Comprehensive Summary
This article studied how adult ADHD symptoms and depression are related to suicidal ideation and whether machine learning can predict suicide risk better than traditional methods. The researchers used data from the 2021 Korean National Mental Health Survey, which included over 5,500 adults. Suicide risk was defined as having ever seriously thought about suicide. ADHD symptoms were measured using the ASRS-6, and depressive symptoms were measured using the PHQ-2. The study found that depression was the strongest predictor of suicidal ideation, but ADHD symptoms, especially inattention-related symptoms like difficulty finishing tasks and procrastination, were also linked to higher suicide risk. Instead of using total ADHD or depression scores, the researchers analyzed each symptom separately to see which ones mattered most. This showed that certain ADHD symptoms added risk even when depression was accounted for. The researchers also compared a traditional logistic regression model to a machine learning random forest model. The machine learning model was much better at identifying people who had suicidal ideation, especially in an imbalanced dataset where fewer people reported suicide risk. Overall, the study shows that combining ADHD and depression screening with machine learning can improve suicide risk detection in the general population.
Outcomes and Implications
This study has important implications for suicide prevention and mental health screening. First, it shows that adult ADHD symptoms should not be overlooked when assessing suicide risk, even if a person does not have a formal ADHD diagnosis. Inattention symptoms in particular may contribute to stress, emotional dysregulation, and feelings of failure, which can increase suicidal thoughts. Therefore, the findings suggest that suicide risk screening should include both depression and ADHD symptom assessments. Clinicians often focus heavily on depression, but this study shows that ADHD symptoms add meaningful risk and can help identify vulnerable individuals earlier. Finally, the study supports the use of machine learning tools in mental health care. The random forest model was better at detecting people at risk than traditional statistical models, without showing major bias. This suggests that machine learning could be useful in large-scale screening programs, digital health tools, or population-level surveys to flag high-risk individuals and guide early intervention.