Investigating the Relationship Between Suicide Risk and Warning Signs Within China and the United States: Using...

This study explores how machine learning can facilitate large-scale content analysis of suicide-related expressions on social media across cultural contexts. Drawing on suicide theory and psychological assessment tools, 3545 English posts and 2326 Chinese posts posted between 2010 and 2019 were collected. A supervised machine learning algorithm was used to screen the posts for suicide-related content and conduct content analysis. Logistic regression analysis of 932 Chinese posts from Zhihu and 877 English posts from Reddit revealed significant cross-cultural differences in suicide risk expressions, such that Chinese posts were more likely to mention sleep issues, feeling tired, eating problems and concentration problems. Notably, the author used binomial distribution to predict the relationship between PHQ items, nationality and suicide risk. Two items of the Patient Health Questionnaire (PHQ-9) - low interest (PHQ-1) and feeling bad about oneself (PHQ-6)-were negatively associated with suicide risk in both samples, suggesting they may reflect general emotional distress or self-reflection rather than imminent risk. These findings underscore the importance of cultural and contextual nuance in interpreting online expressions of suicidality and demonstrate how machine learning can assist researchers in identifying complex patterns in suicidal expression with implications for prevention and treatment.