Predictive Model of Suicidal Ideation Among Children and Adolescents in Pediatric Outpatient Settings

Early identification of suicide risk in pediatric outpatient settings is crucial for preventive interventions. This multicenter study aimed to develop a machine learning model to predict self-reported suicidal ideation among children and adolescents aged 7-17 years visiting pediatric outpatient settings. Least Absolute Shrinkage and Selection Operator and logistic regression analysis were used for model development and feature selection. Of 855 patients recruited, 329 were included in the analysis (mean age 11.0 years, 61.7% males). Frequent suicidal ideation (defined as always thinking life is not worth living during the past week) was reported by 20 (6.1%) patients. Based on six items assessing child-rated psychosocial functioning and physical and psychological symptoms, the selected model achieved an Area Under the Curve (AUC) of 0.81 (95% CI: 0.70-0.92). This brief six-item model can help identify children at risk of suicide in pediatric outpatient settings, potentially facilitating timely intervention.