← allmeta

Workflow · Cochrane Rapid Reviews

Rapid review

A streamlined evidence synthesis for decisions on a deadline. Rapid reviews are explicit, transparent, and reproducible — they trade comprehensiveness for speed by limiting the search, using single-reviewer screening with a verification subset, and standardising the synthesis. Defaults below follow the Cochrane Rapid Reviews Methods Group recommendations (Garritty et al. 2021).

  1. Refine the question with the requester

    Rapid reviews start with a tight, decision-focused question. Negotiate which PICO elements are essential and which can be relaxed (e.g. broaden the population, narrow the comparator). Document scope decisions and the deadline up front.

    Garritty 2021 question-formulation block. The single biggest determinant of feasibility is question scope.

    PICO Formulator → Learn the theory: Rapid Reviews course

  2. Limit the search — but document it

    Search MEDLINE/Embase plus one specialised register. Apply date and language restrictions where defensible. Skip grey literature unless the topic demands it. Record every limit.

    Garritty 2021 search-strategy block. MEDLINE plus one additional database (typically Embase) is the rapid-review floor; trial registers and grey literature are encouraged when the topic warrants. Document every restriction so reviewers can judge bias.

    Search Translator → Citation Chaser → Learn the theory: search strategy

  3. Single-reviewer screening with verification

    One reviewer screens titles/abstracts; a second verifies a sample of excluded records (commonly 20%, per Affengruber 2020 evidence cited within Garritty 2021). At full-text screen, two reviewers in parallel — disagreement adjudication is fast because the set is small.

    Garritty 2021 screening block. Dual screening of full text remains required even when title/abstract is single-reviewer.

    Citation Dedup → PRISMA Screening → RCT Extractor (AI assist) →

  4. Streamlined data extraction & risk of bias

    Single extractor with a second verifier on key outcomes (effect estimate, denominator, definition). Use the appropriate RoB instrument — RoB 2 for RCTs, ROBINS-I for non-randomised — but only on the primary outcome unless time permits more.

    Garritty 2021 extraction & risk-of-bias block. Verification of the effect-estimate row is non-negotiable; the rest is risk-tolerated.

    RoB 2 → ROBINS-I → RoB Traffic Light →

  5. Synthesise pragmatically

    If the studies are clinically homogeneous, run a meta-analysis with REML τ² and the HKSJ small-sample correction (IntHout 2014; floor q*≥1 to avoid CI narrowing when Q<k−1). Otherwise, narrative synthesis with vote-counting by direction of effect — clearly labelled as such.

    Heterogeneity discovery is fast and cheap; pooling decisions follow once you can see the spread.

    Forest Plot Viewer → Heterogeneity Explorer → Meta-Regression →

  6. GRADE the certainty & report transparently

    GRADE the primary outcomes only, but use the full 5-domain reasoning. Render PRISMA-2020 flow. State explicitly which streamlining choices were made — that's what makes a rapid review reproducible.

    Garritty 2021 reporting block: report exactly what was abbreviated. The methods section IS the credibility story.

    GRADE SoF → PRISMA Flow → PRISMA Checklist → Learn the theory: GRADE Certainty course

Defaults follow Garritty C, et al. J Clin Epidemiol 2021;130:13-22 (Cochrane Rapid Reviews Methods Group). Need a full systematic review instead? See systematic review. Need a prognostic-factor review? See prognostic review. Browse the full 26-course collection.