Evidano is an AI-powered qualitative data analysis platform that ingests interview audio, transcripts, and spreadsheets, performs transcription and AI-assisted coding, and generates thematic maps, cross-segment comparisons, and co-occurrence networks. Late childhood cancer diagnosis is a major driver of low survival in low-resource settings. This PLoS Medicine study (published June 29, 2026) reports that a Jan–Sep 2024 multilevel onsite training plus a 6-month mentorship in Northwest Ethiopia lifted provider knowledge and cut the diagnostic interval by a median 54.3% (see PLOS Medicine). For researchers and implementers who work with mixed-methods data, AI-enabled qualitative analysis can accelerate synthesis, surface contextual drivers (leadership, supply shortages, cultural beliefs), and produce reproducible recommendations. In this post we show how to convert the study’s interviews, focus groups, KAP surveys, and chart-review timelines into decision-ready insight using Evidano: thematic maps, cross-segment comparisons, co-occurrence networks, and a short reproducible workflow you can run in two weeks.
Key Takeaways
A Jan–Sep 2024 multilevel onsite training with a subsequent 6-month mentorship in Northwest Ethiopia was associated with large provider knowledge gains and a median 54.3% reduction in the diagnostic interval, per PLOS Medicine.
The study used a mixed-methods design combining structured KAP surveys, 100 pediatric oncology chart timelines, and CFIR-guided qualitative interviews, a corpus well suited to AI-enabled qualitative synthesis.
- The intervention reached 18 Level I clinicians, 29 Level II providers, and 1, 020 HEWs during Jan–Sep 2024 including a 6-month mentorship (Apr–Sep 2024).
- Patient-journey analysis showed a median diagnostic interval decrease of 54.3% (56.5 → 25.8 days) with p < 0.001, and a median patient delay decrease of 9.3% (27.0 → 24.5 days, p = 0.02).
- Mixed-methods synthesis can prioritize modifiable bottlenecks (labs, transport, supply chains) and translate findings into targeted investments and user-centered tools.
Fast take, source & payoff
A Jan–Sep 2024 multilevel onsite training and 6-month mentorship in Northwest Ethiopia was associated with large gains in provider knowledge and a 54.3% reduction in the diagnostic interval, per PLOS Medicine.
- Why analysts should care: the study combines structured KAP surveys, quantitative chart timelines (n=100 charts), and CFIR-guided qualitative interviews, an ideal mixed-methods corpus for AI-enabled qualitative analysis.
- Payoff: convert dispersed transcripts, survey spreadsheets, and charts into thematic drivers, timeline visualizations, and cross-segment comparisons that inform scale-up and procurement priorities.
Findings snapshot
| Metric | Value / Detail | Source / Note |
|---|---|---|
| Publication date | June 29, 2026 | PLOS Medicine |
| Study period | Jan–Sep 2024 (mentorship Apr–Sep 2024) | Methods section |
| Participants trained | 18 Level I clinicians; 29 Level II providers; 1, 020 HEWs | Intervention description |
| Patient charts reviewed | 100 pediatric oncology charts | Outcomes: patient-journey intervals |
| Diagnostic interval change | Median ↓54.3% (56.5 → 25.8 days) | Results (p < 0.001) |
| Patient delay change | Median ↓9.3% (27.0 → 24.5 days) | Results (p = 0.02) |
| Budget | USD 52, 762 (total project cost reported) | Discussion |
What the team did and how they measured it
The team delivered tiered, hands-on training followed by mentorship and measured change with mixed-methods pre–post approaches.
Intervention: tiered, hands-on training (10-day Level I; 7-day Level II; 5-day pictorial HEW module) followed by 6 months of monthly onsite mentorship plus tele-support. Adaptive cascade mini-trainings addressed access gaps during civil unrest and seasonal constraints.
Design and measures: the study used a quasi-experimental pre–post mixed-methods design. Quantitative measures included KAP surveys (normalized 0–100) and patient-journey intervals extracted from charts (onset→first contact→diagnosis→treatment). Qualitative work included 18 in-depth interviews and multiple focus group discussions with HEWs, coded with CFIR and thematic analysis.
- High fidelity: 90% of planned training days were delivered and the study reported a 100% response rate among participants.
- Key limitations: the study had no contemporaneous control group, operational confounders (civil unrest, competing campaigns) affected implementation, and some missing chart dates were handled with multiple imputation.
So what for researchers, UX and policy teams
For implementation researchers
Implementation researchers can use mixed-methods datasets to identify both what changed and why those changes occurred.
Mixed-methods datasets like this expose both 'what changed' (timelines, KAP scores) and 'why' (CFIR themes: leadership, supplies, cultural beliefs). Use AI-enabled thematic synthesis to prioritize modifiable system bottlenecks (for example, reagent supply chains and referral coordination).
Action: extract and code all transcripts, then run co-occurrence and timeline linkage to see which barriers co-occur with long diagnostic intervals.
For UX / product teams
UX and product teams can translate training outputs into usable tools by clustering user feedback and surfacing friction points.
Translating training outputs into usable tools (referral forms, pictorial manuals) benefits from rapid user-testing transcripts. AI can cluster feedback, surface friction points, and produce prioritized product-requirement lists.
Action: map HEW quotes to usability issues and automatically generate design tickets or prioritized feature lists for low-literacy materials.
For health policy / program leads
Health policy and program leads can use quantified reductions and cross-segment analyses to identify where infrastructure investment is most needed.
Quantitative reduction in diagnostic delay (54%) shows training plus mentorship is effective but treatment initiation lagged due to infrastructure; policy investment points are clear (labs, transport, oncology capacity).
Action: use cross-segment frequency analysis to quantify how often supply shortages versus cultural beliefs versus leadership explain delays and use that to allocate resources.
Do more, faster with Evidano (how features map to this study)
Problem: Disparate documents (audio interviews, KAP spreadsheets, charts)
Evidano ingests interview audio, transcripts, and spreadsheets and prepares them for analysis.
Solution in Evidano: ingest interview audio, transcripts, and spreadsheets. Use built-in transcription with custom dictionary (local terms like 'nififit'), PII redaction, and translation to standard English before analysis.
Problem: Manual, inconsistent coding across transcripts
Evidano enables consistent, hierarchical coding with AI assistance and interactive review.
Solution in Evidano: import or build a hierarchical codebook; run AI-assisted coding to apply themes consistently; review and refine with interactive code trees and subcodes.
Problem: Linking themes to timelines and segments
Evidano links themes to timelines and segments to reveal associations between CFIR constructs and patient-journey intervals.
Solution in Evidano: generate thematic and frequency analyses, cross-segment comparisons (Level I vs Level II vs HEWs), and co-occurrence networks that link CFIR constructs to patient-journey intervals.
Problem: Stakeholders need visual, shareable outputs
Evidano creates one-click visualizations and exportable summaries suitable for MOH or donor briefs.
Solution in Evidano: one-click visualizations (word clouds, co-occurrence graphs, hierarchical code maps) plus exportable quotes and executive summaries for MOH or donor briefs. Data is encrypted and never used to train third-party models.
Problem: Need follow-up data collection
Evidano supports autonomous qualitative follow-up and continuous thematic monitoring.
Solution in Evidano: deploy AI-avatar interviewers for autonomous qualitative follow-up (for example, post-training HEW surveys), then run continuous thematic monitoring to detect emerging barriers.
Reproducible 2-week workflow to analyze this corpus in Evidano
The two-week workflow described here ingests, cleans, codes, synthesizes, and visualizes the corpus to produce decision-ready outputs.
Week 1; Ingest & clean
- Day 1: Upload transcripts, audio files, and the KAP spreadsheet. Set custom dictionary entries (local terms like 'nififit') and enable PII redaction.
- Day 2–3: Run batch transcription and translation; review and correct 1–2 spot transcripts.
Week 2; Code, synthesize, visualize
- Day 4–6: Import or build a CFIR-based codebook; apply AI-assisted coding and review inconsistencies.
- Day 7–9: Run thematic frequency analysis, cross-segment comparisons (Level I vs II vs HEWs), and link themes to patient-journey timelines.
- Day 10: Generate visuals (co-occurrence network, hierarchical code map) and export an executive brief with 5 prioritized implementation recommendations.
Deliverable: decision memo and visuals to support targeted investments (for example, lab reagents or tele-support frequency changes) within two weeks.
FAQ: AI-enabled qualitative analysis
What is AI-enabled qualitative analysis and when should I use it?
AI-enabled qualitative analysis is the use of tuned AI to accelerate coding, theme extraction, and cross-segment comparisons over transcripts and documents.
Use AI-enabled qualitative analysis when you have large qualitative corpora or mixed-methods datasets and need reproducible synthesis fast.
How do I compare segments (for example, HEWs vs clinicians)?
You compare segments by importing segment labels and running comparative frequency and quote-sampling analyses.
Import segment labels (role, facility level) into Evidano, run frequency and sentiment-style comparisons, and surface statistically different themes alongside representative quotes.
How secure is the data and is it OK for clinical research?
Evidano encrypts data end-to-end and does not use customer data to train third-party models.
Evidano encrypts data and does not use customer data to train third-party models; this work is for research and implementation (non-diagnostic), so always follow local ethics approvals and consent procedures.
Wrapping up & next steps
The PLOS Medicine study (June 29, 2026) shows that context-tailored training plus mentorship can sharply reduce diagnostic intervals but also highlights persistent infrastructure bottlenecks that require targeted policy action.
- Next move for teams: run a rapid AI-enabled qualitative re-analysis of the study’s transcripts and KAP data to quantify drivers of treatment-delay and produce a prioritized, evidence-backed investment list.
- Try it: upload the corpus and run the 2-week workflow on Try Evidano for free to generate visuals and a one-page decision memo for stakeholders.
- Source: full study at PLOS Medicine.