Childhood Trauma and Illness Phase Drive Suicide Risk and Aggression in Bipolar Disorder

Bipolar disorder is one of the most burdensome psychiatric conditions in the world, and its deadliest consequences are not the mood swings themselves but what follows them: suicide attempts, self-harm, and violence directed at other people or at objects. A new cross-sectional study from Fuzhou Neuropsychiatric Hospital Affiliated to Fujian Medical University, published in BMC Psychiatry, takes a careful look at which factors actually track with these three distinct outcomes. The research team, led by Xiaoting Wang and Saizheng Weng with corresponding author Jiawu Ji, argues that clinicians have too often lumped suicide risk, self-directed aggression, and external aggression together as if they were a single phenomenon. Their findings suggest that this assumption is not only wrong but potentially dangerous, because the strongest predictors of each outcome differ in ways that matter for how patients are assessed and treated.

The study recruited 116 patients aged 14 to 70 years who carried ICD-10 diagnoses of bipolar disorder, drawing 81 participants from inpatient wards and 35 from outpatient clinics at a single psychiatric hospital in Fuzhou, China. Each participant was rated on three separate outcome measures. Suicide risk was assessed with the Nurses’ Global Assessment of Suicide Risk, a structured 15-item instrument that nurses use to quantify an individual’s likelihood of attempting suicide. Aggressive behavior was measured with the Modified Overt Aggression Scale, which the researchers split into two subscores: a self-directed aggression score capturing harm aimed at the self, and an external aggression score summing verbal aggression, aggression against objects, and aggression against other people. This separation is the methodological heart of the paper, allowing the team to ask whether the forces behind turning pain inward are the same as those behind turning it outward.

Beyond the clinical outcomes, the researchers administered a battery of well-validated self-report instruments targeting psychosocial factors that have long been suspected of shaping risk in bipolar disorder. The Childhood Trauma Questionnaire quantified experiences of abuse and neglect before adulthood. The Connor-Davidson Resilience Scale measured psychological resilience, the capacity to adapt to stress. The Perceived Social Support Scale captured how supported patients feel by family, friends, and significant others. The Dysfunctional Attitude Scale, in its Chinese version, assessed rigid, maladaptive beliefs about self-worth and achievement. Demographic and clinical variables, including sex, age group, marital status, occupation, age at onset, illness duration, and current illness phase, were recorded alongside these measures.

The statistical strategy was deliberately layered. Bivariate group differences were tested with independent-samples t-tests and one-way ANOVA, and correlations were computed with Pearson’s or Spearman’s coefficients depending on the distribution of the data. To guard against false positives across the many exploratory comparisons, the team applied Benjamini-Hochberg false-discovery-rate correction to the correlation and group-comparison families. The core of the analysis, however, was a set of hierarchical multiple linear regressions, one each for suicide risk, self-directed aggression, and external aggression, with bootstrap and ordinal-logistic sensitivity analyses run on the suicide risk model to check robustness. Variance inflation factors were monitored to ensure that multicollinearity did not distort the estimates.

In the unadjusted comparisons, nearly every demographic variable seemed to matter. Suicide risk scores differed by sex, age group, marital status, and illness phase. Self-directed aggression scores differed by sex, age group, marital status, occupation, age at onset, and illness phase. External aggression scores differed by age group, illness duration, and illness phase. But when all of these variables were entered into the final hierarchical models, a striking pattern emerged: none of the demographic variables that looked significant in simple comparisons survived. What remained standing were two factors, and only two, that independently tracked with all three outcomes: current illness phase and history of childhood trauma.

The suicide risk model explained a substantial share of the variance, with an R-squared of 0.469 and an adjusted R-squared of 0.424. Within that model, illness phase carried the largest standardized coefficient, with patients in a depressive phase scoring significantly higher on suicide risk than patients in a manic phase: a standardized beta of 0.290, an unstandardized coefficient of 3.318 with a 95 percent confidence interval of 1.420 to 5.215, and a p-value below 0.001. Childhood trauma followed closely, with a standardized beta of 0.252 and an unstandardized coefficient of 0.094 per unit increase in trauma score, confidence interval 0.025 to 0.163, p equal to 0.008. In plain terms, a patient with bipolar disorder who is currently depressed and who carries a heavy burden of early adversity is the profile most likely to register high suicide risk on structured assessment.

The self-directed aggression model performed even better statistically, explaining 48.4 percent of the variance, but the hierarchy of predictors flipped. Here childhood trauma was the dominant correlate, with a standardized beta of 0.305 and an unstandardized coefficient of 0.073, confidence interval 0.027 to 0.119, p equal to 0.002. Illness phase remained significant but with a negative direction, beta of minus 0.176, indicating that self-directed aggression was more pronounced outside the manic phase. The external aggression model was weaker, explaining only 17.6 percent of the variance, yet the same two variables held: illness phase with a beta of 0.273 and childhood trauma with a beta of 0.261, both statistically significant. Manic phases and histories of childhood maltreatment were the correlates most strongly associated with aggression aimed outward at the world.

These asymmetries carry real biological and psychological plausibility. Childhood trauma is thought to sensitize the hypothalamic-pituitary-adrenal stress axis, alter emotional regulation circuitry, and instill the kind of dysfunctional self-directed beliefs that the Dysfunctional Attitude Scale was designed to capture. Such early developmental injuries may bias a person toward turning distress inward, which would explain why trauma was the strongest correlate of self-directed aggression and suicide risk. Illness phase, by contrast, reflects the acute neurochemical and behavioral state of the disorder at the moment of assessment. Depression amplifies hopelessness and suicidal ideation, while mania fuels irritability, impulsivity, and confrontational behavior, which fits the finding that the depressive phase predicted higher suicide risk while manic states contributed to external aggression.

The authors are careful about what their design can and cannot show. Because the study is cross-sectional, it captures a single moment in time and cannot establish that childhood trauma or illness phase causes suicide risk or aggression; it can only demonstrate independent association. The sample of 116 patients from a single hospital, though diverse in age and care setting, limits generalizability, and the reliance on self-report questionnaires introduces the possibility of recall bias, particularly for childhood experiences that may date back decades. The external aggression model’s modest explanatory power also suggests that important drivers of outward violence, perhaps substance use, medication status, or personality factors, were not captured in this analysis. The team explicitly calls for longitudinal studies to confirm the associations and to test whether interventions targeting trauma-related sequelae can reduce risk.

Even with those caveats, the practical message for clinicians is concrete and actionable. Routine assessment in bipolar disorder, the study suggests, should systematically include questions about childhood trauma history and a clear documentation of current illness phase, rather than relying on demographic proxies such as sex, age, or marital status that failed to hold up under multivariable adjustment. Structured tools like the NGASR and the MOAS, separated into self-directed and external components, offer a way to make these assessments reproducible across clinical teams. If future longitudinal work confirms these patterns, screening for early adversity could become as standard in bipolar disorder care as mood charting, giving clinicians a second, developmentally rooted axis of risk to monitor alongside the patient’s current phase of illness.

Subject of Research: Clinical and psychosocial correlates of suicide risk and aggression in bipolar disorder

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

Article References: Clinical and psychosocial correlates of suicide risk, self-directed aggression, and external aggression in bipolar disorder: a cross-sectional study. (n.d.). https://doi.org/10.1186/s12888-026-08702-w

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08702-w

Keywords: bipolar disorder, suicide risk, self-directed aggression, external aggression, childhood trauma, illness phase, cross-sectional study, psychiatry, Modified Overt Aggression Scale, NGASR, mental health, risk factors