The value of integrated data relies upon common data points having a consistent interpretation in each case; to achieve this at scale requires appropriate informatics support. This paper explains how a model-driven approach to software engineering and data management, in which software artefacts are generated automatically from data models, and models are used as metadata, can achieve this. It introduces a simple data modelling language, consistent with standard object modelling notations, together with a set of tools for model creation,maintenance, and deployment. It reports upon the application of this approach in the provision of informatics support for two large-scale clinical research initiatives.