1251 - DEVELOPMENTAL TRAJECTORIES AND PREDICTORS OF NON-SUICIDAL SELF-INJURY IN ADOLESCENTS: A MACHINE LEARNING STUDY

Session: D08S0033 - Suicide, Self-Harm & Risk Behaviours 2
AUTHORS:
Zhang Cai (Collaborative Innovation Center of Assessment for Basic Education Quality, Beijing Normal University ~ Beijing ~ China) , Zhang Zhiqian (Collaborative Innovation Center of Assessment for Basic Education Quality, Beijing Normal University ~ Beijing ~ China) , Wang Yun (Faculty of Psychology, Beijing Normal University ~ Beijing ~ China) , Lei Hanning (Collaborative Innovation Center of Assessment for Basic Education Quality, Beijing Normal University ~ Beijing ~ China) , Ling Furui (Collaborative Innovation Center of Assessment for Basic Education Quality, Beijing Normal University ~ Beijing ~ China) , Chen Fumei (Collaborative Innovation Center of Assessment for Basic Education Quality, Beijing Normal University ~ Beijing ~ China)
Abstract text:
Non-suicidal self-injury (NSSI) refers to the deliberate and direct harm to one's body without suicidal intent. A person-centered approach to tracing the developmental trajectories of NSSI ideation and behavior in early adolescence offers valuable insights into their progression and heterogeneity, thereby enhancing our understanding of the underlying mechanisms. Moreover, as previous research has shown that NSSI is influenced by a combination of individual and environmental factors, this study aims to identify predictors of distinct developmental trajectories in order to improve early detection and enable targeted interventions.
This study followed 11,366 adolescents in Beijing (Mage = 10.72±0.29 years; 48.60% female) from Grade 5 over four years, with annual assessments of NSSI ideation and behavior. Baseline measures included learning attitudes, positive psychological traits, emotional states, and behavioral problems, which were used as predictors. Latent Class Growth Analysis (LCGA) was employed to assess the developmental trajectories of NSSI. Baseline variables were entered into an XGBoost machine learning model, with SHAP values used to evaluate their relative importance.
Five NSSI trajectories were identified: consistently no ideation (58.90%), persistent ideation without behavior (27.04%), transition from ideation to no ideation (10.44%), transition from ideation with behavior to without behavior (2.21%), and transition from ideation without behavior to with behavior (1.41%). The XGBoost model highlighted low mood, self-blame, and family pressure as the strongest predictors of trajectory classification, followed by emotion regulation and self-esteem. Gender differences were also significant, underscoring its importance as a factor. SHAP analysis indicated that predictor importance varied across groups and revealed nonlinear effects. Notably, low mood was the most influential predictor in most of the groups.
This study identified distinct developmental trajectories of adolescent NSSI and key predictors, highlighting the need for early, targeted school and family interventions, and providing a reference for subtype-based early warning models to guide mental health promotion.