one would expect. Their commitment to the hospital as dangerous to others challenges this perception of themselves. Multivariate Results Based on these descriptive findings, we conducted multivariate logistic regression analyses in order to model who would engage in violent acts or threats during the study. The results of these analyses are reported in Table 5.10 Clinical variables proved to be the only significant predictors of who would engage in violent acts, while violent threats were influenced by the social functioning, clinical, previous violence, and social network/social support characteristics of the cohort. Individuals with a diagnosis of schizophrenia were substantially more likely than all others to commit acts of violence toward others, but no more likely to threaten others than people with other diagnoses. Perceived hostility from identified and unidentified others (PERI PH) substantially increased the risk for violent acts, but confused thinking (PERI Confused Thinking) and withdrawal and isolation (PERI Schizoid Personality) decreased the likelihood of acting violently. Being married and working lowered the odds for threatening violence. Diagnosis was not related to threats, but respondents with more previous hospitalizations and more symptoms (higher BPRS score) were more likely to make threats. When respondents perceived hostility from others (PERI PH, SASB Attack 1), they were substantially more likely to threaten others. Confused thinking (PERI Confused Thinking) and withdrawal and isolation (PERI Schizoid Personality) lowered the odds for making threats. Higher concentrations of relatives in the social network increased the odds for threatening others, as did increased network size. Living with unrelated others also increased threatening behavior. Typically, scales like the PERI and SASB are considered to be clinical in nature. Here we consider them also to be important indicators of the quality of relations, or social support. The PERI PH effect is reinforced by the SASB results. People who threatened others described their significant others as attacking and quite hostile, while perceiving themselves to be comparatively friendly toward this person. We included factors in the model that are thought to influence violence among persons with mental illness: demographic; clinical; previous violence; and then added social network and social support. The social network and social support variables model added significantly (.10 < p < .05) to the model chi-square, raising it from 95.45 to 126.78, a difference of 31.33, with 22 degrees of freedom. We think this is persuasive evidence, statistical and descriptive, that considering social network and social support factors increases our understanding of how context affects violence among the target population. Discussion Taken as a whole, the findings support our contention that interpersonal and social context, relationship quality, and subjective experience or clinical condition, as inter-related and independent factors, are important considerations in assessing risks for violence. As suggested at the outset, contextual thinking about the risk for violence considers the social network as the opportunity to engage in violence, and social support or quality of relationships the provocation or perceived need for such behaviors. Perhaps the most persuasive evidence in support of this view are the numerous indications that respondents who were violent felt malice and danger from significant others, and perceived and experienced hostility within their interpersonal networks. The text analysis and two psychological instruments confirm that people who threatened violence felt threatened. People who committed violent acts also perceived hostility from others in their networks, albeit to a lesser extent. In contrast to findings from recent influential studies, 3,41-43 substance abuse (co-morbidity) was statistically unrelated to violence. The measures of threat we used may be more direct indicators of what incites violence. The negative impact of primary symptoms (PERI CT, SP) on the risk for violence, particulary acts, diverges from research findings43 and clinical wisdom. We found that an individual who is cognitively disordered and withdrawn was less likely to be violent than someone who felt threatened. This suggests that violent behavior requires a certain amount of organization, opportunity, and perceived need to defend. This issue of threat perceived by respondents deserves considerably more attention. If individuals with mental illnesses are living in situations which they experience as threatening, in this case mainly families, analytic and treatment frames need to shift. We should be assessing more carefully how fearful and victimized the individual feels (and is) in their household or social network in order to more accurately identify high risk situations. Several recent investigations demonstrate that current domestic violence44 and prior sexual abuse are not uncommon among psychiatric patients, but attention to these experiences by clinicians and researchers is.45-48 We think it is a mistake to categorize people as violent or not, to conceptualize violence as a characteristic of a person, absent equal attention to the underlying or concurrent interpersonal and clinical processes and social contexts that individuals experience over time which vary in their contributions to violence.49-50 The inverse relationship of childhood trauma to violence in our results implies that the timing of the threat or attack from others is crucial. Those who were threatened and attacked in childhood were less likely to threaten or act with violence toward others as adults. Some people who were harmed physically and psychologically by others subsequently harmed themselves, were unlikely to engage in threatening behaviors, and, ironically, may have responded to their abuse with the symptoms and behaviors that resulted in their being diagnosed as mentally ill. It is clear from our investigation and others,5,23,24,29 that mothers bear not only considerable responsibility for caring for relatives with mental illness, but a concurrent risk for being the targets of repeated violence by this relative.19 Mothers were the most frequent related targets of violence, and the most likely targets of repeated violence. All of the mothers who were targets lived with their adult mentally ill children. A high proportion of the sample lived in a household with a parent, thus the opportunities for adult child to parent violence were comparatively high. Households which include a mother and adult child apparently are at some higher risk for violence, especially when there is no other parent present.7,8 It is not premature to propose preventive interventions in these families. First, there are very few residential services such as supported housing for persons with disabling mental illness in the study counties. The opportunity and need for intra-familial violence might be reduced by developing other places of residence, wherein they feel safe. Individuals who listed mental health professionals in their social networks were, on the whole, less likely to be violent during the study (Table 2). This suggests that a trained person who is involved enough with the patient to be considered a social network member may be a deterrent to violence, able to intervene should signals or precursors of violence appear. It is possible that treatment personnel were more willing to be closely involved with individuals who were not threatening or attacking, or that those who were less likely to be violent were for other reasons more involved with treatment personnel. Nevertheless, our findings would support the assignment of intensive case managers trained in violence detection and prevention to individuals diagnosed with schizophrenia, who are isolated and fearful, and who live in related households or with their mothers. We found significant differences in social network characteristics between diagnostic groups (Table 2). In view of the strong association between schizophrenia and violent threats and acts, these distinctions should be investigated further. We need to know much more about the complicated interactions among clinical, social, and social network characteristics in relation to violence.13,42,43,51 Conclusion The most significant conclusion, from the standpoint of intervention, is the combination of mother/adult child co-residence and violence. This suggests that it is the parental activity of the mother, along with her degree of proximity and involvement in daily living, and vulnerability (fathers, even if co-resident, rarely are reported to be attacked or threatened), which creates the opportunity for the violence.23 Our findings suggest that a profile of risk consists of vulnerability, opportunity, mutual threat, and dependency within the social networks of persons with psychiatric disorders. The references that accompany this article are in a separate file. This file came from anonymous ftp sjuvm.stjohns.edu cd MADNESS The MADNESS ftp site is a service of MADNESS, an online discussion on LISTSERV@sjuvm.stjohns.edu Please credit the list if you copy this file.