Rapid Review Methodologies

A rapid review is a structured review that simplifies or leaves out named steps to deliver evidence faster, and states each cut and the bias it may add. It is not a full systematic review done in a hurry. You keep the protocol, risk of bias, and a known-study search check. You may narrow the scope, search fewer databases, and single-screen after a dual-screened agreement check. Cochrane’s Rapid Reviews Methods Group describes a rapid review as speeding up how a structured review is planned, done, or shared, by simplifying or omitting methods the authors define. Those restrictions, and the bias they may introduce, have to be reported. A review that keeps full systematic methods and simply finishes quickly is still a systematic review. For Cochrane rapid reviews, the same guidance says the work should take no longer than six months, and that time alone does not define the method.

What can move, with the risk written in the protocol, is how many interventions and outcomes you keep, how many databases you search, and how far grey literature goes. Screening and extraction can narrow only after a check. What must not move is the protocol, a risk-of-bias assessment, and the test that the search still finds studies you already know are relevant. The guidance is for rapid reviews of intervention effectiveness.

The same decision, two method budgets

This is an illustrative scenario, not a completed review and not a result. The question is whether a discharge support bundle changes unplanned readmission among adults leaving hospital after heart failure. Only the method changes.

Step	Full systematic review	Rapid review change	Risk accepted
Protocol	Written first	Stays, and names the cuts	None from dropping it
Scope	May stay broad	Fewer interventions and outcomes	Cannot answer a wider question
Databases	Wider set, aimed at sensitivity	At least two likely databases; not MEDLINE alone	Studies only in a skipped source
Search check	Peer review and a known-study test	Both stay	Known studies missed if the test is dropped
Screening	Two people decide inclusion independently	A share is dual screened; one person continues only if agreement is good	More missed studies if single screening is unchecked
Outcome data	Two people extract	One extracts; a second checks critical fields	Unchecked fields can be wrong
Risk of bias	A validated tool	Kept; can be limited to decision outcomes	Unrated outcomes are not assessed
Grey literature	Often searched	Limit it, and justify sources you still search	Some topics need it more than databases

The Cochrane rapid review guidance treats the screening share as an example: at least 20% of records, then one screener if agreement is high (for example kappa 0.8 or above). The authors say 20% is experience, not evidence. Poor agreement means dual screening continues. A few hundred records may still be dual screened throughout. Pilot the forms first, for example on 50 to 100 records. That pilot is not the agreement check. Cochrane rapid reviews use RoB 1 or RoB 2 for randomised trials, and ROBINS-I for non-randomised studies of interventions. If extractors still disagree, a third person decides.

When the shortcut removes the sensitivity check

Fewer databases and a more precise strategy are allowed, and both change what can be found. The group’s search guidance says to test the primary strategy on known relevant records from the preliminary search, and to revise if a known study is missing. They suggest even a precision-focused strategy should still find most of those known studies. The figure they offer is at least 80–90%, drawn from benchmarks for sensitivity-and-precision filters, and only if the known set represents the studies you care about. It is not a result of this scenario.

Skipping that check can harm the review, especially when few sources are searched. PRESS peer review is recommended when possible. If it is not, check spelling, operators, and line numbers. The rapid-review guidance adds typos, missed keywords, and Boolean operators. A syntax check can catch a broken line. It cannot show that a tighter concept dropped a known study.

In the scenario the team cuts the databases and tightens the strategy, then skips the known-record test. Two trials from the preliminary search no longer come back, and the result does not say so. That is an illustration, not a published case.

Read the neighbouring methods

Screening strategies is the full-review default: two people, independently. The rapid-review allowance is a checked share of records, with the agreement written down. A living systematic review repeats the search so the review stays current. The protocol in best practice is a step this guidance keeps. If the decision cannot accept the risks in the table, run a systematic review, with a named-batch queue for dual screening and extraction.

Reporting what was cut

The cuts belong in the methods section and in the flow diagram, not only in the protocol. A limited grey literature search still needs a log of each source, date, and hit count; grey literature searching has an illustrative log and a stopping rule. The record counts go into the same PRISMA 2020 flow diagram a full review uses; PRISMA reporting guidelines explained maps each item to the file that fills it, and the dedupe-to-flowchart example shows where database and duplicate counts come from. When fewer databases are searched, deduplicate them anyway, using the multi-database deduplication guide.

Sources

Garritty and colleagues, Cochrane Rapid Reviews Methods Group. BMJ. 2024;384:e076335. Read the guidance. Klerings and colleagues, Cochrane Rapid Reviews Methods Group. BMJ Evidence-Based Medicine. 2023;28:412-417. Read the search guidance.