Complex Network Structure Patterns in Open Internet Communities for People with Diabetes

Abstract

Type 2 diabetes is one of the greatest challenges that continues to grow because of the ageing population, increasing morbid obesity and sedentary lifestyles. Social media such as Facebook and YouTube have transformed the way people interact in general and on the Internet, but the role of social media in healthcare is still not well-understood. Current understanding of the association between user interaction patterns and health outcomes or behaviour change is still limited. In this dissertation I present a framework, based on social network analysis, to explore the nature of patient interactions in online communities. Results show that people with diabetes join online communities typically immediately following diagnosis, with over 80% of the patients having being diagnosed in under 2 years. The networks are very centralized with continually shrinking density and diameter as the the networks grow, and these results directly contrast with current evidence about non-healthcare social networks. Further, using machine learning techniques, I show that we can predict health outcomes such as weight loss performance based on how the patients interact online. The results have practical relevance for understanding the nature of patients interactions, as well as for designing personalized diabetes interventions based on emergent social technologies.