IRB review
How to get a transcription vendor approved by your IRB (what to submit)
IRBs approve protocols, not vendors. To use an outside transcription service, your protocol has to say who will hear the recordings, how the files travel and where they live, who can access them, whether transcripts are de-identified, and what happens to the data afterward, with the vendor’s confidentiality agreement or BAA and security documentation attached and a sentence in the consent form telling participants a transcription service is involved. Landmark supplies the agreement, the documentation and Safe Harbor de-identification, and its security practices have been through university IRB and IT reviews before.
Key takeaways
- IRBs approve protocols, not vendors. A vendor is "IRB-approved" for your study when your data-handling section describes it accurately and its agreement and security documentation are attached.
- Attach three things: the confidentiality agreement (NDA, or a BAA when the data are PHI from a covered entity), a security summary, and a description of de-identification.
- Focus groups add three requirements: consent that covers the group, speakers identified by label rather than name, and de-identification of what participants say about each other.
- Landmark supplies the agreement, IRB-ready security documentation and Safe Harbor de-identification; focus groups are transcribed from $2.69 per audio minute, one-on-one interviews from $1.99.
What to write in the data-handling section
What to attach
| Document | From Landmark |
|---|---|
| Confidentiality agreement: NDA, or a Business Associate Agreement when the data are PHI from a covered entity | Either, on Landmark’s form or your institution’s template; BAAs in place with many universities. |
| Security summary: encryption, access controls, staff training, storage location | IRB-ready security documentation on request; staff complete human subjects protection and HIPAA training. |
| De-identification description | Safe Harbor de-identification of all 18 identifier categories, included at no extra charge. |
Not sure whether your study needs a BAA at all? The HIPAA-compliant transcription guide covers when HIPAA applies and what to require either way.
Consent-form wording that reviewers accept
Adapt to your IRB’s template; the elements reviewers look for are the outside party, the confidentiality agreement, and de-identification:
Audio recordings will be transcribed by a professional transcription service (Landmark Associates) that has signed a confidentiality agreement. Identifying information will be removed from the transcripts, and recordings will be transferred and stored using encrypted systems.
Focus groups: three extra lines in the protocol
- Consent covers the group. Every participant is recorded and transcribed; the consent language applies to all of them, and participants should be reminded that others in the room will hear what they say.
- Speakers by label, not name. Ask for speaker identification (an optional add-on) as Participant 1, Participant 2, or pseudonyms you assign; names spoken in the room are de-identified in the transcript.
- De-identify what people say about each other. Safe Harbor covers third parties’ identifiers as well as the speaker’s own; say so in the protocol.
What to look for in a focus-group transcription service
Speaker identification
Turns labeled Participant 1, Participant 2 or pseudonyms you assign; an optional add-on at Landmark. Without it a group transcript cannot be coded by speaker.
Overlap handled, not dropped
Crosstalk marked as [crosstalk] with each speaker’s words where audible, so heated exchanges survive into the transcript.
Verbatim style chosen for the analysis
Clean verbatim for thematic work; full verbatim when turn-taking and hesitation are the data. See the verbatim guide.
De-identification that covers third parties
Safe Harbor applies to names and places participants mention about each other, not only their own identifiers. Included at no extra charge at Landmark.
Consent language that covers the whole group
Every participant is recorded and transcribed; the consent form names the outside service and its confidentiality agreement.
A rate that anticipates the speaker count
Recordings with three or more speakers are priced as focus groups (Plus, from $2.69 per audio minute), so the budget line is right the first time.
Files ready for your analysis software
Transcripts formatted for NVivo, ATLAS.ti, Dedoose and MAXQDA, with timestamps as an optional add-on.
Trained people, not a consumer app
Transcription by staff who have completed human subjects protection and HIPAA training and signed confidentiality agreements.
Human transcription or an AI app? What the IRB will ask
Reviewers do not rule on the technology; they ask the same four questions of any tool: who or what receives the audio, under what agreement, where it is stored, and who else can use it. A consumer transcription app answers those with its terms of service, which you have to read and attach, and which may permit the provider to retain or use recordings. A human service under a signed NDA or BAA answers them with the agreement itself. Whichever you choose, write the true answer in the data-handling section; that is what gets approved.
Focus groups and recordings with three or more speakers are transcribed at the Plus rate, from $2.69 per audio minute; one-on-one interviews from $1.99. See qualitative research transcription for methods coverage and formats.
Questions researchers ask
- Does an IRB approve transcription vendors?
- Not directly. IRBs approve protocols. Your protocol describes how recordings leave your team, who hears them, how they are protected, and what happens to them afterward, and the IRB approves that description. A vendor becomes "IRB-approved" for your study when those sections are accurate and the vendor’s agreements and security documentation are attached.
- What documents should I attach for the transcription vendor?
- Three things cover most reviews: the signed confidentiality agreement (an NDA, or a Business Associate Agreement if the data are protected health information), a security summary describing encryption, access controls, staff training and where files are stored, and a description of de-identification. Landmark provides all three on request: BAA or NDA, IRB-ready security documentation, and Safe Harbor de-identification included with transcription.
- Do I need a BAA or is an NDA enough?
- A BAA is required when the recordings contain protected health information from a HIPAA covered entity, such as a university medical center or clinic partner. For other studies a confidentiality agreement or NDA is usually what the IRB expects. If you are unsure, ask your IRB or privacy office; Landmark will sign whichever applies, on your institution’s template if you prefer.
- What should the consent form say about transcription?
- Participants should know that a professional transcription service will hear the recording under a confidentiality agreement. Wording researchers commonly use: "Audio recordings will be transcribed by a professional transcription service that has signed a confidentiality agreement. Identifying information will be removed from the transcripts." Adapt it to your IRB’s template.
- Can the vendor de-identify transcripts before I analyze them?
- Yes. Landmark applies HIPAA Safe Harbor de-identification, removing all 18 identifier categories, at no extra charge with transcription. You choose whether identifiers are removed, replaced with placeholders such as [NAME] or [CITY], or left in place for your own coding.
- Can I use an AI transcription app for IRB-approved focus groups?
- That is your IRB’s decision, and it will turn on four questions: who or what receives the audio, under what agreement, where it is stored, and whether the provider may retain or use the recordings. With a consumer app the answers come from its terms of service, which you attach to the protocol; with a human service they come from a signed NDA or BAA. Describe whichever you use accurately in the data-handling section.
- What is different for focus groups?
- Every participant consents, so the consent language applies to the group; transcripts identify speakers by label or pseudonym rather than name (speaker identification is an optional add-on); and de-identification has to cover what participants say about each other, not only about themselves. Focus groups and recordings with three or more speakers are transcribed at the Plus rate, from $2.69 per audio minute.