Clinical Report Writing Skill — Complete Guide

The Clinical Report Writing skill helps you draft and validate clinical documentation: case reports (CARE guidelines), diagnostic reports (radiology, pathology, lab), clinical trial reports (ICH-E3, SAE, CSR), and patient documentation (SOAP notes, H&P, discharge summaries). This guide explains what the skill covers, how to use it, what structure and templates it provides, what references it uses, and what checks the built-in tools run.

What the skill includes

The skill guides four major categories of clinical reports:

1. Clinical case reports for journal publication — CARE (CAse REport) guidelines: title, keywords, abstract, introduction, patient information, clinical findings, timeline, diagnostic assessment, therapeutic interventions, follow-up, discussion, and informed consent.
2. Diagnostic reports — Radiology, pathology, and laboratory report structures with standard sections (e.g. clinical indication, technique, comparison, findings, impression).
3. Clinical trial reports — Clinical Study Reports (CSR) per ICH-E3 (title page, synopsis, TOC, abbreviations, ethics, investigators, introduction, objectives, study patients, efficacy, safety, discussion, tables/figures, references, appendices); SAE reports; DSMB-style summaries.
4. Patient documentation — SOAP notes, history & physical (H&P), discharge summaries, consultation notes.

The skill also enforces regulatory and quality practices: HIPAA de-identification (all 18 identifiers), informed consent, JCAHO “Do Not Use” and ambiguous abbreviations, and optional use of the scientific-schematics skill for at least one figure per report (e.g. timeline, CONSORT flow).

The agent follows the skill’s workflow when you ask it to draft, validate, or refine any of these document types. It uses bundled templates and reference files and runs extension tools for validation and data formatting.

What you can prompt

You can ask the agent to do any of the following (alone or in combination):

Example conversation:

You: I need a case report for journal submission following CARE guidelines. I’ve added my notes and the patient timeline (de-identified).

Agent: I’ll draft a CARE-compliant case report. I’ve read your notes and timeline. A few clarifications: (1) What is the main clinical lesson or novelty you want the conclusion to stress? (2) Do you have consent for publication and have all 18 HIPAA identifiers been removed or generalised?

You: The lesson is early imaging changed management. Yes, consent on file and I’ve stripped identifiers — can you double-check with the de-ID tool?

Agent: I’ve run the de-identification checker; one date pattern was flagged (admission date in the timeline). I’ve generalised it to “Day 0” in the draft. I’ve created output/case-report-draft.md with CARE sections. I also ran the CARE validator — all required elements are present. Want to add a figure (e.g. timeline or algorithm) or refine the discussion?

Add your source materials (notes, timelines, lab values, CSV data) to the project workspace and activate the Clinical Report Writing skill; the agent will guide you step by step.

Created file structure

The skill does not impose a single folder layout. It helps you elaborate content using bundled templates. You (or the agent with your approval) typically create documents in a project folder such as output/ or a name you choose. The skill provides these templates (in the skill’s assets/), which the agent can copy and fill:

Case reports and trial reports

Patient documentation

Diagnostic reports

Checklists

A typical case report the skill helps you produce has sections like: Title, Keywords, Abstract, Introduction, Patient Information, Clinical Findings, Timeline (table or figure), Diagnostic Assessment, Therapeutic Interventions, Follow-up and Outcomes, Discussion, Informed Consent. The agent can run validate_case_report and check_deidentification on the draft and fix gaps iteratively.

What is used as reference

The skill loads the following reference files from its references/ directory when the agent needs them. They are not reproduced in full here; they guide structure and content:

Standards referenced conceptually include CARE, ICH E6(R3) GCP, ICH E3 (CSR), HIPAA (18 identifiers, de-identification), JCAHO “Do Not Use” list, and FDA (study identifiers, adverse event reporting). For exact regulatory wording, use the official standards alongside the skill.

What checks are made

The skill ships with an extension that registers seven tools. The agent can run them on files in your workspace.

1. validate_case_report

Purpose: Validate a clinical case report against CARE guidelines.

What it does:

• CARE checklist — Looks for required elements: title containing “case report” or “case study”, keywords, abstract, introduction, patient information, clinical findings, timeline, diagnostic assessment, therapeutic interventions, follow-up, discussion, informed consent. Reports PASS/MISSING per element and an overall compliance rate (e.g. 90%+ = PASS).
• Quick HIPAA check — Scans for common HIPAA identifier patterns (dates, phone, email, SSN, MRN, etc.) and reports violations.
• Word count — Reports word count (typical case report range 1500–3000).
• References — Counts citation-style references (e.g. [1]) and suggests a minimum (e.g. 10).

Output: A short report with overall status (PASS / NEEDS_REVISION), CARE checklist lines, HIPAA warnings (if any), word count, and reference count.

2. validate_trial_report

Purpose: Validate a Clinical Study Report (CSR) against ICH-E3 structure.

What it does:

• Scans the document for ICH-E3 sections: Title Page, Synopsis, Table of Contents, Abbreviations, Ethics, Investigators and Study Administrative Structure, Introduction, Study Objectives, Study Patients, Efficacy Evaluation, Safety Evaluation, Discussion and Overall Conclusions, Tables/Figures, References, Appendices.
• Reports each section as OK or MISSING and computes a compliance percentage (e.g. 90%+ = PASS).

Output: Compliance percentage, status (PASS / NEEDS_REVISION), and a section checklist.

3. check_deidentification

Purpose: Scan a clinical report for all 18 HIPAA identifiers and age compliance.

What it does:

• 18 HIPAA identifiers — Uses pattern-based checks for: names, geographic data smaller than state, dates (except year), telephone, fax, email, SSN, MRN, health plan numbers, account numbers, certificate/license numbers, vehicle identifiers, device identifiers, URLs, IP addresses, biometric identifiers, full-face photographs, other unique identifying characteristics. Reports severity (CRITICAL, HIGH, MEDIUM) and example matches (up to 5 per type).
• Age compliance — Detects ages over 89; HIPAA requires aggregation to “90 or older”.

Output: Overall status (COMPLIANT / NEEDS_REVIEW / NON_COMPLIANT), number of violation types and instances, and for each type: description, severity, count, and example snippets. Also reports age compliance (e.g. “ages >89 must be aggregated to 90 or older”).

4. validate_terminology

Purpose: Check for prohibited abbreviations, ambiguous abbreviations, and ICD-10 codes.

What it does:

• JCAHO Do Not Use — Looks for: U (→ Unit), IU (→ International Unit), QD (→ daily), QOD (→ every other day), MS, MSO4, MgSO4 (with suggested spell-outs) and reports counts and replacements.
• Ambiguous abbreviations — Flags: cc, hs, TIW, SC, SQ, D/C, AS, AD, AU, OS, OD, OU and suggests spelling out.
• ICD-10 — Detects ICD-10-style codes (e.g. A12.34) and lists them (up to 10).

Output: Lines for Do Not Use violations (with replacement), ambiguous abbreviations (with counts), and ICD-10 codes found.

5. check_compliance

Purpose: Check for regulatory compliance indicators (HIPAA, GCP, FDA).

What it does:

• HIPAA — Looks for: informed/written consent obtained; de-identification or anonymisation mentioned.
• GCP — Looks for: IRB/IEC/ethics committee approval; protocol; informed consent.
• FDA — Looks for: IND/IDE/protocol number; adverse event or SAE reporting.

Output: Per regulation (HIPAA, GCP, FDA), reports OK or MISSING for each check. Does not interpret legal compliance; only reports presence of these markers.

6. extract_clinical_data

Purpose: Extract structured clinical data from a report (demographics, vital signs, medications).

What it does:

• Demographics — Age (e.g. “X year old”), sex (male/female).
• Vital signs — Temperature, BP, HR, RR, SpO2 from common patterns (e.g. “BP: 120/80”, “HR: 72”).
• Medications — Drug, dose, route (PO/IV/SC), frequency from a simple pattern (e.g. “Drug 10 mg PO daily”).

Output: JSON with demographics, vital_signs, and medications. Useful for consistency checks or populating tables.

7. format_adverse_events

Purpose: Convert CSV adverse event data into a markdown summary table by treatment arm.

What it does:

• Expects a CSV with columns: treatment_arm, any_ae, related, serious, fatal, discontinuation.
• Aggregates per arm: total N, any AE, treatment-related AE, serious AE, deaths, discontinuation due to AE (counts and percentages).
• Produces a markdown table suitable for inclusion in a clinical trial report.

Output: Markdown table with one column per treatment arm and rows for each category.

When to run which check: After drafting a case report, run validate_case_report and check_deidentification (and fix issues before submission). For CSRs, run validate_trial_report. Before finalising any clinical document, consider validate_terminology and check_compliance. Use extract_clinical_data when you need structured data from a note or report, and format_adverse_events when you have AE data in CSV form.

Summary

Use the Clinical Report Writing skill when you need to draft or validate clinical documentation and want the agent to follow CARE, ICH-E3, and HIPAA practices with built-in templates, references, and validation tools in one place.