Online Health Communities (OHCs) have become a major source of social support for people with health problems. Using a case study of an OHC among breast cancer survivors, we revealed the types of social support embedded in each post using text mining techniques. Then we aggregated users’ activities in different types of social support and identified different roles that users play in an OHC via unsupervised machine learning techniques. By analyzing how users’ roles change over time, we constructed a transition graph to illustrate the evolution of users’ roles in an OHC. In addition, we discovered that a user’s behavior in receiving social support is correlated with the transition of her role. It was revealed that the types social support received by a user may facilitate or delay her role transitioning. Our research has implications for OHC operators to track users’ behaviors in order to manage and sustain an OHC.