Unifying approaches to TRE fragmentation#
Session 2, Room 2
Chair: Aida Sanchez (UCL)
CONTEXT: We have to share the same survey datasets via several TREs: UK Data Service (survey data + geographical indicators), SAIL DataBank (survey data + Wales admin data), ONS SRS (survey data + linked education data), UK LLC (survey data + linked health data), UK Biobank (survey data + biomedical data). Long separate negotiations with each TRE and different technical requirements (IDs, data transfer, data formats, metadata specifications, version control).
PROBLEM: Very taxing for the local data management team.
QUESTION: Can TREs have a unified technical approach?
CONTEXT: there are many small ‘thematic’ TREs.
PROBLEM: Collaborative projects across ‘thematic’ TREs: researchers cannot transfer research data from one TRE to another, so data canot be pulled together to enable multidisciplinary research and collaboration.
QUESTION: How can data & projects flow between TREs?
CONTEXT: Researchers need to access research data via several TREs.
PROBLEM: Different security requirements, application forms, approval criteria, required researcher accreditation. Very time consuming and admin heavy for researchers and TRE personnel.
QUESTION: Can this be unified or federated? How?
Data discovery across TREs (metadata management): patchy, lack of comprehensive metadata catalogues and standards. Lots of pressure on data management teams to have their local own metadata discovery tool and researchers find it difficult to understand what data are available across TREs. Can this be unified? Who can host this?
Pressure on data management teams to share data via many different TREs with different technical specs:
The different specifications are resulting in lots of overwork for the data management team at UCL, projects waiting months, e.g. depositing/delivering data. Impact: researchers are waiting for a long time
Hard to come up with a unified approach to TREs due to all the different approaches people have
In terms of data movement between TREs the DARE UK federation blueprint is a step towards that, but it’s not a done deal and it’s not a quick fix.
Multidiscliplinary research across different TREs. Is there a future where data can be moved across thematic TREs easily?:
Many thematic TREs will potentially impede a multi-disciplinary approach
Moving toeards R and Python only has had a pushback because of heavy use of SPSS and Stata
In terms of common environments there is (partly) a generational thing going on: older researchers are more comfortable (and have a lot of investment in) tools like SPSS, Stata and SAS, whilst younger researchers are more comfortable with R and Python
Small numner of large TREs (Goldacre recommendation) vs pop-up virtual research TREs (ATI, Francis Crick):
Running both kinds of environment side by side in a single TRE has costs in terms of both licensing and management.
there is room for both models.
ONS IDS and NHS SDEs are the big TRE players: other TREs will need to work around them
How amenable are different institutions to federation?
The emphasis should be on WHO will operate the TREs and what accreditations are needed at organisational level