科研 Skill
质性访谈编码分析
qualitative-research
Codes interview or open-text data into themes using thematic or framework analysis. Use when analyzing qualitative clinical or public-health transcripts.
SKILL.md
Qualitative Research Analysis Skill
Codes and analyzes qualitative data from interviews, focus groups, open-ended survey text, or
field notes, and produces a defensible coding frame, theme structure, and reporting-ready methods
and results text. This is the text-analysis counterpart to analyze-stats (which handles numeric
data) — route a mixed-methods study's quantitative strand to analyze-stats and its qualitative
strand here, then integrate the two narratives in write-paper.
When to use
- Interview or focus-group transcripts, open-text survey responses, or field notes need systematic coding and theme development (thematic analysis, framework analysis, or qualitative content analysis).
- A qualitative or mixed-methods manuscript needs a COREQ (interviews/focus groups) or SRQR (any qualitative approach) compliant Methods section.
- An existing coding frame or theme structure needs an inter-coder reliability check before the team treats it as final.
When NOT to use
- Purely numeric/statistical data →
analyze-stats. - A systematic review of qualitative studies (meta-synthesis) → this skill can produce the
per-study coding, but
meta-analysis/review-paperown the review-level synthesis structure. - Survey-instrument psychometrics (reliability/validity of a closed-ended scale) →
survey-instrument-validation. - Reporting-guideline item-by-item audit of a finished qualitative manuscript →
check-reporting(SRQR.md / COREQ.md checklists already live there).
Prerequisites
- De-identified transcripts or text (run
deidentifyfirst if the raw data may contain names, identifying details, or other PHI/PII — qualitative transcripts are especially prone to identifying quotes; redact or generalize any detail that could re-identify a participant even after names are removed, e.g. a rare job title plus a small town). - A stated analytic approach chosen before coding begins: reflexive thematic analysis (Braun & Clarke), framework analysis (Ritchie & Spencer / NatCen), or conventional/directed qualitative content analysis (Hsieh & Shannon) — each has a different relationship to pre-existing theory and a different audit trail; do not blend them silently.
- Python 3.10+ (stdlib +
pandasfor the code-frequency matrix) or R withtidyverse; a dedicated CAQDAS tool (NVivo, MAXQDA, Atlas.ti, or the open-sourceTaguette) is optional — this skill produces plain-text/CSV coding artifacts that work with or without one. - At least two coders for any manuscript reporting inter-coder agreement; this skill supports a single-coder workflow (common for theses, scoping work) but must label the output as such.
Workflow
Phase 1 — Set the analytic frame
- Confirm the approach (reflexive thematic / framework / content analysis) and the epistemological stance (e.g., for reflexive thematic analysis: is coding inductive (data-driven), deductive (theory-driven), or both?). State this up front; it determines whether a fixed codebook is built before or after first-pass coding.
- Confirm the unit of analysis (whole transcript, speaker turn, meaning unit/sentence) and the sample: how participants were selected (purposive, maximum-variation, theoretical, etc.) and whether data collection continued to information power / saturation — name the stopping rule used, because "we stopped at N interviews" needs a justification in SRQR/COREQ reporting.
Phase 2 — First-pass (open) coding
- Read all transcripts in full before coding any of them — do not code the first transcript in isolation; early codes should already be informed by a sense of the whole dataset.
- Apply line-by-line or meaning-unit codes. For inductive work, let codes emerge from the data; for deductive/framework work, apply codes from the pre-specified framework and add new codes only when the data clearly does not fit the existing frame (track these as frame deviations).
- Produce a codebook: for each code, a short definition, an inclusion/exclusion rule, and one exemplar quote. Never invent an exemplar quote — every quote in the codebook and in the manuscript must be copy-pasted verbatim from a transcript, with its source ID.
codebook.csv columns: code_id, code_label, definition, inclusion_rule, exclusion_rule,
example_quote, example_source_id
Phase 3 — Code consolidation and theme development
- Group related codes into candidate themes (patterns of shared meaning across the dataset, not just topic summaries) or, for framework analysis, chart coded data into a framework matrix (rows = participants/cases, columns = codes/subtopics, cells = summarized data with a pointer back to source quotes).
- For reflexive thematic analysis, iterate: review themes against the coded extracts, then against the whole dataset, refining theme names and boundaries — this is iterative by design, not a single pass.
- Check for negative/deviant cases — participants whose account does not fit the emerging theme — and report how they were handled (incorporated by refining the theme, or reported as a genuine exception) rather than discarded silently.
Phase 4 — Inter-coder reliability (when ≥2 coders)
- Have each coder independently code a shared subset (commonly 10-20% of the corpus, or more for a small sample) using the same codebook.
- Compute a chance-corrected agreement statistic per code or per coding decision:
from sklearn.metrics import cohen_kappa_score
kappa = cohen_kappa_score(coder1_codes, coder2_codes)
Use Cohen's kappa for two coders on categorical code presence/absence, Fleiss' kappa for three or more coders, or Krippendorff's alpha when units have missing codes or ordinal structure. 3. Report the statistic and its interpretation band (e.g., <0.40 poor, 0.40-0.75 fair-to-good,
0.75 excellent — cite the scale used), and resolve disagreements by discussion, recording the final decision rule (consensus, arbitration by a third coder, or majority).
- A single high overall kappa can mask poor agreement on a specific, substantively important code — report per-code agreement for codes central to the paper's argument, not only the aggregate.
Phase 5 — Reporting
- Write the Methods section to match the declared approach's own reporting checklist: COREQ
(32-item, for interviews/focus groups — covers research team, reflexivity, data collection,
analysis) or SRQR (21-item, broader than interviews). Route the item-by-item compliance audit of
the drafted manuscript to
check-reporting. - Write reflexivity content honestly: the researchers' position relative to participants (insider/outsider, clinical role, prior assumptions) and how it may have shaped data collection or interpretation — this is a required COREQ/SRQR item, not optional color.
- Present themes with a thin description (the theme) tied to thick description (multiple illustrative verbatim quotes, each attributed to a de-identified participant ID) — a theme asserted with no quote, or a quote with no interpretive commentary, both fail a qualitative peer review.
Standard Output
## Qualitative Analysis Report
Approach: reflexive thematic | framework | content analysis (inductive/deductive/both)
Unit of analysis: ... | Sample: N participants, selection strategy, stopping rule
### Codebook summary
- N codes, N themes/categories
### Themes
1. Theme name — definition
- Supporting codes: ...
- Exemplar quotes: "..." (participant ID), "..." (participant ID)
- Deviant case(s): ...
### Inter-coder reliability (if ≥2 coders)
- Statistic, value, interpretation band, resolution method
### Reflexivity statement
- ...
### Reporting checklist
- COREQ / SRQR item coverage: route to /check-reporting
Quality checks
- Every exemplar quote traces to a real, verbatim transcript passage with a source ID — never paraphrase a quote and present it as a direct quote.
- The analytic approach is named and held constant through the whole analysis; do not describe the work as "thematic analysis" in Methods while actually running a fixed deductive framework, or vice versa.
- Any direct quote in the manuscript is screened for re-identification risk (rare job titles, specific locations, unique clinical details) even though the transcript itself was already de-identified — a verbatim quote can re-identify someone the ID-stripped transcript could not.
- Saturation/information-power claims are backed by a stated rule (e.g., no new codes in the last two transcripts), not asserted without evidence.
Limitations
- This skill supports coding and theme development; it does not replace the researchers' own interpretive judgment, and reflexive thematic analysis in particular treats the researcher's perspective as a resource, not a bias to be eliminated — do not present kappa statistics as "proof" of objectivity for an approach that explicitly rejects that framing.
- Automated code suggestions (if the user asks for them) are a starting point only; every AI-suggested code must be verified against the actual transcript text before it enters the codebook — language models can produce a plausible-sounding code that does not match what the participant said.
- Works on text already in hand; it does not conduct interviews, recruit participants, or obtain
informed consent — those remain the researcher's responsibility under
write-protocol's ethics section. - Cross-cultural or cross-language corpora need a translation and back-translation protocol with explicit handling of untranslatable concepts; this skill does not perform translation.
示例提示词
帮我对这批患者访谈录音转写稿做主题分析并写出编码框架。
所属插件
安装该插件会一并获得同领域的 58 个 Skill / MCP。
同插件的其他科研 Skill
- 医学AI论文检索优化Optimizes a medical AI paper’s title, abstract and metadata for AI search engines. Use when a paper should be findable and citable by RAG tools and AI search.
- 临床数据统计分析与建模Generates reproducible Python or R code for clinical statistics and tables. Use when collected data needs Table 1, regression, survival or agreement analysis.
- 作者发表记录与策略分析Analyzes a PubMed author record to classify study types and authorship patterns. Use when planning a publication strategy from a researcher’s track record.
- 批量队列暴露结局分析Generates one cohort-analysis script per exposure-outcome pair from one template. Use when a validated method must run across many variable combinations at once.
- 样本量估算与检验效能Walks a decision tree to the right test and computes sample size with IRB text. Use when planning how many patients a study needs before data collection.
- 因果推断方法设计Runs DAG-based, instrumental-variable, difference-in-differences or RD analyses. Use when an association in observational data must support a causal claim.
- 报告规范逐条核查Audits a manuscript item by item against 49 reporting and risk-of-bias checklists. Use when checking a draft against STROBE, CONSORT, TRIPOD or similar guidelines.
- 临床数据清洗与画像Profiles a clinical dataset and generates cleaning code in staged, approved steps. Use when a CSV or Excel dataset needs missing-value and outlier checks before analysis.
- 跨国队列数据协调对比Harmonizes variables and runs parallel weighted analyses across national surveys like KNHANES and NHANES. Use when comparing one exposure-outcome link across countries.
- 变量定义与截断值溯源Maps exposure, outcome and covariate definitions to a cited guideline and DB columns. Use when variable cutoffs need a citable basis before the study protocol.
- 本地化临床数据脱敏Runs a local, network-free script that detects and removes identifiers from clinical data. Use when data may contain PHI and must be de-identified before LLM analysis.
- AI对标专家基准研究设计Plans the arms, rubrics, calibration probes and reviewer panel for an AI-vs-expert benchmark. Use before data collection when designing an AI-vs-human rating study.
- 放射影像AI研究设计审查Reviews a study’s analysis unit, cohort logic, leakage risk and reporting-guideline fit. Use when checking a clinical or medical-AI study design before drafting.
- 传染病仓室模型拟合Builds and fits SIR, SEIR and SEIHR compartmental models with Rt estimation. Use when an outbreak or epidemic’s spread needs mathematical modeling in Python.
- ICMJE利益冲突声明填写Clones a pre-filled synthetic ICMJE COI form, replacing only date, name and title. Use when each author on a submission needs a disclosure form filled quickly.
- 机构Word表单内容填充Fills an institutional Word form, such as an IRB protocol, without breaking its layout. Use when drafted content must be rendered into a CJK-aware .docx template.
- 队列数据库选题挖掘Profiles a longitudinal cohort database and ranks research-topic proposals with gap evidence. Use when looking for a question an available cohort or registry can answer.
- 投稿期刊匹配推荐Matches a manuscript abstract against curated journal profiles and ranks candidates. Use when choosing where to submit, or re-targeting after a rejection.
- DOI全文批量获取Batch-downloads open-access PDFs for a list of DOIs via Unpaywall, PMC and OpenAlex. Use when full texts are needed for a screening set such as a meta-analysis.
- 数据集编码本自动生成Profiles every variable in a tabular dataset into a structured, citable codebook. Use when a CSV, Excel or Stata dataset needs a data dictionary before analysis.
- 全球卫生数据获取分析Accesses and analyzes global health data from WHO GHO, IHME GBD and Our World in Data. Use when DALYs, age-standardized rates or choropleth maps are needed.
- 医学AI项目基金申请书撰写Structures a grant proposal’s significance, innovation, approach, aims and milestones. Use when drafting a research funding proposal, including a Korean grant plan.
- 卫生经济学评价建模Builds Markov cohort models with QALYs, PSA and cost-effectiveness planes. Use when a health intervention needs a formal cost-effectiveness evaluation.
- AI写作痕迹识别改写Scans a manuscript for AI writing patterns and rewrites flagged passages carefully. Use when a draft or response letter reads as obviously AI-written.
常见问题
质性访谈编码分析能做什么?
Codes interview or open-text data into themes using thematic or framework analysis. Use when analyzing qualitative clinical or public-health transcripts。
如何安装质性访谈编码分析?
在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「质性访谈编码分析」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。
质性访谈编码分析需要哪些权限?
该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。
质性访谈编码分析经过官方审核吗?
是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。