Social Network Analysis and Modeling of Online Health Communities


Student thesis: Doctoral Thesis

View graph of relations



Awarding Institution
Award date31 Aug 2017


Mental health problems have become increasingly prevalent during the past decades. With the advance of Internet and Web 2.0 technologies, social media have provided novel platforms for users with mental health problems to form online health communities. These online health communities have been used as tools for communicating, sharing knowledge and exchanging social support. The interactions among community members contain valuable information to understand their mutual help behaviors and their mental health condition. In this dissertation, we analyzed the self-disclosure content of users in online health communities through linguistic analysis, and characterized the dynamic structure of user interactions through social network analysis. From the results of both linguistics and social networks, the readers can understand the expression of depression on a large scale and the dissemination of depression-related information in highly mutually connected communities. After the empirical analysis, we also developed a set of generative network models to capture the evolution of the social networks mapped from the real-world online health communities. The proposed models could provide some insights for healthcare providers to understand the underlying mechanisms of social interactions and to facilitate better social support in online health communities.
We first characterized user interactions in an online health community for depressive patients. Linguistic analysis revealed the intensive use of self-focus words and negative emotion content, which were consistent with previous psychology research. In terms of the interaction structures among community members, we found that the social networks possessed small-world and scale-free properties, with a much higher reciprocity and clustering coefficient value. We also observed a number of interesting associations between the social network properties and linguistic properties of depressive community members. Furthermore, we investigated the dynamic patterns of user interactions in a set of online health communities for more mental illnesses. The evolutionary trends revealed the distinctive growth patterns and unique social interaction structures of the depression-related communities, particularly for the member addition and mutual interaction patterns.
Finally, based on the findings in the empirical study, we derived two possible generative mechanisms that led to the formation of the unique interaction structures in the depression-related online health communities. Three generative models were proposed based on these two mechanisms. There were four key types of network properties we aimed to capture with the proposed models: the reciprocal properties, the clustering properties, the degree distributions, and the bow tie structure. Besides, we kept the models simple with a small set of parameters in the generation processes for generality and easy explanation. A comprehensive set of simulation experiments demonstrated that, by appropriate tuning of the parameters, the proposed models could reproduce most of the unique network properties observed in the depression-related online health communities. Therefore, the derived mechanisms make sense of the formation of the unique social network structures and provide some insights in facilitating social support in health-related communities. Besides, the proposed generative models are general in simulating other social networks of varying network properties through appropriate setting of the parameters.