Association between hypomania and self-harm behaviors in Chinese children and adolescents with depressive symptoms

Bipolar disorder (BD) is associated with a higher risk of self-harm (SH) when compared with depression. Therefore, it is reasonable to suspect that the state of mania or hypomania may independently contribute to increased SH risk. However, for hypomania, its association with SH remains less known. We intend to investigate this hypothesis in a large sample of Chinese children and adolescents with depressive symptoms. Based on a two-stage simple random cluster sampling method with probability proportionate to sample size (PPS) design, a total of 4,858 children and adolescents aged between 10 and 17 years were surveyed in southwestern China, Yunnan Province, by using self-administered questionnaires. Among them, 1,577 respondents with depressive symptoms were screened out and included in the final analysis. Descriptive statistics were calculated to illustrate the major characteristics of the study subjects. Multivariate logistic regression models were fitted to evaluate the adjusted association between hypomanic symptoms and SH. The prevalence of SH in children and adolescents with depressive symptoms was 63.92% (95% CI: 58.70-69.00%). The two hypomanic factors, which measure "active/elated" (factor I) and "risk-taking/irritable" (factor II), were significantly and discordantly associated with SH: after adjustment, every one-point increase in factor I and factor II scores was associated with 0.94-fold (95% CI: 0.91-0.97) and 1.25-fold (95% CI: 1.15-1.36) of odds ratio (OR) in SH prevalence. Further analyses based on quartiles of the two factors revealed a more prominent dose-response relationship between factor II and SH prevalence, SH repetition, and SH severity. The results of this study may suggest that, for hypomanic children and adolescents, individuals with elevated factor II score are probably of greater urgency for SH intervention. Major limitations of this study include inability of causal inference, risk of information bias, and limited results extrapolation.