Clinical and psychosocial correlates of suicide risk, self-directed aggression, and external aggression in bipolar disorder: a cross-sectional study

Abstract Purpose Suicide risk and aggressive behavior are leading causes of morbidity in bipolar disorder (BD), yet suicide risk, self-directed aggression and external aggression are distinct constructs that are often analyzed interchangeably. This single-center cross-sectional study examined the demographic, clinical and psychosocial variables associated with each of these outcomes in patients with BD. Methods We recruited 116 patients aged 14–70 years with ICD-10 diagnoses of BD from the inpatient ( n = 81) and outpatient ( n = 35) departments of a single psychiatric hospital. Suicide risk was rated with the Nurses’ Global Assessment of Suicide Risk (NGASR), and aggression with the Modified Overt Aggression Scale (MOAS), separated into self-directed (MOAS-S) and external (MOAS-E) aggression. Childhood trauma (CTQ), resilience (CD-RISC), perceived social support (PSSS) and dysfunctional attitudes (DAS) were assessed by self-report. Group differences were examined with independent-samples t -tests and one-way ANOVA, bivariate associations with Pearson’s or Spearman’s correlations, and independent associations with NGASR, MOAS-S and MOAS-E with hierarchical multiple linear regression; bootstrap and ordinal-logistic sensitivity analyses were performed for the NGASR model. Benjamini–Hochberg false-discovery-rate correction was applied to the exploratory correlation and group-comparison families. Results In unadjusted comparisons, NGASR scores differed by sex, age group, marital status and illness phase; MOAS-S scores differed by sex, age group, marital status, occupation, age at onset and illness phase; and MOAS-E scores differed by age group, illness duration and illness phase (all p < 0.05). In the final hierarchical model for suicide risk (R² = 0.469, adjusted R² = 0.424), illness phase had the largest standardized coefficient (depression versus mania: β = 0.290; B = 3.318, 95% CI 1.420–5.215; p < 0.001), followed by childhood trauma (β = 0.252; B = 0.094, 95% CI 0.025–0.163; p = 0.008). In the final model for self-directed aggression (R² = 0.484, adjusted R² = 0.424), childhood trauma had the largest standardized coefficient (β = 0.305; B = 0.073, 95% CI 0.027–0.119; p = 0.002), followed by illness phase (β = -0.176; B = -1.298, 95% CI -2.538 to -0.058; p = 0.040). The final model for external aggression explained less variance (R² = 0.176, adjusted R² = 0.131), with illness phase (β = 0.273; B = 2.606, 95% CI 0.824–4.388; p = 0.005) and childhood trauma (β = 0.261; B = 0.081, 95% CI 0.022–0.140; p = 0.008) as significant correlates. Demographic variables significant in unadjusted comparisons did not remain significant in any final model. Conclusion In this cross-sectional sample, childhood trauma and current illness phase were independently associated with suicide risk, self-directed aggression and external aggression: illness phase was the strongest correlate of suicide risk and external aggression, whereas childhood trauma was the strongest correlate of self-directed aggression. Clinical assessment in BD may benefit from systematically evaluating childhood trauma history and illness phase, but longitudinal studies are needed to confirm these associations and inform intervention strategies.