Eating disorders are complex mental disorders and responsible for the highest mortality rate among mental illnesses. Traditional research methods on these diseases mainly rely on personal interview and survey, which are often expensive and time-consuming to reach large populations. Recent studies show that user-generated content on social media provides useful information in understanding these disorders. However, most previous studies focus on analyzing content posted by people who discuss eating disorders on social media. Few studies have explored social interactions among individuals who suffer from these diseases over social media, while social networks play an important role in influencing and shape individual behavior and health. This thesis aims to provide insights into eating disorders and their related communities from a network perspective, particularly to understand how individuals interact with one another, and the interplays between online social networks and individual behaviors. To this end, we first develop a snowball sampling method to automatically gather individ-uals who self-identify as eating disordered in their profile descriptions, as well as their social connections on Twitter, and verify the effectiveness of our sampling method by both computational analysis and manual validation. Second, we examine a large com-munication network of individuals suffering from eating disorders on Twitter to explore how social media shape community structures and facilitate interactions between com-munities with different health-related orientations. Third, we propose to use multilayer networks to model multiplex interactions among individuals and explore how activities of a set of actors in one type of communication correlate and influence activities of the actors in other types of communication. Finally, leveraging the longitudinal data on posting activities in our user samples spanning 1.5 year, we investigate characteristics of dropout behaviors among eating disordered individuals on Twitter and to estimate the causal effects of personal emotions and social networks on dropout behaviors. Our find-ings contribute to understanding of development and maintenance of healthy behaviors and cognition online, and have practical implications for designing network interventions that can promote organizational well-being in online health communities.