Using social media content to identity mental health problems: The case of depression in Sina Weibo

Abstract

This paper aims to propose a logistic regression model that would predict depression by relying on content-related factors observed through social media. Specifically, we propose that by examining: 1) emotions embedded in Weibo posts, 2) user’s tendency of self-disclosure, and 3) user characteristics, it could be determined whether the person is likely to be suffering from depression or not. Data was gathered from 48 individual users of Sina Weibo, a popular microblogging platform in mainland China, and was then subjected to quantitative analysis based on several factors. The sample consisted of 5,354 Weibo posts from the 48 individuals, 11 of whom were suffering from depression. We found that some, but not all, negative emotions (e.g., sadness, disgust) were positively related to depression. Furthermore, the two indicators of self-disclosure, number of followers and length of self-introduction in user profile, were negatively related to depression. In addition, user characteristics such as gender and location could determine the likelihood of depression. Finally, we addressed some theoretical and practical implications and suggested several directions for future study on related topics.