AI-enabled Qualitative Analysis: Childhood Cancer Training

Fast, robust qualitative synthesis matters when implementation teams must explain why a training reduced diagnostic delay by 54.2% and what blocked treatment start. A 2024 quasi-experimental mixed-methods project in Northwest Ethiopia (Jan–Sept 2024; published 29 June 2026) trained 1, 020 health extension workers and dozens of clinicians and reported major knowledge gains and a 54.2% drop in diagnostic interval. Read the PLoS Medicine source here: PLoS Medicine. This post shows how AI-enabled qualitative analysis can turn interviews, focus groups, and implementation logs from that study into reproducible insights and program-ready recommendations, using tools like Evidano to speed coding, triangulate with quantitative charts, and produce stakeholder-ready visualizations.

Key Takeaways

Evidano is an AI-powered qualitative data analysis platform that can convert the Gondar study’s interviews, FGDs, and implementation logs into reproducible insights in days rather than weeks.

The Gondar intervention (Jan–Sept 2024) trained 1, 020 HEWs and reported median knowledge gains from 54.6 to 90.9 and a 54.2% reduction in diagnostic interval, published 29 June 2026 in PLoS Medicine.
Qualitative data (18 IDIs and FGDs) explained drivers of change: leadership support, supply shortages, cultural beliefs, and referral coordination, and these drivers map to concrete fixes.
Using Evidano for transcription, CFIR-seeded coding, cross-segment triangulation, and one-click visualizations can produce stakeholder-ready reports, co-occurrence networks, and quote packs within 48 hours of upload.

Findings snapshot

Metric	Value	Source	Implication
Study period	Jan–Sept 2024	PLoS Medicine	Intervention + 6-month mentorship window
Published	29 June 2026	PLoS Medicine	Peer-reviewed results available for replication
HEWs trained	1, 020	PLoS Medicine	Community referral capacity expanded
Knowledge (median)	from 54.6 → 90.9 (primary level)	PLoS Medicine	Large short-term learning gains
Diagnostic interval	decreased 54.2%	PLoS Medicine	Faster diagnosis; treatment bottlenecks remain
Patient delay	decreased 9.3% (27.0 → 24.5 days)	PLoS Medicine	Small but measurable change in care-seeking

What happened (methods in plain English)

This section explains the study methods and operational context in plain language.

The University of Gondar implemented tiered, on-site training: 10-day immersive sessions for higher-level clinicians, 7-day courses for primary providers, pictorial modules for HEWs, plus a 6-month mentorship phase combining monthly onsite visits and remote supervision.

Mixed-methods evaluation used pre/post KAP surveys, chart review of 100 pediatric cancer records, and qualitative interviews and FGDs analyzed with CFIR.
Operational constraints included active regional conflict (since April 2023) and seasonal rains, which required cascade 'mini-trainings' and increased tele-support mid-study.
Limitations included a quasi-experimental pre–post design without a contemporaneous control, which limits causal claims; qualitative data were used to triangulate and explain mechanisms.

So what for qualitative teams: where thematic analysis adds value

For implementation researchers

This subsection shows how thematic coding explains outcome differences across tiers and timelines.

Use thematic coding to map CFIR constructs (leadership, available resources, networks) to patient-journey intervals: this explains why diagnostic interval dropped 54% while treatment initiation lagged.

Compare codes across facility tiers to identify transfer points (for example, referral-form errors concentrated at a specific level).

For program managers / policy teams

This subsection explains how managers can convert qualitative findings into operational priorities.

Prioritize actionable barriers surfaced in interviews: reagent shortages, transport checkpoints, and cultural beliefs, each aligns with a different funding or operational response.

Turn qualitative quotes into decision memos and QI priorities for district health offices within days, not months.

For funders and evaluators

This subsection explains why qualitative evidence is essential for interpreting fidelity and adaptations.

Qualitative evidence shows fidelity and adaptations (cascade trainings, tele-support) that explain heterogeneous outcomes, use rapid thematic summaries to guide scale-up investments.

Document unintended consequences (HEW workload trade-offs) so scale-up plans include workload mitigation.

Do more, faster with Evidano: operationalizing the study’s qualitative synthesis

Ingest raw materials

This subsection explains what raw assets to collect and upload for reproducible synthesis.

Upload interview and focus-group audio, translated transcripts, mentorship logs, and chart-review notes to Evidano.

Use Evidano transcription with a custom dictionary (local terms like 'nififit') to reduce mistranscription and speed preprocessing.

Automated coding + CFIR mapping

This subsection explains how to seed and lock a CFIR-based codebook with AI assistance.

Run AI-assisted code suggestions seeded with CFIR constructs, then review and lock a codebook once inter-coder agreement stabilizes.

Evidano generates code to subcode hierarchies and applies them across all documents for consistent thematic counts.

Triangulation & cross-segment analysis

This subsection explains how to link themes to facility-level and timeline differences.

Automatically cross-tab themes by facility level (primary/secondary/tertiary), month, or mentor intensity to recreate dose–response patterns reported in the study.

Link thematic clusters to quantitative outcomes, for example facilities with high 'leadership engagement' codes show larger drops in diagnostic interval.

Stakeholder-ready outputs

This subsection explains the outputs you can produce for decision-makers from the coded corpus.

One-click visualizations include co-occurrence networks, hierarchical code trees, and quote packs for briefings.

Export reproducible reports and an AI chat interface to ask project-specific questions over the corpus.

Security & compliance

This subsection states the platform security and data-use posture relevant to de-identified narratives.

Data encrypted in transit and at rest, Evidano models are proprietary and data are never used to train third-party models, important when handling de-identified patient narratives.

Checklist: Reproduce the study’s qualitative synthesis in 7 steps

This checklist lists the concrete steps to reproduce the Gondar study’s qualitative synthesis using the described workflow.

Step 1: Collect raw assets (audio, transcripts, FGDs, mentorship logs, referral forms).

Step 2: Upload to Evidano; run transcription with a custom dictionary for local terms.

Step 3: Seed an initial CFIR-based codebook in Evidano and apply AI-assisted coding across documents.

Step 4: Review and refine codes with 2–3 human validators; lock codebook.

Step 5: Run cross-segment analysis (by facility level, mentor frequency, timeline) and link themes to patient-journey metrics.

Step 6: Generate co-occurrence networks and quote packs for decision-makers.

Step 7: Produce an executive brief and a 90-day operational plan mapping qualitative barriers to concrete actions (supplies, transport, community messaging).

FAQ: AI-enabled qualitative analysis of implementation research

Can AI preserve nuance in translated interviews?

Yes, AI can preserve nuance when paired with custom dictionaries and human review.

Use Evidano’s custom translation dictionary and keep human-in-the-loop review for cultural idioms, for example local disease names.

How do you validate AI-assisted codes?

You validate AI-assisted codes with inter-coder reliability checks and provenance tracking.

Run inter-coder reliability checks on a sample, refine the codebook, and re-apply; Evidano tracks changes and provenance for auditability.

Is the platform secure for de-identified patient narratives?

Yes, the platform can be configured to meet ethical governance and data security needs.

Evidano encrypts data and does not share customer data with third-party LLM trainers, suitable for ethically governed research.

Wrapping up: next steps and a concrete offer

This section summarizes the study implications and offers a practical next step for teams with qualitative data.

The Gondar study (Jan–Sept 2024; published 29 June 2026) shows how multilevel training plus mentorship can change diagnostic timelines, and qualitative evidence explained the how and why.

See the full study: PLoS Medicine.
Try a pilot: Evidano, upload a small corpus (transcripts or audio) and get a themed report, co-occurrence network, and executive quote pack within 48 hours, or Try Evidano for free.

Topics

AI qualitative analysis
implementation research Ethiopia
Gondar childhood cancer training
CFIR thematic coding

Keep reading

Browse all articles