Fast payoff: If you analyze interviews, field notes, or survey comments about TB care, this June 30, 2026 PLoS One study shows how a culturally adapted instrument (HELIA‑TB) and mixed-methods validation (Mar 2024–Mar 2025, Junagadh, Gujarat) produced reliable, actionable measures of health literacy (n = 393; overall α = 0.86; ICC = 0.88) that predicted treatment adherence (Cohen’s d = 0.89). Read the original study at PLoS One. Ethics note: this summary is research-focused and non-diagnostic; the original study received ethics approval and informed consent.
Key Takeaways
This post shows how an AI-enabled qualitative workflow can reproduce the HELIA‑TB adaptation and speed the pipeline from interviews to psychometrics-ready data while preserving conceptual fidelity.
- HELIA‑TB, adapted and validated in Junagadh district (Mar 2024–Mar 2025), showed good reliability (overall α = 0.86), high test–retest stability (ICC = 0.88), and predicted adherence (Cohen’s d = 0.89).
- Embedding cognitive interviews and frontline worker input before large-scale validation preserves comprehension for low-literacy groups, then psychometrics confirm structure (CFA: CFI 0.946; RMSEA 0.051; SRMR 0.041).
- AI-enabled platforms can centralize transcripts, automate transcription and translation, support AI-assisted coding, export CFA-ready datasets, and maintain encryption and PII redaction for ethics compliance.
Evidano is an AI-powered qualitative data analysis platform that centralizes transcripts, automates transcription and translation, supports AI-assisted coding, and enforces data privacy.
Evidano is an AI-powered qualitative data analysis platform that centralizes transcripts, automates transcription and translation, supports AI-assisted coding, and enforces data privacy.
Evidano preserves local terms with custom dictionaries, supports PII redaction, provides hierarchical code trees and co-occurrence analysis, and exports item-level responses for CFA and reliability testing.
Findings snapshot
| Metric | Value | Notes / Implication |
|---|---|---|
| Publication | June 30, 2026 | PLoS One (open access) |
| Study period | Mar 2024–Mar 2025 | Sequential qualitative → quantitative adaptation |
| Setting | Junagadh district, Gujarat, India | Urban, rural, tribal mix |
| Validation sample | n = 393 adults with TB | CFA, predictive and convergent validity tested |
| Scale | HELIA‑TB (33 items, 5 domains) | Access, Reading, Understanding, Appraisal, Decision‑making |
| Internal consistency | α = 0.82–0.89 (overall 0.86) | Good reliability across domains |
| Test–retest | ICC = 0.88 | High temporal stability (30‑day interval) |
| Content validity | S‑CVI/Ave = 0.92 | Strong expert agreement |
| Predictive validity | Cohen's d = 0.89 (adherence) | Large effect: higher literacy → better adherence |
| CFA fit | CFI 0.946; RMSEA 0.051; SRMR 0.041 | Five‑factor structure confirmed |
What they did (plain English)
The authors adapted the Health Literacy Instrument for Adults (HELIA) to TB (HELIA‑TB) using an exploratory sequential mixed-methods design: expert review, forward translation to Gujarati, cognitive interviews (5 patients, 8 health workers), back-translation, pilot (30), then psychometric validation (393 participants).
- Qualitative phase: item contextualization (for example, replace generic items with TB-specific prompts) and simplify terminology for low-literacy populations.
- Quantitative phase: assessed face and content validity, internal consistency (Cronbach’s alpha), test–retest reliability (ICC), CFA, predictive validity against recorded treatment adherence, and convergent validity with self-rated health.
- Key numeric anchors: S‑CVI/Ave = 0.92; overall α = 0.86; ICC = 0.88; Cohen’s d (adherence) = 0.89.
So what for qualitative researchers and program teams
For qualitative researchers
For qualitative researchers, this study is a model for instrument adaptation: integrate cognitive interviews with frontline workers and patients before large-scale validation to preserve conceptual fidelity while improving comprehension for low-literacy groups.
Use mixed evidence: thematic findings should inform item wording and response options before psychometric testing, not after.
For TB program managers and policy teams
For TB program managers and policy teams, HELIA‑TB identifies patients at risk of non-adherence with a large effect size, so embedding a validated tool into routine monitoring can prioritize education, adherence support, and resource allocation.
Because the instrument performed across literacy strata, the instrument supports equity-focused screening in community programs.
For UX & product teams
For UX and product teams, map themes from cognitive interviews to microcopy and decision flows, and use tested term swaps (for example, 'physician' to 'doctor') as practical UX fixes validated by patients.
The study’s adaptation steps are reproducible UX actions that improve comprehension for low-literacy users.
Do more, faster with Evidano (mapping tasks → features)
Ingest & centralize all materials
Ingest and centralize all materials by uploading transcripts, PDFs, and spreadsheets into Evidano, which maintains one searchable corpus with access controls and encrypted storage.
Problem: interviews, pilot notes, field forms and scoring sheets are scattered. Evidence: centralization reduces manual file handling and supports reproducibility.
Translate & transcribe for low‑literacy contexts
Translate and transcribe multilingual interviews using automated tools with custom dictionaries to preserve local terms such as ASHA and DOT, and apply PII redaction for ethics compliance.
Problem: multilingual interviews (Gujarati) and inconsistent terminology slow analysis. Evidano provides automated transcription and translation with custom dictionaries and PII redaction to speed coding and anonymize responses.
From themes to validated measures
Link qualitative item changes to psychometrics by generating exportable item-level responses and codebooks that map to HELIA domains so teams can run CFA-ready analyses.
Problem: linking qualitative item changes to psychometrics is manual and error-prone. Evidano provides AI-assisted thematic coding, hierarchical code trees, co-occurrence networks, and exportable codebooks you can map to instrument items.
Segment & predictive analysis
Compare literacy scores across adherence groups and literacy strata with cross-segment analysis that outputs frequency tables and effect-size metrics for reporting.
Problem: it is hard to compare literacy scores across adherence groups and literacy strata. Evidano supports cross-segment analyses (for example, adherent vs. non-adherent; literate vs. no formal education).
Stakeholder-ready visuals & storytelling
Produce short, actionable briefs with one-click visualizations and AI chat over your documents to generate executive summaries and policy briefs.
Problem: field teams need short, actionable briefs. Evidano provides one-click visualizations (word clouds, co-occurrence graphs, hierarchical themes) and AI chat to draft stakeholder deliverables.
Secure, auditable workflow
Maintain secure and auditable workflows with end-to-end encryption, access logs, and a policy that customer data are not used to train third-party models to meet research ethics expectations.
Problem: sensitive health data and ethics approvals require secure handling. Evidano offers encryption, access logs, and explicit nonuse of customer data for third-party model training.
Checklist: Reproduce HELIA‑TB adaptation as an AI-enabled workflow
This 9-step checklist shows how to go from raw transcripts to segment-level insights and a psychometrics-ready dataset.
- 1) Collect materials: expert notes, forward and back translations, cognitive interview transcripts, pilot responses, and scoring guides.
- 2) Upload to Evidano: documents and survey sheets, retaining original language files and translations.
- 3) Run automated transcription and translation with a custom dictionary to preserve local terms and acronyms.
- 4) Auto-suggest initial codes from cognitive interviews and review to import a draft codebook aligned to HELIA domains.
- 5) Apply AI-assisted coding across the corpus and review a sample of coded excerpts for reliability.
- 6) Generate thematic frequency tables and co-occurrence networks and flag items with low coherence for rewrite (for example, Q32 in HELIA‑TB).
- 7) Export item-level responses and metadata for CFA and reliability analysis in CSV or SPSS formats.
- 8) Produce stakeholder deliverables such as a brief, visuals, and recommended intervention targets for low-literacy subgroups.
- 9) Archive source files and the audit trail for ethics and reproducibility.
FAQ: AI for qualitative analysis of health literacy
When should I adapt an instrument versus build a new one?
+
Adapt when a validated tool exists and requires contextual tweaks, because adaptation saves time and preserves comparability.
The HELIA to HELIA‑TB adaptation is an example where contextualizing items for TB preserved the original constructs while improving relevance.
How do I ensure low‑literacy respondents understand items?
+
Combine cognitive interviews with frontline workers and patients, simplify wording, pilot timing, and track comprehension flags in transcripts to ensure understanding.
The HELIA‑TB process included cognitive interviews, term swaps, and pilot testing to improve comprehension for low-literacy populations.
Is AI safe to use with sensitive health data?
+
Use platforms with end-to-end encryption, access logs, and clear data usage policies to protect sensitive data.
The HELIA‑TB workflow expectations align with platforms that support PII redaction and explicit nonuse of customer data for third-party model training.
Wrapping up & next steps
The HELIA‑TB study (June 30, 2026) provides concrete evidence that careful cultural adaptation plus rigorous psychometrics yields a tool that predicts adherence and works across literacy levels.
- Try a pilot: upload one round of cognitive interview transcripts and your draft items into Evidano and run a thematic and cross-segment analysis in hours, not weeks.
- If you want help reproducing the HELIA‑TB workflow or running a pilot in your language, Try Evidano for free.
Topics
- qualitative analysis of health literacy
- HELIA-TB adaptation
- AI-assisted qualitative coding
- psychometrics workflow
Keep reading
- Commentary on NewsAI-assisted qualitative coding: Qualcoder MCP serverExplainer of the Qualcoder MCP server for AI-assisted qualitative coding, including the PyPI release date, privacy flow to Anthropic, safety steps for researchers, and practical next steps.
- Commentary on NewsBoosting Recall: Qualitative analysis of clinical validationUse PLOS ONE (Jul 20, 2026) findings to guide qualitative analysis of clinical validation: methods, codebook tips, AI-era QA, reproducible item-level workflows and reporting.
- Commentary on NewsBoost Recall: Qualitative Analysis of Clinical ConsultationsPerceived clinician validation improves recall in clinical consultations (n=245, Jul 20, 2026). Practical AI-enabled qualitative workflow and QA steps.