Mental Illness and Violence: Lessons From the Evidence

Abstract

The debate about addressing mental illness and violence often ignores key facts. Many people experience mental illnesses, so having had a diagnosed illness is not a very specific predictor of violent behavior. This means that many proposed policy approaches, from expanded screening to more institutionalization, are unlikely to be effective. Expanded access to effective treatments, although desirable, will have only modest impacts on violence rates. Most people with mental health problems do not commit violent acts, and most violent acts are not committed by people with diagnosed mental disorders.

WHEN A HORRIFIC AND INEXplicable act of violence occurs—as in the school shooting in Newtown, Connecticut, the movie theater killings in Aurora, Colorado, or the shoving murders in the New York City subway system—people, quite reasonably, recognize the perpetrator’s behavior as abnormal. Journalists, investigating the story, seek and frequently find long-standing symptoms of a mental disorder: the perpetrator had been depressed, was a loner, or had sought treatment of a mental health problem. Advocates for people with mental illness respond by pointing out the often weak association between mental illness and violence, and pleading that isolated violent acts not lead to further stigmatization and loss of civil liberties for those with mental health problems. Some commentators call for widespread screening to identify those in need of mental health treatment. Others take a different tack, urging a return to an era when people exhibiting disturbed and disturbing behavior were more frequently institutionalized. Policymakers struggle with an appropriately balanced response. Yet much of this story is missing or incomplete. A better understanding of the epidemiology, history, and treatment of mental illness can help improve the policy responses to these appalling events.

The journalist’s search for a mental illness explanation of aberrant acts will almost always succeed. Epidemiological research suggests that nearly half the population—whether or not involved in crime—experience some symptoms of mental illness over the course of their lifetimes. The most recent population estimate of the lifetime prevalence of major mental illnesses meeting diagnostic criteria among US adults is 46%, and 9% meet criteria for a personality disorder.^1,2 Seeking mental health treatment is hardly less common: the literature suggests that about one fifth of the US population report seeking professional care for a mental health problem in a year and nearly one third do so over their lifetimes.³ The very high lifetime prevalence of illness and treatment seeking helps explain why virtually every story of a violent act can be linked to some clues of psychological abnormality or mental health treatment—even though the rate of violent behavior of any type among people who meet diagnostic criteria for mental illnesses is estimated to be only about twice as high as the rate among those who never experience a mental illness.^3–5 Mental illness is simply not a very specific predictor of violence.

As is often the case with low specificity predictors, the high prevalence of symptoms of mental illness and the relatively low odds of violence associated with most of these disorders mean population-level screening would probably do more harm than good. Population screening would identify far more people than could be diagnosed or treated effectively, diverting resources from those in distress and yielding far too many false positives.

Although mental health problems are widespread, a much smaller share of the population—fewer than 2%—meet diagnostic criteria for severe and persistent mental illness (SPMI).³ The subway pushers in New York City had been previously diagnosed with these conditions.⁶ The adolescent and young adult perpetrators of mass shootings often meet criteria for these conditions at the time of the shootings. Because the initial onset of these conditions often occurs at just this age, the antisocial and peculiar behaviors preceding the violent act may not have been well enough established to allow a diagnosis of serious mental illness to be made, even if the behaviors had been properly assessed.

A recent review of multiple studies suggests that there is a relationship between severe forms of mental illness and the more typical acts of violence that can be captured in surveys, and that this relationship is strongest in subgroups of the population with SPMI.⁷ Adults with SPMI who had a childhood conduct disorder (exhibiting defiant and impulsive behavior) are more likely to be violent, as are those with SPMI who also have a substance use disorder.^5,7 Even within the highest-risk subgroups of the SPMI population, however, most people do not go on to commit the more typical acts of violence against others—indeed, people with SPMI are much more likely to be victims of violent crime than perpetrators.⁵

The association between SPMI and violence has led some commentators to urge a return to an era that involuntarily institutionalized more people with SPMI in psychiatric hospitals. These commentators implicitly assume that higher institutionalization rates in the past were associated with lower levels of violence attributable to people with SPMI. It is certainly true that institutionalization rates have fallen: in 1955, about 27% of those with an SPMI were living in institutions (three quarters of these in psychiatric hospitals), whereas today only about 7% of the SPMI population live in institutions (about two thirds of these in jails and prisons, reflecting the much higher overall incarceration rate today).³ However, a recent review of the literature and a more recent study find no clear relationship between these patterns of deinstitutionalization and patterns of violence or incarceration among people with SPMI.^7,8

The surprising absence of a relationship is likely attributable to 3 factors. First, even in 1955, the vast majority of people with an SPMI were not institutionalized; they lived in the community, as most people with SPMI do today.³ Second, although risk of violence has always been a criterion for commitment, clinicians in 1955 were less able than clinicians today to predict which people with SPMI were most likely to commit violent acts in the short run.⁹ Third, at least half of those with an SPMI living in the community in 1955 received no treatment whatsoever because there were very few treatment resources available in outpatient settings.³ Today, most people with an SPMI who live in the community receive treatment, largely voluntarily.³ Thus, in practice, despite deinstitutionalization, people with an SPMI living in the community probably have no greater risk of violence today, in part because they are much more likely to be receiving effective treatment than was the case in 1955. Returning to the era of institutionalization is unlikely to have much effect and would require a wholesale denial of established civil liberties. It would also be pragmatically impossible, just as it was in 1955, to institutionalize significant segments of the SPMI population of more than 3.5 million in the United States.³

Instead of addressing the connection between SPMI and violence with broad gestures, policymakers would do better to focus on targeting identifiable subpopulations at elevated risk of violence, identifying effective treatments, and determining where policy interventions could reduce barriers to these treatments. Clinical research suggests that there are promising, evidence-based interventions that address conditions most closely linked to violence but that these interventions have not yet been broadly diffused.⁵ Indeed, studies find that many perpetrators had already been in contact with mental health service systems, suggesting the need for improvements in the effectiveness of treatment.^10,11

Several studies document the effectiveness of treatments for conduct disorder in children and youths, including evidence that these treatments significantly reduce rates of violence.^5,12 Effective interventions for substance use disorder and for co-occurring mental illness and substance use treatment also exist.¹³ One important step would be to encourage evidence-based screening and treatment of substance use disorder among people with severe mental disorders—for example, by making application of the evidence-based Screening, Brief Intervention, and Referral to Treatment for people with SPMI a quality metric in treatment programs.¹³ It is also critical to develop, evaluate, and disseminate additional interventions that address SPMI conditions associated with violence.

Targeting funding to interventions for conduct disorders, substance use disorders, and patterns of SPMI that have been linked to later criminal activity would likely come with many benefits, including improved education and labor market outcomes—and it would constitute a down payment on reduced violence. Yet mental health interventions alone are unlikely to have much impact on the overall level of violence in our society. Most people with mental health problems do not commit violent acts, and most violent acts are not committed by people with diagnosed mental disorders.