Faster Insights: qualitative analysis of THEO

Fast take: A new protocol published in BMJ Open (Aug 1, 2025) describes a multicentre quasi-experimental study testing a nurse-led Therapeutic Optimisation (THEO) intervention across two NHS Trust older-person wards (study window Feb 2025–May 2026). Read the protocol: www.bmjopen.bmj.com/content/15/8/e102529. If your team will analyse emotional touchpoints, observations and dyadic interviews from THEO, this post explains a reproducible, audit-friendly approach to qualitative analysis of THEO intervention data, mapped to AI-enabled tools (transcription, thematic coding, cross-segment comparison and visualisation) you can run in www.evidano.com. In the next sections you’ll get a brief study snapshot, the core qualitative methods used, a 7-step workflow to go from audio/files to themes, and concrete ways Evidano speeds each step while protecting sensitive data (E2E encryption; no third-party model training). Ethics note: this is research-focused guidance, not clinical advice.

Findings snapshot & key dates

Item	Value	Note / Source
Protocol published	1 August 2025	BMJ Open protocol
Study window	Feb 2025 – May 2026	Implementation staggered across sites
Intervention length (per site)	12 months	THEO delivery per trust
Quantitative data extraction	1 Jan 2015 → +30 days post-intervention	Control period to account for COVID effects
Qualitative sample (per ward per Trust)	Up to 15 staff + 15 patients	Semistructured + dyadic interviews
Workshops	4 workshops (≈5 participants each per trust)	Coproduction guided discussions
Policy context	Excess bed day cost £2, 089–£2, 532/day	Motivation to reduce length of stay

What the THEO protocol collects (methods in plain English)

The THEO protocol combines participatory action research (PAR) with an embedded convergent mixed-methods process evaluation. Core qualitative inputs include:

Values Clarification Exercises, leadership and workplace culture assessments (staff co-researchers).
Emotional touchpoint interviews (staff and up to 3 patients per iteration) and semistructured one-to-one and dyadic interviews (up to 15 staff and 15 patients per ward per trust).
Observations of care (1–3 short non-participant observations per ward) and monthly fourth-generation evaluation workshops (up to 15 sessions).
Process workshops (coproduction) and diaries/engagement logs fed into the process evaluation framed by realist evaluation (context–mechanism–outcome).
Qualitative analysis plan: verbatim transcription, NVivo coding, six-step reflexive thematic analysis with independent coders and triangulation against quantitative indicators.

Why AI-enabled qualitative analysis matters for THEO

Qualitative datasets described in the protocol are multi-modal (audio, notes, workshop transcripts, observation sheets) and longitudinal. Manual handling creates delays, inconsistent coding and missed cross-segment signals (pre vs during intervention).

AI speeds verbatim transcription and supports consistent PII redaction, crucial for sensitive hospital data.
Automated thematic extraction and cross-segment frequency comparisons highlight signal (themes that move from pre→during) faster than manual line-by-line coding.
Visualisations (co-occurrence networks, hierarchical code maps) make CMO (context–mechanism–outcome) patterns visible for coproduction workshops and exec briefs.

How to operationalise qualitative analysis of THEO intervention with Evidano

Ingest & secure: one place for mixed inputs

Upload audio, observation notes, workshop recordings, and the extracted admin tables into Evidano. Files are encrypted at rest and in transit; Evidano does not use your data to train external models.

Use file tags (site, month, WP1/3/4) so you can later run segment comparisons: pre vs during, ward A vs ward B, staff role vs patient.

Transcription & translation with governance

Auto-transcribe interviews with custom dictionaries (medical terms, ward-specific names) and automated PII redaction. This reduces manual QA time from days to hours.

If multilingual participants appear, run translation with the same custom dictionary to keep term consistency across languages.

Thematic coding at scale

Import your codebook or let Evidano propose an initial code hierarchy from the corpus. Apply batch AI-assisted coding, then review edge cases with a small coder team to lock themes (rigorous audit trail recorded).

Use hierarchical codes→subcodes to represent CMO structures (e.g., 'staffing' → 'experienced RN presence' → 'leadership visibility').

Cross-segment & frequency analysis

Run frequency and co-occurrence analyses to spot themes that shift post-THEO and quantify mentions by cohort (patients, registered nurses, allied health).

Overlay administrative indicators (length of stay, falls) to map qualitative signals to outcomes for a convergent mixed-methods write-up.

Visuals & stakeholder artifacts

Generate word clouds, co-occurrence networks, and timeline visuals for the coproduction workshops and final participatory evaluation, ready for slides and reports.

Export interactive lists of quotes linked to source files so co-researchers can verify interpretation during member-checking.

Iterative sensemaking with AI chat & follow-ups

Use AI chat over your uploaded corpus to draft theme summaries, code memos, and suggested CMO statements. Then assign targeted follow-up (AI avatar interviews) for clarifying questions without re-burdening clinicians.

Keep a reproducible log: every AI prompt, human review, and revision is saved for audit and methods transparency.

7-step checklist: from audio to decision-ready themes

Follow this week-by-week plan to analyse THEO qualitative data and feed the process evaluation:

1) Tag & upload all files (site, WP, month) to Evidano; confirm encryption and access controls.
2) Run auto-transcription with custom dictionary + PII redaction; review 10% for QA.
3) Auto-suggest initial codes; import any existing codebook from past ward studies.
4) Apply AI-assisted batch coding; assign 2 human coders to resolve disagreements and finalize theme labels.
5) Run cross-segment frequency & co-occurrence analyses (pre vs during; staff vs patients).
6) Produce visuals (word cloud, co-occurrence map) and an AI-drafted executive summary for the coproduction workshop.
7) Save full audit trail; prepare member-checking packets (quotes + context) and export for publication or stakeholder briefs.

FAQ: qualitative analysis of THEO intervention

Q: How do I compare themes across sites and time?

A: Tag transcripts by site and time period on ingest. Use Evidano cross-segment analysis to compute mention rates, normalized by transcript length, then visualise shifts with side-by-side heatmaps.

Q: Can AI handle dyadic interviews and emotional touchpoints?

A: Yes, transcribe dyads, then use speaker-attribution to separate voices. Emotional touchpoint words can be coded and frequency-tracked to map sentiment trends across the intervention.

Q: Is patient data safe?

A: Use built-in PII redaction at transcription, role-based access, and E2E encryption. Evidano explicitly does not use your data to train third-party models.

Conclusion, next steps

If you’re preparing to analyse THEO protocol data (or a similar nurse-led ward intervention), combining disciplined qualitative methods with AI-assisted tooling will cut synthesis time and improve reproducibility.

Start by uploading a pilot tranche (one site, one month) into www.evidano.com to validate transcription, code suggestions and visual outputs against your NVivo plan.
Want help mapping the THEO codebook to an AI workflow or automating cross-segment reports for your process evaluation? Book a demo or pilot on www.evidano.com and bring your protocol to analysis-ready in days.

Keep reading

Browse all articles