Groundbreaking study unveils biomarkers for predicting suicide risk in depression patients

Suicide is a devastating outcome of severe depression, and its prediction remains one of the most challenging tasks in psychiatry. Currently, risk assessment relies heavily on clinical interviews and self-reporting, which can be subjective. A recent study published in General Psychiatry explored the clinical value of routinely tested biomarkers for predicting suicide risk and developed a prognostic model with potential clinical utility using machine learning.

Based on a retrospective analysis of comprehensive data on patients with depression who were first hospitalized between 2016 and 2023, this study included 3,143 individuals from four specialized mental health institutions. Researchers found that 21% were at high risk of suicide. To predict this risk, they developed and compared several computer models, among which the XGBoost model demonstrated the best performance, achieving an accuracy of 82% in identifying high-risk patients.

To better understand the factors most influential in the prediction model, researchers analyzed the key variables that contributed to its decision. The most influential factors included both clinical features—such as prior Electroconvulsive Therapy (ECT), use of physical restraints, and presence of psychotic symptoms—and biochemical biomarkers identified in blood tests, including Thyroid-Stimulating Hormone (TSH) and uric acid levels. The analysis further indicated that these factors interact rather than work in isolation; for example, the effect of ECT on suicide risk was influenced by both age and TSH levels. By categorizing patients based on these key features, the model demonstrated how different combinations of factors contribute to the suicide risk collectively.

When biomarkers were integrated with other types of data—such as treatment history, psychological assessment scores, and life stress factors—the model's performance improved consistently and substantially compared to using these factors alone.

“For decades, the assessment of suicide risk has relied almost exclusively on subjective evaluation,” said Dr. Zhu, the lead author of the study. “Our findings represent a tangible step toward a more objective and precise method. This could fundamentally change how we approach prevention in vulnerable populations.”