After beloved actor Robin Williams died by suicide in August of 2014, a tragic pattern followed: a suicide spike, eventually estimated at more than 1800 additional deaths over 4 months in the United States. Such suicide contagion, as it’s called, is a well-known phenomenon, especially after the death of a celebrity. Smaller clusters can also occur in high schools and other communities, though such contagion can be blunted with certain interventions.
Still, much remains mysterious, including how suicidal thinking and behavior spread, and what fuels transmission. A new study published today in Science Advances explores those questions by modeling suicide much like a pathogen, such as flu or COVID-19. The authors stress that the novel strategy is at least as important as the initial results their model generated. They hope in the long run such modeling could advance understanding of suicide contagion, help predict suicide risk, and better target preventive efforts.
Although suicide isn’t transmitted the way a virus is, the model the team produced is “very convincing,” says Nick Allen, a clinical psychologist who studies adolescent suicide prevention at the University of Oregon and who was not involved in the work. He’s keen for more granular detail—for example, he says, it “would be great … to understand more about different patterns in the community, whether there are certain groups who are more susceptible to this kind of contagion.”
This initial model, though, begins with a broad sweep. Columbia University climate and infectious disease modeler Jeffrey Shaman, along with Columbia psychiatric epidemiologist Katherine Keyes and colleagues, first aimed to establish background rates of suicidal thinking and behavior in the U.S. population. They relied on birth and death rates, national surveys that include questions about suicidal thinking (known as suicidal ideation), calls to the national suicide hotline, and other data. They also quantified three categories of people: those who weren’t thinking about suicide, those who were having thoughts of suicide, and those who had died by suicide in recent days. The latter two groups, Shaman notes, can in turn affect suicidal thinking and behavior in other people.
The resulting model was a series of mathematical equations that represent those three groups and how they influence each other. The scientists then observed what happened when the model was disrupted—when a celebrity died by suicide. They studied the effects of Williams’s death, as well as those of fashion designer Kate Spade and chef Anthony Bourdain, which occurred 3 days apart in 2018 and were examined in the study as a single event.
To capture the effects of those deaths on contagion, the scientists incorporated additional information: first, call volume and call times from what’s now the 988 Suicide and Crisis Lifeline, which served as an estimate for suicidal ideation; and second, mortality data from the National Center for Health Statistics. After Williams’s death, their model suggested that, for people who were not ideating, the risk of starting to do so briefly increased 1000-fold. For those who were already having ideations, suicide became three times as “contagious” as it was before—in other words, the risk of dying by suicide was three times greater than before Williams’s death.
The suicide contagion patterns were similar following the deaths of Spade and Bourdain, though not quite as pronounced—perhaps because media coverage had become more responsible, Keyes says. (Her team previously published research on contagion following Williams’s death, and she suspects some of the suicide spike was due to news stories that violated expert guidance on suicide coverage, including glorifying the death and describing the method used.)
It’s also possible that fewer people in the general public felt as deeply connected to Spade and Bourdain. “I remember when Robin Williams died, I was crushed,” Shaman says. “It was what you could call a gut punch. … I don’t think I was ideating, but I could feel the effect.”
In both 2014 and 2018, the model showed, the spike in contagion subsided quickly, dropping to near baseline within a month or so. Many of those in the ideating category transitioned back to nonideation.
There are limitations to the data, Allen and others note. For example, calls to the suicide hotline may increase because of greater suicide ideation, Allen says, but also because the hotline is publicized in news stories after a celebrity death.
Ultimately, all in the field share the same goal: preventing more deaths. One way to do that, says Ronald Kessler, a psychiatric epidemiologist at Harvard University, is by analyzing subgroups impacted by a particular suicide death—men versus women, for example, or different age groups. This would allow for prevention efforts, such as direct outreach to those at risk, to be more targeted. He calls the results of this model “reasonable but limited,” in part because they’re so broad.
Keyes agrees the model is “a first pass.” She and Shaman hope to incorporate more information over time, such as hospitalization data from suicide attempts. Shaman also wants to understand how much of the contagion effect is due to the original death versus the secondary effect from contact with those impacted—again, similar to the spread of pathogens.
With refinements, Keyes hopes the model could be used to forecast and respond to contagion risk after a new event, slowing the spread of harm and heartache.
