rapid-evidence-summary-method

Writing a Rapid Evidence Summary for Decision-Makers

Policy teams, hospital committees and research funders often need to know what the evidence says about a question within days, not months. A full systematic review is impossible on that timescale, but an unstructured literature trawl produces something worse: a confident-sounding summary that nobody can check. The rapid evidence summary sits between these extremes. It uses explicit, simplified methods to answer a focused question quickly, and it is honest about what the shortcuts might have missed. This guide sets out a practical method, with notes on how an evidence-traceable research workbench such as Yila can make each claim in the summary checkable.

Agree the question and the deadline

Rapid summaries fail most often because the question was never pinned down. Before searching, agree with the decision-maker on four points: the population or setting, the intervention or issue, the outcomes that matter for the decision, and the date by which the answer is needed. Write these down in two or three sentences and get explicit agreement. A question such as "Does extending clinic opening hours reduce emergency department attendance for adults in urban areas?" is answerable in a week. "What works for emergency department pressures?" is not.

Also agree on what kind of answer is useful. Some decision-makers need a yes-or-no judgement with a confidence rating; others need a short list of options with evidence for each.

Use a simplified but explicit search

A rapid summary does not need to search every database, but it should state what it did search. A sensible default is: one or two major bibliographic databases, a search for existing systematic reviews, and a targeted search of key organisations' reports. Record the search terms, the databases and the date. Limit by date or language if necessary, and say so.

Prioritise existing reviews. If a recent, good-quality systematic review answers the question, the summary's job may be mainly to report and update it.

Screen with one reviewer, check a sample

Full dual screening is usually impossible on a rapid timescale. A common compromise is single-reviewer screening with a second reviewer checking a random sample of decisions, typically ten to twenty percent. Record the agreement rate. If disagreement is high, widen the check or tighten the criteria.

Keep a short log of exclusions with reasons. Even in a rapid summary, someone may later ask why a particular study was left out.

Extract only what the decision needs

Rapid summaries save time by extracting less. Agree in advance which data matter for the decision, typically design, setting, population, intervention, outcome, effect size and a brief quality note, and extract only those.

Every extracted item should carry its source location. This is where an evidence-traceable assistant speeds the work without sacrificing checkability. Uploading each included paper to Yila and asking "What was the effect on emergency attendances, in which population, and where is this reported?" returns an answer linked to the exact passage or table. The analyst confirms it, records it, and moves on. When the decision-maker or a reviewer later asks where a number came from, the answer is one click away.

Appraise proportionately

Quality appraisal in a rapid summary is usually lighter than in a full review, but it should not be absent. Use a short checklist appropriate to the main designs, and record a one-line judgement for each study: for example, "randomised, adequate concealment, high attrition in control arm". The aim is to let the reader see which findings rest on stronger designs.

Synthesise in plain language

Decision-makers read the first page and sometimes nothing else. Put the answer there. A useful structure is:

• Bottom line: one or two sentences answering the question, with a confidence level.
• What the evidence shows: three to five bullet points, each tied to the studies that support it.
• What we do not know: important gaps or uncertainties.
• How this summary was done: two or three sentences on methods and shortcuts.

Use numbers where they help, and avoid jargon. "Two trials in comparable cities found attendance fell by about eight percent" is more useful than "evidence suggests a modest effect".

Be explicit about shortcuts

The credibility of a rapid summary depends on honesty about its limits. State which shortcuts were taken: limited databases, single screening, date restrictions, lighter appraisal. Where possible, say how these might affect the conclusion. If the evidence base is small and the shortcuts could plausibly have missed important studies, say so plainly.

Make every claim traceable

A rapid summary will often be read by people who did not commission it and who may disagree with its conclusion. The best defence is traceability. Every bullet point in the synthesis should cite specific studies, and every number should be traceable to a location in a source. Keeping located evidence from the extraction stage makes this straightforward. A summary whose claims link back to passages is far more persuasive in a contested meeting than one that asks to be taken on trust.

A realistic five-day plan

• Day 1: agree the question, design the search, run it.
• Day 2: screen titles and abstracts; retrieve full texts.
• Day 3: full-text screening and extraction with located evidence.
• Day 4: appraisal and synthesis; draft the summary.
• Day 5: internal check of claims against sources; revise and deliver.

The internal check on day five matters. Have someone other than the main analyst pick ten claims at random and confirm them against their cited locations. With located evidence, this takes under an hour.

Updating the summary

Decisions often come back. Keep the search strategy, the screening log and the extraction table so the summary can be updated quickly when new evidence appears or the decision is revisited. If the source documents and their located extractions are stored together in a workbench, an update means adding new studies, not starting again.

Summary

A rapid evidence summary answers a focused question within days using explicit, simplified methods: an agreed question, a recorded search, single screening with a checked sample, targeted extraction with source locations, proportionate appraisal, a plain-language synthesis and honest reporting of shortcuts. Traceability is what separates it from an opinion piece. An evidence-traceable research workbench like Yila helps by tying every extracted number to the passage it came from, so a fast answer can still be a checkable one.