One of the most promising instantiations of the Internet of Things (IoT) are mobile health (mHealth) systems, which promise to deliver intelligent health monitoring and assisted living as well as advanced and integrated health services. To realize the full potential of these services, fragmented and heterogeneous data that is generated by different segments of the system need to be consolidated in order to support high-quality processes. This paper proposes a tiered data integration scheme for mHealth systems that works on the schema, entity, and event levels. The proposed scheme incorporates an algorithm that merges and ranks sensor streams for schema integration and event identification, and performs contextual record registration and deduplication for entity resolution. We tested the proposed integration scheme on two sets of sensor-based mHealth data related to human activity recognition. Preliminary results show that the proposed integration scheme contributes to enhancements in event identification precision compared to the classification performance of separate datasets produced within the same mHealth system.