Evidano is an AI-powered qualitative data analysis platform that supports transcription, thematic coding, cross-segment analysis, and secure reporting. This post shows a reproducible qualitative analysis workflow that turns the PLoS case study of Zhuhai (published June 5, 2026) into actionable recommendations for researchers and public-health teams. The introduction maps problems in the paper (n=41 experts invited; 31 responses; 75.6% response rate) to Evidano capabilities (transcription, thematic coding, cross-segment frequency analysis, and secure reporting) and gives a 7-step checklist you can run on your first dataset. Read the original study at PLOS ONE and try these patterns in Evidano.
Key Takeaways
This post shows how the Zhuhai PHEIS modified e-Delphi converted expert input into compact operational indicators and how Evidano can automate transcription, thematic coding, and cross-segment analysis to reproduce that pipeline. The Zhuhai study (published June 5, 2026) delivered city/district and city/district/town indicator frameworks and supported handling 536 outbreaks between Jan 1 and Nov 30, 2024. Use a small pilot to check time savings and traceability before scaling.
- The researchers produced two frameworks: city/district (5 first-level, 21 second-level) and city/district/town (4 first-level, 17 second-level) from a modified e-Delphi process.
- The Delphi panel invited 41 experts and received 31 responses (75.6%), reporting an expert authority coefficient Cr = 0.72 and operational validation over 536 outbreaks (Jan 1–Nov 30, 2024).
- Evidano automates ingest, AI-assisted coding, and cross-segment frequency/co-occurrence analysis so teams can prioritize indicators and export traceable supporting quotes.
Fast take + source
Researchers used a modified e-Delphi to build a public health emergency information system (PHEIS) framework for Zhuhai city, China. The study was published June 5, 2026 and reports a city/district framework (5 first-level and 21 second-level indicators) and a city/district/town framework (4 first-level and 17 second-level indicators. Source: PLOS ONE.
- Why researchers care: the paper converts stakeholder knowledge (41 invited experts; 31 respondents; Cr = 0.72) into a lean indicator set that supported handling 536 outbreaks between Jan 1 and Nov 30, 2024.
- Payoff for analysts: learn how to move from qualitative expert text and Delphi scores to prioritized modules (emergency response, preparedness, monitoring) and operational tasks you can automate with Evidano.
Findings snapshot
| Item | Value | Source / note |
|---|---|---|
| Publication date | June 5, 2026 | PLOS ONE |
| Population (Zhuhai) | 2.4941 million (end 2023) | Paper background |
| Experts invited / responded | 41 invited; 31 responded (75.6%) | Delphi rounds |
| Final frameworks | City/district: 5 first-level, 21 second-level; City/district/town: 4 first-level, 17 second-level | Post-Delphi results |
| Operational test | 536 outbreaks handled (Jan 1–Nov 30, 2024) | System use & redesign |
What the Delphi actually did (method in plain English)
The Delphi converted a candidate indicator list into compact, operational indicator sets using a two-round modified e-Delphi. The team combined literature review and stakeholder workshops to build the candidate list, then ran a two-round modified e-Delphi using Excel questionnaires, keeping indicators with mean importance ≥ 3.5 and CV ≤ 0.25 while retaining a small number of conceptually critical items as documented exceptions.
- Inputs: policy documents, provincial standards, and frontline staff suggestions were used to form candidate indicators.
- Output: compact, operational indicator sets prioritized for emergency response, preparedness, and monitoring.
- Validation: the researchers measured expert authority (Cr = 0.72), used summary feedback between rounds so panelists could revise ratings, and validated the framework in live use during 2024 with 536 logged outbreaks.
So what for researchers and public-health analysts: implications
For qualitative researchers
Qualitative researchers should convert Delphi text responses and free-text suggestions into reproducible themes, then triangulate those themes with numeric scores (importance, feasibility, CV). The paper shows that Delphi open comments are rich but noisy, so convert open comments into reproducible themes and use cross-segment analysis to detect where frontline and managerial priorities diverge, for example ease-of-use versus model-driven decision tools.
For UX/implementation teams
UX and implementation teams should prioritize high-utility modules with low friction, because the paper shows emergency response scored highest. The study suggests prioritizing smart forms, QR case entry, and auto-reporting, and keeping decision-analysis features modular to reintroduce once data completeness improves.
For policy or evaluation teams
Policy and evaluation teams should embed periodic Delphi-style reviews into rollout plans to support iterative re-weighting of indicators as system maturity changes. The study recommends measuring operational outcomes such as response time and data completeness, these are the metrics that unlock re-introduction of advanced analytics.
Do this with Evidano: map problems→features
Problem: messy Delphi inputs (mixed Likert scores + free text)
Messy Delphi inputs combine Likert scores and open text, making prioritization hard. Evidano ingests Excel ballots and transcripts, auto-extracts Likert distributions, and runs thematic coding on open comments so you can link theme frequency to importance scores and prioritize indicators quantitatively.
Problem: inconsistent coding and cross-level comparisons
Inconsistent coding and unclear cross-level comparisons obscure where city, district, and street staff diverge. Evidano lets you import an initial codebook, apply AI-assisted coding to new rounds, and run cross-segment frequency and co-occurrence networks to reveal those divergences.
Problem: reports take too long and lack traceability
Slow reporting and weak traceability reduce stakeholder trust. Evidano provides one-click export of coded quotes, visual co-occurrence maps, and downloadable executive tables that show which expert comments supported each retained indicator, stored encrypted and not used to train third-party models, available at Evidano.
Problem: multilingual or audio inputs from field
Multilingual and audio inputs complicate analysis when transcripts and notes are in different formats. Evidano offers automated transcription with custom dictionaries and PII redaction, plus translation with custom terms so QR-entered case notes or voice interviews can be analyzed together.
Problem: need for follow-up data
Lack of follow-up data prevents understanding why some indicators were deprioritized. Evidano enables deployment of AI-avatar interviewers for autonomous qualitative data collection to fill gaps identified in the Delphi, for example reasons decision-analysis indicators were deferred.
Checklist: 7-step qualitative workflow to reproduce Zhuhai-style analysis
This checklist gives seven steps to reproduce Zhuhai-style analysis using your Delphi rounds, meeting notes, and system logs.
- Step 1: Gather inputs, policy documents, Delphi ballots (Excel), free-text suggestions, and usage logs (for example 536 outbreaks).
- Step 2: Ingest into Evidano, upload spreadsheets, transcripts, and system logs, and apply custom dictionaries for domain terms.
- Step 3: Auto-code and map, run AI-assisted thematic coding, then align themes with Likert scores and CVs.
- Step 4: Run cross-segment analysis, compare city versus district versus street/township frequencies and co-occurrence patterns.
- Step 5: Produce traceable outputs, extract supporting quotes per retained indicator and generate one-page executive summaries.
- Step 6: Pilot operational changes, prioritize high-utility modules (emergency response, preparedness), and instrument metrics such as response time and data completeness.
- Step 7: Re-run Delphi or rapid feedback rounds, repeat rounds and use AI-avatar interviews to collect missing perspectives.
Ethics & safeguards (brief)
The Zhuhai study had IRB approval and a waiver of written consent where appropriate. When analyzing PHEIS data, treat outputs as research-only, redact PII, follow local ethics guidance, and use tools that support encrypted storage; Evidano supports PII redaction and encrypted storage.
FAQ: qualitative analysis of public health emergency systems
How do you combine Likert scores with open-text Delphi feedback?
You quantitatively link mean importance and feasibility scores and coefficient of variation with thematic counts. Flag items with high mean but high CV for discussion and use documented exception rules for conceptually essential items, as the Zhuhai paper did.
When should decision-analysis modules be postponed?
Postpone decision-analysis modules when feasibility scores are low and onboard data is incomplete. The Zhuhai team deferred some decision-analysis indicators until data completeness improved, so measure readiness before investing in predictive models.
How do we validate indicators after deployment?
Validate indicators by tracking operational KPIs and repeating short Delphi rounds. Track response times, number of cases captured, and user adoption, and re-run Delphi-style reviews after major incidents or annually.
Wrapping up: next moves
To convert a Delphi study, meeting notes, or system logs into prioritized, auditable indicators, start with a small pilot that imports one Delphi round and corresponding usage logs into Evidano and produces a one-page decision brief. Run a two-week pilot to check time savings and signal clarity, and for hands-on help, Try Evidano for free or visit Evidano to request a demo or upload a sample Excel/CSV.