Qualitative analysis of breastfeeding training

The Population–Concept–Context (PCC) framework makes study selection reproducible, useful when you want to reproduce counts (how many simulation-based programmes vs web-based modules) or compare LMIC vs HIC implementations.

The protocol explicitly includes qualitative studies and programme reports. That breadth demands a workflow that preserves nuance (quotes, context) while enabling cross-study theme counts, a classic AI-assisted qualitative research job.

By including literature since 2000 and planning searches through 2025, the review captures digital, hybrid and simulation-based training innovations introduced or accelerated since COVID-19.

Export search results (RIS/CSV) and batch-download PDFs. Use a platform that ingests mixed inputs (PDFs, Word, web pages).

Evidano fit: bulk import of PDFs and scraped web reports; metadata parsing into a single corpus (title, year, country, setting).

Extract methods, intervention descriptions, and outcome sections as text; translate Chinese records with a custom dictionary to preserve technical terms.

Evidano fit: built-in translation with custom dictionary to keep terms like “Baby‑Friendly Hospital Initiative” and clinical skills labels consistent.

Seed codes from the protocol’s PCC headings (format, pedagogy, outcomes, contextual factors) and add emergent codes from a 10% sample.

Evidano fit: import a codebook or generate one automatically from frequent phrases; iterate with human-in-the-loop editing.

Run AI-assisted coding to label passages; then produce frequency tables by region, setting (NICU vs maternity), and training type.

Evidano fit: thematic, content frequency, and cross-segment analyses (click to compare Chinese vs English studies, simulation vs web modules).

Manually review coded excerpts, extract representative quotes, and track provenance for PRISMA-ScR reporting.

Evidano fit: clickable source-to-quote links and exportable evidence tables for appendices.

Create co-occurrence networks, hierarchical code→subcode trees, and simple counts (n programs by pedagogy).

Evidano fit: one-click visualizations (word cloud, co-occurrence network, hierarchical coding) to share with stakeholders.

Export tables (study characteristics), PRISMA flow diagram data, and the narrative synthesis; include limitations about language bias (English/Chinese) and grey literature coverage.

Evidano fit: export reproducible tables and an audit trail to support method transparency.

Manual translation and normalization of technical terms slows synthesis.

Evidano: automated translation with custom dictionary + unified ingestion of PDFs and web reports to keep terminology consistent across English and Chinese sources.

Manual thematic coding is time-consuming and hard to reproduce.

Evidano: AI-assisted coding with editable codebooks, frequency counts, and exportable code→quote provenance for PRISMA-ScR compliance.

Government and NGO reports often live on websites; harvesting them is manual.

Evidano: website/social scraping to ingest training manuals and hospital reports, then include them in the same analytic pipeline as journal articles.

Stakeholders need concise tables and visuals, not raw excerpts.

Evidano: generate visual summaries (co-occurrence networks, hierarchies) and downloadable evidence tables for presentations and policy briefs.

If your corpus includes sensitive evaluations or interview transcripts, ensure research ethics and consent.

Evidano: enterprise-grade encryption and a policy that your data is never used to train third-party models (secure-by-design). Note: this post focuses on research workflows and is non-diagnostic.

A: AI accelerates initial coding and surface patterns, but human validation (especially for clinical nuances like latch techniques) is essential.

A: Segment the corpus by metadata (country, setting, year) and run cross-segment frequency and co-occurrence analyses to spot differences and implementation gaps.

A: The BMJ protocol limits to English and Chinese; if you need broader coverage, add targeted translation and local grey-literature scraping early in the pipeline.