Childhood maltreatment and adolescent non-suicidal self-injury: cross-sectional conditional process and network analyses of depressive symptoms and physical activity

Abstract

Background Childhood maltreatment is associated with adolescent non-suicidal self-injury (NSSI), but the roles of depressive symptoms and physical activity remain unclear. We examined these relationships using variable-level and network approaches. Methods This cross-sectional study included 2,354 adolescents recruited by convenience sampling from 30 secondary schools in six Chinese provinces. The 12-item behavior dimension of the Adolescent Non-Suicidal Self-Injury Assessment Questionnaire was the primary outcome. Covariate-adjusted indirect-effect and conditional process analyses used HC3 standard errors and 5,000 bootstrap samples; an 18-node regularized partial-correlation network was also estimated. Results Childhood maltreatment correlated positively with depressive symptoms (r = .365) and NSSI behavior (r = .366), while depressive symptoms correlated positively with NSSI behavior (r = .531; all p < .001). Physical activity was negatively associated with depressive symptoms and NSSI behavior. Depressive symptoms showed a significant cross-sectional indirect association between childhood maltreatment and NSSI behavior (effect = .084, 95% bootstrap CI [.072, .098]). The depressive symptoms × physical activity interaction was statistically significant but very small (B = −.015, 95% CI [−.021, −.010]; ΔR2 = .013; f2 = .019), indicating limited practical significance despite the large sample size. Results remained stable after excluding the PHQ-9 self-harm item. Network analysis identified emotional neglect, sleep problems, and physical neglect as prominent positive bridge nodes, while the single NSSI node showed high cross-community connectivity; physical activity showed only weak cross-community connections after accounting for the remaining nodes. Conclusions Depressive symptoms statistically accounted for part of the childhood maltreatment–NSSI association. Although physical activity significantly interacted with depressive symptoms, the interaction explained only 1.3% of additional variance and had limited practical significance. The weak physical activity connections in the network further suggested little unique symptom-level association after shared variance was controlled. Because the data were cross-sectional, temporal mediation and causal inferences cannot be established, and clinical benefit cannot be inferred.