Curating qualitative research data for TREs: the UKDS perspective

Curating qualitative research data for TREs: the UKDS perspective 1 Dr. Sharon Bolton, UK Data Service ISDFPN Meeting, 10 December 2025 2 Qualitative data curation at the UKDS • Around 2,000 studies in the UKDS catalogue with a qualitative element (mixedmethods or fully qualitative) • Mostly subject to Safeguarded access (UKDS End-User-Licence or Special Licence Access) • Mainly sets of interview transcripts • A few instances of social media data • Audio and video data generally accepted for archival purposes only (difficult and resource-intensive to anonymise, large size) Qualitative curation – the first stage • Process begins at acquisition stage before data arrive; dialogue with data owner: • Advise on disclosure control – if early enough in project, discuss informed consent and participants’ agreement to archiving • Level of detail in materials; emphasis strongly on data owner and their research team to undertake necessary disclosure-related edits before deposit • Don’t remove too much detail and reduce research utility – adjust access control instead • Once data arrive, Curation team carry out assessment prior to processing: • Disclosure review of all materials. Check that agreed pseudonymisation and other disclosure control edits with the data owner have been done, and applied consistently • Basic formatting checks, ensure collection is complete and all materials have been sent • Ensure documentation is sufficient and comprehensive enough to enable secondary user to make informed analysis 3 Assessing disclosure risk in qualitative data: standard access Disclosure review: • Direct identifiers, including file properties information • personal names, place and company names, street or other location names, phone numbers, email addresses, etc. • Indirect identifiers: elements of text that could reveal interviewee’s identity, either alone or combined with social media or other data • detailed information on employment/workplaces, educational institutions/qualifications, occupations of other family members, personal circumstances and detailed geographical locations, e.g. ‘I won the National Lottery. I was the only winner in North East England last year.’ ‘I live in the converted windmill in the centre of the village. It’s the only windmill painted red in the whole of East Anglia.’ • Remove or anonymise? 4 Assessing disclosure risk in qualitative data in the UKDS TRE • For quantitative secure access data, generally only direct identifiers are removed – data are ‘de-identified’- so in theory we do not remove indirect identifiers • Does this approach work for qualitative data? How does it fit in with the 5 Safes? • Safe projects: research projects are approved by data owners for the public good  • Safe people: researchers are trained and authorised to use data safely  • Safe settings: a SecureLab environment prevents unauthorised use  • Safe outputs: screened and approved outputs that are non-disclosive  • Safe data: data are treated to protect confidentiality ? • For the examples given, some anonymisation may still be needed for qualitative data held in a TRE, but how much? • Qualitative data suitable for a TRE may be sensitive: personal/organisational • Consider informed consent: are participants happy that the other safeguards are strong enough to allow fully detailed data to be available? 5 Qualitative data in the UKDS TRE: starting the journey Long history of archiving qualitative data, preparing to trial it in UKDS TRE. (Other TREs may have made more progress, happy to discuss and learn.) • Resource-intensive processing • AI solutions for anonymisation? Must be offline, local installation • Trial Textwash/Famtafos - most interview transcripts in Word, still supports .txt only. Conversion takes time and loses formatting • Training needed for output checking • Working with depositors… watch this space 6 7 Thank you for listening! Any questions? UK Data Service Data Curation team: [email protected]