Large amount of data related to health care are stored in heterogeneous data sources. Independently, social media provides information about people's environment and activities, such as family relationships or patient's habits and social interaction. This information could be used to complement patients medical profiles to improve patient's care. Providing expert users with mechanisms to integrate and query such sources becomes crucial to retrieve information allowing to improve the analysis of patient's situations. This work contributes to facilitating visualization and querying of data coming from such sources. We adopt a graph data model at the conceptual level as it facilitates the integration of structured and semi-structured data. Our purpose is to go a step forward by providing a conceptual query language intended to allow end users, medical domain experts, to retrieve data from heterogeneous data sources by ad hoc queries. In this paper we introduce a set of operators to query data by transforming a graph and we analyze how they fulfill some design features of the conceptual language. These operators allow successive graph transformation to generate subgraphs with filtered data and to derive new relations representing information that is implicit or that is sparse in the data.