medical-meta-analysis-writing

name	medical-meta-analysis-writing
description	Write medical systematic reviews and meta-analyses using a strict stepwise workflow aligned with PRISMA 2020, Cochrane, MOOSE, GRADE, and JAMA-style manuscript expectations. Use this skill when planning, screening, extracting, synthesizing, grading evidence, and drafting a publication-grade medical meta-analysis.

Medical Meta-analysis Writing

When to Use

Use this skill when the task is to plan, execute, write, revise, or audit a medical or clinical systematic review with or without quantitative meta-analysis. This skill is designed for publication-grade work rather than generic literature summaries.

Typical triggers:

• Draft a medical SCI meta-analysis
• Plan a systematic review and meta-analysis protocol
• Build PRISMA-compliant screening and exclusion logic
• Extract study data and assess risk of bias
• Choose effect measures and synthesis structure
• Write a JAMA-style title, Key Points, abstract, and manuscript
• Stress-test a meta-analysis before submission

Default Policy

Apply these defaults unless the user explicitly overrides them:

• English-language full-text literature only
• Exclude grey literature, conference abstracts, dissertations, protocols, and preprints
• Default journal style: JAMA-like medical manuscript structure and tone
• Default review type: pairwise medical or clinical systematic review and meta-analysis
• Default interpretation style: conservative, evidence-proportionate, non-promotional
• Observational evidence should not be framed causally unless clearly justified

Workflow Gatekeeping Rules

Do not skip these gates:

• Do not proceed to statistical synthesis or manuscript drafting before protocol-level decisions are stable. Load prompt-module-1-protocol.md.
• Do not produce PRISMA narrative text until full-text screening and exclusion logging are complete. Load prompt-module-2-search-screening.md.
• Inclusion in the systematic review does not automatically imply inclusion in quantitative synthesis. Load prompt-module-3-extraction-bias.md.
• Do not start pooled synthesis until candidate synthesis sets and pooling eligibility are defined. Load prompt-module-4-statistical-synthesis.md.
• Do not write strong conclusions before certainty of evidence is assessed. Load prompt-module-5-grade-displays.md.
• If quantitative synthesis is not justified or not performed, downgrade the article type and front matter from meta-analysis to systematic review.

Workflow Map

Follow the workflow in order:

1. Topic feasibility, question framing, and protocol drafting
2. Search strategy, title/abstract screening, and full-text exclusion logging
3. Data extraction, risk-of-bias assessment, and pooling eligibility decisions
4. Effect measure selection, primary synthesis, heterogeneity, subgroup, sensitivity, and publication-bias analysis
5. GRADE certainty assessment, Summary of Findings logic, and figure/table package design
6. JAMA-style front matter and full manuscript drafting
7. Submission audit, reviewer simulation, and cover letter preparation

For the full dependency map, load workflow-overview.md.

Which Reference File to Load

Load only the files needed for the current stage:

• Methodological hard rules: methodology-core.md
• Workflow dependencies and stage inputs/outputs: workflow-overview.md
• Topic framing and protocol prompts: prompt-module-1-protocol.md
• Search, screening, and PRISMA prompts: prompt-module-2-search-screening.md
• Extraction, bias, and pooling-eligibility prompts: prompt-module-3-extraction-bias.md
• Statistical synthesis prompts: prompt-module-4-statistical-synthesis.md
• GRADE, Summary of Findings, and display prompts: prompt-module-5-grade-displays.md
• Title, abstract, IMRAD, and conclusion prompts: prompt-module-6-manuscript-writing.md
• Submission audit, reviewer simulation, and cover letter prompts: prompt-module-7-submission-review.md
• JAMA-facing style constraints: journal-style-jama.md

Output Standards

Unless the user asks otherwise:

• Final manuscript prose should be in English
• Working explanations and methodological reasoning may be in Chinese
• Report effect estimates with uncertainty, not p-values alone
• Distinguish pre-specified analyses from exploratory analyses
• Distinguish narrative-only inclusion from pooling-ready inclusion
• Keep claims aligned with risk of bias, heterogeneity, and GRADE certainty
• Prefer honest downgrading over overstated synthesis