Using machine learning to identify suicide risk: A classification tree approach to prospectively identify adolescent suicide attempters

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, Mage = 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.