Machine learning-based nomogram for non-suicidal self-injury among depressed adolescents: a multicentre study

Non-suicidal self-injury (NSSI) is a major clinical concern among adolescents with depression, but reliable tools for individualized risk assessment are limited. This study aimed to develop and validate a clinically applicable nomogram for estimating the probability of NSSI in this population.

Data were obtained from a nationwide multicenter cohort of 2, 343 adolescents with depression recruited from 14 hospitals in China. Participants were randomly divided into training and validation sets. Variables associated with NSSI were selected by utilizing Random Forest integrated with SHAP values combined with logistic regression. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), the Hosmer-Lemeshow test, calibration curves, and decision curve analysis (DCA).

Eight variables were identified, including depression score, sleep medication use, difficulty identifying feelings, age, perceived family support, female, hallucination and externally oriented thinking. The nomogram performed well in both the training and validation cohorts, as evidenced by AUC values of 0.754 (95% CI: 0.726-0.781) and 0.748 (95% CI: 0.707-0.789), together with the calibration curves and DCA.

A nomogram integrating eight clinical and psychosocial variables was developed and validated to estimate NSSI risk among adolescents with depression. This tool may help clinicians estimate the current probability of NSSI and support further psychosocial assessment and individualized clinical management.