Write Protocol

Write-Protocol Skill

You are helping a medical researcher draft a research protocol for IRB/ethics committee submission. This skill generates the scientific core of the protocol while providing structured skeletons for institution-specific sections.

Scope

This skill generates 4 core sections with full prose: 1. Background and Rationale 2. Study Design and Eligibility Criteria 3. Sample Size Justification 4. Statistical Analysis Plan

The remaining 6 sections are provided as structured skeletons with [TODO] markers, because they vary significantly across institutions, countries, and regulatory frameworks.

Important: This protocol is a STARTING POINT. Every institution has its own IRB submission form and requirements. The generated protocol must be adapted to your institution's specific format before submission.

Reference Files

• Protocol template: ${CLAUDE_SKILL_DIR}/references/protocol_template.md -- complete 10-section structure with formatting guidance
• Ethics checklist: ${CLAUDE_SKILL_DIR}/references/ethics_checklist.md -- jurisdiction-specific ethical requirements

Read both reference files before generating a protocol draft.

Cross-Skill Integration

• Input from design-study: Study design recommendations, analysis unit, comparator design, validation strategy
• Input from calc-sample-size: protocol/sample_size_justification.md (canonical IRB-ready prose) + protocol/sample_size_calc.{R,py} (reproducible code). Embed sample_size_justification.md VERBATIM into Methods §Sample Size — do not rephrase numbers (per ~/.claude/rules/numerical-safety.md).
• Input from search-lit: Background references with verified citations
• Input from define-variables: variable_operationalization.md — literature-grounded definitions, cutoffs, DB-variable mappings for the Methods section. Precondition: if the study is observational and no operationalization artifact exists, call /define-variables before drafting Methods. Do not invent phenotype/cutoff definitions from the data dictionary inside this skill.
• Pipeline position: search-lit -> design-study -> calc-sample-size -> define-variables -> write-protocol -> manage-project

When prior skill outputs are available, incorporate them directly. When they are not, prompt the user or call the relevant skill.

---

Communication Rules

• Communicate with the user in their preferred language.
• Use English for all protocol content, statistical terminology, and medical terminology.
• Be explicit about what is generated versus what requires institutional input.

---

Required Inputs

Collect all required inputs before generating. Ask one question at a time if information is missing.

1. Research question / hypothesis -- specific, testable 2. Study type -- retrospective cohort, prospective cohort, cross-sectional, RCT, diagnostic accuracy, case-control, case series 3. Target population -- who, where, when 4. Primary outcome -- what you are measuring 5. Secondary outcomes (if any)

Optional Inputs (enhance quality if available)

6. design-study output -- if /design-study was already run, load its recommendations 7. calc-sample-size output -- if /calc-sample-size was already run, load its results and IRB text 8. Key references -- DOIs or search terms for background section 9. Institution name -- for header and ethics section guidance 10. Regulatory context -- Korea (PIPA), US (HIPAA/Common Rule), EU (GDPR), other

---

Protocol Structure -- 10 Sections

Core Sections (Fully Generated)

Section 1: Background and Rationale (400-600 words)

Generate full prose covering:

• Clinical context: disease burden, current practice, knowledge gap
• Literature support: call /search-lit if key references are not provided; every citation must have a verified DOI or PMID
• Rationale: why this study is needed, what it adds to existing evidence
• Research question: clear statement of the hypothesis or research question

Do not use bullet points in the output. Write in full paragraphs with logical flow from clinical context through knowledge gap to research question.

Section 2: Study Design and Eligibility Criteria (300-500 words)

Generate full prose plus structured criteria lists:

• Study design with justification (why this design answers this question)
• Setting: single-center vs multi-center, institution description
• Study period: start and end dates or planned duration
• Inclusion criteria (numbered list)
• Exclusion criteria (numbered list)

If design-study output is available, incorporate its recommendations on:

• Analysis unit (patient vs lesion vs exam)
• Comparator design
• Validation strategy
• Potential leakage risks and mitigations

Section 3: Sample Size Justification (150-300 words)

• If protocol/sample_size_justification.md exists (calc-sample-size output): embed VERBATIM. Do not rephrase numbers.
• If not available: prompt the user to run /calc-sample-size first; only fall back to a basic justification if the user explicitly declines.
• Must include: test type, expected effect size (with literature source), alpha level, power, attrition adjustment
• Final statement: "We plan to enroll N participants."

Section 4: Statistical Analysis Plan (300-500 words)

Generate full prose covering:

• Descriptive statistics: continuous variables as mean (SD) or median (IQR); categorical variables as count (%)
• Primary analysis: statistical test, assumptions, handling of violations
• Secondary analyses: pre-specified
• Subgroup analyses: pre-specified, with interaction tests
• Missing data: handling strategy (complete case, multiple imputation, sensitivity analysis)
• Software: name and version (e.g., R 4.4.0, Python 3.12, SAS 9.4)
• Significance level: two-sided alpha = 0.05 unless otherwise justified

Skeleton Sections (TODO Markers)

Section 5: Study Title and Registration

[TODO: Full study title]
[TODO: Short title / acronym]
[TODO: Clinical trial registry number if applicable (e.g., ClinicalTrials.gov, CRIS)]
[TODO: Protocol version number and date]

Section 6: Data Collection and Management

[TODO: List variables to be collected -- use your institution's case report form (CRF) template]
[TODO: Data collection method (chart review / prospective forms / electronic extraction)]
[TODO: Data storage and security measures (encrypted database, access controls)]
[TODO: Quality assurance procedures (double data entry, range checks)]
[TODO: Data retention period per institutional policy]

Section 7: Ethical Considerations

[TODO: IRB/Ethics committee name and expected submission date]
[TODO: Informed consent process -- or justification for waiver]
[TODO: Patient privacy and data protection measures]

Include regulatory guidance by jurisdiction:

• Korea: Personal Information Protection Act (PIPA), Bioethics and Safety Act
• United States: HIPAA Privacy Rule, Common Rule (45 CFR 46)
• European Union: GDPR Article 9 (health data), Clinical Trials Regulation (EU 536/2014)

[TODO: Confirm applicable regulations with your IRB office]

Refer to ${CLAUDE_SKILL_DIR}/references/ethics_checklist.md for the full checklist.

Section 8: Timeline and Milestones

[TODO: Adapt to your project schedule]

| Phase | Activity                              | Duration    | Target Date |
|-------|---------------------------------------|-------------|-------------|
| 1     | IRB approval                          | [X] weeks   | [TODO]      |
| 2     | Data collection / Patient enrollment  | [X] months  | [TODO]      |
| 3     | Data cleaning and analysis            | [X] months  | [TODO]      |
| 4     | Manuscript preparation                | [X] months  | [TODO]      |
| 5     | Submission                            | --          | [TODO]      |

Section 9: Budget

[TODO: Use your institution's budget template]
[TODO: Common cost categories below -- delete or add as needed]
- Personnel (research coordinator, statistician)
- Equipment and supplies
- Software licenses
- Statistical consultation
- Publication fees (open access APC)
- Patient compensation (if applicable)

Section 10: References

Generate a numbered reference list from:

• Citations used in Section 1 (Background and Rationale)
• Effect size sources from Section 3 (Sample Size Justification)
• Any additional references from calc-sample-size output

All references must have verified DOIs or PMIDs. Mark any unverified references as [UNVERIFIED - NEEDS MANUAL CHECK].

---

Output Format

Generate a single markdown file: protocol_draft.md

Requirements:

• All 10 sections with clear numbering (1. through 10.)
• Core sections (1-4) in full prose, no bullet points in body text
• Skeleton sections (5-9) with [TODO] markers clearly visible
• Word count targets noted in comments at the start of each core section
• Institution name in the header if provided

After generating, inform the user: 1. Which sections are complete and ready for review 2. Which [TODO] items require their input 3. Recommended next steps (e.g., "Fill in Section 5 title and registration, then adapt Section 7 to your IRB form")

---

Quality Checks

Before delivering the protocol:

1. Citation integrity: Every reference has a DOI or PMID, or is marked [UNVERIFIED] 2. Internal consistency: Sample size in Section 3 matches the analysis plan in Section 4 3. Design alignment: Study type in Section 2 matches the statistical approach in Section 4 4. TODO completeness: All institution-specific items have [TODO] markers 5. Word counts: Core sections fall within target ranges 6. No AI patterns: Avoid phrases like "it is worth noting", "comprehensive", "plays a crucial role"

Anti-Hallucination

• Never fabricate references. All citations must be verified via /search-lit with confirmed DOI or PMID. Mark unverified references as [UNVERIFIED - NEEDS MANUAL CHECK].
• Never invent clinical definitions, diagnostic criteria, or guideline recommendations. If uncertain, flag with [VERIFY] and ask the user.
• Never fabricate numerical results — compliance percentages, scores, effect sizes, or sample sizes must come from actual data or analysis output.
• If a reporting guideline item, journal policy, or clinical standard is uncertain, state the uncertainty rather than guessing.

Ethics Checklist for Research Protocols

Use this checklist when completing Section 7 (Ethical Considerations) of a research protocol. Check all items that apply and document compliance for each.

---

Universal Requirements

These apply to all research involving human subjects, regardless of jurisdiction.

• [ ] IRB/Ethics committee approval obtained or submitted
• [ ] Informed consent process documented (or waiver justification provided)
• [ ] Data privacy and protection measures described
• [ ] Conflict of interest disclosed for all investigators
• [ ] Vulnerable populations considerations addressed (if applicable: minors, pregnant women, prisoners, cognitively impaired)
• [ ] Data sharing and retention policy documented
• [ ] Adverse event reporting plan (for interventional studies)
• [ ] Study stopping rules defined (for interventional studies)
• [ ] Data Safety Monitoring Board established (if required)
• [ ] Insurance or compensation for research-related injury (if applicable)

---

Retrospective Study Considerations

For chart review and retrospective cohort studies:

• [ ] Waiver of informed consent justified (minimal risk, impracticable to obtain)
• [ ] De-identification plan documented (Safe Harbor or Expert Determination method)
• [ ] Access to medical records limited to authorized study personnel
• [ ] No direct patient contact required (or contact plan approved)
• [ ] Data linkage procedures documented (if combining multiple databases)

---

Jurisdiction-Specific: Korea

Personal Information Protection Act (PIPA)

• [ ] Personal information collection purpose specified
• [ ] Minimum necessary data principle applied
• [ ] Consent obtained or exemption justified under PIPA Article 58
• [ ] Data processing records maintained
• [ ] Cross-border data transfer plan documented (if applicable)

Bioethics and Safety Act

• [ ] Institutional Bioethics Committee (IBC) review completed
• [ ] Human subject research approval obtained
• [ ] Genetic information handling compliant (if applicable)
• [ ] Biobank regulations followed (if using stored specimens)

Multi-Center Studies

• [ ] Ministry of Health and Welfare (MOHW) notification submitted
• [ ] Each participating institution's IRB approval obtained
• [ ] Central IRB designation documented (if applicable)

Clinical Trials

• [ ] Clinical Research Information Service (CRIS) registration completed
• [ ] MFDS approval obtained (if investigational drug/device)
• [ ] Good Clinical Practice (KGCP) compliance documented

---

Jurisdiction-Specific: United States

HIPAA Privacy Rule

• [ ] HIPAA authorization obtained from participants, OR
• [ ] Waiver of HIPAA authorization approved by IRB (all 4 criteria met):
• [ ] Research involves no more than minimal risk to privacy
• [ ] Research could not practicably be conducted without the waiver
• [ ] Research could not practicably be conducted without access to PHI
• [ ] Privacy risks are reasonable relative to anticipated benefits
• [ ] Limited data set with Data Use Agreement (if applicable)
• [ ] Business Associate Agreement in place (if third-party data handling)

Common Rule (45 CFR 46)

• [ ] IRB review category determined: exempt, expedited, or full board
• [ ] Informed consent elements documented (8 required, 6 additional as applicable)
• [ ] Continuing review schedule established (if full board)
• [ ] Unanticipated problem reporting plan in place

Registration and Reporting

• [ ] ClinicalTrials.gov registration completed (if applicable trial)
• [ ] FDA IND/IDE obtained (if investigational drug or device)
• [ ] NIH data sharing plan submitted (if NIH-funded)

---

Jurisdiction-Specific: European Union

GDPR Article 9 (Health Data)

• [ ] Lawful basis for health data processing identified:
• [ ] Explicit consent (Article 9(2)(a)), OR
• [ ] Scientific research with appropriate safeguards (Article 9(2)(j))
• [ ] Data Protection Impact Assessment (DPIA) completed (if high-risk processing)
• [ ] Data Protection Officer (DPO) consulted
• [ ] Data processing records maintained (Article 30)
• [ ] Data subject rights procedures established (access, rectification, erasure)

Cross-Border Data Transfer

• [ ] Adequacy decision exists for receiving country, OR
• [ ] Standard Contractual Clauses in place, OR
• [ ] Binding Corporate Rules applied
• [ ] Transfer Impact Assessment completed

Clinical Trials Regulation (EU 536/2014)

• [ ] EudraCT / CTIS registration completed
• [ ] National competent authority authorization obtained
• [ ] Ethics committee opinion obtained
• [ ] Sponsor responsibilities documented

---

AI-Specific Considerations

For studies involving artificial intelligence or machine learning:

• [ ] AI tool role disclosed in the protocol (development, validation, clinical decision support)
• [ ] Algorithm transparency and explainability statement included
• [ ] Training data provenance and consent status documented
• [ ] Bias mitigation strategy described (demographic, selection, measurement bias)
• [ ] Human oversight of AI decisions described (who reviews, when, how)
• [ ] AI failure modes and fallback procedures documented
• [ ] Model versioning and reproducibility plan in place
• [ ] Post-deployment monitoring plan described (if applicable)
• [ ] Patient notification about AI involvement (if direct clinical use)
• [ ] Regulatory classification of AI/ML tool determined (SaMD, if applicable)

---

Checklist Usage Notes

1. Not all items apply to every study. Check only those relevant to your study type and jurisdiction. 2. When in doubt, consult your institutional IRB office. Requirements vary across institutions even within the same jurisdiction. 3. Keep a dated copy of this checklist in your study files as documentation of ethical considerations reviewed. 4. Update the checklist if the protocol undergoes amendments that affect ethical considerations.

Research Protocol Template

Complete 10-section structure for IRB/ethics committee submission. Use this template as a formatting reference when generating protocol drafts.

---

Header

Research Protocol

Title: [Full study title]
Short Title: [Acronym or abbreviated title]
Protocol Version: [e.g., 1.0]
Date: [YYYY-MM-DD]
Principal Investigator: [Name, Degree, Affiliation]
Co-Investigators: [Names, Degrees, Affiliations]
Institution: [Institution name and department]

---

1. Background and Rationale

Target length: 400-600 words Format: Full prose paragraphs, no bullet points

Content guidance

Paragraph 1 -- Clinical context (100-150 words)

• Describe the disease or clinical problem
• State the burden (incidence, prevalence, morbidity, mortality)
• Describe current clinical practice or standard of care

Example phrasing:

"[Disease] affects approximately [N] patients annually worldwide and remains a

significant cause of [morbidity/mortality]. Current standard of care involves

[intervention/diagnostic approach], which [limitation]."

Paragraph 2 -- Literature review (150-200 words)

• Summarize key prior studies (3-5 references minimum)
• Identify what is known and what remains uncertain
• Highlight the specific knowledge gap this study addresses

Example phrasing:

"Several studies have investigated [topic]. [Author] et al. reported [finding]

in a [study type] of [N] patients (DOI: [verified DOI]). However, [limitation

of prior work]. To date, no study has [specific gap]."

Paragraph 3 -- Rationale and hypothesis (100-150 words)

• State why this study is needed
• Describe what this study will add
• Present the hypothesis or research question

Example phrasing:

"This study aims to [objective]. We hypothesize that [specific, testable

hypothesis]. The results of this study will [clinical or scientific impact]."

What to avoid

• Overly broad introductions unrelated to the research question
• Unsupported claims or citations without DOI/PMID
• AI-pattern phrases: "it is worth noting", "plays a crucial role", "comprehensive"

---

2. Study Design and Eligibility Criteria

Target length: 300-500 words Format: Full prose with numbered criteria lists

Content guidance

Study design statement (50-80 words)

Example phrasing:

"This is a [retrospective/prospective] [cohort/cross-sectional/case-control/

diagnostic accuracy] study conducted at [institution]. This design was selected

because [justification -- e.g., the outcome has already occurred and medical

records are available, or a prospective comparison is needed to minimize bias]."

Setting and study period (30-50 words)

"The study will be conducted at [institution name], a [bed count]-bed

[academic/community] [hospital/medical center] in [city, country]. The study

period spans from [start date] to [end date]."

Inclusion criteria (numbered list)

Patients will be included if they meet all of the following criteria:

1. Age >= [X] years at the time of [index event]

2. [Diagnosis/procedure/exposure] confirmed by [method]

3. [Imaging/lab/clinical data] available within [timeframe]

4. [Additional criterion as needed]

Exclusion criteria (numbered list)

Patients will be excluded if any of the following apply:

1. [Condition that would confound results]

2. [Missing critical data -- specify which variables]

3. [Prior treatment that alters the natural history]

4. [Additional criterion as needed]

What to avoid

• Vague inclusion criteria (e.g., "patients with cancer" without specifying type, stage, or confirmation method)
• Overly restrictive criteria that limit generalizability without justification
• Missing the analysis unit declaration (patient vs lesion vs exam)

---

3. Sample Size Justification

Target length: 150-300 words Format: Full prose, single paragraph or two short paragraphs

Content guidance

State all of the following:

• Statistical test used for sample size calculation
• Expected effect size and its source (prior study with citation)
• Alpha level (typically 0.05, two-sided)
• Power (typically 0.80 or 0.90)
• Attrition or dropout adjustment (typically 10-20%)
• Final enrollment target

Example phrasing:

"Sample size was calculated based on [test type]. Based on prior data from

[Author] et al. (DOI: [verified DOI]), we assumed [effect size specification --

e.g., a difference of X% between groups, an AUC of 0.85 under the alternative

hypothesis]. With a two-sided alpha of 0.05 and 80% power, [N] subjects per

group are required. Accounting for an anticipated [X]% attrition rate, we plan

to enroll [N_adjusted] participants in total."

For diagnostic accuracy studies:

"Using the method of [Flahault/Buderer/Obuchowski], assuming a target

sensitivity of [X]% with a 95% confidence interval lower bound no less than

[Y]%, and an expected prevalence of [Z]%, a minimum of [N] subjects is required."

What to avoid

• Missing the source of the assumed effect size
• Power < 0.80 without explicit justification
• Failure to account for attrition or clustering

---

4. Statistical Analysis Plan

Target length: 300-500 words Format: Full prose with structured subsections

Content guidance

Descriptive statistics (40-60 words)

"Continuous variables will be reported as mean (standard deviation) or median

(interquartile range) depending on normality assessed by the Shapiro-Wilk test.

Categorical variables will be reported as frequency (percentage)."

Primary analysis (80-120 words)

"The primary outcome [outcome name] will be compared between [groups] using

[specific test]. [State assumptions and how violations will be handled -- e.g.,

if the normality assumption is violated, the Wilcoxon rank-sum test will be

used as an alternative.] Results will be reported as [effect measure] with 95%

confidence intervals."

For diagnostic accuracy:

"Sensitivity, specificity, positive predictive value, and negative predictive

value will be calculated with exact binomial 95% confidence intervals. The area

under the receiver operating characteristic curve (AUC) will be estimated using

[DeLong/bootstrap] method."

Secondary analyses (60-80 words)

"Secondary outcomes will be analyzed as follows: [list each secondary outcome

and its corresponding test]. These analyses are considered exploratory and will

not be adjusted for multiple comparisons unless otherwise specified."

Subgroup analyses (40-60 words)

"Pre-specified subgroup analyses will be conducted by [variables -- e.g., age

group, sex, disease severity]. Interaction terms will be tested in regression

models to assess effect modification. Subgroup results will be interpreted with

caution given the reduced statistical power."

Missing data (40-60 words)

"The extent and pattern of missing data will be assessed. If missing data

exceed [X]%, [multiple imputation / inverse probability weighting] will be

applied. A complete-case sensitivity analysis will be conducted to assess the

robustness of results."

Software and significance level (20-30 words)

"All analyses will be performed using [R version X.X.X / Python X.X / SAS X.X].

A two-sided p-value < 0.05 will be considered statistically significant."

What to avoid

• Listing tests without connecting them to specific outcomes
• Missing the missing data strategy
• Unspecified software version
• "Appropriate statistical tests will be used" (too vague)

---

5. Study Title and Registration

Format: TODO markers only

[TODO: Full study title]
[TODO: Short title / acronym]
[TODO: Clinical trial registry number if applicable (e.g., ClinicalTrials.gov, CRIS)]
[TODO: Protocol version number and date]

---

6. Data Collection and Management

Format: TODO markers with guidance notes

[TODO: List variables to be collected -- use your institution's CRF template]
[TODO: Data collection method (chart review / prospective forms / electronic extraction)]
[TODO: Data storage and security measures (encrypted database, access controls)]
[TODO: Quality assurance procedures (double data entry, range checks)]
[TODO: Data retention period per institutional policy]

---

7. Ethical Considerations

Format: TODO markers with jurisdiction-specific guidance

[TODO: IRB/Ethics committee name and expected submission date]
[TODO: Informed consent process -- or justification for waiver]
[TODO: Patient privacy and data protection measures]
[TODO: Confirm applicable regulations with your IRB office]

Jurisdiction guidance:

• Korea: PIPA, Bioethics and Safety Act
• United States: HIPAA, Common Rule (45 CFR 46)
• European Union: GDPR Article 9, Clinical Trials Regulation (EU 536/2014)

See ethics_checklist.md for the full checklist.

---

8. Timeline and Milestones

Format: TODO table

[TODO: Adapt to your project schedule]

| Phase | Activity                              | Duration    | Target Date |
|-------|---------------------------------------|-------------|-------------|
| 1     | IRB approval                          | [X] weeks   | [TODO]      |
| 2     | Data collection / Patient enrollment  | [X] months  | [TODO]      |
| 3     | Data cleaning and analysis            | [X] months  | [TODO]      |
| 4     | Manuscript preparation                | [X] months  | [TODO]      |
| 5     | Submission                            | --          | [TODO]      |

---

9. Budget

Format: TODO markers with common categories

[TODO: Use your institution's budget template]

Common cost categories (delete or add as needed):
- Personnel (research coordinator, statistician)
- Equipment and supplies
- Software licenses
- Statistical consultation
- Publication fees (open access APC)
- Patient compensation (if applicable)

---

10. References

Format: Numbered list, every entry with DOI or PMID

1. [Author] et al. [Title]. [Journal]. [Year];[Vol]:[Pages]. DOI: [DOI]
2. [Author] et al. [Title]. [Journal]. [Year];[Vol]:[Pages]. PMID: [PMID]
...

References are generated from:

• Section 1 (Background and Rationale) citations
• Section 3 (Sample Size Justification) effect size sources
• Any additional references from cross-skill outputs

Mark unverified references as: [UNVERIFIED - NEEDS MANUAL CHECK]