Fast take: The SOUND‑BITES pilot (published 14 July 2026) embeds hearing checks into Meals on Wheels deliveries in metropolitan Sydney to test feasibility and acceptability. The study plans n=60 volunteers and ~600 clients/household members, mixed methods with optional audio interviews coded to the Theoretical Domains Framework (TDF), and timelines through April 2027. Read the protocol on PLOS ONE. If you are a qualitative researcher or program evaluator, this post shows an AI-enabled workflow to move from audio to thematic, frequency, and cross-segment insight faster and reproducibly.
Key Takeaways
Evidano is an AI-powered qualitative data analysis platform that accelerates transcription, framework coding, and stakeholder-ready reporting for mixed-methods pilots like SOUND‑BITES.
The SOUND‑BITES pilot (published 14 July 2026) pairs audiology students with Meals on Wheels NSW volunteers to deliver home-based otoscopy and a short tablet hearing screen, with optional audio interviews coded to the Theoretical Domains Framework (TDF).
Applying an AI-enabled workflow can cut weeks from the manual transcription-to-report timeline while preserving auditability and the two-coder reliability workflow required by the protocol.
- SOUND‑BITES plans n=60 volunteers and ~600 clients/household members, with qualitative interviews typically ~4 weeks post-program.
- The protocol specifies deductive TDF coding in NVivo with two coders and a third for arbitration, and targets saturation ranges cited in the protocol.
- Common bottlenecks are manual transcription, transcript checking, inter-coder drift, segment comparisons, and stakeholder-ready reporting.
- Evidano automates verbatim transcription with custom dictionaries and PII redaction, supports AI-assisted framework coding, cross-segment comparisons, and one-click visuals for evidence packs.
- A practical step is to pilot AI-assisted workflows on the first 10 interviews to validate transcripts, coder agreement, and visual outputs.
Study snapshot
The Study snapshot lists key protocol items, values, and sources for SOUND‑BITES.
| Item | Value | Source / Note |
|---|---|---|
| Publication date | 14 July 2026 | SOUND‑BITES protocol |
| Ethics approval | 11 Dec 2025 (Macquarie Univ., ID:16818) | Protocol |
| Target sample | 60 volunteers; ~600 clients/household members | Recruitment across 8 Meals on Wheels NSW sites |
| Qualitative data | Audio-recorded interviews (optional) | 4-weeks post-program; phone or Zoom |
| Primary qualitative framework | Theoretical Domains Framework (TDF) | Deductive coding in NVivo described by authors |
| Planned timelines | Recruitment complete by 12/10/2026; data done 30/04/2027; preliminary results 01/09/2027 | Protocol |
What the SOUND‑BITES protocol actually does
The SOUND‑BITES protocol pairs Master of Clinical Audiology students with Meals on Wheels NSW volunteers to deliver home-based otoscopy and an 8–10 minute Sound Scouts tablet hearing screen.
Visits include education, referrals where needed, and a baseline plus 6-month follow-up survey. Optional interviews capture participant and volunteer perspectives; transcripts are checked manually and coded to the TDF in NVivo. The protocol explicitly aims to evaluate acceptability, feasibility, and preliminary help-seeking outcomes.
- Screening equipment: Arclight otoscope and Sound Scouts app on iPad with Sennheiser HD 400S headphones.
- Interview timing: approximately 4 weeks post program for qualitative semi-structured interviews.
- Qualitative plan: Deductive TDF coding, two coders plus a third for arbitration, with target saturation ranges cited in the protocol.
So what for qualitative researchers and evaluators
Qualitative researchers and evaluators should focus on transcript processes because the work done with interview transcripts determines whether findings are actionable for funders and partners.
The protocol uses standard rigor, including double coding, participant transcript checks, and TDF, but the manual steps described in the protocol add weeks of work. Common bottlenecks in the workflow include:
- Turnaround delays caused by manual transcription and transcript checking before coding.
- Inter-coder drift when maintaining consistency across coders using framework coding.
- Difficulty in quickly comparing segments across sites, volunteer roles, or device referral outcomes.
- Reporting burdens: extracting quotes, visualizing co-occurrence, and producing stakeholder-ready summaries.
Do more, faster with Evidano
1) Fast, audit-ready transcription with privacy controls
Evidano ingests audio and produces intelligent verbatim transcripts with a custom dictionary and automatic PII redaction.
That transcription replaces the manual Word-plus-listen cycle described in the protocol and preserves an audit trail required for ethics reviews.
2) Framework coding at scale
Evidano imports finalized transcripts and applies AI-assisted coding mapped to a TDF codebook.
Evidano suggests subcodes, flags low-confidence segments for human review, and supports the two-coder reliability workflow while reducing initial coding time.
3) Thematic, frequency & cross-segment analysis
Evidano runs thematic summaries, frequency tables, and cross-segment comparisons in minutes rather than weeks.
Examples include counting how often referral barriers appear by site and comparing volunteer versus client perspectives or referral recipients versus non-recipients.
4) Visualizations & evidence packages for stakeholders
Evidano produces one-click visuals such as hierarchical code trees, co-occurrence networks, and filtered quote lists for stakeholder reporting.
Evidano exports assembled evidence packs tailored for Meals on Wheels, funders, or ethics committees.
5) Surveys & mixed-methods joins
Evidano ingests spreadsheets to link HHIE‑S, SSQ‑12, and EQ‑5D‑5L scores to qualitative cases for integrated mixed-methods narratives.
Evidano enables integrated statements like: clients with HHIE‑S > X report Y barrier more often, using REDCap or CSV exports.
6) Secure, research-first data handling
Evidano encrypts data and uses LLMs that are tuned for qualitative research, and uploaded data is not used to train third-party models.
That data governance model supports ethics requirements and participant trust in health research.
A 7-step AI-enabled workflow to reproduce SOUND‑BITES qualitative outputs
This 7-step checklist shows how to go from audio to stakeholder report using the protocol methods and AI-enabled steps.
- 1) Collect audio with consent and metadata (site, role, date).
- 2) Upload recordings to Evidano; apply project-level PII redaction and a custom dictionary for domain terms.
- 3) Auto-transcribe; reviewer listens to flagged low-confidence segments and signs off.
- 4) Import codebook (TDF) and run AI-assisted initial coding; assign two human coders to review divergent items.
- 5) Generate thematic summaries, code frequency tables, and cross-segment comparisons by site, volunteer experience, and help-seeking outcome.
- 6) Create visuals (co-occurrence network, hierarchical codes to subcodes) and pull vetted quotes into an evidence pack.
- 7) Export a concise decision memo for Meals on Wheels and funders: acceptability rate, common barriers, recommended SOP changes, and cost-utility inputs linked to EQ‑5D‑5L and help-seeking data.
FAQ: common questions about applying AI to this protocol
Does AI replace human coders?
No, AI does not replace human coders for this protocol.
Evidano accelerates the initial coding pass, surfaces low-confidence segments, and enforces reproducible codebook application, while human review remains critical for validity, especially with the TDF.
How many interviews before saturation?
Empirical saturation is often reached around 9–17 interviews for many studies, according to the protocol's citation.
The protocol cites Hennink & Kaiser (2022) for that saturation range, and AI tools can help identify saturation earlier by tracking new theme emergence over time.
Is this ethical for health research?
Yes, AI can be used ethically for health research when consent, PII redaction, and data governance are followed.
Analyses must remain research-focused and non-diagnostic, and clinical referrals stay the responsibility of qualified clinicians.
Wrapping up & next steps
The SOUND‑BITES pilot has clear qualitative procedures but manual steps that slow time-to-insight, and an AI-enabled research platform can cut weeks from the timeline while preserving auditability.
- Start small: pilot Evidano on the first 10 interviews to validate transcripts, coder agreement, and visual outputs.
- Evidano integrates with REDCap/CSV exports for linking HHIE‑S and other scales to qualitative cases.
- For a walkthrough or demo of how this maps onto SOUND‑BITES methods, see Evidano or request a pilot project to try the 7-step workflow above; alternatively, Try Evidano for free.
Ethics note: Use AI tools only within approved protocols and with participant consent; outputs are for research and program evaluation, not clinical diagnosis.