Using Machine Learning to Identify Suicide Risk: A Classification Tree Approach to Prospectively Identify Adolescent Suicide Attempters

Abstract

This study applies classification tree analysis to prospectively identify suicide attempters among a large adolescent community sample, to demonstrate the strengths and limitations of this approach for risk identification. Data were drawn from the National Longitudinal Study of Adolescent to Adult Health. Youth (n = 4,834, M_age = 16.15, SD = 1.63, 52.3% female, 63.7% White) completed at-home interviews at Wave 1 and a measure of suicide attempts 12 months later, at Wave 2. Results indicated two classification tree solutions that maximized risk prediction, with 69.8%/85.7% sensitivity/specificity and 90.6%/70.9% sensitivity/specificity, respectively. Classification trees provide a technique for identification of individuals at-risk for suicide attempts. Classification trees produce easy-to-implement decision rules and tailored screening approaches that can be adapted to the goals of a particular organization.

Keywords:

• adolescent
• classification tree analysis
• machine learning
• suicide attempt

DISCLOSURE STATEMENT

The authors declare no conflicts of interest.

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Notes on contributors

Ryan M. Hill

Ryan M. Hill, Department of Pediatrics, Baylor College of Medicine, Houston, TX, USA.

Benjamin Oosterhoff

Benjamin Oosterhoff, Department of Psychology, Montana State University, Bozeman, MT, USA.

Calvin Do

Calvin Do, Department of Pediatrics, Baylor College of Medicine, Houston, TX, USA.