1 - (Sym 120) Understanding Real-world Personalized Risk Factors of Self-harm Through Online Information-seeking Behavior of Individuals

Suicide continues to be among the top 10 leading causes of death in the US. Despite decades of research, our ability to predict when someone might be at the highest risk of self-harm suicide has not significantly improved. Furthermore, < 1% of the studies look at individualized and proximal risk factors. Personalized social media and online search history data, by contrast, could provide an ongoing real-world datastream revealing internal thoughts and personal states of mind. 

We conducted this study to determine the feasibility and acceptability of using personalized online information-seeking behavior to identify the risk for suicide attempts. Google search data was collected from a subset of participants and de-identified to remove sensitive search queries to maintain participants' privacy. The raw search data were featurized to generate quantitative (#searches, time of day of searches detect) and semantic features using NLP approaches. Individualized nonparametric association analysis was used to assess the magnitude of difference in web search data features derived from periods proximal to the suicide attempt versus participants' typical (baseline) search behavior.

A total of 62 participants who had attempted suicide agreed to participate in the study. The top 3 search constructs associated with attempts were online searching patterns, semantic relatedness of search queries to suicide methods, and anger. Changes in online search behavior proximal to suicide attempts were evident up to 60 days before the attempt. Participants (68%) indicated that use of this personalized web search data for prevention purposes was acceptable with noninvasive potential interventions such as connection to a real person (e.g., friend, family member, or counselor); however, concerns were raised about detection accuracy, privacy, and the potential for overly invasive intervention. 

Using analytical approaches applied to real-world online information-seeking behavior could offer an unparalleled opportunity for researchers to understand the novel and proximal risk factors associated with an individual’s underlying thoughts of killing themselves and potentially provide an early intervention opportunity.