Due to the large development of medical information systems over the last few years, there is today a strong need for an infrastructure that uniformly integrates the distributed and heterogeneous collections of patient data to deliver value-added information to healthcare professionals at the points of care. The adoption of Electronic Health Records (EHRs) and Web services as a software infrastructure has become an extremely important prerequisite for patient data integration. In this paper we propose a semantic-enabled architecture for the automatic composition of EHR (Electronic Health Record) DaaSs (Data-as-a-Service). In our architecture, DaaSs are selected and composed automatically to resolve the user queries (i.e. queries posed by physicians, nurses, etc) using a query rewriting approach. Our proposed approach can also handle the semantic conflicts of data exchanged among component services in an EHR DaaS composition by deriving and applying automatically the necessary data conversions.