This post shows how researchers and policy/health teams can run a rigorous qualitative analysis of post-COVID care and operationalize LCovB-style mixed-methods with AI-enabled workflows. Fast-moving mixed-methods studies like the LCovB protocol (published 9 July 2026) generate routine claims, surveys and interview transcripts that must be triangulated to form usable care recommendations. The LCovB team combines data from 44 statutory health insurers (7.8 million insured; routine data expected to include ~13, 500 PCS patients), a hybrid survey (2, 854 contacted; target ≥200 responses), ~15 semi-structured interviews and ~15 home medical assessments, all feeding a Delphi expert panel (PLoS One). In this post we show how researchers and policy/health teams can run a rigorous qualitative analysis of post-COVID care, avoid common bottlenecks, and operationalize findings with AI-enabled workflows (see Evidano).
Key Takeaways
Evidano is an AI-powered qualitative data analysis platform that accelerates transcription, thematic coding, and integration across transcripts, surveys, and claims.
This post explains how the LCovB mixed-methods study (registered Aug 2025, protocol published 9 July 2026) maps post-COVID care trajectories from 44 SHIs and how researchers can reproduce its convergent analysis to produce guideline-ready recommendations by Nov 2026.
- LCovB combines claims from 44 statutory health insurers covering 7.8 million insured, expecting ~13, 500 PCS cases for cohort and cost analyses.
- LCovB integrates a hybrid survey (2, 854 contacted; target ≥200 responses), approximately 15 semi-structured interviews, and approximately 15 home assessments, triangulated via a Delphi panel.
- A reproducible 7-step pilot pipeline (2–4 weeks) can validate linkages: transcribe audio, map identifiers, auto-code, run joint displays, and prepare expert-ready outputs using Evidano.
Fast take + source
Fast summary: LCovB is a convergent mixed-methods study registered Aug 2025 that maps care trajectories and gaps for post-COVID syndrome and aims to produce guideline-ready recommendations by Nov 2026.
- Primary source: PLoS One (Published 9 July 2026).
- Why qualitative analysis matters: interviews (n≈15) supply contextual barriers, coping strategies and patient-reported gaps that claims data alone cannot show.
- Evidano fit: use AI-assisted transcription, thematic coding, and cross-segment analysis to integrate transcripts, surveys and claims-derived variables (Evidano).
Findings snapshot
| Item | Value / design detail | Source / implication |
|---|---|---|
| Publication date | 9 July 2026 | PLoS One study protocol |
| Routine data base | 44 SHIs; 7.8 million insured | Enables population-level care-trajectory analysis |
| Expected PCS cases in routine data | ~13, 500 | Sufficient for subgroup and cost analyses |
| Survey | 2, 854 contacted; target ≥200 responses | Hybrid (postal + online) to reduce selection bias |
| Interviews | ≈15 semi-structured interviews (60 min) | Depth: patient experiences, stigma, barriers |
| Home assessments | ≈15 severely affected patients (Berlin area) | Clinical validation; therapy plans for PCPs |
| Integration method | Convergent mixed-methods + Delphi panel | Joint displays to triangulate quantitative + qualitative |
What the LCovB study does (plain English)
Plain-English summary: LCovB runs four parallel modules that together map prevalence, patient-reported outcomes and clinical validation to produce care recommendations.
- Module 1: complete claims extraction (01.01.2021–31.12.2023 / 2024 depending on submodule), focus on ICD codes U07.x and U09.9.
- Module 2: hybrid survey including validated scales (SF-36, PHQ, FUNCAP) and project-specific questions.
- Module 3: ~15 recorded and transcribed interviews analyzed with Kuckartz qualitative content analysis.
- Module 4: physician-led home assessments with monthly monitoring for 6 months.
How to scale qualitative analysis of post-COVID care
Problem: fragmented inputs
The problem is fragmented inputs: claims data give trajectory and cost signals but lack patient voice, surveys provide structured measures but limited narratives, and interviews are rich but small-N.
LCovB’s convergent design needs reproducible methods to connect these layers.
Practical steps for researchers
Practical steps start with standardizing metadata and mapping clinical identifiers across sources.
1) Standardize metadata across sources (age bands, region, ICD flags).
2) Transcribe interviews with a research-safe pipeline and custom dictionaries for clinical terms.
3) Use a shared codebook that maps interview themes to survey items and claims-derived events.
4) Produce joint displays (theme × cohort × costs) to present convergent/divergent signals to experts.
Ethics note
Ethics summary: LCovB was approved by Charité Ethics Committee and results should be treated as research-only and non-diagnostic.
Researchers should maintain consent, anonymization and data minimization.
So what for UX researchers, policy & health teams
Implication statement: Researchers, UX teams and policy stakeholders should prioritize actionable themes linked to care pathways and use qualitative excerpts to illustrate cost impacts.
- Prioritize themes that link to actionable pathways (e.g., 'GP first contact but no referral' → design referral triggers).
- Use qualitative excerpts to illustrate cost-impact findings in briefings for payers and policymakers.
- Segment analysis (age, gender, severity, region) uncovers where rehab or specialized clinics are most needed.
Do more, faster with Evidano
From raw transcripts to themes
This subsection explains how Evidano processes raw transcripts into themes with research-grade tools.
Evidano ingests audio/video, applies research-grade transcription (custom dictionary for ME/CFS, PEM, U09.9) and redacts PII.
Automated thematic extraction accelerates Kuckartz-style content analysis while preserving a human-in-the-loop codebook.
Linking qualitative themes to claims and survey variables
This subsection explains how to link themes to structured variables for co-occurrence analysis.
Import CSVs of claims or survey outputs; run cross-segment frequency and co-occurrence analyses to show which patient-reported barriers co-occur with long work disability or higher cost trajectories.
Expert-ready outputs
This subsection explains the outputs Evidano produces for expert panels and policy stakeholders.
Generate joint displays, word co-occurrence networks and hierarchical codes→subcodes for Delphi panels.
Export visual reports for clinicians and policy stakeholders without manual rework (reduces synthesis time from weeks to days).
Security & research governance
This subsection explains Evidano’s security posture and governance alignment with SHI constraints.
Evidano uses encrypted storage and proprietary LLMs that are not used to train third-party models, and aligns with the data protection constraints LCovB encountered when handling SHI data.
Checklist: 7-step workflow to reproduce LCovB-style integration
This checklist gives a 7-step run-book you can apply in 2–4 weeks for a small pilot (n≈200 survey + 10 interviews).
- 1) Collect and catalogue inputs (claims extract metadata, survey CSV, raw audio).
- 2) Transcribe audio with a custom clinical dictionary; run QA.
- 3) Import all sources into Evidano and map shared identifiers/segments.
- 4) Draft an initial codebook from literature + 2 pilot interviews; use AI-assisted auto-coding.
- 5) Run thematic frequency and cross-segment analyses; generate joint displays.
- 6) Prepare 10–15 top quotes and visualizations for an expert workshop.
- 7) Iterate codes based on expert feedback and finalize recommendations.
FAQ: qualitative analysis of post-COVID care
Q: How do I compare interview themes to claims-based outcomes?
A: Create linking variables in claims data and run co-occurrence and subgroup frequency analyses against coded themes.
Create linking variables (e.g., 'work disability >100 days') in claims data and run co-occurrence and subgroup frequency analyses against coded themes.
Q: What sample sizes are appropriate for saturation?
A: LCovB plans approximately 15 interviews based on saturation principles, and heterogeneous populations may require 15–25 interviews.
LCovB plans ≈15 interviews based on saturation principles; for heterogeneous populations expect 15–25 interviews, monitoring emergent themes iteratively.
Q: How to preserve privacy when sharing quotes?
A: Remove direct identifiers, paraphrase rare details, and keep identifiable linkage only within secured study teams.
Remove direct identifiers, paraphrase rare details, and keep identifiable linkage only within secured study teams; Evidano supports PII redaction at transcription.
Wrapping up: next steps
Wrapping up: Prioritize a reproducible pipeline that links transcripts, surveys and claims into joint displays for decision-makers when synthesizing mixed-methods post-COVID data like LCovB.
- Try a pilot: ingest one claims extract + 10 interviews + survey CSV to validate linkages and produce an expert brief.
- Get started with Evidano to accelerate transcription, thematic coding and cross-segment synthesis, or contact us to map this workflow to your data.
- Try Evidano for free to run a 2-week pilot and deliver a joint display and policy brief that mirrors LCovB’s integration approach.