Suicide is hard to predict

Warning signs are subtle and fleeting. For Americans aged 10 to 34, suicide is the second-leading cause of death.

Most models predict too far out

Typical suicide-prediction models focus on months, years, or even decades. Harvard researchers wanted a much tighter time frame.

Real-time surveys tracked 600 people

Matthew Nock led a study of high-risk adults and teens who had presented to hospitals with suicidal thoughts or behavior. They received smartphone surveys up to six times daily.

75% of attempts predicted accurately

The team predicted 75% of suicide attempts and 87% of suicide-related events in the week before they occurred.

Agitation beats depression as a signal

For every additional point on the agitation scale, odds of attempting suicide increased by 11%. Agitation proved a stronger indicator than depression.

“90% of people who survived a suicide attempt reported feeling an urge to alleviate psychological agitation and pain that turned out to be temporal.”

Alerts enabled immediate intervention

A real-time alert system intervened when someone showed intent to attempt. The goal is to build a scalable system to keep people safe before events occur.

Wearables could track more signals

Nock's lab is exploring wearable sensors to monitor sleep, heart rate, and voice signatures. Interventions could remind people of therapy tools or prompt them to reach out.