gms | German Medical Science

Background: The i2b2 data warehouse (DWH) with i2b2 webclient as query tool are internationally established software tools in the field of bio- and medical informatics. For data integration, i2b2 serves as DWH to store and connect data from different sources like electronic medical records, laboratory information systems and clinical studies. Users access the cumulative data via the i2b2 webclient query tool. Using their web browser, users can combine query criteria to form complex queries and retrieve information ranging from patient counts to breakdowns (e.g. age) and comprehensive timelines. In Germany, i2b2 is an integral part of the medical informatics initiative's data integration centers [1] and also used in several research projects for data integration and access to local data [2], [3], [4].

While the i2b2 webclient is ideal for querying local DWHs, its use for large research networks as query tool for overarching analyses spanning multiple institutions is limited by three large hurdles: (1) The i2b2 webclient does not support external authentication methods (e.g. OAuth) which prevents integration into IT infrastructures. (2) Query distribution: The official i2b2 SHRINE platform for distributing queries is limited to simple patient counts and requires direct network connections. (3) The local i2b2 installations have no approval workflows for distributed queries to allow approval/rejection of queries on individual basis.

The use of the i2b2 webclient for distributed queries lays outside of the scope of the original purpose of i2b2 – therefore adaptions by its original authors in this regard cannot be expected.

Methods: As the i2b2 webclient is continuously maintained and improved by its authors, an ideal solution to these limitations should not impose changes to the original source code to preserve compatibility to future versions.

To overcome these issues, we devised solutions which don't require changes to the original source code.

Results: For (1) authentication, external login can be performed outside the webclient (e.g. OAuth). Direct login via webclient is disabled and all i2b2 server passwords are set to empty strings. Once the user is externally authenticated, a server side script obtains a token via the i2b2 backend (DWH or façade [5]). Javascript code is attached to the webclient which uses the obtained token to skip the original login screen.

For (2) query distribution and result aggregation, the li2b2 SHRINE component [6] is extended to support arbitrary result breakdowns. Queries are distributed to the federated DWHs and the generated breakdown result is aggregated by the li2b2 broker. The central i2b2 webclient shows the aggregated result breakdown to the user.

For (3) approval workflows, i2b2 workplace folders are used. All federated DWHs use three additional workplace folders: “Approved queries”, “Rejected queries” and “Open queries”. Once a query is distributed to federated DWHs, it is placed in the “open” folder. Users can then move the query to the either folder after review.

Conclusion: The presented extensions to the i2b2 webclient are sufficient to enable its use for federated query infrastructures without requiring changes to the original sourcecode. For local i2b2 DWH installations or data integration centers, no additional user interfaces are needed – users use the regular i2b2 frontend for the approval process.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.