Harvard study predicts most suicide attempts a week in advance

As a leading researcher on suicide, Matthew K. Nock knows the subject’s grim statistics by heart. Suicide is the second-leading cause of death for Americans aged 10 to 34, trailing only accidents. It claims more lives around the world each year than war, genocide, homicide, and all other violence combined. And 50 percent of people who died by suicide saw a clinician in their final month, according to multiple studies in recent decades, including Nock’s own research.

For Nock, that last statistic is both agonizing and motivating. Some might take it to mean that the warning signs of suicide are imperceptible or think that suicide is somehow inevitable for those in the grips of extreme suffering. Nock, who is both a researcher and a licensed clinician, reads the statistic as a tragic gap in a clinical-care system rooted in scheduled therapy appointments, which can fail to detect the highly variable — and fleeting — nature of suicidal thoughts.

But two new statistics make Nock, the Edgar Pierce Professor of Psychology, cautiously optimistic: In a multiyear study of 600-plus high-risk adults and adolescents, his lab predicted 75 percent of suicide attempts and 87 percent of suicide-related events in the week before they occurred. (Suicide-related events refer to attempts as well as hospitalizations to prevent attempts.)

Those findings, forthcoming in the October issue of the Journal of Psychopathology and Clinical Science, are extraordinary in a field in which the best prediction models have typically focused on the likelihood of a future suicide event for a given individual in the next six months, year, or even the next decade. Data from this traditional modeling has drawn primarily from self-reporting that asks participants if they have attempted suicide since the last survey.

“We’re not good at predicting suicide attempts and suicide death. We need to get better,” Nock said. “We need to understand what puts people at risk, and the ebb and flow of suicidal thoughts — how to better measure them when they occur, and how to better predict increases in risk, so that we can then provide more support, resources, and prevention to help keep people safe.”

A 2011 MacArthur Foundation “genius” grant recipient, Nock pioneered the technique of surveying people in real time — using personal digital assistants, at the dawn of the smartphone age — to demonstrate that suicidal thoughts fluctuate moment to moment.

“For the vast majority of people, suicidal thoughts and suicide risk states are transient. They move into these states where they want to escape from them, and they consider, in some cases, dying by suicide as a way of escape,” Nock said. Many manage to navigate through and live a fruitful life. “What guts me is knowing that a lot of people don’t make it.”

But recording real-time thoughts about suicide and predicting future suicide attempts are two different things. Nock wondered if sophisticated surveying could help fine-tune the prediction process. The answer, in a word, was yes.

“I’m very self-critical, very concerned about us getting this right and being able to predict and have our findings be real and valid,” Nock said. “I was pleasantly surprised at how strong our sensitivity was here.”

Participants were recruited from two groups: adults who had received emergency room-based psychiatric treatment, and young people aged 12 to 19 who had been treated at an inpatient clinic for suicidal thoughts or behavior. Both groups began receiving surveys on an app immediately after release from the hospital or clinic, with six optional surveys a day for the first three months, and one a day for the next three months.

The surveys presented 20 questions with a 0–10 slider — three gauging aspects of suicidal thinking (urge, intent, and ability to resist suicidal urges) and 17 asking about “affective states” (emotions and moods such as negative, hopeless, trapped, isolated, angry, agitated, worried, fatigued, energetic, and positive). Nearly 500 unique participants completed at least one full survey with the affective questions, while inclusively answering a total of more than 77,000 surveys.

In addition to crunching the answers, Nock’s team parsed revealing metadata from the ways people engaged with the surveys, including how long they took to start a survey after receiving a prompt and how long they took to complete it.

The study included a real-time alert system for intervening if a person indicated a high suicidal intent. “That’s the purpose of this, to build a system that’s scalable and reproducible — and accurate — so we can intervene before these events occur, to keep people safe,” Nock said.

Among the affective states, agitation was a far greater indicator of suicide risk than depression, with an 11 percent increase in the likelihood of a suicide attempt for every additional point of agitation a person indicated on the sliding scale. That’s consistent with other recent research from Nock’s lab, which found that 90 percent of people who survived a suicide attempt reported that they had been feeling an urge to alleviate psychological agitation and pain that turned out to be temporal — like “being in a burning room” or wanting to “shut everything off for a couple of days,” Nock said.

As principal investigator, Nock designed and led the study and published the paper with 20 co-authors, including researchers from his own lab as well as colleagues and clinicians from multiple Harvard-affiliated hospitals, the Harvard T.H. Chan School of Public Health, and other institutions.

It’s part of a wider set of research initiatives from his lab aimed at real-time monitoring for suicide risk, including the use of wearable sensors to monitor sleep, heart rate variability, skin conductance, and even voice signatures — as well as “just-in-time” interventions that would remind someone to use tools they’ve learned in therapy or reach out to a loved one or their clinical team to keep them safe at times of elevated risk.

Nock likened monitoring mood fluctuations to tracking blood sugar or heart attack risk while acknowledging that there are additional privacy layers to navigate with mental health. “What excites me about this line of work is with advances in technology, we now have the ability to bring care to the person’s natural environment,” he said. “We have to be careful in how we do it, and we have to work hand in hand with clinicians and with the people with the illnesses and conditions we’re trying to treat, but there’s great possibility here.”

Research described in this story was partially supported by federal funding from the National Institute of Mental Health (U01MH116928).