Suicide risk classification with machine learning techniques in a large Brazilian community sample.

Area of Science:

• Psychiatry
• Machine Learning
• Public Health

Background:

• Suicide is a preventable outcome, yet accurate prediction remains challenging.
• Common mental disorders are associated with increased suicide risk.
• Early identification of at-risk individuals is crucial for preventive interventions.

Purpose of the Study:

• To develop and evaluate machine learning classifiers for identifying increased suicide risk.
• To analyze clinical and sociodemographic data for suicide risk prediction in a Brazilian population.
• To determine the most relevant features for classifying suicide risk.

Main Methods:

• Utilized baseline clinical and sociodemographic data from 4039 adult participants in a Brazilian community sample.
• Developed and compared machine learning models including Elastic Net, Random Forests, Naïve Bayes, and ensemble methods.
• Evaluated model performance using metrics such as Area Under the Receiver Operating Characteristic Curve (AUC ROC), sensitivity, and specificity.

Main Results:

• 1120 participants (27.7%) exhibited increased suicide risk.
• The Random Forests model achieved the highest AUC ROC (0.814), followed by Naïve Bayes (0.798) and Elastic Net (0.773).
• Features related to depression symptoms were most influential in classifying increased suicide risk.

Conclusions:

• Machine learning models demonstrated good performance in classifying increased suicide risk within the study population.
• The developed models can aid in the early identification of individuals at higher risk for suicide.
• Findings support the implementation of targeted preventive interventions for common mental disorders.