SR Phase 9: Manuscript Writing and PRISMA 2020 Reporting

Your Role

Produce a complete, publication-ready SR manuscript following PRISMA 2020 standards, tailored to EM and critical care journals. Generate supplementary materials including the PRISMA checklist, cover letter, and response to reviewers.

Prerequisites

All prior phase outputs (PICO, protocol, search strings, screening flow, extraction data, RoB, synthesis results, GRADE assessments)
If any phase is missing data, flag: "Missing data from Phase [X]. This section will be incomplete without it."

Workflow

Step 1: PRISMA 2020 Checklist Auto-Population

GENERATE the PRISMA 2020 checklist (27 items) pre-populated from prior phases:

| Section | Item | Description | Location in Manuscript | Status |
|---------|------|-------------|----------------------|--------|
| Title | 1 | Identify as a systematic review | Title page | ✓ Auto-filled |
| Abstract | 2 | Structured summary | Abstract | ✓ Auto-filled |
| Introduction | 3 | Rationale | Intro ¶1-2 | Needs writing |
| Introduction | 4 | Objectives / PICO | Intro last ¶ | ✓ From Phase 1 |
| Methods | 5 | Eligibility criteria | Methods | ✓ From Phase 1/2 |
| Methods | 6 | Information sources | Methods | ✓ From Phase 3 |
| Methods | 7 | Search strategy | Methods / Appendix | ✓ From Phase 3 |
| Methods | 8 | Selection process | Methods | ✓ From Phase 4 |
| Methods | 9 | Data extraction | Methods | ✓ From Phase 5 |
| Methods | 10 | Risk of bias | Methods | ✓ From Phase 6 |
| Methods | 11 | Effect measures | Methods | ✓ From Phase 7 |
| Methods | 12 | Synthesis methods | Methods | ✓ From Phase 7 |
| Methods | 13 | Reporting bias | Methods | ✓ From Phase 7 |
| Methods | 14 | Certainty assessment | Methods | ✓ From Phase 8 |
| Results | 15 | Study selection | Results ¶1 | ✓ From Phase 4 |
| Results | 16 | Study characteristics | Results + Table 1 | Needs writing |
| Results | 17 | Risk of bias | Results | ✓ From Phase 6 |
| Results | 18 | Results of individual studies | Results | Needs writing |
| Results | 19 | Synthesized results | Results | ✓ From Phase 7 |
| Results | 20 | Reporting bias | Results | ✓ From Phase 7 |
| Results | 21 | Certainty of evidence | Results | ✓ From Phase 8 |
| Discussion | 22 | Interpretation | Discussion ¶1 | Needs writing |
| Discussion | 23 | Limitations | Discussion ¶5 | Needs writing |
| Discussion | 24 | Conclusions | Conclusion | Needs writing |
| Other | 25 | Registration & protocol | Methods | ✓ From Phase 2 |
| Other | 26 | Funding | Declarations | Needs user input |
| Other | 27 | Conflicts of interest | Declarations | Needs user input |

FLAG any missing items with "⚠️ Needs data."

Step 2: Generate Manuscript Sections

Work through each section, generating text from accumulated data:

Title Page

Generate 3-5 options combining descriptive title + PICO + "systematic review and meta-analysis":

Option 1: "[Intervention] for [Population]: A Systematic Review and Meta-Analysis"
Option 2: "[Intervention] Compared to [Comparator] for [Outcome] in [Population]: A Systematic Review"
Option 3: "[Outcome] in [Population] Treated with [Intervention]: A Systematic Review and Meta-Analysis"

Include: author list placeholder, affiliations, corresponding author, word count, keywords (MeSH terms from Phase 1).

Structured Abstract (250-350 words)

Background: [2-3 sentences on clinical problem]
Objective: [PICO-aligned one sentence]
Methods: [Databases, dates, study designs, primary outcome, synthesis method]
Results: [Studies included, sample size, main pooled estimate with 95% CI, I²]
Conclusion: [One sentence on what the evidence shows and its clinical implication]
Registration: [PROSPERO ID]

Introduction (400-600 words)

¶1: Clinical burden (incidence, mortality, cost) — cite global figures
¶2: Current standard of care and its limitations
¶3: The controversy or evidence gap the review addresses
¶4: Previous reviews and their limitations
¶5: Aim (research question from Phase 1)

Methods (800-1200 words)

Follow your protocol. Generate sections with bullet points from prior phases:

Eligibility criteria: From Phase 1 PICO Information sources: From Phase 3 search log Search strategy: "Full search strings are provided in the supplementary appendix." Selection process: "Two independent reviewers screened... with kappa >0.70." Data extraction: "One reviewer extracted... second reviewer verified 20%." Risk of bias: "[Tool name] was used... disagreements resolved by consensus." Synthesis methods: "[Model], [software], [heterogeneity measures]." Publication bias: "Funnel plot + Egger's test if ≥10 studies." Certainty of evidence: "GRADE was used to rate certainty for each outcome."

Results (1000-1500 words)

¶1: Search results (from Phase 4 PRISMA flow) "The search retrieved X records. After deduplication, X records were screened. Following full-text review of X articles, X studies met inclusion criteria (Figure 1)."

Table 1: Study characteristics (from Phase 5 extraction) Generate table summarizing study design, population, setting, intervention, comparator, outcomes.

¶2: Risk of bias (from Phase 6) "Of X studies, Y were low risk, Z had some concerns, and W were high risk (Figure 2)."

¶3-N: Outcome results (from Phase 7, one paragraph per outcome) "Pooled analysis of X studies (n=X) showed [outcome] was [higher/lower] in the intervention group (RR X.XX, 95% CI X.XX-X.XX; I²=X%; τ²=X.XX). The 95% prediction interval ranged from X to X, suggesting [interpretation]."

Subgroup analyses: "Subgroup analysis by [variable] showed [results] (p-interaction = X.XX)."

Sensitivity analyses: "Results were robust to exclusion of high-risk studies [and/or] the use of fixed-effect models."

Publication bias: From Phase 7 results.

GRADE results: From Phase 8 SoF table. Can be Table 2 or inline.

Discussion (600-900 words)

¶1: Summary of main finding: Do NOT repeat numbers. Synthesize the clinical meaning.
¶2: Comparison with existing reviews: "Previous reviews found [X]. Our review adds [Y] by including [Z studies/meta-analysis]."
¶3: Biological/clinical plausibility: Why does this effect make sense?
¶4: Heterogeneity explanation: What clinical/methodological factors might explain variation?
¶5: Limitations: Be direct and specific:
- "The main limitation is the inclusion of observational studies with residual confounding."
- "Heterogeneity in outcome definitions limits precision."
- "Short follow-up in most ED-based studies."
- "Publication bias could not be assessed (fewer than 10 studies)."
¶6: Clinical and research implications:
- Clinical: "Current evidence supports [practice implication]."
- Research: "Future trials should [specific recommendation: longer follow-up, standardized outcomes, multicenter]."

Conclusion (50-100 words)

One paragraph: what the evidence shows, certainty level, one specific research recommendation. Do NOT overstate.

Example: "Moderate certainty evidence suggests that [intervention] reduces [outcome] in [population]. Further research is needed to determine [specific gap]."

Step 3: Journal-Specific Formatting

ASK the user: "Which journal are you targeting?"

Adjust formatting for:

| Journal | Key Requirements | |---------|-----------------| | Annals of Emergency Medicine | ~3,000 word limit, structured abstract, 30 references max, specific heading format | | Intensive Care Medicine | ~3,500 words, structured abstract, 50 references, PRISMA + SoF table required | | Critical Care Medicine | ~4,000 words, structured abstract, no limit on references, specific line spacing | | Resuscitation | ~3,500 words, structured abstract, 40 references, Specific to cardiac arrest/resuscitation | | Emergency Medicine Journal | ~4,000 words, structured abstract, 40 references | | Systematic Reviews (BMC) | No word limit, PRISMA checklist submission required, protocol citation mandatory | | BMJ Open | ~4,000 words, structured abstract, 60 references, patient and public involvement | | PLOS ONE | No word limit, structured abstract, PRISMA checklist required, data availability mandatory |

GENERATE: "Based on your target journal [name], this manuscript should be adjusted to [specific requirements]."

Step 4: Generate Supplementary Materials

PRISMA 2020 Checklist

Generate the full 27-item checklist filled in from accumulated data as a separate file (required by most journals).

PRISMA 2020 Flow Diagram

Generate from Phase 4 PRISMA flow data as a text-based diagram or table.

Full Search Strings

Generate appendix with all database strings from Phase 3.

Data Extraction Form

Include as appendix from Phase 5.

GRADE SoF Table

Include from Phase 8.

Step 5: Generate Cover Letter

Dear Editor,

We are submitting our manuscript entitled "[Title]" for consideration in [Journal].

This systematic review and meta-analysis [summarize main finding in 1-2 sentences].

Key strengths:
- PRISMA 2020 compliant
- Pre-registered on PROSPERO (ID: [ID])
- Comprehensive search of X databases
- GRADE assessment of all outcomes
- [Other strengths]

We believe this manuscript will be of interest to your readers because [relevance].

All authors have approved the submission and declare no conflicts of interest.

Sincerely,
[Author Name]
[Affiliation]
[Contact]

Step 6: Generate Response to Reviewers Template

## Response to Reviewers

We thank the reviewers for their constructive feedback. We have addressed each comment as follows:

### Reviewer 1

**Comment 1:** [paste comment]
**Response:** [template]
**Changes:** [Location in manuscript]

**Comment 2:** [paste comment]
**Response:** [template]
**Changes:** [Location in manuscript]

...

Include common SR review concerns pre-populated:

"We have now included a PRISMA 2020 checklist as supplementary material."
"We have added the full search strings as an appendix."
"We have discussed the impact of heterogeneity on our findings."
"We have added a limitations section addressing [specific issue]."

Scripts

`scripts/prisma_checklist.py`

#!/usr/bin/env python3
"""Generate PRISMA 2020 checklist from accumulated phase data."""

prisma_2020 = [
    ("1", "Title", "Identify the report as a systematic review.", "Title page"),
    ("2", "Abstract", "See PRISMA 2020 for Abstracts checklist.", "Abstract"),
    ("3", "Rationale", "Describe the rationale for the review.", "Introduction"),
    ("4", "Objectives", "Provide an explicit statement of objectives.", "Introduction"),
    ("5", "Eligibility criteria", "Specify inclusion/exclusion criteria.", "Methods"),
    ("6", "Information sources", "List databases, registers, dates.", "Methods"),
    ("7", "Search strategy", "Present full search string for ≥1 database.", "Methods/Appendix"),
    ("8", "Selection process", "Describe screening process.", "Methods"),
    ("9", "Data extraction", "Describe extraction process and forms.", "Methods"),
    ("10", "Risk of bias", "Describe RoB tools used.", "Methods"),
    ("11", "Effect measures", "Specify effect measures.", "Methods"),
    ("12", "Synthesis methods", "Describe synthesis methods.", "Methods"),
    ("13", "Reporting bias", "Describe reporting bias assessment.", "Methods"),
    ("14", "Certainty assessment", "Describe GRADE or other certainty method.", "Methods"),
    ("15", "Study selection", "Report number of studies screened.", "Results"),
    ("16", "Study characteristics", "Present study characteristics.", "Results"),
    ("17", "Risk of bias", "Present RoB assessment results.", "Results"),
    ("18", "Results of individual studies", "Present results per study.", "Results"),
    ("19", "Synthesized results", "Present results of synthesis.", "Results"),
    ("20", "Reporting bias", "Present reporting bias assessment.", "Results"),
    ("21", "Certainty of evidence", "Present GRADE results.", "Results"),
    ("22", "Discussion", "Interpret results in context.", "Discussion"),
    ("23", "Limitations", "Discuss limitations at study and review level.", "Discussion"),
    ("24", "Conclusion", "Provide general interpretation.", "Conclusion"),
    ("25", "Registration", "Provide registration details.", "Methods"),
    ("26", "Funding", "Report funding sources.", "Declarations"),
    ("27", "Conflicts of interest", "Declare competing interests.", "Declarations"),
]

def generate_checklist(filled_items):
    for num, section, desc, location in prisma_2020:
        status = filled_items.get(num, "Not yet completed")
        print(f"| {num} | {section} | {desc} | {location} | {status} |")

Outputs to Generate

sr-manuscript.md — Full manuscript
sr-abstract.md — Abstract variants (structured + plain language)
sr-prisma-2020-checklist.md — Completed checklist
sr-cover-letter.md — Journal-specific cover letter
sr-response-to-reviewers.md — Template for revisions
sr-supplementary-appendix.md — Search strings, extraction form, SoF table

Guardrails

Do NOT overstate conclusions — "Our findings suggest" not "Our findings prove." Avoid causal language for observational studies.
PRISMA checklist REQUIRED — Almost all journals now require it at submission.
Report all pre-specified analyses — Even those with no data: "[Subgroup] could not be assessed due to insufficient data."
Prediction interval mandatory — Report for every pooled estimate with ≥3 studies.
Data availability statement — Journals increasingly require data sharing. "All data extracted are available in the supplementary materials."
Do NOT hide limitations — Be direct. Reviewers will find them anyway. Acknowledged limitations show methodological rigor.

Edge Cases

| Situation | Response | |-----------|----------| | User is not a native English speaker | Recommend: write, then use the humanizer skill for language polishing. Suggest journal editing services. | | Review found no eligible studies | "This systematic review found no studies meeting the eligibility criteria, highlighting a critical evidence gap." — Still publishable in some journals. | | Target journal has specific PRISMA extension | Check: PRISMA-DTA (diagnostic), PRISMA-S (search), PRISMA-P (protocol), PRISMA-ScR (scoping). Use the correct extension. | | Word count exceeds journal limit | Flag: "This section is [N] words over. Suggest cutting [specific paragraphs]." | | Figures needed at specific resolution | Recommend: "Export forest plots at 300 DPI TIFF for [journal] submission." |

Handoff

When writing is complete:

Pass: completed manuscript + PRISMA checklist + supplementary materials
Summarize: "Manuscript complete and formatted for [journal]. Total word count: [N]. PRISMA checklist: 27/27 items addressed."
Reference: "The next phase is Phase 10: Update & Maintenance — for keeping your review current after publication."