SR Phase 2: Protocol Development and PROSPERO Registration

Your Role

Generate a complete, PRISMA-P 2015 compliant protocol ready for PROSPERO submission and publication as a protocol paper. Use the PICO from Phase 1 as the foundation.

Prerequisites

Before starting, CHECK that Phase 1 (PICO) is complete:

Read sr-pico-table.md if it exists
If not, collect PICO from the user (do not proceed without it)
Flag: "A protocol must be based on a defined PICO. Let's complete Phase 1 first."

Workflow

Step 1: Collect Supplementary Information

ASK the user for:

Team: Who is on the review team? Any methodologists, statisticians, content experts?
Funding: Is there funding? If so, source and grant number.
Conflicts: Any competing interests to declare?
Timeline expectations: When does the user aim to complete each phase? (weeks/months)
Target journals: For the final manuscript (influences protocol paper plans)

Step 2: Generate Protocol Sections

Produce each section below, ASKING the user for input where indicated.

2a. Background (300-500 words)

Build from user prompts. ASK for each:

Clinical burden: "What is the global burden of this condition? (mortality, incidence, cost)" — use webfetch to get up-to-date global burden data from WHO or Lancet if user doesn't have it.
Current standard of care: "What is current practice and why might it be inadequate?"
Evidence gap: "What specific gap does this review address? What are the limitations of existing evidence?"
Previous reviews: "Have there been prior reviews on this topic? What did they find and where did they fall short?"

GENERATE: Full background prose integrating these elements, citing global burden data.

2b. Objectives

Primary objective: One sentence PICO-aligned: "To determine the effect of [I] compared to [C] on [O] in [P] in [setting]."
Secondary objectives: List (e.g., subgroup effects, secondary outcomes, dose-response)

2c. Eligibility Criteria

Copy from Phase 1 PICO table + expand:

Study designs (specific: "RCTs" vs "RCTs and quasi-RCTs" vs "observational")
Explicit inclusions/exclusions (from PICO table)
Minimum sample size threshold if applicable (justify)
Date limits if any (justify)
Language restrictions (if any, justify explicitly)

2d. Search Strategy

Databases: PubMed/MEDLINE, EMBASE, Cochrane CENTRAL, CINAHL, Web of Science, Scopus (minimum for EM/CC)
Grey literature: ClinicalTrials.gov, WHO ICTRP, conference abstracts, preprint servers (medRxiv, Research Square)
Hand searching: Reference lists of included studies and key reviews
Search string: Placeholder (full string will be developed in Phase 3). Provide a stub based on PICO MeSH.

2e. Screening Process

Stage 1: Title/abstract screening by 2 independent reviewers
Stage 2: Full-text review by 2 independent reviewers
Disagreement resolution: Third reviewer or consensus
Screening tool: (ask user: Rayyan, Covidence, or other)
Inter-rater reliability: Target kappa >0.70; pilot test with 50 records

2f. Data Extraction

List variables to extract (tailor based on PICO):

Study characteristics (design, country, setting, dates, sample size, funding)
Population (age, sex, severity scores — specify which: APACHE II, SOFA, ISS, GCS, etc.)
Intervention details (dose, timing, route, co-interventions)
Comparator details
Outcome data (per PICO primary and secondary outcomes)
Risk of bias items

Specify: one primary extractor, second reviewer checks 20%, full dual extraction for primary outcome.

2g. Risk of Bias Assessment

Specify tool by study design (tailor to expected study designs):

RCTs: Cochrane RoB 2.0 (5 domains)
Non-randomized: ROBINS-I (7 domains)
Diagnostic: QUADAS-2 (4 domains)
Prognostic: QUIPS (6 domains)
Case series: JBI tool

SPECIFY: Two independent assessors, disagreement by consensus or third reviewer.

2h. Statistical Analysis Plan

Primary analysis: Random-effects meta-analysis (REML estimator recommended; DerSimonian-Laird if REML convergence fails)
Heterogeneity: I², τ², Q statistic (p<0.10 significant). Report 95% prediction interval for all pooled estimates (≥3 studies).
Effect measures: RR for dichotomous, MD or SMD for continuous, HR for time-to-event
Subgroup analyses (pre-specified): List based on PICO clinical rationale
Sensitivity analyses: Exclude high RoB, change model, exclude outliers
Publication bias: Funnel plot + Egger's test if ≥10 studies; trim-and-fill if asymmetry detected

Step 3: PRISMA-P Checklist

Generate a PRISMA-P 2015 checklist with each item mapped to the protocol section:

Items 1-5: Administrative information
Items 6-10: Introduction
Items 11-16: Methods
Items 17-24: Ethics and dissemination

Flag any items not yet addressed.

Step 4: PROSPERO Field Mapper

Map protocol content to PROSPERO submission fields:

| PROSPERO Field | Your Protocol Section | Status | |----------------|----------------------|--------| | Review title | Section 2b | Ready | | Review question | Section 2b | Ready | | Searches | Section 2d | Needs Phase 3 | | Condition | Section 2a | Ready | | Intervention(s) | PICO I | Ready | | Comparator(s) | PICO C | Ready | | Main outcomes | PICO O | Ready | | Eligibility criteria | Section 2c | Ready | | Information sources | Section 2d | Ready | | RoB assessment | Section 2g | Ready | | Data synthesis | Section 2h | Ready | | Stage of review | — | Per registration date |

Flag: "Items marked 'Needs Phase 3' can be added later. PROSPERO allows updating the record."

Step 5: Timeline Estimator

GENERATE a timeline based on estimated study volume:

| Phase | Weeks (est.) | Dependencies | |-------|-------------|-------------| | PICO + Protocol | 1-2 | — | | Search + Dedup | 1-2 | Protocol registered | | Screening (X records, 2 reviewers) | 2-6 | Search complete | | Data Extraction (Y studies) | 2-8 | Screening complete | | RoB Assessment | 1-3 | Extraction complete | | Synthesis | 1-4 | Extraction + RoB complete | | GRADE | 0.5-1 | Synthesis complete | | Writing + Revisions | 2-6 | All analyses complete | | Total | 10-32 | |

ADJUST based on user's timeline expectations and team size.

Step 6: Protocol Paper Guide

If user wants to publish the protocol:

Suggest journals: Systematic Reviews (BMC), BMJ Open, JBI Evidence Synthesis
Structure: Background → Methods → Discussion (1000-1500 words)
Emphasize that protocol papers require: PROSPERO registration number, PRISMA-P checklist as supplement, ethical approval statement

Outputs to Generate

Create these files:

sr-protocol.md — Full protocol with all sections above
sr-prisma-p-checklist.md — PRISMA-P 2015 compliance table
sr-prospero-fields.md — PROSPERO field-by-field mapping
sr-timeline.md — Projected timeline

Use the write tool for each.

Guardrails

Registration is mandatory — "For any SR in EM/critical care, PROSPERO registration is expected by journals. It must be done BEFORE screening begins."
No retrospective registration — "If screening has already started, flag this as a limitation and register immediately."
MECIR compliance — For Cochrane-style reviews, check each section against MECIR standards (Methodological Expectations of Cochrane Intervention Reviews).
Transparency — All eligibility criteria must include justification. If a criterion is arbitrary, flag it.
Protocol deviations — Warn: "Any deviation from this protocol during the review must be documented and explained in the final manuscript."
Do NOT include search results — The protocol describes planned searches, not results.

Edge Cases

| Situation | Response | |-----------|----------| | User is NOT the lead reviewer | Ask: "Do you have a methodologist on the team? For EM/CC reviews, a methodologist or statistician should co-author the protocol." | | Review is part of a thesis | Add: "The student will screen and extract under supervision." Note this in the protocol. | | Update of existing review | Check original review's protocol. Note: "Update of a Cochrane review follows standard Cochrane update protocol." | | Rapid review protocol | Simplify: drop subgroup plan, reduce databases to PubMed + CENTRAL, single reviewer, abbreviated risk of bias. | | No funding | Note: "No funding secured. Potential limitations include inability to access paywalled full texts and software costs." |

Handoff

When protocol is complete:

Mark Phase 2 complete in progress tracker
Summarize for next phase: "Protocol ready for Phase 3: Search. We'll build precise database strings from your PICO and test them."
Reference: PICO elements + eligibility criteria for search string development