Using machine learning and cohort-sequential modeling to predict suicide attempts among Colorado adolescents
Date
2025
Journal Title
Journal ISSN
Volume Title
Abstract
Suicide has become a leading cause of death across the United States with adolescents posed at particular risk for engaging in self-harm and suicidal ideation, plans, and attempts. As the number of suicide attempts increases, the greater the likelihood that someone will continue attempting suicide, incur an injury during one of their attempts, or die by suicide also increases. Although researchers have identified individual predictors of suicide, very few have investigated the intersectional and interacting variables that predict suicide attempts while differentiating the predictors of multiple suicide attempts from predictors of single attempts and those who have not attempted suicide. The present study utilized the exploratory classification trees to identify these predictors of multiple suicide attempts across the 2015, 2017, and 2019 Healthy Kids Colorado Surveys, which assesses the health and safety of Colorado adolescents. The present study sought to identify if the predictors of multiple suicide attempts change over time and for participants with expansive transgender identities and/or sexual orientation. Models identified 26 predictors of multiple suicide attempts with creating a plan for suicide in the last year as the most predictive of multiple suicide attempts, followed by the number of times participants used heroin in their lifetime, the number of times in the past month when participant misused prescription medications, and the number of days in the last month when participants smoked cigarettes. Results support the use of classification trees in identifying risk factors for multiple suicide attempts though replication is necessary to support these findings.
Description
Rights Access
Subject
machine learning
adolescents
suicide