Predictors of trajectories of suicidal thoughts in Texas youth depression and suicide research network youth

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Background: Suicidal ideation among youth is a public health crisis, creating a need to identify features that can prognosticate suicide-related outcomes. This study characterized trajectories of suicidal thoughts in youth with depression/suicidality and evaluated features predicting trajectories. Methods: Participants in the Texas Youth Depression and Suicide Research Network with self-reported suicidal ideation data available for baseline and ≥2 other time points were included ( n = 1,449). Group mixture models were used to identify trajectories of suicidal thoughts at four follow-up assessments during the succeeding 6 months. Random Forest analyses evaluated baseline clinical and sociodemographic variables that predicted trajectory group membership after splitting into training (two-thirds) and validation (one-third) samples. Results: Mean (standard deviation) age was 15.6 (2.7) years, and 72.8% were female. Three trajectories were identified: persistently no (43.3%), persistently low (44.2%), and persistently high (12.4%) suicidal thoughts. In Random Forest analyses, self-reported severity of suicide propensity (including pessimism, helplessness, perceived lack of social support, and despair) plus a clinical interview of suicidal ideation predicted these trajectories. The area under the curve in training and validation samples for predicting persistently no suicidal thoughts was 0.83 and 0.80, respectively, and for persistently high suicidal thoughts was 0.87 and 0.87, respectively. Conclusions: Three distinct trajectories of suicidal thoughts were identified. A combination of baseline self-reported measures of depression, pessimism, helplessness, perceived lack of social support, and despair, along with a clinical interview of suicidal ideation, predicted no or high suicidal thoughts over 6 months with high accuracy, providing a clinically useful predictive model of suicide risk.

This publication has 37 references indexed in Scilit: