Sensitive qualitative data and pseudonymisation

Arin Tham Savran

OS in the Swedish Context

Göteborg, 2026-04-06

^{| Göteborgs universitet - Chalmers tekniska högskola - Karolinska Institutet - Kungliga Tekniska högskolan - Lunds universitet - Stockholms universitet - Sveriges lantbruksuniversitet - Umeå universitet - Uppsala universitet}

Detta verk är licensierat under en Creative Commons Erkännande 4.0 Internationell Licens.

Session plan

Photo Matthew Henry on Burst

Types of qualitative ”sensitive data collection”

• In-depth interviews
• Focus groups
• Observations
• Field notes
• Diaries/journals
• Audio, video, photo elicitation
• Online forum or social media analysis
• Open-ended questionnaries or surveys
• Document analysis

Svensk nationell datatjänst

3

Sensitive data

• Personal information: ethnicity, political opinion, union membership, religious and philosophical beliefs, health & sexual orientation, genetic info, biometrical info

​

• But also other details ex. criminal history & sanctions record, and financial info

​

• Geographical location of humans, animals, a physical site/entity

​

• And probably some more…

​

Svensk nationell datatjänst

4

GDPR (Dataskyddsförordningen)

• Appropriate measures must be taken to protect sensitive information…
• Pseudonymising or anonymising are important tools

​

• All research data must be stored securely (during and after project finishes) using technical and organisational measures
• University Data Office, IT, and Archive services usually help with this

​

• Remember to ask for help if you don’t know how!

Svensk nationell datatjänst

5

Pseudonymisation & Anonymisation: �In a nutshell

• Create codes or aliases

​

• Use a code key or log

​

• Subject to GDPR as long as the code key exists

​

• If code key is deleted, data is anonymised and GDPR not applicable. But…

Svensk nationell datatjänst

6

Here’s the thing though…

• Swedish Archive's Act

Svensk nationell datatjänst

7

RE-IDENTIFICATION �Indirect identifiers and background variables

• Gender
• Age
• Income
• Occupation
• Municipality of residence
• And so forth…

​

​

Svensk nationell datatjänst

8

1. Remove the ”obvious” and generalise identifying details

• Removing or masking: �”Hello, my name is Anders Andersson!” → ”Hello, my name is Anders Andersson!”
• Generalising: �”I live in Kalix” →. ”I live in a small town in northern Sweden”. �Though perhaps masking more suitable?

Make sure there’s nothing that could identify individuals, study participant or otherwise! Including via re-identification…

Exceptions? Yes, in some cases people want to be identified. But up to PI to ultimately judge the appropriateness!

​

Svensk nationell datatjänst

9

2. Create code/alias key

Use fake names, participant ID’s, or role-based labels.

​

Also, keep the key separately from the research data, stored securely, and accessible only to authorised individuals.

​

Ex:

​

​

Svensk nationell datatjänst

10

3. Pseudonyms for roles, organisations, geographical regions

• [Teacher 1]
• [Teacher 2]
• [School 1]
• [School 2]
• [Region 3]
• [University]
• [major newspaper]
• [municipality]

​

​

​

Svensk nationell datatjänst

11

4. Pseudonymise places, organisations/companies, and networks

​

Original:

“I worked at Volvo Torslanda under my foreman Anders at T3.”

​

Replace with something like:

“I worked for a large manufacturing company in a major Swedish city under a senior manager.”

​

Svensk nationell datatjänst

12

5. Modify dates

Dates can be identifying, especially around incidents, complaints, hospitalisation, legal proceedings, or media-covered events.

​

​

​

​

​

Source: ChatGPT Edu

Svensk nationell datatjänst

13

6. Generalise and categorise background information

​

Examples:

• converting age to age brackets
• recoding municipality into county
• classifying income as low, medium, or high
• recoding political affiliation as right or left
• grouping occupation or workplace into public or private sector

Source: www.researchdata.se

Svensk nationell datatjänst

14

ACTIVITY 1

3-5 min

ACTIVITY 1: Pseudonymised examples from ChatGPT

Svensk nationell datatjänst

16

Pseudonymised examples from ChatGPT (cont.)

Svensk nationell datatjänst

17

Read more…

https://researchdata.se/en/manage-data/data-containing-personal-information/handbook-data-containing-personal-information/methods/methods-for-qualitative-data

​

​

Svensk nationell datatjänst

18

ACTIVITY 2

10 min

CLASS DISCUSSION

10 min

Also, to think about…

• Square [brackets] for substitutions
• Double-check masking function you use
• Save in PDF/A format
• Editable and machine-readable version, such as UTF-8 TXT, DOCX/ODT, or XML, might be relevant, depending on type of research as well as repository requirements
• Before depositing, check for identifiers outside the visible transcript text (next slide)

Svensk nationell datatjänst

21

Swedish National Data Service

22

Questions?

• Was there anything that I didn’t cover? Or didn’t cover enough?

Swedish National Data Service

23

Thanks for listening

Arin Tham Savran

arin.tham@snd.se

www.researchdata.se

^{| University of Gothenburg - Chalmers University of Technology - Karolinska Institutet - KTH Royal Institute of Technology - Lund University - Stockholm University - Swedish University of Agricultural Sciences - Umeå University - Uppsala University}