Chronic disease care, e.g., care of type 2 diabetes mellitus, is a long-term, complex process involving collaboration and coordination among multiple healthcare providers. To facilitate and accelerate the process, it is key to understand the information flow and identify the information dependency (e.g., temporal dependency of co-occurrence and sequential occurrence) during care provision, which is also the objective of this work. Since most health interventions and decisions are made in outpatient encounters for chronic patients, in this paper, we propose an approach to mine temporal information dependency in outpatient encounter records using sequential pattern mining techniques. By exploring the real data of over 10,000 type 2 diabetes patients from three hospitals, the proposed approach effectively works out sets of meaningful information dependency patterns for different patient groups. The discovered information dependency can be used to guide the information sharing between different health providers, and optimize the chronic disease care coordination.