Rapid, defensible qualitative analysis of child drowning case narratives can change where programs invest prevention and training. This post refracts a patient-journey mapping study from the Sundarbans (published 1 August 2025) and shows a repeatable, AI-enabled workflow to extract actionable findings from transcripts, interview notes, and code maps. You’ll learn what the authors found (delay zero + three-delay mapping), which signals matter most for intervention design, and a concrete 7-step way to run the same analysis in Evidano (www.evidano.com) on your corpus. Original study: www.bmjopen.bmj.com/content/15/8/e103099.
Fast take + source
What happened: A qualitative patient-journey mapping study in the Sundarbans identified an extra temporal bottleneck ("delay zero" (recognition/rescue)) in addition to the classic three delays (decide, reach, receive). The paper is open access and reports 18 case studies based on 37 interviews collected 1 Oct 2024–5 Feb 2025 (published 1 Aug 2025). Source: www.bmjopen.bmj.com/content/15/8/e103099.
- Primary finding: recognition/rescue often took 30 seconds to >30 minutes; rescue by untrained bystanders; traditional resuscitation attempts delayed formal care.
- Consequence for researchers: narratives contain short, high-signal episodes (rescue, first-aid decisions, transport choice) ideal for thematic extraction and timeline reconstruction.
Findings snapshot
| Date / Metric | Value | Source / Note |
|---|---|---|
| Data collection | 1 Oct 2024 – 5 Feb 2025 | Interview window reported in study |
| Sample | 18 drowning cases, 37 interviews | Adults present during/rescue (ages of victims 1–15 yrs) |
| Key delays | Delay 0 + three-delay framework | Delay 0 = recognition & rescue; then decide, reach, receive |
| Transport | Hospitals up to 30 km | Reported transport distances; bikes & rental cars used |
| Outcomes | 4 hospital deaths among those taken; many never taken | Attributed to late rescue or incorrect resuscitation |
What happened: methods and analytic framing
The authors used purposive sampling across eight Gram Panchayats in the Sundarbans and performed in-depth interviews with adults who witnessed rescue and postevent decisions. Bengali interviews were transcribed and translated into English; delays were coded in Excel against an extended three-delay model (0–3). Data collection ran Oct 1, 2024 to Feb 5, 2025 and analysis produced 16–20 intended maps, with saturation reached at 18 cases.
- Delay 0 (new): time to detect and rescue (30s → >30min; obstacles: vegetation, night, lack of supervision).
- Delay 1: decision to seek formal care, often postponed while traditional techniques were attempted.
- Delay 2: reaching a facility, transport constraints, distance up to 30 km, reliance on bikes/rental vehicles.
- Delay 3: receiving treatment, formal facilities generally responsive when reached, but local informal doctors often lacked CPR skills.
Implications for researchers, UX teams & program designers
For qualitative researchers
Target short, time-stamped narrative fragments (rescue start, first-aid attempts, transport decision) when building codebooks, these are high signal for causal inference and intervention mapping.
Collect metadata: exact time of day, distance to nearest facility, who performed rescue, and whether CPR training was known, these fields enable cross-segment frequency and co-occurrence analysis.
For program and policy teams
Delay zero reframes prevention: invest in supervision and physical barriers (fenced ponds, supervised childcare) rather than solely expanding hospital capacity.
Training targets: informal local practitioners and bystanders for CPR and safe-rescue techniques; mobile clinics and predictable staffing schedules to reduce Delay 2/3 impact.
Do more, faster with Evidano
Ingest & prepare transcripts
Problem: multilingual interviews and translation noise. Evidano ingests audio, applies transcription and translation with custom dictionaries and PII redaction, so Bengali interviews can be harmonised for pooled analysis.
Automated thematic + timeline extraction
Problem: manual coding of rescue timelines is slow. Evidano generates thematic codes, extracts time-relevant phrases ("rescue started", "found after torch was fetched") and builds timeline visualisations so delay zero events are flagged automatically.
Cross-segment analysis & visual evidence
Problem: hard to compare cases by distance, time-of-day, or rescuer type. Evidano produces cross-segment frequency tables, co-occurrence networks and hierarchical code→subcode reports to show which combinations (e.g., night + vegetation + no supervision) predict longer rescue time.
Stakeholder-ready outputs
Problem: translating transcripts into policy asks. Evidano exports clickable quotes, figure-ready visualisations and a reproducible codebook so you can hand a 2-page policy brief to health planners.
Security & provenance
Evidano encrypts data, supports controlled sharing, and does not use your corpus to train third-party models, critical for sensitive health interviews.
7-step checklist: reproduce the Sundarbans mapping in Evidano
Quick run-book to go from raw audio/transcripts to decision-ready insight:
- 1) Import audio and transcripts into Evidano; apply the Bengali→English custom dictionary during transcription/translation.
- 2) Create fields for event metadata: timestamp, location, rescuer type, transport mode, outcome.
- 3) Run automated thematic extraction; seed codebook with "delay 0/1/2/3" codes and local rescue techniques (e.g., "spin upside down").
- 4) Use timeline visualiser to surface rescue durations and flag cases >6 minutes submerged.
- 5) Run cross-segment analysis (time-of-day × rescuer type × distance) to prioritise interventions.
- 6) Generate a stakeholder deck with co-occurrence networks and quoted evidence for each recommended action.
- 7) Export reproducible codebook and share a secured project link with partners.
Ethics & a short research note
This post summarises peer-reviewed qualitative research and is research-focused, non-diagnostic, and not clinical advice. The original study limited transcript release for ethical reasons; Evidano supports secure, consent-aligned workflows for storing and sharing sensitive transcripts.
Wrapping up: next moves
If you analyse field interviews or plan interventions based on patient journeys, start by extracting rescue/timeline fragments and running cross-segment frequency checks, these reveal where prevention beats remediation.
- Try a pilot: import a small batch (10–20 interviews) and run the 7-step checklist in Evidano to reproduce the delay-zero mapping in hours, not weeks.
- Learn more and start a secure trial at www.evidano.com. Read the original BMJ Open paper: www.bmjopen.bmj.com/content/15/8/e103099.
Keep reading
- Commentary on NewsTwo Definitions: Climate Change Acceptance for UndergradsHow a PLoS One Delphi study (Aug 25, 2026) defined climate change acceptance for undergraduate science students, and how AI-enabled qualitative analysis applies it.
- Commentary on NewsResearcher-in-the-loop: AI-enabled UX researchHow the researcher-in-the-loop model governs AI-enabled UX research. Learn practical governance, stats from the August 2026 piece, and how Evidano supports this workflow.
- Commentary on NewsResearcher-in-the-Loop: Governance for AI UX ResearchGovern AI in qualitative UX research with the researcher-in-the-loop model from Jennifer L. Bowie (Aug 25, 2026): practical rules, risks, and tool mappings.