When Should Participant Identifiers Be Separated From Research Data?

At What Point Should Names and Other Identifiers Leave the Working Dataset?

A study may legitimately begin with identifiable information. Researchers recruit named participants, obtain consent, schedule interviews, link records, or follow people over time.

But identity does not necessarily need to accompany the data through every later stage.

Should identifiers be separated immediately after enrolment? After data collection? Before transcription? Before analysis? Only after longitudinal follow-up ends?

There is no single moment that fits every study. The useful principle is to separate direct identifiers once they are no longer necessary for the particular research operation, while preserving controlled linkage when the study still legitimately requires it.

Separate Identifiers as Early as the Study Can Function Without Them

Separation Should Follow the Research Workflow

Separation Means Breaking the Routine Proximity Between Identity and Research Content

Separating identifiers does not necessarily mean destroying them.

A research team might remove names, email addresses, telephone numbers, medical record numbers, or other direct identifiers from the working dataset and replace them with a study code. A separate, more restricted file then preserves the relationship between the code and the participant when authorised re-identification remains necessary.

This is commonly a form of coding or pseudonymization rather than anonymization.

Under the UK GDPR framework, pseudonymization means processing personal data so that they can no longer be attributed to a particular person without additional information, provided that additional information is kept separately and subject to technical and organisational measures. The data remain personal data.

The Best Time Is Usually Before People Who Do Not Need Identity Begin Working With the Data

One practical trigger for separation is a change in function.

Recruitment staff may need names and contact details. Once survey responses move to the analysis team, those identifiers may serve no analytical purpose. Separating them before routine analysis prevents identity from accompanying the data merely because it was needed earlier.

This principle can be expressed simply: if the next person or process does not need to know who the participant is, consider removing direct identity before the information reaches that person or process.

That supports role-based access to identifiable research information rather than relying only on promises that authorised users will ignore columns they do not need.

Some Identifiers Can Be Separated Immediately After Collection

Consider a one-time questionnaire in which names are collected only to document enrolment or administer incentives. If no later research procedure requires researchers to connect a participant's identity with their responses, direct identifiers may be separable very early.

In some designs, they may not need to be collected with the responses at all.

Researchers should therefore ask during protocol development whether identifying information can follow a different data path from substantive research information from the beginning.

Longitudinal Studies Often Need the Link for Longer

A longitudinal study must usually recognise that the person who completed Wave 1 is the same person completing Wave 2. The research team may also need to send reminders and schedule later assessments.

That does not require every longitudinal dataset to contain names.

A participant can be assigned a stable study code. The analysis dataset uses that code to connect observations across time, while a separate file connects the code to contact information or identity for authorised follow-up.

The identity link remains, so the data are not necessarily anonymous. Separation instead reduces routine exposure while preserving the longitudinal function.

Identifiers Used Only for Contact Can Often Follow a Separate Path

Email addresses and telephone numbers frequently exist for recruitment, scheduling, reminders, or incentives rather than analysis.

When that is the case, researchers can consider storing participant contact information separately from research responses.

The key design question is whether the contact information needs to remain linkable to responses. Follow-up studies may require that connection. An anonymous survey collecting a separate prize-draw email may not.

Transcription Creates Its Own Separation Decision

Qualitative studies often begin with inherently identifiable source material. An interviewer knows whom they interviewed, and an audio recording may contain a recognisable voice, names, workplaces, relationships, locations, or other identifying details.

A research team may decide to create a working transcript in which unnecessary direct identifiers are removed or replaced with pseudonyms before coding and thematic analysis. The original recording and any identity key can then receive more restricted access than the working transcript.

This does not guarantee anonymity. Narrative details can still identify participants without their names appearing. Separation nevertheless reduces unnecessary exposure of obvious identity information during routine qualitative analysis.

Record Linkage May Require Temporary Access to Identifiers

Some research deliberately connects participant information with health records, educational records, administrative databases, registries, or other sources.

Linkage may require identifiable variables during the matching process. Once the authorised linkage has been completed, the analytical dataset may no longer need those direct identifiers.

One possible architecture therefore separates functions:

The exact process should follow the approved protocol, applicable law, institutional requirements, and data-provider agreements.

Separation Should Happen Before External Sharing Whenever Direct Identity Is Unnecessary

Collaborators and repositories should not receive direct identifiers merely because the originating research team possesses them.

Before data move to another person or organisation, researchers should determine what the recipient actually needs. Depending on the purpose and governing requirements, sharing may involve coded, pseudonymized, de-identified, anonymized, aggregated, or otherwise appropriately prepared information.

The terminology matters because separating direct identifiers does not necessarily remove re-identification risk from indirect identifiers and linked information.

Separate the Key From the Coded Data in a Meaningful Way

If the code key is stored beside the coded dataset with identical access permissions, the files may be technically separate while providing little practical access separation.

Meaningful separation can involve different permissions, storage locations within approved institutional systems, role restrictions, or other technical and organisational controls appropriate to the study.

The aim is to prevent every routine user of the research data from automatically acquiring the information needed to restore identity.

Under OHRP guidance, a code key is precisely what permits coded private information to be linked back to individuals. In certain secondary-research circumstances, whether investigators can obtain that key can determine whether they can readily ascertain identities for purposes of the US Common Rule.

Separation and Permanent Removal Are Different Decisions

A study may separate identifiers early while retaining the linkage key for years because follow-up, withdrawal, safety monitoring, audit requirements, or another approved function remains necessary.

Later, the research team may reach a point at which that connection is no longer needed. Permanently destroying the linkage mechanism is a more consequential step because it may prevent future re-identification by the research team.

The question of when identifiers should be permanently removed therefore comes after, and is distinct from, deciding when they should stop accompanying routine research data.

Separation Can Occur at Different Times for Different Identifiers

There is no requirement that every identifier follow the same schedule.

A home address needed only to mail study materials may become unnecessary after enrolment. An email address may remain necessary through follow-up. A participant code may remain throughout analysis. A signed consent record may have a separate retention requirement.

Thinking of "the identifiers" as one indivisible package can therefore be misleading. Each category can have its own purpose and lifecycle.

The Right Timing Depends on What Would Be Lost by Breaking the Link

Separating identifiers from the working dataset is often reversible because the protected key remains. Permanently eliminating the connection is not.

Before changing the architecture, researchers should identify what participant-specific functions remain necessary.

Separating Identifiers at Different Points in a Longitudinal Study

The identifiers were separated early even though the identity link was retained. Separation reduced routine exposure without sacrificing functions the longitudinal design still required.

Common Mistakes About Separating Participant Identifiers

Use the Next Research Task to Decide Whether Identity Still Belongs With the Data

Instead of choosing one arbitrary project date for "de-identification," examine each transition in the data workflow.

The practical rule is straightforward: direct identity should not accompany a research operation merely because an earlier operation needed it. Treat identity as a functional requirement that must continue to justify its presence at each stage.

Identify the Earliest Practical Point for Separation

Frequently Asked Questions About Separating Participant Identifiers

Separate Identity When the Next Task No Longer Needs It

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.