Profiling Generalized Anxiety Disorder (GAD) on Social Networks (Preprint)

Abstract

Despite a dramatic increase in the number of people struggling with mental health diseases, a significant number still do not seek professional help. That makes mental health problems one of the major causes of disability worldwide, reducing the quality of life (QoL) in sufferers. With the growth in popularity of social media platforms, individuals have become more willing to share their feelings and express emotions through these channels. Therefore, social media data have become a viable digital biomarker for mental health conditions.

This study aims to investigate and analyze GAD sufferers' social media post contents and associated behavioral patterns. These investigations could reveal their mental health status signals to help with digital interventions to promote help.

We collect data from Twitter for users with and without self-reported GAD. Several preprocessing steps are done. For linguistic-based content analysis, we define three measurements derived from the impairments associated with GAD based on the LIWC (Linguistic Inquiry and Word Count). GuidedLDA is also used to identify the themes present in the tweets. In addition, users' behaviors are analyzed based on Twitter metadata to perform an in-depth analysis. Finally, we study the correlation between these themes and their behaviors.

The results of the linguistic analysis indicate differences in cognitive style, personal needs, and emotional expressiveness for GAD sufferers compared to those without GAD. In addition, topic modelling identify four primary themes, i.e., symptoms, relationships, life problems, and feelings. We find that all themes are significantly higher for GAD sufferers than those without GAD. Moreover, studying users’ activities based on Twitter metadata, including hashtag participation, volume, interaction pattern, social engagement, and reactive behaviors, reveals some digital markers of GAD. We find that the individuals without GAD are more active because of higher volumes of friends, followers, listed count, retweets, hashtag participation, and mentions, except replying, which may be due to reactive responses. Furthermore, some interesting correlations between these themes and users' behaviors are also identified.

Several digital makers for GAD have been uncovered through content and behavioral analyses. The findings of these investigations may contribute to developing a robust and reliable assessment tool that psychologists could utilize for the initial diagnosis of GAD or the detection of an early signal of worsening mental health states of GAD sufferers via social media posts.