Construction and verification of nomogram prediction model for non-suicidal self-injury in adolescents with depression

Background Accurate identification of adolescents with depression at high risk of non-suicidal self-harm and implementation of appropriate interventions are critical to reducing the risk of self-harm. This study used the nomogram technique to develop a predictive model for predicting non-suicidal self-injury (NSSI) behavior in adolescents with depression. Methods Convenience sampling was used to examine 596 adolescents with depression. Data from one hospital were used as training and internal validation sets (n = 455) and data from two other hospitals as external validation sets (n = 144). In the former, nine optimal predictors were identified by combining the results of one-way analysis, LASSO regression, and binary logistic regression. These predictors were used to construct a graphical column-line model of the NSSI. The receiver operator characteristic (ROC) curve, the area under the curve (AUC), the Hosmer-Lemeshow test, and calibration curves were used to test the discrimination and calibration of the training and internal and external validation sets of the predictive model, and decision curve analysis (DCA) was used to measure clinical applicability. Results Predictors included in the delivery model, history of peer NSSI, parental psychiatric history, sleep duration, social life events, self-esteem, psychological resilience, social support, and degree of depression (P < 0.05). The area under the ROC curve of the model was 0.880 (P < 0.001), and the Hosmer-Lemeshow test (χ2 = 7.19, P = 0.516), showed a maximum Jordan index of 0.668, a specificity of 0.765 and a sensitivity of 0.933. For clinical applicability analyses, the nomogram prediction model had a significant net gain for almost all risk threshold probabilities. In particular, there was a significant net gain for threshold probabilities of 0.1-1.0. Both internal and external validation of the model showed good discrimination, calibration, and clinical applicability. Conclusion The constructed nomogram risk prediction model performs well and can effectively predict the occurrence of NSSI in adolescents with depression, which is scientifically instructive and worthy of further promotion.