A rapid review is a form of evidence synthesis that follows the logic of a systematic review but deliberately streamlines specific steps so the whole project can be completed in weeks rather than a year or more. The streamlining is not improvisation: since 2024 the Cochrane Rapid Reviews Methods Group has published explicit recommendations on which shortcuts are acceptable, which must be verified by a second person, and which quietly wreck the review. Done well, a rapid review trades a small, declared amount of certainty for a large amount of speed, and reports that trade openly.
Why decision-makers needed a faster method
The method matured because the people who commission evidence could not wait for it. Health ministries choosing between competing policies, guideline panels facing a fixed publication date, and hospital formulary committees deciding on a new drug all need a defensible answer inside a planning cycle, and how long a conventional systematic review takes is typically twelve to twenty-four months. The pandemic made the pressure impossible to ignore: hundreds of urgent questions arrived at once, and teams worldwide discovered that an unmethodical quick summary and a full review were not the only two options. The Cochrane Rapid Reviews Methods Group responded by codifying practice, first in interim guidance and then in its updated 2024 recommendations, which now function as the closest thing the field has to a rulebook. The recommendations were built by surveying what rapid review producers actually did, testing which abbreviations changed conclusions in methodological studies, and then keeping only the shortcuts whose measured cost was small. The central idea is that speed should come from pre-agreed methodological concessions chosen with the commissioner, never from silently skipping steps and hoping nobody asks. That single principle separates a rapid review from the quick literature summaries that circulate inside organisations without methods sections attached.
The shortcuts the 2024 Cochrane recommendations endorse
The recommendations work through the review pipeline stage by stage and say, for each one, what may be abbreviated and what safeguard must be kept. The logic pairing every shortcut with a verification step is deliberate: errors of omission, a missed study or a mistyped number, are far more likely under time pressure, so the second reviewer’s reduced role concentrates on catching exactly those. Four endorsed shortcuts do most of the time-saving.
Limit the number of databases searched
Instead of the five or more sources a full review might cover, a rapid review can restrict itself to a small core, typically MEDLINE or PubMed, Embase, and the Cochrane Central Register of Controlled Trials, chosen for the topic. The guidance also allows limiting grey literature and supplementary searching to what the question genuinely needs. Knowing which databases reviews normally search is what makes this cut principled rather than lazy: the team documents what was excluded and why, so a reader can judge what might have been missed. Methodological studies comparing full and restricted searches find that a well-chosen small set of sources captures the large majority of the studies that end up driving a review’s conclusions, which is why this is the least controversial of the endorsed shortcuts.
Single-reviewer screening with verification
Full reviews screen every record in duplicate. The rapid review compromise is single-reviewer screening after a calibration exercise, with a second reviewer checking all excluded records at the title and abstract stage, or at minimum a defined percentage of them. At full text, the recommendation is stricter: one reviewer screens, and a second reviewer verifies the excluded full texts, because a wrongly discarded eligible study is the most damaging screening error a review can make.
Single data extraction, verified
Rather than two people extracting everything independently, one reviewer extracts and a second checks the extraction against the source papers, with particular attention to outcome data that will feed any synthesis. This extract-and-verify model catches most transcription errors at a fraction of the cost of full duplicate extraction. The verification step is not optional garnish: unverified single extraction is one of the few abbreviations shown to change review conclusions, because a transposed event count or a misread denominator flows straight into the pooled result.
Focused critical appraisal
Risk of bias assessment is retained, because dropping it altogether is not endorsed, but it can be focused: appraising only the outcomes that matter for the decision, using a single assessor with verification, and declining to appraise study designs that contribute little to the conclusions. The judgements still feed rating certainty with GRADE where the commissioner needs a certainty statement.
The shortcuts that are not endorsed
The recommendations are equally clear about the false economies. A rapid review should not abandon a written protocol: the question, eligibility criteria, and planned shortcuts are fixed in advance, ideally with the knowledge user, even when the document is short. It should not skip critical appraisal entirely, because an unappraised evidence summary cannot support a decision. It should not silently narrow the question after seeing the results, and it should not present itself as a full systematic review. Restricting searches to English and to recent years is permitted only when the topic justifies it and the restriction is declared. The pattern across every stage is the same: the shortcut is legitimate when it is pre-specified, applied with verification, and reported; it becomes a defect the moment it is hidden. A second pattern is worth noticing too: none of the endorsed shortcuts touches the question formulation or the eligibility criteria. The intellectual work of deciding what counts as relevant evidence is never abbreviated, only the labour of processing it.
Involve the knowledge user from the first week
One recommendation runs through the whole 2024 guidance and is easy to miss because it is not a shortcut at all: involve the knowledge user, the person or body who will act on the review, from the start and at every pivot. The commissioner helps set the question, agrees the eligibility criteria, signs off the planned abbreviations, and is consulted if the evidence base turns out larger or messier than expected. This is partly about relevance, since a rapid review that answers a slightly different question from the one the decision-maker holds is worthless at any speed. But it is also about legitimacy: when the shortcuts were agreed in advance by the people relying on the answer, the review cannot later be accused of cutting corners unilaterally. In practice this means a short kick-off meeting, a one-page protocol the commissioner countersigns, and a standing agreement about how mid-project surprises will be handled. Teams that skip this step tend to discover the question was wrong at the only point where the discovery is unaffordable, which is the end.
Rapid review, scoping review, or full systematic review
The three designs are separated by purpose, not just by effort. A rapid review answers a focused effectiveness or policy question quickly, accepting a declared risk of missing some evidence. A scoping review maps the breadth of a literature, asking what evidence exists rather than what it shows; the practicalities are covered in our guide to conducting a scoping review, and the boundary with full reviews in the systematic review and scoping review comparison. A full systematic review remains the reference standard when the stakes are high and time allows, working through the full sequence of systematic review steps with duplicate screening, duplicate extraction, and comprehensive searching. A useful test: if a wrong answer would be expensive or harmful and the deadline is negotiable, escalate to a full review; if the question is broad and descriptive, scope it; if the question is narrow and the deadline is real, a rapid review is the honest middle path. The three also differ in what they promise the reader. A full review promises completeness, a scoping review promises coverage of the landscape, and a rapid review promises a defensible answer by a date, with the residual uncertainty spelled out rather than hidden.
What a realistic timeline looks like
Published rapid reviews typically run from four weeks to six months, with eight to twelve weeks a common commission. A workable eight-week schedule allocates the first week to the protocol and search design, weeks two and three to searching and title and abstract screening, weeks four and five to full-text screening and extraction, week six to appraisal, and the final fortnight to synthesis and writing. Two things stretch the schedule more than anything else: an unfocused question that generates tens of thousands of records, and commissioners who revise the question mid-project. Both are managed at the scoping stage, which is why the first week matters more than any other. A pre-search that estimates the likely yield before the protocol is signed lets the team tighten the population, the date range, or the outcome list while tightening is still cheap. It is also why experienced teams quote a range rather than a date until the scoping search has run: promising six weeks before seeing the record count is how rapid reviews become late reviews. Teams that maintain standing search strategies and calibrated screeners, as groups producing how a Cochrane review is built will recognise, can compress the schedule further without touching the safeguards.
Reporting the review and declaring its limits
There is no separate reporting standard for rapid reviews, so the expectation is to report against the PRISMA 2020 reporting guideline and to state explicitly, item by item, where the methods were abbreviated. Synthesis follows the same rules as any review: if the included studies are clinically and methodologically similar enough, a meta-analysis is legitimate, and if they are not, the honest route is a structured narrative summary rather than forced pooling. Speed is never a reason to pool studies that should not be pooled. The limitations section carries unusual weight. It should name each shortcut taken, say what the plausible consequence is, and avoid the temptation to soften: single-reviewer screening can miss eligible studies, a two-database search can miss unpublished or non-indexed evidence, and a date restriction can miss older trials that still dominate a meta-analysis. Stating this plainly is not a weakness. It is the feature that distinguishes a rapid review from a quick, unaccountable summary, and it is what lets the reader calibrate how much weight the conclusions can bear.
When a rapid review is the wrong tool
Some questions should not be answered quickly. A rapid review is the wrong choice when the decision is effectively irreversible, such as a national screening programme or a regulatory judgement, where the cost of a missed study dwarfs the cost of waiting. It is wrong when the evidence is known to be scattered across grey literature, non-English sources, or many small studies, because those are precisely the corners the endorsed shortcuts stop you searching. It is wrong when the review is intended as the definitive statement on a contested question, where critics will probe every methodological concession. It is usually wrong when effect sizes are expected to be small and finely balanced, because the marginal studies a restricted search misses are exactly the ones that could tip a borderline conclusion. And it is wrong as a thesis substitute: examiners assess methodological competence, and a deliberately abbreviated method demonstrates less of it. In each of these cases the honest options are a full systematic review, or a rapid review explicitly framed as an interim product with an update already planned. The method rewards teams that respect its boundary: fast, focused, verified, and candid about what was traded away.