Qualitative Data Management and Coding

Introduction

Qualitative health research produces rich records: interview audio, field notes, observations, photographs, documents, reflective journals, transcripts, translations, and analytic memoranda. These materials are not merely files to be stored. They are records of relationships, experiences, meanings, and contexts. Poor management can expose participants, obscure the provenance of findings, or cause researchers to lose the contextual detail on which qualitative interpretation depends.

Qualitative data management is therefore both a technical and interpretive activity. It includes the secure collection, naming, storage, transcription, anonymization, translation, organization, coding, retrieval, documentation, and eventual archiving or destruction of data. Coding is the systematic process of attaching meaningful labels to segments of data so that patterns, contrasts, cases, and explanations can be developed. Neither management nor coding is neutral: decisions about what is retained, translated, condensed, coded, or excluded influence the knowledge produced.

For health-care managers and researchers, competence in this area is essential. Sensitive data may concern diagnoses, reproductive health, disability, mental health, violence, poverty, migration, professional misconduct, or institutional failure. A defensible workflow must protect confidentiality while preserving enough context for credible analysis and audit.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Design a secure and traceable workflow for qualitative health data.
2. Distinguish identifiers, pseudonyms, anonymization, confidentiality, and data minimization.
3. Evaluate transcription and translation choices in relation to the research question.
4. Construct a codebook and apply inductive, deductive, descriptive, process, and interpretive codes.
5. Use CAQDAS platforms such as NVivo, ATLAS.ti, and MAXQDA appropriately.
6. Develop audit trails, coding protocols, reflexive records, and quality checks.
7. Identify ethical, legal, technological, and interpretive risks in qualitative data management.

1. The Data Life Cycle in Qualitative Health Research

A useful way to conceptualize management is as a life cycle rather than a single storage decision:

1. Planning: determine what data are necessary, who will access them, how long they will be retained, and how risks will be managed.
2. Collection: record interviews, observations, or documents using approved procedures.
3. Transfer and ingestion: move files from recording devices to secure institutional storage.
4. Preparation: transcribe, translate, de-identify, verify, and organize the material.
5. Analysis: code, compare, memo, retrieve, display, and interpret data.
6. Quality assurance: document decisions, check consistency, examine negative cases, and preserve provenance.
7. Sharing, archiving, or destruction: follow consent, ethics approval, institutional policy, and applicable law.

These stages overlap. For example, a decision made during planning about whether verbatim speech is needed affects transcription, coding, storage capacity, and the possibility of data sharing.

A data management plan should specify:

• the types and volume of data;
• the sensitivity of each data type;
• ownership and custodianship;
• access permissions and role separation;
• encryption during transfer and storage;
• backup and recovery procedures;
• naming conventions and version control;
• transcription and translation procedures;
• de-identification rules;
• retention and destruction dates;
• procedures for data requests, breaches, and withdrawal;
• the relationship between raw data, processed data, analytic files, and outputs.

1.1 Data classes and separation

A practical classification separates the following:

The key principle is separation of identity from meaning. A participant list and a transcript should not ordinarily be stored in the same folder or accessible to the same people. A linking file may connect a participant ID to a name, but it should be encrypted, separately permissioned, and used only when justified.

2. Secure Recording, Transfer, and Storage

2.1 Recording practices

Before recording, the researcher should verify consent for the specific medium. Consent to participate does not automatically imply consent to audio recording, video recording, quotation, or secondary use. The interviewer should have a non-recorded contingency procedure when a participant withdraws recording permission.

Recording devices should be institutionally approved. Consumer cloud applications, automatic transcription services, and personal messaging platforms may transfer data to unknown jurisdictions or retain data for service improvement. Their use requires explicit governance review, contractual safeguards where appropriate, and participant consent compatible with the actual processing.

Field notes should distinguish observed material from interpretation. A useful structure includes:

• date, time, location, and activity;
• descriptive account of what occurred;
• relevant quotations or near-verbatim language;
• contextual features;
• researcher reflections and preliminary interpretations;
• unresolved questions and follow-up actions.

The separation is not absolute, but labelling prevents a later reader from mistaking an interpretation for an observation.

2.2 Transfer and storage controls

After collection, files should be transferred as soon as practical from the device to an approved encrypted location. The original should not be edited. A controlled copy may be created for processing. Secure practice commonly includes:

• encryption at rest and during transfer;
• strong authentication, preferably multi-factor authentication;
• least-privilege access;
• institutional rather than personal accounts;
• automatic screen locking and secure device configuration;
• tested backups in a separate location;
• malware protection and security updates;
• documented access logs;
• no unapproved USB devices or personal cloud folders.

A backup is useful only if it can be restored. Recovery tests should therefore be conducted periodically. A simple naming convention might be:

StudyID_ParticipantID_DataType_Date_Version.ext

For example, HCM07_P014_Interview_2025-03-18_v01.wav. Names should not contain the participant’s real name. Dates should be used consistently, and version numbers should not be overwritten.

2.3 Metadata and file integrity

Metadata give a file its provenance: who created it, when, in what language, under what conditions, and what processing has occurred. A metadata register may include file ID, participant ID, collection date, data type, language, consent status, sensitivity level, transcription status, translator, analyst, and current location.

Where institutional policy permits, a cryptographic hash can help identify whether a file has changed. A hash is not encryption and does not make a file confidential; it is an integrity check. It should be used only as an adjunct to, not a substitute for, access control and encryption.

3. Transcription as an Analytic Decision

Transcription converts speech into a searchable and analysable representation. It is never a perfectly transparent copy. Choices about punctuation, pauses, overlap, laughter, crying, silence, emphasis, pronunciation, and non-verbal events affect interpretation.

3.1 Levels of transcription

• Intelligent verbatim: preserves the substance while removing repetitions and fillers. It is suitable for many service evaluations and topic-focused studies.
• Full verbatim: preserves repetitions, fillers, false starts, and notable speech features. It is useful when interaction, hesitation, or emotion matters.
• Conversation-analytic transcription: records timing, overlap, emphasis, intonation, and sequential detail using a specialized notation system.
• Summary transcription: records the content without reproducing exact speech. It may be efficient for preliminary screening but is weaker for quotation and close interpretation.

The choice should follow the research question. A study of barriers to medication access may require accurate content and contextual pauses, whereas a study of clinician–patient interaction may require detailed turn-taking and overlap. An analyst should never claim to have conducted verbatim analysis if the material was substantially edited.

3.2 Transcription protocols

A protocol should define spelling, punctuation, speaker labels, inaudible sections, overlapping talk, non-verbal events, code-switching, and redaction. For example, [inaudible 00:14:22] is more transparent than silently omitting unclear speech. Timestamps enable verification against audio.

Quality assurance may involve a second person checking a sample against the audio, or the interviewer reviewing the transcript. Automatic speech recognition can accelerate first drafts, but it performs unevenly with accents, clinical terminology, background noise, multiple speakers, and multilingual speech. Human review remains necessary when accuracy affects interpretation or participant safety.

4. Anonymization, Pseudonymization, and Disclosure Risk

Pseudonymization replaces identifiers with codes while retaining a key that permits re-identification. Anonymization aims to make re-identification reasonably infeasible in the relevant context. In small communities, complete anonymization may be impossible because combinations of details identify a person even after names are removed.

De-identification should therefore address both direct and indirect identifiers. Examples include:

• replacing names with participant IDs;
• generalizing exact ages into bands;
• changing precise dates to months or relative time;
• removing employer names and rare job titles;
• aggregating locations;
• altering non-essential details in published cases;
• suppressing quotations that reveal identity through distinctive wording.

Alterations to quotations can create ethical and analytic problems. If wording is changed, the researcher should document the change and avoid presenting an edited passage as exact verbatim speech. In some studies, a composite account may be appropriate, but it must be labelled as composite rather than attributed to one participant.

A disclosure review should consider the mosaic effect: several apparently harmless details may jointly identify an individual. Risk increases when data are linked with public registers, social media, news reports, or institutional knowledge. Consent forms should explain foreseeable uses without promising absolute anonymity where that promise cannot be kept.

5. Translation and Multilingual Data

Translation is both linguistic and interpretive. Words may have no exact equivalent; culturally embedded expressions may carry social meanings that disappear in a literal translation. The original-language material should be retained whenever ethically and technically possible.

A robust multilingual workflow may include:

1. transcribing in the source language;
2. recording culturally significant terms and idioms;
3. translating for analysis or dissemination;
4. documenting translator decisions and uncertainties;
5. reviewing a purposive sample through back-translation, bilingual review, or team discussion;
6. coding with attention to both translated meaning and original wording.

Back-translation is not a guarantee of correctness. A literal reverse translation may appear different because several translations are semantically valid. Bilingual researchers should discuss concepts, register, politeness, gender, power, and culturally specific meanings rather than merely compare sentences.

A translation log can record source phrase, literal gloss, selected translation, alternatives considered, rationale, and analyst comments. Where coding is conducted only in English, a bilingual glossary can preserve key concepts in the source language. This is particularly important when participants use metaphors, humour, stigma-related terms, or local categories of illness.

6. Coding: From Segments to Meaning

Coding is the process of assigning a label to a meaningful segment of data. A segment may be a word, phrase, sentence, turn, paragraph, event, case, or whole document. The appropriate unit depends on the question and analytic tradition.

Coding is not simply counting words. It links data to concepts and supports comparison. A quotation may receive several codes because it can simultaneously express a practical barrier, an emotional response, and a relationship with a health professional.

6.1 Major coding approaches

• Descriptive coding: identifies the topic or subject, such as “transport difficulties.”
• In vivo coding: uses a participant’s own words, such as “the hospital forgets us.”
• Process coding: captures action, often with an -ing form, such as “negotiating appointment times.”
• Emotion coding: identifies feelings such as fear, shame, or relief.
• Values coding: captures beliefs, attitudes, and priorities.
• Structural coding: organizes data according to interview questions or domains.
• Deductive coding: applies concepts derived from theory, literature, policy, or the research objectives.
• Inductive coding: develops codes from engagement with the data.
• Interpretive or pattern coding: groups initial codes into higher-order explanations or patterns.

Most applied health studies use a hybrid approach. A framework may provide initial domains, while open coding permits unanticipated meanings and contradictions to emerge.

6.2 A codebook

A codebook is a working specification of codes and their application. It is not a substitute for judgement; it makes judgement visible and discussable. A useful codebook contains:

A weak code is broad, vague, or merely repeats the interview question. For example, “barriers” is usually less useful than “unaffordable transport,” defined as references to travel costs that prevent, delay, or condition service use. The codebook should also state whether a code is applied to explicit statements only or to reasonable interpretation of implied meaning.

6.3 Coding stages

Familiarization involves reading or listening to the data repeatedly, noting initial impressions without prematurely fixing categories.

Initial coding labels relevant segments. Coding should remain close enough to the data to preserve participants’ meanings, while being sufficiently analytic to support the research question.

Focused coding identifies codes that are frequent, significant, conceptually rich, or useful for comparison. Similar codes may be merged; overly broad codes may be split.

Pattern or axial analysis examines relationships among conditions, actions, consequences, contexts, and cases. The precise terminology varies by methodology, so researchers should not combine incompatible procedures without justification.

Thematic development moves from codes to candidate themes, reviews their coherence against coded extracts and the full dataset, defines their scope, and names them. A theme is more than a topic. “Waiting” is a topic; “waiting as a demonstration of institutional disregard” is an interpretive theme if supported by the data and analysis.

Case comparison examines similarities and differences by age, gender, service site, diagnosis, professional role, or other analytically relevant characteristics. Comparisons should not imply statistical representativeness unless the design supports that claim.

7. Reflexivity, Team Coding, and Quality

Qualitative quality does not depend on eliminating the researcher’s influence. It depends on recognizing and documenting it. Reflexive memos should record assumptions, emotional responses, prior professional knowledge, relationships with participants, and moments when the analyst’s interpretation changed.

In team research, coding consistency can be useful, but mechanical agreement is not the sole standard of rigour. Researchers may code the same material differently because they attend to different meanings. Discussion can expose assumptions and refine the codebook. A consensus process should not erase minority interpretations or force agreement before disagreements are understood.

If numerical agreement is calculated, the statistic must match its purpose and assumptions. Simple percentage agreement can be misleading because agreement may occur by chance. Cohen’s kappa, for example, is designed for two coders classifying the same units into categorical codes, but it assumes a defined set of units and categories and can be affected by prevalence and marginal distributions. It should not be treated as a universal measure of qualitative quality. Interpretive depth, reflexivity, negative-case analysis, and analytic transparency remain important.

Quality strategies include:

• peer debriefing;
• deviant or negative-case analysis;
• participant or stakeholder feedback where appropriate;
• triangulation across data sources or analysts;
• maintaining an audit trail;
• documenting data saturation or information power claims cautiously;
• preserving links between claims, coded extracts, and source files;
• seeking an alternative interpretation before finalizing a theme.

8. CAQDAS Platforms and Responsible Use

Computer-Assisted Qualitative Data Analysis Software (CAQDAS) supports organization, retrieval, annotation, coding, memoing, querying, visualization, and team coordination. NVivo, ATLAS.ti, and MAXQDA provide broadly similar functions, although interfaces and licensing arrangements differ. The software does not perform the intellectual work of interpretation; it makes selected analytic operations more manageable and traceable.

8.1 Common functions

• importing text, PDF, audio, video, images, and spreadsheets;
• attaching codes to segments;
• hierarchical code systems;
• analytic memos and annotations;
• case classifications and attributes;
• Boolean, proximity, and text searches;
• matrices and code-by-case comparisons;
• links between documents, quotations, codes, and memos;
• export of codebooks and reports;
• user accounts and project merging in some configurations.

8.2 A safe implementation workflow

1. Create a project on approved encrypted storage.
2. Establish folder structures, naming rules, roles, and a version convention.
3. Import de-identified working copies while retaining originals securely.
4. Record document metadata and case attributes without unnecessary identifiers.
5. Develop and pilot the codebook on a small purposive sample.
6. Record code changes in a decision log rather than silently replacing earlier versions.
7. Use memos to distinguish description, interpretation, and methodological decisions.
8. Back up the project and export essential codebooks, memos, and reports at intervals.
9. Check that quotations exported for reports remain de-identified.
10. Archive or destroy project components according to the approved plan.

Software labels can create a false sense of objectivity. A frequency report shows how often a code was applied, not how important a meaning is. A word cloud may privilege common words and overlook silence, metaphor, or context. Automated sentiment analysis is especially risky with clinical language, negation, irony, multilingual data, and culturally specific expressions.

9. Audit Trails and Analytic Provenance

An audit trail allows another researcher, supervisor, ethics committee, or auditor to understand how the project moved from raw material to claims. It need not make qualitative analysis reproducible in the same way as a laboratory measurement; interpretation is situated. It should, however, make decisions traceable.

An audit trail may include:

• the approved protocol and amendments;
• consent and access records;
• data-management plan;
• file inventory and metadata register;
• transcription and translation protocols;
• anonymization decisions;
• codebook versions;
• coding assignments and team discussion records;
• reflexive and analytic memos;
• theme-development tables;
• negative-case records;
• links between findings and supporting extracts;
• dissemination and disclosure checks.

A decision log might state: “Code ‘staff disrespect’ divided into ‘dismissive communication’ and ‘discriminatory treatment’ after review of 12 extracts; the distinction is based on whether participants described general interpersonal disregard or treatment linked to a social identity.” This is more informative than merely recording that the codebook was “updated.”

10. Applied Case: Continuity of Care After Hospital Discharge

The following is a fictional illustrative case.

Background

A regional health system commissioned a qualitative study of why older adults experienced difficulty maintaining care during the first month after discharge. The team conducted 30 interviews with patients, family caregivers, nurses, and primary-care staff across three facilities. Interviews were conducted in two languages.

Problem

Early files contained real names, exact home locations, and detailed diagnoses. Interviewers used inconsistent labels for the same facility, and an automated transcription service produced errors in medication names. The initial code “discharge problems” was too broad to support useful comparison.

Analysis and approach

The team created participant IDs and stored the identity key separately. Exact addresses were removed, facility names were generalized in working files, and rare combinations of age, diagnosis, and occupation were reviewed for disclosure risk. Source-language transcripts were retained; translations included a log of culturally specific terms.

A pilot codebook distinguished “unclear medication instructions,” “transport dependence,” “fragmented responsibility,” “caregiver improvisation,” and “patient advocacy.” Two analysts independently coded six interviews, then discussed differences. They found that “non-adherence,” initially used by one analyst, sometimes blamed patients for system failures. It was replaced by more precise codes describing missed doses, conflicting instructions, affordability, and negotiated treatment decisions.

Memos and matrices compared accounts by patient, caregiver, and professional role. Negative cases were actively sought: several patients reported excellent continuity because a named nurse coordinated follow-up. This led the team to refine its emerging explanation from “discharge causes fragmentation” to “continuity depends on whether responsibility is visibly assigned and supported across organizational boundaries.”

Outcome

The final report used carefully reviewed quotations, omitted identifying details, and distinguished patient experience from institutional interpretation. The health system introduced a named transition coordinator, a medication reconciliation checklist, and a follow-up call protocol. The qualitative findings did not prove that these interventions would improve outcomes, but they explained mechanisms and conditions that could guide implementation and subsequent evaluation.

Learning points

1. Data security and analytic quality are connected: poor identifiers and poor metadata can both create risk and weaken interpretation.
2. Codes should describe experiences and mechanisms rather than reproduce stigmatizing institutional labels.
3. Multilingual analysis requires preserving original meanings, not merely producing fluent translations.
4. Negative cases can improve an explanation rather than threaten it.
5. An audit trail supports defensible managerial action by showing how recommendations were derived.

11. Ethics, Governance, and Contemporary Developments

Qualitative data management must comply with the approved research protocol, institutional policy, and applicable data-protection law. Relevant principles generally include purpose limitation, data minimization, confidentiality, secure processing, transparency, participant rights, and proportional retention. The exact legal requirements vary by jurisdiction; researchers must consult the relevant ethics committee, data-protection officer, and institutional policy rather than assume that one international framework applies everywhere.

Governance challenges include secondary use, data linkage, cross-border cloud storage, withdrawal after anonymization, and requests for data sharing. A participant may be able to withdraw identifiable data before analysis, but withdrawal may be technically impossible after data have been irreversibly anonymized or incorporated into aggregate findings. This possibility should be explained during consent.

Artificial intelligence can assist with transcription, translation, document classification, and preliminary coding. However, risks include undisclosed data reuse, model retention, inaccurate transcription, hallucinated text, bias against dialects, and loss of contextual or emotional meaning. AI-generated output should be treated as an unverified draft. Human researchers must review it against the source, document its use, avoid uploading restricted data to unapproved systems, and ensure that the final interpretation remains accountable to the research team.

Contemporary best practice increasingly emphasizes data sovereignty, community governance, culturally safe research, and meaningful involvement of participants in decisions about access and reuse. In Indigenous, minority, or otherwise historically exploited communities, institutional consent alone may be insufficient. Community protocols may govern storage, interpretation, quotation, and dissemination.

Challenges and Limitations

First, de-identification can reduce contextual richness. Removing a small community, occupation, or diagnosis may make a quotation safer but less analytically meaningful. The solution is not to ignore risk; it is to document trade-offs and use the minimum necessary detail.

Second, coding can fragment narratives. When a long account is divided into many coded pieces, chronology, tone, and relationships may disappear. Researchers should preserve whole-case files, use case summaries, and retrieve surrounding context before interpreting extracts.

Third, team consistency can become an unhelpful obsession. High agreement may reflect a shallow codebook, while disagreement may reveal productive ambiguity. Coding procedures should serve the methodology and question.

Fourth, software can encourage overcoding, premature quantification, and dependence on visible features of text. Silence, absence, embodied experience, and power relations may require close reading and reflexive interpretation beyond software functions.

Finally, retention creates a tension between future scholarly value and ongoing privacy risk. A defensible retention decision weighs consent, scientific value, sensitivity, legal obligations, community expectations, and the feasibility of safe access.

Best-Practice Checklist

Before analysis, confirm that:

• consent covers recording, transcription, translation, quotation, storage, and proposed reuse;
• raw files are encrypted and identity keys are separated;
• the team has an approved naming, backup, and access procedure;
• transcription and translation choices are documented;
• de-identification addresses indirect as well as direct identifiers;
• the codebook has definitions, boundaries, examples, and version history;
• analysts maintain reflexive and analytic memos;
• CAQDAS files are backed up and access-controlled;
• quotations are checked for disclosure risk;
• the audit trail links data, coding, interpretation, and reporting;
• AI and cloud services have been approved for the sensitivity of the data.

Chapter Summary

Qualitative data management is an integrated life-cycle process encompassing planning, secure recording, transfer, storage, transcription, translation, anonymization, analysis, reporting, and archiving or destruction. Separation of identity data from research data, least-privilege access, encryption, tested backups, clear metadata, and version control protect participants and preserve provenance.

Transcription and translation are interpretive decisions. Their level of detail and documentation must match the research question and linguistic context. Pseudonymization is not the same as anonymization, and indirect identifiers can create a mosaic effect. Coding should move deliberately from meaningful segments to focused codes, patterns, themes, and explanations. A codebook makes analytic decisions transparent, while reflexivity, negative-case analysis, triangulation, and peer discussion support rigour.

CAQDAS platforms improve organization and retrieval but do not replace interpretation. Their outputs must be read in context, and automated transcription, translation, or coding requires human verification and governance. An audit trail preserves the pathway from source material to claims. In health research, secure management is therefore not administrative overhead: it is an ethical condition and a foundation of credible knowledge.

Key Terms

Audit trail: A documented record of methodological, data-management, coding, and interpretive decisions.

CAQDAS: Computer-Assisted Qualitative Data Analysis Software used to organize, code, retrieve, and memo qualitative materials.

Code: A label assigned to a meaningful segment of qualitative data.

Codebook: A structured record of code names, definitions, boundaries, examples, and versions.

Deductive coding: Applying concepts derived from theory, literature, policy, or research objectives.

De-identification: Removing or modifying information that could identify a participant.

Disclosure risk: The possibility that a person can be identified from data alone or in combination with other information.

Inductive coding: Developing codes through engagement with the data rather than imposing a predetermined framework.

In vivo code: A code using a participant’s own words.

Metadata: Information describing a file’s origin, content, processing, context, and status.

Negative case: Data that contradict, qualify, or complicate an emerging interpretation.

Pseudonymization: Replacing identifiers with codes while retaining a re-identification key.

Reflexivity: Critical examination of how the researcher’s position, assumptions, and relationships influence the research.

Source data: Original recordings, notes, documents, images, or other collected materials.

Thematic analysis: A method of developing and interpreting patterned meanings across qualitative data.

Translation log: A record of linguistic choices, alternatives, uncertainties, and rationales.

Verbatim transcription: Transcription that preserves speech features according to a defined protocol.

Version control: A system for tracking changes to files, codebooks, and analytic products.

Review Questions

Short Answer Questions

1. Distinguish pseudonymization from anonymization.
2. Why should the identity key be stored separately from transcripts?
3. State three elements that should appear in a codebook.
4. Why is translation an interpretive activity rather than a purely mechanical one?
5. What is the purpose of an audit trail?

Long Answer Questions

1. Explain the qualitative data life cycle and identify the governance risks at each stage.
2. Compare intelligent verbatim, full verbatim, and conversation-analytic transcription, giving an appropriate health-research use for each.
3. Discuss how a research team can combine deductive and inductive coding without losing unexpected meanings.
4. Evaluate the benefits and limitations of using NVivo, ATLAS.ti, or MAXQDA in a multilingual health study.
5. Explain how reflexivity, negative-case analysis, and team discussion contribute to qualitative rigour.

Analytical / Application Questions

1. A transcript contains a participant’s name, rare occupation, exact village, and a distinctive quotation. Design a de-identification and disclosure-review procedure that preserves as much analytic context as possible.
2. Two analysts disagree about whether extracts describing missed medication doses should be coded as “patient non-adherence” or “conflicting clinical instructions.” How should the team investigate and resolve the disagreement?
3. An institution proposes uploading sensitive interview recordings to a free online AI transcription service. Prepare a governance assessment covering consent, confidentiality, accuracy, data location, human verification, and alternatives.

Multiple Choice Questions with Answers

Question 1

Which practice best separates participant identity from research meaning?

A. Store names inside every transcript for easy retrieval

B. Store the identity key separately from coded research files

C. Use personal email accounts for all project files

D. Publish exact locations with participant quotations

Correct Answer: B. Store the identity key separately from coded research files

Question 2

What is the primary purpose of a codebook?

A. To replace the need for interpretation

B. To calculate sample size

C. To document how codes are defined and applied

D. To encrypt audio recordings

Correct Answer: C. To document how codes are defined and applied

Question 3

Which statement correctly describes pseudonymization?

A. It guarantees that re-identification is impossible

B. It replaces identifiers while retaining a possible linking key

C. It permanently destroys all source data

D. It converts qualitative data into quantitative data

Correct Answer: B. It replaces identifiers while retaining a possible linking key

Question 4

Why should original-language material generally be retained in multilingual research?

A. It prevents the need for any translation review

B. It preserves meanings that may be lost in translation

C. It makes anonymization unnecessary

D. It guarantees identical interpretations by all analysts

Correct Answer: B. It preserves meanings that may be lost in translation

Question 5

Which is an example of a negative case?

A. An extract that exactly repeats the dominant pattern

B. A file with a missing date

C. Evidence that contradicts or qualifies an emerging theme

D. A transcript without a codebook

Correct Answer: C. Evidence that contradicts or qualifies an emerging theme

Question 6

What does CAQDAS primarily support?

A. Automatic proof that themes are true

B. Organization, retrieval, coding, and memoing of qualitative data

C. Replacement of informed consent

D. Elimination of researcher interpretation

Correct Answer: B. Organization, retrieval, coding, and memoing of qualitative data

Question 7

Which transcription choice is most appropriate for studying turn-taking and overlapping speech?

A. Summary transcription

B. Intelligent verbatim transcription only

C. Conversation-analytic transcription

D. Keyword extraction without transcripts

Correct Answer: C. Conversation-analytic transcription

Question 8

Why can a word-frequency report be misleading in qualitative analysis?

A. Frequent words are always deleted by software

B. Frequency necessarily proves causal importance

C. Important meanings may be expressed through silence, metaphor, or uncommon words

D. Software cannot store text files

Correct Answer: C. Important meanings may be expressed through silence, metaphor, or uncommon words

Question 9

Which control most directly limits access to sensitive qualitative data?

A. Least-privilege permissions

B. Using identical passwords for all team members

C. Removing version numbers

D. Publishing the file inventory

Correct Answer: A. Least-privilege permissions

Question 10

How should AI-generated transcription generally be treated in sensitive health research?

A. As a final authoritative record without review

B. As an unverified draft requiring human checking and governance

C. As a substitute for participant consent

D. As automatically anonymized output

Correct Answer: B. As an unverified draft requiring human checking and governance