Talking to people — what an interview or focus-group file has to carry before anybody speaks

A qualitative file is approved on four documents: the guide as it will actually be spoken, the plan for recording and transcribing, consent with its limits stated honestly, and an account of what becomes of the audio. Every worry a reviewer has lives inside one of those four.

Isabel Rourke, PhD · filed 2026-08-23

In short. File the questions in the words you will use, the recording and transcription arrangements, and consent stating the real limits. Discretion can be asked of everybody in a group; it cannot be guaranteed, so do not print that it is.

What has to be in the file before anybody is interviewed?

The conversation itself is the instrument, so the board reads it the way it would read a questionnaire. That means the guide goes in — opening framing, the domains you intend to cover, the main questions in the words you plan to use, and the probes you expect to reach for. Reviewers are not looking for a script to be read robotically; they are looking to see what a participant will actually be asked, because that is what the risk assessment rests on.

Alongside it travel the recruitment materials in the form they will really take, the consent, and any instrument you are borrowing. Aspen’s handbook is specific on that last point: if you intend to modify somebody else’s instrument or protocol, the author’s permission has to be included, and some measures require registration even where no fee applies. The same logic reaches transcription. If audio is going to a service, the reviewer will want that named among the people with access, since a third party listening to a recording is part of your data plan whether or not you thought of it that way.

Two things are worth settling before the file leaves. First, whether the conversation touches anything Aspen treats as sensitive — its handbook lists topics including sexual matters, substance use, mental health and traumatic experience such as combat — because that pushes the file toward fuller scrutiny. Second, who your participants are: a group drawn from a vulnerable population also moves the review level, and neither judgement belongs to you — the handbook reserves that call for the IRB chair or a designee. What each level means in practice is worked through in our piece on the three review levels.

How much can a semi-structured guide drift once it is approved?

Within its approved territory, freely. Beyond it, not at all without going back. Aspen’s change-request form exists for exactly this: any alteration to an approved project — the instrument, the consent, the location, the way people are recruited — must be approved before you proceed, and the handbook notes that the type of review a change gets depends on how big it is and what it does to the balance of risk and benefit.

The practical way through is to write the guide so that ordinary qualitative responsiveness is already inside the approval. Name the domains rather than only the sentences. State in the application that follow-up probes will stay within those domains and will not introduce new topics. Then, if the third interview reveals something genuinely outside them and you want to pursue it, you know the answer: a change request first, questions afterwards. Doing it in that order costs a pass; doing it the other way round is a deviation, and deviations get reported.

Recording: what must consent actually say?

That it is happening, in the document and again out loud before the recorder starts. Guidance for qualitative work is consistent here: disclose the use of recording devices in the consent form and in the consent conversation, and tell people what the recording is for. Beyond that, three details separate a clean file from a queried one.

Recording as a separate choice. Where the project can survive it, let somebody take part while declining to be recorded, and say in the consent what happens then — contemporaneous notes, usually. A single yes covering both is a weaker consent than two.
A withdrawal window after the session. Focus-group guidance suggests giving participants the chance to withdraw their consent once the discussion has finished, when they know what they actually said. Write the window into the consent and honour it in the data plan; a recording nobody can pull back from is a promise you have already broken.
A named path for the audio. Aspen’s application asks for the whereabouts of the data, the people who can open it, the protection around it and the moment it is deleted. In qualitative work, answer that for recording and transcript separately, because they usually have different lifespans — and say when identifiers come out of the transcript.

One habit prevents most later trouble: instruct people at the top of a recorded session to avoid names — their own, colleagues’, patients’ — and to keep other people’s details out of the account. Guidance on group work asks for exactly that, and it is easier to observe than to repair, since a name spoken aloud is in the audio permanently, whatever the transcript later says.

What can a focus group honestly be promised?

Less than most consent forms claim, and the gap is the single most common finding in qualitative files. In a one-to-one interview the confidentiality chain runs through you and whoever you name in the data plan. In a group it runs through everybody in the room, and you have no authority over any of them once they leave. Guidance across boards says the same thing: the nature of a group is such that confidentiality cannot be guaranteed, and participants should be told both that they must not repeat what they hear and that nobody can promise they won’t.

The promise, the mechanism, and what to write instead
What files often promise	Can the mechanism deliver it?	What the consent should say
“Your participation is anonymous”	No — the others can see who came	Participation is known to the group; your identity is not attached to anything reported
“Everything said here stays confidential”	Only for your side of it	We ask everyone to keep the discussion private, and we cannot guarantee that everyone will
“Nothing will be quoted”	Rarely true of qualitative work	Short quotations may appear, without names or details that would identify the speaker
“Nobody at your workplace will know”	Not where the session runs on site	Where and when the session happens, and any steps taken to keep attendance discreet
“The recording is destroyed immediately”	Only if the plan actually says so	The point at which audio is deleted, and how long the de-identified transcript is kept

The federal consent elements make room for this honesty rather than punishing it. Among them sits a requirement to state how far — if at all — confidentiality of identifying records will be kept. That qualifier is doing deliberate work. A file that describes the true limit and the controls around it reads as competent; a file that promises the impossible reads as one nobody has thought through, and boards commonly route recorded group work on sensitive topics to a fuller level of review for precisely that reason.

What breaks confidentiality whatever the consent says?

Duties that exist outside your project. A registered professional hearing about abuse, neglect, or a serious risk of harm to somebody may be obliged to act, and no consent form can suspend that. Where the guide could plausibly lead there — questions about errors, about unsafe staffing, about a colleague’s practice — say so in the consent in a single plain sentence, and describe what you would do. Reviewers are looking for the sentence; its absence in a project about clinical incidents is itself the finding.

The same paragraph should say what happens if somebody discloses distress. A named stopping point, a pause, and a route to support cost nothing to write and answer the question a reviewer will otherwise ask. Where the topic sits on Aspen’s sensitive list, expect that question to be asked carefully.

Where interviews meet your own workplace

If the people you plan to interview are colleagues, everything in our companion piece on surveying the people you work with applies here with more force, because an interview is not anonymous and cannot be made so. Someone outside the reporting line should invite; sessions should be scheduled so attendance is not on display; and the room matters, since a glass office on the main corridor tells everyone who took part. Written permission for the employer’s space, list or platform travels with the site permission letter rather than being assumed.

The qualitative file, item by item

The guide: framing, domains, main questions in the words you will use, and the probes you anticipate.
Recruitment materials in their final form — the poster people will read, the message that will actually be sent, the words you will say when inviting somebody.
Consent stamped to the current protocol version, carrying the required elements, the recording disclosure, the true confidentiality limit and the reporting duties.
For groups: the ground rules you will read aloud, and the withdrawal window after the session ends.
The recording and transcription plan, naming every person and service with access to the audio.
Storage, access, de-identification and destruction, answered separately for audio and transcript.
Site permission from somebody with authority, plus written permission for any employer resource in use.
Author permission where an instrument or published protocol is being adapted, and current training certificates.

Every one of those is read against the others before anything is filed — the consent against the protocol version, the guide against the risk description, the data answers against the transcription arrangement. That cross-read is the whole of how we work.

What to do next

Read your consent form as a participant who is about to say something they would not want repeated on the unit. If it promises more than the room can deliver, the promise is the finding — and it is a cheap one to cure before submission, expensive afterwards.

Send whichever exists so far — topic guide, consent, recording arrangements. All three are read as a reviewer would, and the findings come back written down, for nothing.

Request the free application review

Everything after that reading can be handed over too: the category settled, the documents built, the file lodged, the board’s questions answered one by one. What stays with you is the work itself and what it finds. See the questions answered first.