Voices of Suicidal Distress Online: Developing a Typology of Suicide-Related Disclosures Using a Mixed-Methods Approach with Natural Language Processing and Qualitative Analysis

The lack of understanding of how individuals communicate suicidal stress hinders global suicide intervention plans and practices. This study identifies patterns of online suicidal disclosure by applying natural language processing and thematic analysis to 128,587 narratives written by individuals at risk of suicide on an international social media platform between 2021 and 2022. Six unique themes emerged, reflecting heterogeneous behavior patterns: repetitive suicidal ideation disclosure (36.4%), relational stress (31.9%), suicide attempts and negative healthcare experiences (9.9%), abuse (8.8%), contextual stress (7.2%), and philosophical or informational discussion (5.8%). Across these clusters, three common patterns also emerged-social disconnection, manifestations of stress, and feelings of burdensomeness-highlighting shared underlying experiences. Integrating psychological theory and clinical guidance, this study introduces a typological framework that links data-driven insights with online behavioral patterns related to suicide. Our findings contribute to informing global suicide prevention strategies by providing real-life examples of diverse suicidal expressions.