Backend federation #2
Description
The mechanism by which data from heterogenous patient-identifiable patient sources can be safely combined, analysed and discarded.
For example, it should be possible to write a study definition that uses TPP and EMIS patient data and ONS death data, where the framework handles:
- Identifying which backend can supply each column of data
- Routing requests to that backend
- Coordinating secret-sharing for hashing patient identifiers
- Coordinating upload of each component to a shared area
- Combining the data
- Making it available to later steps in the pipeline
- Deleting this data as early as possible (there is a tradeoff between caching data for iterative development, and only keeping it for as a long as is strictly necessary)
This system requires a trusted (cloud?) infrastructure for the combination and storage of temporary cohorts and the execution of pipeline actions against them.