Establishment and verification of a diagnostic prediction model for non-suicidal self-injury behaviors in adolescents with depressive disorder

ObjectiveTo establish a diagnostic prediction model for non-suicidal self-injury （NSSI） behaviors in adolescents with depressive disorder， in order to provide references for early identification of NSSI behaviors in them.MethodsRetrospective analysis was performed on the clinical data of adolescents with depressive disorder （n=366） who were admitted to the Pediatric Department of Shenzhen Kangning Hospital from January 1 to December 31， 2021. According to the Diagnostic criteria of Diagnostic and Statistical Manual of Mental Disorders， fifth edition （DSM-5） diagnostic criteria for NSSI， the patients were divided into comorbid NSSI group （n=289） and non-NSSI group （n=77）. The selected adolescents were randomly divided into a training set （n=258） and a verification set （n=108） in a 7∶3 ratio. Logistic regression analysis was used to screen the independent risk factors for NSSI behaviors in adolescents with depressive disorder， which served as the basis for prediction model. Finally， the receiver operating characteristic （ROC） curve was established and the area under curve （AUC） was calculated to evaluate the discrimination in the training set and validation set. Calibration curve was applied to evaluate the calibration degree of the model. The Homser-Lemeshow （HL） test was conducted to evaluate the goodness of fit of the model. And decision curve analysis （DCA） was performed to evaluate the clinical benefit of the model.ResultsGender （β=1.734， OR=5.561， 95% CI： 2.678~11.964）， education level （β=0.864， OR=2.737， 95% CI： 1.174~4.795）， history of suicide attempts （β=0.932， OR=2.539， 95% CI： 1.253~5.144）， being an only child （β=0.745， OR=2.106， 95% CI： 1.029~4.311） and depression severity （β=0.056， OR=1.058， 95% CI： 1.025~1.092） were independent risk factors related to NSSI behaviors in adolescents with depressive disorder （P<0.05 or 0.01）. The AUC was 0.808 （95% CI： 0.746~0.870） in the training set， and was 0.722 （95% CI： 0.581~0.864） in the validation set. The prediction model showed good calibration with the HL test （P=0.561）.ConclusionGender， education level， suicide attempt history， being an only child and depression severity are independent risk factors for NSSI behaviors in adolescents with depressive disorder， furthermore， the diagnostic clinical prediction model constructed using above factors for NSSI behaviors in adolescents with depressive disorder has displayed good sensitivity and specificity.