PRISMA 2020 Explained: Checklist, Flow Diagram, and How to Comply

What PRISMA 2020 Is — and What It Isn't

PRISMA — Preferred Reporting Items for Systematic Reviews and Meta-Analyses — is the reporting guideline most journals now require for systematic reviews. The current version, PRISMA 2020, was published as Page MJ et al., BMJ 2021;372:n71, and replaces the 2009 statement. Its job is narrow and specific: to ensure that when you write up a systematic review, readers can see exactly what you did, what you found, and where every number came from.

The word to hold onto is reporting. PRISMA tells you what to report, not how to conduct the review. Decisions about search design, screening procedure, risk-of-bias tools, and synthesis methods belong to methodological guidance such as the Cochrane Handbook. A review can follow PRISMA perfectly and still be methodologically weak — complete reporting of a bad search is still a bad search. The two work together: the Handbook shapes the review; PRISMA shapes the manuscript.

PRISMA 2020 is indexed on the EQUATOR Network, the central library of reporting guidelines, and the checklist, the abstracts checklist, and editable flow diagram templates are freely available at prisma-statement.org.

What Changed from PRISMA 2009 to 2020

If you learned PRISMA from the 2009 version, treat the 2020 update as a rewrite rather than a patch. The checklist still has 27 items, but their content, wording, and structure changed substantially to reflect a decade of advances in review methods — including the rise of automation tools in screening and stronger expectations around reproducibility.

• ·A revised 27-item checklist. Items now explicitly cover certainty-of-evidence assessment, how automation tools were used during selection, competing interests, and — new as item 27 — the availability of extracted data, analytic code, and other materials.
• ·A redesigned flow diagram. Databases and registers are counted separately at identification, records removed by automation tools get their own line, and dedicated templates exist for review updates, with boxes carrying forward studies included in the previous version.
• ·An abstracts checklist. PRISMA 2020 includes a 12-item checklist for abstracts, so the abstract is formally in scope rather than an afterthought.
• ·Formalized terminology. The distinction between records, reports, and studies is now explicit — and the flow diagram depends on it.

The wider PRISMA family — extensions such as PRISMA-P for protocols and PRISMA-S for search reporting — continues alongside the main statement, but for a standard systematic review of interventions, the 2020 checklist, the abstracts checklist, and the flow diagram are the core compliance set.

The 2020 Flow Diagram, Box by Box

The flow diagram is the most-checked figure in any systematic review, and 2020 restructured it. Where PRISMA 2009 had four phases — identification, screening, eligibility, included — PRISMA 2020 merges the old eligibility phase into screening, so the diagram now runs through three: identification, screening, and included. The screening phase, however, contains more boxes than before, and each has a precise meaning.

Identification

Report the number of records identified, with databases and registers (such as ClinicalTrials.gov) counted separately. Then report records removed before screening, broken into duplicate records removed, records marked as ineligible by automation tools, and records removed for other reasons.

Screening

This phase now covers three steps. First, records screened — typically titles and abstracts — with the number excluded. Second, reports sought for retrieval, with the number not retrieved (full texts you could not obtain). Third, reports assessed for eligibility, with excluded reports itemized by reason — wrong population, wrong study design, wrong comparator, and so on — each reason with its own count.

Included

The final phase reports the number of studies included in the review and the number of reports of those studies. If you are updating an earlier review, the 2020 template adds boxes for studies included in the previous version, feeding into a combined total.

A parallel arm of the diagram handles studies identified through other methods — websites, organizations, and citation searching — which flow through their own retrieval and eligibility boxes before merging into the included set. The distinction between the three units of counting is what keeps the whole figure coherent:

Checklist Items Journals Actually Reject On

All 27 items matter, but in practice a handful account for most editorial pushback. These are the ones to get right before submission.

• ·Item 7 — search strategy. The full, line-by-line strategy for every database and register searched, not one "representative" example. This usually belongs in a supplementary appendix, together with the date each search was run.
• ·Item 8 — selection process. How many reviewers screened, whether they worked independently, and how disagreements were resolved. "Two reviewers screened independently; conflicts were resolved by a third" is reportable; "studies were screened for relevance" is not.
• ·Item 11 — risk of bias. Name the tool, state how many assessors applied it and how, and report results per study rather than as a single summary sentence.
• ·Items 15 and 22 — certainty of evidence. State how confidence in the body of evidence was assessed and report it per outcome. Most journals expect the approach developed by the GRADE working group.
• ·Item 16 — study selection. The flow diagram itself, plus the often-missed item 16b: studies that appeared to meet the criteria but were excluded at full text, cited individually with the reason for exclusion.
• ·Item 24 — registration and protocol. Report the register and registration number, or state explicitly that the review was not registered. Prospective registration in PROSPERO is free. PROSPERO does not accept reviews once data extraction is underway, and registrations submitted after screening has started can be flagged as retrospective — so register before screening begins.
• ·Item 27 — data and code availability. New in 2020: state where extracted data, analytic code, and other materials can be found — or that they are not publicly available, and why.

Do not forget the abstract. The 12-item abstracts checklist is part of PRISMA 2020, and a structured abstract missing the number of included studies and participants, or the registration number, is an easy fix that reviewers nonetheless flag constantly.

Keep the Counts as You Go

The most error-prone part of PRISMA compliance is the flow diagram — not because it is hard, but because teams try to reconstruct it months after the fact. By then the deduplication log is gone, exclusion reasons live in three different spreadsheets, and the arithmetic no longer closes. The fix is procedural: treat the diagram as a living document you update at each phase of the review.

1. 1.Log search exports on the day you run them. For each database and register, record the interface, the date, the exact strategy, and the number of records exported. This single habit fills the entire identification phase.
2. 2.Deduplicate once, and record before-and-after counts. Whatever tool you use, note how many records went in and how many unique records came out; the difference is your "duplicates removed" number.
3. 3.Capture the exclusion reason at full-text stage, every time. The diagram requires excluded reports itemized by reason, and item 16b expects near-miss studies cited individually. A dropdown of predefined reasons in your screening sheet is enough.
4. 4.Give every study an ID and link its reports to it. The records-reports-studies distinction is where counts silently diverge — one trial with a journal article, a registry entry, and a conference abstract is three reports but one study.
5. 5.Snapshot the numbers at each phase transition. Freeze the counts when title-abstract screening closes and again when full-text assessment closes. Screening platforms such as AutoEvidence maintain this audit trail automatically and can export the counts in flow-diagram form, but a disciplined spreadsheet achieves the same thing.
6. 6.Rerun searches before submission, and update rather than overwrite. If months pass between the last search and submission, editors often ask for an update. Keep the original counts and add the new ones, so both search dates remain reportable.

A Pre-Submission Compliance Pass

Treat the checklist as an audit tool, not a formality. Work through all 27 items with the manuscript open, entering the actual page or section where each item is addressed. Many journals require the completed checklist as a supplementary file, and a checklist filled with optimistic page numbers is worse than none — it tells reviewers exactly where to look for what is missing.

• ·Download the current checklist and flow diagram template from prisma-statement.org rather than reusing files from an old project — 2009-era templates still circulate.
• ·Complete the abstracts checklist separately; the main checklist does not cover it.
• ·Reconcile every number in the flow diagram against your screening log one final time, including the by-reason exclusion counts.
• ·Cite the statement itself (Page MJ et al., BMJ 2021;372:n71) so editors know which version you followed.

PRISMA 2020 rewards reviews that were run transparently from the start. If the protocol was registered, the searches were logged, and every screening decision was recorded with a reason, compliance is an afternoon of transcription. If not, it becomes archaeology — which is precisely the failure mode the guideline exists to prevent.