Harvard study predicted 75% of suicide attempts a week before they occurred

A Harvard-led research team has done something suicide-prevention experts have long considered out of reach: predicted three-quarters of suicide attempts before they happened — not months in advance, but within the week leading up to the event.

Harvard psychology professor Matthew Nock led the multiyear study, which tracked more than 600 high-risk adults and adolescents who had come to a hospital with suicidal thoughts or behavior; results are due in the October issue of the Journal of Psychopathology and Clinical Science. By combining frequent smartphone surveys with machine-learning analysis, researchers were able to flag 75% of suicide attempts and 87% of suicide-related events — attempts plus hospitalizations taken to prevent attempts — up to seven days in advance.

That precision matters because existing prediction models typically assess risk over months, years, or even decades — timeframes too broad to trigger a timely intervention. The new approach compresses that window dramatically, raising the possibility of reaching someone in the days or hours when the danger is most acute.

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.— Matthew Nock, lead author and Harvard psychology professor

The study enrolled two groups: adults receiving emergency room-based psychiatric treatment and teenagers between 12 and 19 years old who were treated at an inpatient clinic for suicidal thoughts or behavior. Once discharged, participants were sent optional 20-item smartphone surveys — delivered as many as six times a day during the first three months and once daily in the three months that followed — with responses recorded on a sliding scale from zero to ten. Questions covered suicidal thoughts, intent, resistance to the urge to attempt, and emotional states including hopelessness, isolation, anger, worry, energy levels, and positivity.

One finding stood out above the rest: agitation. Each one-point rise on the agitation scale was associated with an 11% higher likelihood of a suicide attempt, a stronger predictive signal than depression. That result echoes earlier research from Nock's lab showing that 90% of people who survived a suicide attempt reported feeling an urge to relieve psychological agitation and pain that turned out to be temporary.

The team also analyzed metadata — how long participants took to open a survey and how long they spent completing it — as an additional signal. A real-time alert system was built into the study to flag participants who indicated intent to attempt, allowing for immediate intervention.

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.— Matthew Nock, lead author and Harvard psychology professor

Nock said the team's aim is to build a system that's scalable and reproducible — and accurate — so they can intervene before these events occur, to keep people safe. Potential interventions the team envisions include automated reminders of coping tools learned in therapy, or prompts to contact a loved one or clinical team.

This research forms one strand of a wider initiative within Nock's lab, which is also exploring wearable sensors capable of monitoring health metrics such as sleep, heart rate, and voice signatures — data that could supplement self-reported survey responses and lessen dependence on what participants choose to disclose.

The stakes are significant. Suicide is the second-leading cause of death for Americans aged 10 to 34, and the fourth-leading cause for those aged 35 to 44. Studies have found that approximately 50% of people who died by suicide had seen a clinician in their final month — a window that existing tools have largely failed to convert into life-saving action.

Nock pioneered the real-time smartphone survey technique the study relies on. The approach addresses a known weakness in prior research: much of the earlier data on suicide attempts came from self-reporting in periodic surveys, where participants were asked whether they had attempted suicide since their last check-in — a method that can miss the texture and timing of a crisis as it builds.

Why it matters — If the system can be scaled and validated beyond this high-risk cohort, it could give clinicians and loved ones a days-long window to intervene before a suicide attempt — transforming crisis care from reactive to anticipatory.

Reported by harvard.edu, current.fas.harvard.edu, nypost.com, news.ycombinator.com