Can teenagers’ smartphone messages reveal suicide risk before hospitalization?

The language used by teenagers in their messages and smartphone searches could help detect an increased risk of suicide in the days leading up to a crisis. A recent analysis examined the written exchanges of five young people hospitalized for suicidal thoughts or actions. For six months, their texts were studied to identify signs of distress or references to suicide.

The results show that four out of the five adolescents used more suicide-related words and expressed more negative feelings in the ten days before their hospitalization. Expressions directly related to suicide were particularly frequent in the five days preceding the crisis, while negativity in messages peaked between five and ten days before. These signals did not appear only during periods of acute risk, making it difficult to distinguish between temporary distress and genuine danger.

Clinicians also noted interpersonal conflicts or academic difficulties in the exchanges—elements that computer tools did not always identify. These tools struggle to grasp the context of conversations, such as the history of exchanges or the tone used. For example, a seemingly innocuous sentence taken in isolation may appear neutral, yet it could reflect deep suffering in its proper context.

The study also highlights that certain themes, such as substance use or references to psychiatric treatments, were easier to detect automatically. However, family tensions or relational problems, often mentioned by clinicians, eluded computer analysis. This is due to the difficulty algorithms face in understanding nuances and implications, especially in adolescent language rich in abbreviations, emojis, or colloquial expressions.

Although the data collected from smartphones is extensive and rarely missing, interpreting it remains a challenge. Current methods can identify clear alerts, such as suicide-related words or a gloomy mood, but they are insufficient for personalizing risk assessment. To improve detection, it would be necessary to better integrate the context of discussions and cross-reference this information with other indicators, such as sleep patterns or movements.

This approach could one day help prevent crises by identifying at-risk youth earlier while respecting their privacy. It also raises ethical questions about digital surveillance and the protection of sensitive data. The goal would be to use these tools to complement, rather than replace, clinical judgment, in order to provide faster and more tailored assistance.

Bibliographie

Source de l’étude

DOI : https://doi.org/10.1038/s44277-026-00057-0

Titre : In their own words: case studies of adolescent smartphone language preceding suicide-related hospitalizations

Revue : NPP—Digital Psychiatry and Neuroscience

Éditeur : Springer Science and Business Media LLC

Auteurs : Isaac N. Treves; Paul A. Bloom; Samantha Salem; Katherine Durham; Valerio Zaccaria; Jamaal Spence; Peter S. Dayan; Lauren S. Chernick; Ashley Blanchard; Jaclyn S. Kirshenbaum; Esha Trivedi; David A. Brent; Nicholas B. Allen; Jamie Zelazny; Karla Joyce; Giovanna Porta; David Pagliaccio; Randy P. Auerbach

Source de l’image

Source de l’image non disponible