What your WGU capstone records about people decides almost everything else
Of everything a WGU IRB application asks, one answer moves more of the file than the rest combined: what will exist, after collection, that points back at a particular person. Not who you spoke to. Not how careful you intend to be. What is written down, and whether you can still reach the thing that connects it to a name.
Identifiability decides your branch. If nothing you keep leads back to a person — and no key you can reach does either — a WGU capstone usually lands on a determination or a light category. If something does, the consent file, data plan and site authorisation are all read closely.
Where does the WGU application ask this?
Directly, and in more than one place. WGU's published Request for Approval to Conduct Research asks the applicant to describe the steps taken to ensure the anonymity or confidentiality of participants and of the collected data, and specifically how personal identifiers will be handled — then adds a follow-up most people skim past: if participants will not take part anonymously or confidentially, explain why that design is necessary.
That follow-up is the tell. The office is not asking for reassurance about your intentions; it is asking you to justify a design choice. Two earlier questions on the same form circle the same ground: one asks what instruments you will use, including any plan to record audio or video, and another asks for the relationship between you and the participants — teacher and learner, supervisor and employee, colleagues on a unit. Recording and relationship are both identifiability questions wearing different clothes.
Why does one answer move the whole file?
Because the regulation builds its central definition on it. At 45 CFR 46.102(e), someone becomes a human subject in one of two ways: an investigator interacts or intervenes with them, or the project obtains, uses, studies, analyses or generates identifiable private information about them. Private information is information given in a setting where no observation was expected, or given for a specific purpose with a reasonable expectation it stays private — a medical record being the textbook case. It is identifiable when the identity is known or may readily be ascertained.
The exemption categories are built on the same hinge. Survey, interview and observation work sits inside 46.104(d)(2) when either identity is not readily ascertainable from what was recorded, or a leak of those answers could not realistically hurt someone's job, money, legal position or good name. Where identities are recorded, that route stays open only through a limited IRB review making the privacy and confidentiality finding at 46.111(a)(7). One answer, and the reading of the file changes shape. Where each branch ends lays the four landings out side by side.
What actually counts as an identifier?
More than names, and more than the eighteen items most people half-remember from a privacy training. The working test is not "is this a name" but "could a reasonable person with this material work out who it was".
| What you keep | How a reviewer reads it | Effect on the file |
|---|---|---|
| Names, initials, employee or record numbers | Direct identifiers, no argument available | Identifiable branch |
| A linkage key you retain, wherever it is stored | Re-identification remains possible by you | Identifiable branch |
| Dates of service joined to one small unit | Indirect identifier; the combination narrows to a person | Usually identifiable |
| Audio, video or photographs | Identifiers in themselves, before transcription | Identifiable, and consent is read closely |
| Small cells — one night-shift educator, two charge nurses | Deducible identity even without a name | Treated as identifiable in practice |
| Distinctive free-text quotes | A colleague could say who said it | Identifiable unless genuinely de-identified in the write-up |
| Survey platform metadata: IP address, device, email | Collected by default unless switched off | Silently identifiable — the most missed row here |
The last row deserves a moment. Online survey tools routinely record respondent IP addresses and, where a link was emailed individually, tie responses to addresses. An application promising anonymous responses while the platform quietly stores both is a contradiction living inside the file, and it is the kind of thing that gets found in the appendix rather than in the narrative. Anonymous-response settings exist; confirm they are on and say so in the data plan.
Does storing the key somewhere else de-identify the data?
No, and this is the single misconception that redirects the most capstones. Published WGU capstone guidance treats a project as not human-subjects research when the information used is de-identified and neither the identifier key nor participant identities are within the candidate's reach. The condition is access, not location. A code list in a locked drawer you hold the key to, a spreadsheet in a folder only you can open, a tab you promise not to look at — each of them keeps re-identification within your reach, and so keeps the data identifiable.
OHRP says the same thing from the other direction: where investigators cannot readily ascertain identity and did not obtain the material through interaction or intervention for research purposes, the analysis does not involve human subjects at all. The determinative words are cannot readily ascertain. Not "will not". Not "would rather not".
The practical consequence is a decision to make deliberately at design time, not by drift. Either the link is genuinely severed — someone outside the project pulls and strips the data, and you never hold the key — or you accept the identifiable branch and build the file it needs. What does not work is claiming the first while operating the second.
What about charts and records you already have access to?
Access through your role is not the same as access for a capstone, and reviewers separate the two cleanly. The questions worth answering before you touch anything: who pulls the data, what fields leave the source system, whether the extract carries dates and unit identifiers, and where the file lives once it is out.
A retrospective extract stripped of identifiers before it reaches you, pulled by someone whose role already permits it, sits very differently from the same extract with medical record numbers, pulled by you, sitting on a laptop. The clinical question is identical. The IRB branch is not.
How do you keep the answer true to the last appendix?
By writing the data plan as a set of facts rather than assurances, and then making every other page obey it.
- Name the collection instrument's settings. Anonymous responses on, IP collection off, no individually tracked links — stated, not implied.
- Name where the data lives. Which encrypted drive or institutional account, not "securely".
- Name who can reach it. A list of people, not a category.
- Name the destruction trigger. The event that ends retention, so the plan does not rely on a date nobody will remember.
- Check the appendices against the claim. If the instrument has a name field, or the recruitment email is addressed personally, or the analysis plan mentions matching pre and post responses by person, the claim in the narrative has already been contradicted.
That last item is where most otherwise clean files fall down. Matching pre and post responses for the same individual requires something that links them — which is a design decision with a branch attached, and it is better made in the open. The folder, document by document covers how the pages are kept in agreement.
What if the project genuinely needs identifiers?
Then it needs them, and saying so is the professional answer. Plenty of legitimate capstone work requires following individuals over time, linking a chart to an outcome, or interviewing named staff. The regulation does not forbid identifiable work; it asks for protections proportionate to it, and WGU's own form invites you to explain why the design requires it.
What costs you is pretending otherwise. An identifiable project described as anonymous does not become lighter; it becomes a file whose narrative and appendices disagree, which is the one thing guaranteed to slow a reader down. If the honest answer is identifiable, the honest move is the fuller folder — and often a conversation about whether a truthful redesign would serve the same aim on a lighter branch.
What to do next
Write down every field you will hold after collection — every column in the spreadsheet, every attribute of every recording, every piece of metadata the platform stores. Then ask, of the list as a whole, whether anyone could work out who a given row belongs to. That list, not your intention, is the answer to the identifiers question.
If it is close, send it to us before you collect. The walk shows what we take off the desk, the questions page covers what candidates ask most, and a free application review names the side of this fork your capstone sits on — including the times the honest answer is that your draft is already clean.
Sources
- Western Governors University Institutional Review Board — Request for Approval to Conduct Research (published copy of the WGU form, PDF)
- WGU Policy Handbook — Institutional Review Board (IRB), cm.wgu.edu
- 45 CFR part 46 (the Common Rule), full regulatory text — see 46.102(e) and 46.104(d), govinfo.gov
- OHRP — Quality Improvement Activities FAQs, hhs.gov
Settle the identifiers answer before you collect anything.
Send the instrument and the data plan as they stand. One of the three consultants reads them against the fork and tells you which side you are on, in writing, at no cost.
Request the free application review- whether anything you keep is identifiable in practice
- whether a small change moves the branch honestly
- what the data plan has to state to hold up