Faster Qualitative Analysis of Hearing Pilot

This post shows how to map the SOUND‑BITES pilot methods to a reproducible, AI-assisted qualitative workflow that moves from audio to insight in days rather than months.

The study targets 60 volunteers and about 600 clients, uses optional 1:1 audio interviews with expected saturation at 9–17 interviews, and operates under Macquarie University HREC approval (ID: 16818, 11 Dec 2025).

• Standardise metadata and batch audio to ensure consistent transcription settings and faster downstream coding.
• Use the Theoretical Domains Framework (TDF) as the deductive coding frame, with a machine-readable codebook (CSV/JSON) for AI-assisted coding and auditability.
• Plan iterative waves of analysis (for example, 5–7 interviews per wave) and log emergent themes to determine saturation early (expected 9–17 interviews).

The SOUND‑BITES pilot pairs Master of Clinical Audiology students and Meals on Wheels NSW volunteers to deliver otoscopy, tablet hearing screening, and education to older clients at home.

SOUND‑BITES is a 2026 pilot that pairs Master of Clinical Audiology students and Meals on Wheels NSW volunteers to deliver otoscopy (Arclight), tablet hearing screening (Sound Scouts) and education to older clients at home.

• Ethics: Macquarie University HREC (ID: 16818), approved 11 Dec 2025.
• Recruitment timeline: students began 01/03/2026; participant recruitment expected complete 12/10/2026; full data collection by 30/04/2027.
• Key qualitative input: optional post-program audio interviews (telephone/Zoom), concurrent analysis to saturation (expected 9–17 interviews).

The qualitative arm collects optional audio interviews by phone or Zoom, transcribes intelligent verbatim, returns transcripts for participant review, and maps codes to the Theoretical Domains Framework.

Interview data are optional, audio‑recorded, and conducted by phone or Zoom.

Transcripts are produced as intelligent verbatim, manually checked against audio, returned to participants for review, then imported into NVivo for deductive content analysis mapped to the Theoretical Domains Framework (TDF).

Coding involves two coders with a third adjudicator for disagreements; data collection and analysis run concurrently to identify saturation (9–17 interviews).

• Interview content: program value, usability of Sound Scouts/Arclight, volunteer comfort, barriers to help‑seeking.
• Analytic frame: deductive coding into TDF domains for implementation evaluation.
• Outcome: acceptability, feasibility, and preliminary help‑seeking behaviours at 6 months.

This checklist describes a two-week batch workflow to process 10 interviews end-to-end using the steps aligned with the study protocol.

Use this checklist to process a batch of 10 interviews end‑to‑end:

• 1) Ingest audio + metadata (site, volunteer/student, date).
• 2) Auto‑transcribe with custom dictionary and enable PII redaction; do quick manual QC.
• 3) Import existing TDF codebook (CSV) and run AI pre‑coding.
• 4) Two researchers reconcile AI tags inside Evidano; track agreements/disagreements.
• 5) Run thematic frequency and cross‑segment reports (site, age band, referral outcome).
• 6) Produce quote packs and visualizations (co‑occurrence network, hierarchy) for stakeholders.
• 7) Export a decision memo and append raw transcripts/QA logs for the ethics file.

Quick QA checks such as sampling AI codes and keeping an audit trail support reproducibility and ethics review.

Small checks that save time:

• Randomly sample 10% of AI codes for manual verification each wave.
• Keep an audit trail linking audio → transcript → code changes for reproducibility and ethics review.
• Log environmental noise metadata and exclude recordings below a set SNR threshold from automated extraction (or treat them as manual jobs).

Teams preparing to analyze SOUND‑BITES interviews should standardise metadata, export the TDF codebook, batch audio for consistent transcription, and use AI to pre‑code so humans can focus on synthesis.

If you’re preparing to analyze SOUND‑BITES‑style interviews, start by standardising metadata, exporting the TDF codebook, and batching audio for consistent transcription settings.

Use AI to pre‑code and surface frequency patterns, then focus human effort on synthesis and interpretation.

To try the exact workflow above and accelerate your qualitative analysis, see Evidano or Try Evidano for free. For the original protocol and full methods, consult the study: PLOS ONE study.

Ethics note: this post focuses on research workflows and does not provide clinical advice. Maintain participant consent, transcript review, and PII safeguards when handling health data.