Key-linked on-line databases for clinical research.

Separating patient identification data from clinical data and/or information about biomaterial samples is an effective data protection measure, especially in clinical research employing "on-line", i.e., web-based, data capture. In this paper, we show that this specialised technique can be generalised into a network architecture of interconnected on-line databases potentially serving a variety of purposes. The basic idea of this approach consists of maintaining logical links, i.e., common record keys, between corresponding data structures in pairs of databases while keeping the actual key values hidden from clients. For client systems, simultaneous access to corresponding records is mediated by temporary access tokens. At the relational level, these links are represented by arbitrary unique record keys common to both databases. This architecture allows for integration of related data in different databases without replicating or permanently sharing this data in one place. Each participating on-line database can determine the degree of integration by specifying linkage keys only for those data structures that may be logically connected to other data. Logical links can de designed for specific use cases. In addition, each database controls user access by enforcing its own authorisation scheme. Another advantage is that individual database owners retain considerable leeway in adapting to changing local requirements without compromising the integration into the network. Beyond protecting individual subject identification data, this architecture permits splitting a cooperatively used data pool to achieve many kinds of objectives. Application examples could be clinical registries needing subject contact information for follow-up, biomaterial banks with or without genetic information, and automatic or assisted integration of data from electronic medical records into research data.