Machine learning-enhanced mapping of suicide risk in Bipolar Disorder: A multi-modal analysis

Area of Science:

• Psychiatry
• Immunology
• Endocrinology

Background:

• Bipolar disorder (BD) presents a significant suicide risk, yet the contributing factors are not fully understood.
• Comprehensive analysis of demographic, clinical, and biological factors is needed to understand suicide risk in BD.
• Developing an integrated suicide risk assessment model for BD patients is crucial.

Purpose of the Study:

• To analyze demographic, clinical, and biological factors linked to suicide risk in bipolar disorder patients.
• To develop and validate a novel suicide risk assessment model for bipolar disorder.
• To identify potential biomarkers for suicide risk stratification in bipolar disorder.

Main Methods:

• Cross-sectional study of 152 bipolar disorder patients across four suicide-risk groups.
• Utilized Mini-International Neuropsychiatric Interview (M.I.N·I.), HAMD-24, YMRS, MADRS, and BSSI for clinical assessments.
• Evaluated thyroid function, inflammatory markers, and lymphocyte subsets (e.g., CD3+ T-cells), alongside machine learning models for risk prediction.

Main Results:

• Depressive symptoms significantly increased odds of medium and high suicide risk (P < 0.01).
• Lower free thyroxine 4 (FT4) levels correlated with increased low and medium suicide risk (P < 0.05).
• Machine learning models achieved 87.1% accuracy, identifying depressive symptom scales, FT4, and interferon-γ as key predictors.

Conclusions:

• Depressive symptoms and thyroid function are critical for suicide risk assessment in bipolar disorder.
• Thyroid autoimmunity and T cell-mediated immunity show potential as biomarkers for risk stratification.
• These findings suggest new personalized intervention strategies for suicide risk in bipolar disorder.