The relationship between mental health and violence is complex and often misunderstood, with a number of misperceptions around risk of violence and victimization. Speakers discussed the stigma and discrimination that people with mental illness experience, particularly in the media. They also discussed the need for additional research on the intersection of mental health and violence, noting that the detection and the assessment of risk of violence are imprecise. Speakers also examined how a better understanding of the pathways for and the risk factors of violence could yield more effective interventions.
Daniel Fisher of the Riverside Community Mental Health Center opened the panel titled “Experiences and Perspectives Related to Mental Health and Violence” by describing the importance of language. He noted that “patient” and “consumer” are not preferred terms within the community, and “survivor” is imprecise, and he suggested “people with lived experience” as a more inclusive and less discriminatory term. These issues, he stated, are not laboratory or clinical issues, but rather community and cultural issues. He emphasized that a more nuanced and accurate perspective of these lived experiences would help reduce stigma and provide better treatment options. For example, a peer-support recovery movement, similar to Alcoholics Anonymous, exists to provide nonmedical options for those with lived experience to connect and empower each other on the path to recovery.
Panelists further explored these issues within the context of the use of mechanical restraints, misperceptions around violence and its association with mental illness, as well as the history of deinstitutionalization and the failure to transition to community-based care.
Use of Restraints1
Because there is a misperception that people with mental illness are more prone to violence, it is a common practice to use mechanical restraints in institutional settings. But Elyn Saks asserted that the use of restraints, though well intentioned, is itself a violent act. Restraints when used over a length of time are extremely painful and degrading and cause feelings of helplessness. They can also be retraumatizing for those with posttraumatic stress disorder (PTSD) or for other survivors of trauma. She noted that, in her personal experience, the use of restraints was not necessarily due to her own behavior. In fact, the literature supports the idea that the use of restraints has more to do with the institutional ethos than other factors, such as patient characteristics or patient–staff ratio.
In exploring why restraints are used, Saks noted that there are studies that indicate that restraints help those being restrained feel safer. However, she observed that in her experience, she had never heard anyone express that sentiment, and that emergency fatalities do not lessen with the use of restraints. A second, more legitimate reason is that restraints can be protective for health care and service providers. For those patients who might become imminently violent, there is legitimate justification for restraints. However, she noted four reasons why the use of restraints as protection might be problematic:
- Restraints are often abused, despite statutes intended to prevent such abuse. She gave the example of one client she knew whose chart suggested restraints were used more for discipline than imminent violence.
- Imminent danger is difficult to predict, and patient and physician perspectives on a patient’s own dangerousness vary widely on this.
- While well meaning, the use of restraints is an act of violence and can be more dangerous than not using them. Most staff injuries occur in the restraint process, which could indicate that the use of restraints itself causes people to be violent.
- There are often less restrictive alternatives available, such as the use of a padded cell.
1 This section summarizes information presented by Elyn Saks, University of Southern California.
Saks suggested that restraints might cause more deaths than lives they save. In a series of articles in the Hartford Courant, a Harvard University statistician estimated that one to three people die each week in restraints—aspirating in their own vomit, strangling, or having heart attacks. She stated that since there are other means of protecting people, it is not clear whether restraints cost or save lives. Restraint-reduction efforts have resulted in lowered use of restraints without increased violence in Philadelphia and Massachusetts. The United Kingdom by and large does not use extreme restraints, and has not done so for 20 years.
In cases where the use of restraints might be justified (e.g., transporting a violent person or when a medical professional needs to be in close quarters), Saks recommended several enhanced procedural steps:
- Requiring 15-minute checks or an attempt to remove the restraints every hour
- Changing the liability scheme to make harm caused by restraints more liable (and harm caused by lack of restraint less liable)
- Providing guidelines to patients on behavior that will result in the removal of restraints
- Videotaping all restraint episodes
- Forbidding “spread eagle” restraint
In the discussion following Saks’s presentation, workshop participants raised questions around the use of chemical restraints, particularly in the older population, and alternatives to restraints. Saks spoke of additional considerations around elderly populations, such as dementia and risk of falling. However, she stated that restraints are not always the answer—a person sitting with the patient could also provide assistance. Another audience member mentioned the use of a hospital bed programmed to alert staff if the patient tried to get up.
Impact of Violence on People with Mental Illness2
Harvey Rosenthal spoke about the impact of violence on the community of people with mental illness, particularly the fallout after horrific episodes of “active shooter” violence. The stigma and the misinformation around the role mental illness plays in violence is often heightened after incidents of mass violence, with resounding repercussions. Rosenthal mentioned that policies are often promoted in government that seek to respond to these incidents, but instead result in depersonalization and
2 This section summarizes information presented by Harvey Rosenthal, New York Association of Psychiatric Rehabilitation Services, Inc.
criminalization of people with mental illness, and threaten to undo progress in promoting recovery, dignity, and integration.
One challenge that Rosenthal raised is that mental illness is loosely defined, particularly in the general population. Some people consider autism a mental illness (it falls instead under the rubric of developmental disability). A recent Kansas ruling in the Supreme Court places sex offenses under mental illness. And others consider sociopathy and substance abuse to be mental illnesses. At the same time, the common perception that one must be “crazy” to commit horrific acts confuses matters further and feeds into the misperception that mental illness is a risk factor for violence.
Rosenthal reiterated the fact that people with mental illness are at most marginally more violent than the general public. He noted that only 4 percent of violent crimes are affiliated with mental illness, and that 1 in 70,000 people with mental illness are committing murder of strangers (Swanson, 2015). To put it another way, of the 140,000 people in New York who are deemed “seriously and persistently mentally ill,” two of them are at risk of committing murder. Additional research has concluded that there is no clear relationship between psychiatric diagnosis and mass murder, and that most mass murderers are young men with no diagnosis of psychosis (Fox, 2015). Despite these facts, the perception of the link between mental illness and violence has policy consequences, such as campaigns to force medication and other treatment on people. In New York, Rosenthal stated, the law now mandates that mental health professionals report if a patient who owns a gun expresses anger.
On the other hand, people with mental illnesses are 11 times more likely to be victims of violence and 5 times more likely to be murder victims. And, yet, Rosenthal asserted, the public discourse still revolves around the harm potentially committed by people with mental illness, and not the potential harm faced by the vulnerable. In particular, there are efforts to promote forced-treatment laws, despite the previously cited lack of evidence on a link between mental illness and violence.
Rosenthal did share some positive items that have come about recently regarding mental illness and violence. In particular, the Associated Press (AP) created guidelines for its reporters on writing about incidents in which people with mental illness might be involved:
- Do not describe an individual as mentally ill unless pertinent to the story and the diagnosis is properly sourced.
- Do not use derogatory terms such as “insane,” “crazy,” “nuts,” or “deranged.”
- Do not assume that mental illness is a factor in violent crime.
- Do not use descriptions that denote pity, such as “afflicted with.”
He closed by discussing the implications of a “broken system,” a commonly used term with different meanings for different stakeholders. For consumers, it is disempowering, dependency fostering, and overly focused on medication. For families, it is a lack of assistance with their loved ones. For others, it means more forced treatment. And for the media and much of the general public, it means unchecked violence. To fix this, he suggested that a new narrative needs to be created—one that is focused on both the facts and the nuance around mental health and violence, and one that gives voice to people with mental illness.
Reflecting on Mental Health and Violence3
Robert Bernstein observed that people with mental illness have endured a long history of segregation and discrimination. In 1990, the Americans with Disabilities Act (ADA) was passed with the intention of “mainstreaming” people with disabilities, including mental illness. The ADA was a sea change in the treatment of people with mental illness, away from the previous approach of mandatory institutionalization and custodial care toward a focus on community inclusion and multimodal treatment. In 1999, the U.S. Supreme Court ruled that, per the ADA, unwarranted institutional confinement was a form of segregation and that public systems have an obligation to provide integrated services where feasible.
This is of course an ongoing discourse, Bernstein stated, but there are profound reforms occurring to improve situations for people with mental illness. In particular, deinstitutionalization was the cornerstone for a mental health civil rights movement that preceded the ADA, and one that is still an important element today. In the 1960s and the 1970s, deinstitutionalization was the first wave of reform and was based on the terrible conditions in state hospitals, where patients were not only incarcerated but also routinely put in restraints. Given that most people with mental illness are not a danger to themselves or others, and that the institutions themselves raised other problematic issues, Bernstein asserted that deinstitutionalization was a positive goal.
However, he was careful to point out that integration of patients with communities was intended to be accompanied by a comprehensive community mental health movement, in which services are community based rather than hospital based. Yet, this movement never materialized, and as a result of poor funding, the mental health domain today exists as a crisis system. People with serious mental illness have suffered, and perceptions of mental illness have suffered as well.
3 This section summarizes information presented by Robert Bernstein, Judge David L. Bazelon Center for Mental Health Law.
The community mental health movement was intended to provide services to anyone with mental illness, but today it operates solely for those who pose a danger. Because of a lack of funding, Bernstein stated, there is little investment in prevention and early intervention with a primary focus on emergency response. The current mental health system, he noted, is an upstream system failure—the development of a mental health crisis being evidence of a lack of early intervention. He closed by noting that because of this systematic lack of funding where it is most needed for long-term solutions, there is a perverse incentive now to capitalize on public perceptions of mental health and violence if it means greater resource allocation for mental health.
Following the presentations, panelists and workshop participants discussed additional issues raised, including challenges outside the United States. In Latin America and the Caribbean, the movement toward recovery and integration is not nearly as robust. Fisher and Rosenthal both noted that the recovery movement has its roots in the United States, and it is important that, even while expanding it outside the United States, continued work and sustained commitment is maintained at its origin.
Additionally, Eric Caine of the University of Rochester Medical Center expanded on the issue of community mental health, observing that one of the reasons for its lack of prioritization and funding was a change in the way mental health was structured and treated. Caine went on to note that, previously, community mental health fell under the purview of the National Institute of Mental Health, but currently the Substance Abuse and Mental Health Services Administration (SAMHSA) provides block grants to states, with individual counties developing systems and allocating funds. At this level of granularity, he postulated, grassroots and peer-led organizations have an important role in shaping community mental health.
Detection of risk of violence is currently an imprecise science, speakers observed, that could benefit from greater study and refinement. One of the major challenges faced is that violence is not a high-probability event, and statistically the percentage of violent people is low. Risk factors for violence are also varied, and it is unclear which factors and in which combination might result in a violent act. With current instruments, this results in inaccurate assessment of risk, which carries implications for those with mental illness and those with a propensity for violence. Speakers discussed these challenges and approaches to develop more refined instruments.
Violence Risk Assessment4
Seena Fazel spoke about risk assessment for interpersonal violence by providing an overview, synthesizing evidence, and reflecting on implications and next steps. Violence assessments, he stated, range from unstructured clinical opinion to validated instruments that use tools as proxies for clinical judgment. There are some 200 of these instruments in existence, and they are widely used in forensic psychiatric services and criminal justice settings. Often, they are used to make decisions about sentencing, parole, and probation. Assessments that combine elements of structured and unstructured approaches, such as actuarial instruments to calculate a probability score, or categorizing risk as high, medium, or low based on a predetermined checklist of risk factors, are commonly used.
Fazel and his colleagues examined the literature and located 40 systematic reviews and meta-analyses on commonly used tools to determine their evidence base. They found a number of problems with these studies, including a failure to exclude duplicates (resulting in overestimation of effects) or to explore heterogeneity (resulting in wide variance). They found six studies that actually examined predictive validity; five of those explored only the PCL-R (Psychopathy Checklist), and one looked at another instrument.
Because of this lack of comprehensive evidence, they ran their own meta-analysis, which also included previously unpublished data. This resulted in a large study of 73 samples with about 24,000 individuals who underwent risk assessment by 1 of the 9 most commonly used tools. The outcomes were presented in a few different ways, but Fazel singled out the positive predictive value (PPV) in particular, because of its clinical usefulness. PPV is an assessment of how well the instrument identifies true positives—that is, “if an instrument determines high risk, how many of those people go on to violently offend or sexually offend?” he explained. Per the meta-analysis, the PPV for “violent offending” was 0.41, meaning the majority of those determined to be high risk did not, in fact, go on to commit violence. At the same time, the analysis showed a PPV of 0.91 for the low-risk group determination, suggesting the tools were better at assessing those who would not go on to commit violence5 (Fazel et al., 2012).
Fazel noted that how well the instruments perform is highly dependent on their use. The PPV for determining violent offense suggested that they were not great at predicting violence and therefore were not suitable for decisions such as sentencing or release from hospital. Looking at other
4 This section summarizes information presented by Seena Fazel, University of Oxford, United Kingdom.
5 A PPV of 0.41 indicated that of those in the high-risk group, only 41 percent went on to commit violence. However, a PPV of 0.91 indicated that 91 percent of the time, the test was correct in determining an individual was in the low-risk group (Fazel et al., 2012).
outcomes from the meta-analysis, Fazel noted there was some evidence that the instruments could inform treatment and management plans and could be used to screen out low-risk individuals. In comparing them to other tools, Fazel observed that they fared poorly compared with diagnostic tools but were more similar to existing prognostic tools from other medical disciplines. However, the consequences of moderately useful tools in violence prevention are different: There are costs in terms of extended detention, as well as costs of staff training and time.
In a second review, Fazel and his colleagues looked more closely at different tools specifically designed for populations with mental illness. He noted that they were disappointing due to wide variation in their predictive ability. Additionally, only two studies looked at schizophrenia, which would normally be considered a risk factor for violence. In looking at the content of the tools, he and his team determined there is a wide variation in what is included; for example, the instruments included a wide variety of factors related to criminal history, failing to converge on what that entailed. And in another recent study by Jeremy Coid and his colleagues, these instruments were found to fare even more poorly with psychopathy than they do with mental illness (Coid et al., 2013).
Fazel concluded his remarks with a summary of his findings. The risk assessment tools he examined had limited value in predicting risk of reoffending but could be useful in identifying different risk groups for management. More importantly, he argued that the tools should be used differently: to screen out low-risk people as a means of focusing resources on the remainder. He also felt that the research could be better improved—by independent funding, validation by impartial experts, and higher standards of evidence—toward the development of better assessment tools.
Strategies for Preventing Youth Violence6
Serious violence, Dustin Pardini observed, peaks in adolescence. Most youth who engage in violence cease over time, and only a small percentage persist into adulthood. In his presentation, he focused on programs implemented during elementary school, before children display seriously violent behavior. Universal interventions are delivered to an entire population of youth, while selected programs target youth with population-level or demographic risk factors, such as living in a high-crime neighborhood. Indicated interventions, which represent a large percentage of interventions, focus on children who exhibit early forms of violent behavior, such as physical fighting, or characteristics of oppositional defiant disorder and conduct disorder.
6 This section summarizes information presented by Dustin Pardini, University of Pittsburgh.
In a recent meta-analysis, effect sizes demonstrated that indicated interventions produced the greatest reduction in aggressive behavior, most likely because these youth are already showing high levels of aggression (Wilson et al., 2003). The effects get smaller as the intervention becomes more general, though they still remain significant. So which approach is better? On the one hand, the identification of high-risk individuals who will commit violence is difficult. Instead, the focus should be on reducing the risk in the population as a whole. The benefit per individual would be small, but everyone would be included in the intervention. On the other hand, some research indicates that a small number of juvenile offenders actually commit a large amount of youth violence. Focusing on those adolescents would have optimal impact on the overall amount of crime.
Pardini noted that both approaches have their advantages and disadvantages, but he suggested that further exploration of how best to implement targeted interventions was important, particularly the process of screening youth for the programs. An effective screening instrument needs to be brief, psychometrically reliable, precise, and administered across multiple settings. Most importantly, it should significantly predict future violence and have evidence to show such. Because violence is a relatively low-probability event, a risk-screening instrument will generate a higher number of false positives.
There are several practical implications of making errors with the instrument, Pardini emphasized. Where the line is drawn between high and low risk, or when there is a false positive, makes the difference between a child being placed in the intervention or not. This could have negative ramifications for the child because of the labeling, as well. Also, there is some evidence to suggest that grouping children into such interventions could result in deviancy training, in which there is take-up of adverse outcomes instead of prevention. False positives are also a poor use of funds, and false negatives reduce the impact of the program.
Currently, there are no standardized empirically based risk-assessment tools for screening youth to refer them to targeted programs. There are a few ad hoc tools, but none that are available for the general population. These ad hoc tools are based on the idea that early conduct problems are strong predictors for future violent behavior, an association seen in longitudinal studies. In an analysis of these risk-assessment tools, Pardini and his colleagues determined that they were mediocre at accurately identifying high-risk youth, with a large number of false positives, especially among girls. This is not surprising, because as Pardini previously explained, violence peaks in adolescence and ceases over time. Only a small number of violent youth persist in their violence.
The Pittsburgh Youth Survey, which began in 1986, was an attempt to develop a more accurate risk-assessment tool (van Wijk et al., 2005).
Researchers followed a sample of children from public schools in Pittsburgh over time. The children were questioned at specific points about their violent behavior, and official criminal records were also collected. Among the sample, the rates of violence were high, echoing concerns about a disproportionate impact of violence on minority youth. Pardini and his colleagues looked at all of the risk factors at the first assessment point—not just behavioral issues, but also family conditions, peer influence, and neighborhood characteristics, among others. They identified 51 risk factors, and using statistical regression, found the 11 strongest, ranging from academic issues to physical aggression to family poverty. These risk factors demonstrate better sensitivity and specificity than many adult assessment tools, but Pardini felt there was room for improvement. He proposed examining multiple datasets across the country and conducting comparable analyses to replicate factors measured by parents, teachers, and the children themselves. Once they have identified those factors that consistently predict violence, they plan to develop standardized item content to assess each risk domain. The final step would be the development of a psychometric tool with as brief a measure as possible that can be administered across multiple settings.
Impact of Bullying and Mental Health7
Dieter Wolke opened his presentation by remarking that bullying has a definition in common language, but also a scientific construction that goes beyond conduct problems. He emphasized that while conflict among children teaches them how to resolve conflicts, bullying is not about conflict resolution but about power and intentional harmdoing. There are different types of bullying, such as overt bullying, relational bullying, and cyberbullying.
Wolke identified four groups associated with bullying:
- Pure bully, who perpetrates the aggression but never becomes a victim
- Pure victim, who gets bullied but never bullies others
- Bully-victim, who bullies at times and is bullied at other times
- Neutral child, who can be a bystander or a defender
From where does bullying stem? In evolutionary biology, bullying could be a means of accessing resources and gaining dominance in a hierarchy. In fact, bullying could be protective against having to fight all the time. If this is true, Wolke posited, then it should be seen in all socioeconomic status
7 This section summarizes the information presented by Dieter Wolke, University of Warwick, United Kingdom.
groups, but would be more frequent the scarcer the resources. In a recent meta-analysis, Wolke and his colleague concluded that, indeed, bullying is found in all classes and segments in society (Tippett and Wolke, 2014). In another meta-analysis, researchers discovered that bullying is more prevalent in more unequal societies, so that inequality as a proxy for scarcity is in fact correlated with bullying occurrence (Elgar et al., 2009).
Adverse consequences of bullying have been explored in the literature, and Wolke shared some examples. In one study in primary school, Wolke and his colleagues looked at physical and emotional health problems in the four previously mentioned bully groups. He noted that the most strongly affected group is the bully-victims, who are somewhat socially defeated. Those with the lowest problems are the pure bullies, who are not victims of bullying themselves (Wolke et al., 2001). In a longitudinal study on bullying history, researchers found that incidence of bullying is not the only factor—chronic bullying has add-on effects. Those who are currently being bullied fare worse than those who were bullied in the past, but those bullied currently and in the past do worst of all (Bogart et al., 2014).
In another study with far-reaching implications, researchers in Britain discovered that bullying in elementary school was associated with self-harm with intent to commit suicide at age 17, with a population-attributable fraction of 20 percent (Lereya et al., 2013). This means that if bullying were eliminated, Wolke explained, 20 percent of adolescent self-harm cases could be prevented. He emphasized the importance of this by noting that, by comparison, obesity, which commands significant resources for its prevention, only accounts for 3 percent of heart attacks. Other research supports similar findings; another study found that chronic bullying before age 11 increases risk of psychotic experiences threefold (Wolke et al., 2014).
In studies done in adults who experienced bullying as children, researchers again found health problems, particularly psychological conditions, in those who were pure victims, but also bully-victims. They had poorer psychosocial outcomes, as well, including difficulty maintaining employment and relationships (Copeland et al., 2013; Wolke et al., 2014). In another study looking at inflammatory responses to C-reactive protein, the stronger or more chronic the bullying, the higher the response. The largest change was for victims, followed by bully-victims, but the lowest was in pure bullies (Copeland et al., 2014).
Wolke closed his remarks by summarizing the findings from the literature: being bullied has wide-ranging effects on mental health, from increasing risk for psychopathology to adverse psychosocial and social outcomes. The chronically bullied and bully-victims have the worst long-term outcomes. Bullies do not experience these adverse outcomes but do tend to show lower empathy and higher rates of manipulation as adults.
Bullying is highly prevalent, affecting 15 to 20 percent of the population, and affects all social strata. In the United Kingdom, the majority of children who have not attended school for a whole year have not done so because of bullying. Bullying’s impacts extend beyond the individual and his or her long-term outcomes, but also has an impact on society through workplace productivity. Policies that address bullying, Wolke asserted, will have a universal impact.
In response to questions regarding mental illness risk factors for youth violence, Pardini noted that the main driving predictors are conduct disorders and oppositional defiant disorder; others, such as depression and anxiety, do not have a strong relationship. Additionally, when he and his colleagues analyzed other potential factors, such as trauma and physical abuse and neglect at home, they were also not as significant in predicting future violent behavior.
A related topic raised in discussion between panelists and the audience was the role of the family in bullying. In response to one such question, Wolke noted that over time, children spend more time with their peers than their family members, emphasizing the importance of peer acceptance. He went on to explain that while violence by parents is detrimental to a child’s well-being, most violence experienced is by peers and siblings. However, violence among siblings is rarely considered abuse or bullying. But Wolke noted that sibling violence has adverse effects, particularly in regard to bullying—those who are victimized in their own home by a sibling are 4 to 12 times more likely to be a victim at school, as well. And those who bully their siblings are three times more likely to bully others.
Participants at the workshop also further explored challenges raised by screening, including the important distinction between a diagnosis and a positive screen, the latter of which has been shown to decrease productivity. Additionally, screening in schools raises issues around data protection. The combination of a false-positive screen and potential privacy concerns has profound negative implications for individuals.
Finally, in a discussion around criminalization of sibling abuse, bullying, and other violent youth behavior, panelists and several audience participants raised skepticism around the effectiveness of criminalizing people. While some bullying and family violence would under other circumstances be considered crimes, families and schools are often reluctant to report such incidents. In addition, several participants noted that rehabilitation and treatment have a greater positive impact than criminalization in both other-directed and self-directed violence.
Bogart, L., M. Elliott, D. Klein, S. Tortolero, S. Mrug, M. Peskin, S. Davies, E. Schink, and M. Schuster. 2014. Peer victimization in fifth grade and health in tenth grade. Pediatrics 133(3):440–447.
Coid, J., S. Ullrich, and C. Kallis. 2013. Predicting future violence among individuals with psychopathy. British Journal of Psychiatry 203(5):387–388.
Copeland, W. E., S. Wolke, and A. Angold. 2013. Adult psychiatric and suicide outcomes of bullying and being bullied by peers in childhood adolescence. JAMA Psychiatry 70(4):419–426.
Copeland, W. E., D. Wolke, S. T. Lereya, L. Shanahan, C. Worthman, and E. J. Costello. 2014. Childhood bullying involvement predicts low-grade systemic inflammation into adulthood. Proceedings of the National Academy of Sciences of the United States of America 111(21):7570–7575.
Elgar, F. J., W. Craig, W. Boyce, A. Morgan, and R. Vella-Zarb. 2009. Income inequality and school bullying: Multilevel study of adolescents in 37 countries. Journal of Adolescent Health 45(4):351–359.
Fazel, S., J. P. Singh, H. Doll, and M. Grann. 2012. Use of risk assessment instruments to predict violence and antisocial behavior in 73 samples involving 24,827 people: Systematic review and meta-analysis. British Medical Journal 345:e4692.
Fox, J. 2015. Extreme killing. Thousand Oaks, CA: SAGE Publications.
Lereya, S. T., C. Winsper, J. Heron, G. Lewis, D. Gunnell, H. Fisher, and D. Wolke. 2013. Being bullied during childhood and the prospective pathways to self-harm in late adolescence. Journal of the American Academy of Child & Adolescent Psychiatry 52(6):608–618.
Swanson, J. 2015. Mental illness and reduction of gun violence and suicide: Bringing epidemiologic research to policy. Annals of Epidemiology 25(5):366–376.
Tippett, N., and D. Wolke. 2014. Socioeconomic status and bullying: A meta-analysis. American Journal of Public Health 104(6):e48–e59.
van Wijk, A., R. Loeber, R. Vermeiren, D. Pardini, R. Bullens, and T. Doreleijers. 2005. Violent juvenile sex offenders compared with violence juvenile nonsex offenders: Explorative findings from the Pittsburgh Youth Study. Sexual Abuse: A Journal of Research and Treatment 17(3):333–352.
Wilson, S. J., M. W. Lipsey, and J. H. Derzon. 2003. The effects of school-based intervention programs on aggressive behavior: A meta-analysis. Journal of Consulting and Clinical Psychology 71(1):136–149.
Wolke, D., S. Woods, K. Stanford, and H. Schulz. 2001. Bullying and victimization of primary school children in England and Germany: Prevalence and school factors. British Journal of Psychology 92:673–696.
Wolke, D., S.T. Lereya, H.L. Fisher, G. Lewis, and S. Zammit. 2014. Bullying in elementary school and psychotic experiences at 18 years: A longitudinal, population-based cohort study. Psychological Medicine 44(10):2199–2211.
This page intentionally left blank.