Harvard Study Predicts 75% of Suicide Attempts One Week in Advance

Researchers at Harvard University have developed a method to predict 75% of suicide attempts and 87% of suicide-related events—including hospitalizations to prevent attempts—within a one-week window. The study represents a fundamental shift in psychiatric forecasting by focusing on short-term fluctuations rather than static, long-term risk factors.

Led by Matthew K. Nock, the multiyear study tracked nearly 500 participants, including youth aged 12 to 19 from inpatient clinics and adults from emergency room-based psychiatric treatment. To capture the "ebb and flow" of mental states, the team utilized a real-time app-based surveying system. Participants completed up to six optional surveys daily for the first three months, followed by one daily survey for the subsequent three months, resulting in more than 77,000 total data points.

A Shift in Methodology

Historically, suicide prediction has relied on retrospective self-reporting or static risk factors measured over windows ranging from six months to a decade. Nock’s approach treats suicidal ideation as a transient state rather than a constant trait. By using ecological momentary assessment—surveying individuals in their natural environments—the research captures acute mood shifts that traditional clinical appointments often miss.

One of the study's most critical findings is the role of psychological agitation. The data revealed that agitation is a stronger indicator of immediate risk than depression. Specifically, every one-point increase on the agitation scale was associated with an 11% increase in the likelihood of a suicide attempt. According to the Harvard Gazette, 90% of survivors of suicide attempts described an urge to alleviate this psychological pain and agitation, comparing the experience to being in a "burning room."

Implications for Intervention

This ability to narrow the prediction window to a single week enables "just-in-time" interventions. The urgency is underscored by the fact that 50% of individuals who die by suicide visited a clinician in their final month but were not intercepted. By identifying acute spikes in intent and agitation via scalable technology, clinicians can provide targeted support at the moment of highest risk.

The Path Forward

Despite these gains, Nock emphasizes that the field still has significant ground to cover, stating, "We’re not good at predicting suicide attempts and suicide death. We need to get better." Future efforts will likely focus on how to integrate these real-time digital signals into standard clinical workflows to ensure that a predicted spike in risk triggers an immediate, life-saving response.