School shootings that lead to death and injuries, such as the one that occurred at Columbine High School in Littleton, Colorado, in April 1999, are horrifying to both the public and the medical profession. Each time a shooting occurs at a school, psychiatrists across the country, particularly those in emergency and crisis settings, are asked to assess potentially disturbed children for the risk of violence. Can we predict whether an individual youth will turn violent? If we cannot, what other things can we do to reassure the public or to help make our schools safe?

The first challenge that confronts us as we attempt to evaluate children is that of public expectation. One way to assess this expectation is to examine coverage of school shootings in the popular press (+1). Most writers in national publications, such as the New York Times, understand that a psychiatrist's ability to predict violence is much better in hindsight (+2) and that child psychiatrists, like everyone else, have a difficult time sorting out which children pose a real threat (+3). Still, as one writer pointed out after the events in Littleton, people are divided between American ideas about freedom and "the faith that social science can protect society by classifying potential malefactors so they can be detected and isolated before they can do any damage" (+4).

Although, in general, popular writers have not specifically blamed psychiatrists for missing dangerous signs among adolescents before they become violent at school, the mental health system—along with families and teachers—has been criticized for being a part of the problem (+5). Retrospective accounts of killings that media reporters are able to assemble often reveal information that, in hindsight, seems to provide abundant clues that the family, teachers, and psychiatrists, should have been able to act on. This pattern of reporting by the media affects current practicing psychiatrists because families or school representatives frequently come to evaluations with expectations generated by retrospective analyses of past events. Families and schools often demand "clearance" for troubled children before they will allow the youths to return to a home or school setting. However, emergency psychiatrists often do not have the time, resources, or information that allow them to confidently say that they have prevented another school shooting or that an individual child is not a danger to himself or herself or to others.

The second challenge that faces us is how to evaluate children's feelings and behavior in order to screen for violence, particularly in an emergency setting. Research on violent acts among children has suggested that demographic characteristics, such as sex and socioeconomic status, as well as socialization and family dynamics may be important clues as to why some youths become violent (+6). In the wake of a rash of school shootings in the past decade, researchers have been trying to identify factors that might indicate a propensity toward school violence, particularly shooting sprees. A study that examined four school shootings identified many traits that the perpetrators had in common. Instead of being seen as troublemakers, these children were thought of as quiet and isolated, were often bullied by their peers, and experienced humiliation in school. These children lacked skills of conflict resolution and anger management, and their interaction with peers stimulated a need for power and control. In contrast, children were less likely to become violent if they respected authority, were responsible, were able to think independently, and were able to communicate well (+7).

Additional clues to violence among children are their access to guns and their exposure to violent media, such as television, videos, and video games. Strong evidence links violence and gun availability (+8). In addition, for decades experts have known about the correlation between media violence and childhood aggression. The Surgeon General's report in 1972 found some causal associations between viewing violent television and short- and long-term aggressive behavior (+9). Since then many studies of media violence have been conducted. Simple frequency of television viewing has an impact on violent behavior. A longitudinal study completed by Huesmann and colleagues (+10) demonstrated a relationship between viewing television violence and behaving aggressively a decade later. Even moderate levels of watching music videos may result in a significant exposure to violence and have been linked to aggressive behavior among college males (+11). Also, some evidence exists that violent video games may have a negative impact on today's youth (+12,+13).

Although it is more difficult to address in an emergency setting, a child's social environment is clearly an important factor in assessing the risk of violence. Schools in which teachers and students have low achievement expectations have had a greater risk of violence, whereas schools with cooperative learning and strong teacher-student relationships tend to be protected from violence. Maternal-paternal conflict, abuse, neglect, and lack of discipline are risk factors in families, whereas affection, support, discipline, rule enforcement, and a high expectation regarding behavior tend to be protective against school violence. Likewise, poor communities with schools that have high drop-out rates, firearms, crime, and lack of positive role models were at higher risk, whereas schools with good role models who helped students were more immune to violence. Although these variables may not necessarily help us predict which children will become violent, they allow for a model of what schools, families, and communities must aspire to be if they are to decrease the likelihood of school violence (+7).

Research on the ability of psychiatrists to predict violent behavior among persons with mental illness is not reassuring. A number of studies have shown that clinicians are poor predictors of violence (+14), particularly because the relative infrequency of violent events makes risk factors difficult to evaluate (+15). However, although there is little evidence to support a psychiatrist's ability to predict a youth's potential for school violence, we will still be asked to assess the risk and must attempt to do so. Complete assessment must address the youth's social functioning, family situation, access to weapons, and exposure to violent media. Recent changes in the youth's behavior—including alienation from peers, use of drugs or alcohol, secretiveness, and increased time spent away from the house—are red flags that clinicians should be aware of (+7).

Perhaps a more important role that psychiatrists can play is in the political and community public health realms. We must be strong advocates of gun control, and we must assist efforts to decrease children's exposure to media violence. We must support community efforts to help shield children from chaotic family settings, and we must educate school counselors and social work staff about the importance of watching for the bullied or isolated child who is fascinated with violence, weapons, and death. School violence is a societal problem, and society is asking for our help. Although it is laudable to try to assess an individual's risk of violence, we are not good at it. To make an impact we must instead look at the things that are leading to this behavior in our communities and work with our leaders, schools, and families to change them.

Dr. Glick is affiliated with the department of psychiatry at the University of Michigan Medical School, C5124 Medical Science I/0611, 1301 Catherine Street, Ann Arbor, Michigan 48109-0611 (e-mail, rglick@umich.edu). Dr. Hirshbein is with the department of psychiatry at the University of Michigan Medical School in Ann Arbor, and Dr. Patel is a recent graduate of the university. Douglas H. Hughes, M.D., is editor of this column.