How do I fill out a PRISMA flow diagram and the PRISMA 2020 checklist?

What is a PRISMA flow diagram?

PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) is the reporting guideline for systematic reviews. The current version, PRISMA 2020, was published by Matthew Page and colleagues in the BMJ in March 2021, with a separate explanation and elaboration paper. It replaced the 2009 statement.

The flow diagram is how you report study selection (checklist item 16a). It lets a reader see, in one figure, how many records your searches produced, how many you screened out and why, and how many studies ended up in the synthesis. Editors and peer reviewers look at it before almost anything else in a systematic review.

What do records, reports and studies mean?

The 2020 diagram uses three units, and mixing them up is the most common error. Definitions follow the PRISMA 2020 explanation and elaboration paper.

Screening happens on records; full-text eligibility is assessed on reports; and the review includes studies. That is why the final box has two numbers: studies included and reports of included studies.

How do I fill out the PRISMA flow chart? Box by box

The official templates are free Word files on prisma-statement.org. For a new review that searched only databases and registers, there are three stages.

1. Identification: records identified. Enter the number of records from each database and register before any deduplication. PRISMA recommends listing each source separately where feasible (for example, MEDLINE, Embase, CENTRAL, ClinicalTrials.gov) rather than one total.
2. Records removed before screening. Three lines: duplicate records removed; records marked as ineligible by automation tools; records removed for other reasons. Enter 0 where a line does not apply rather than deleting it.
3. Screening: records screened (title and abstract), and records excluded. If automation tools excluded any records at this stage, say how many were excluded by humans and how many by tools.
4. Reports sought for retrieval: the records that passed title and abstract screening, now as full texts to obtain.
5. Reports not retrieved: full texts you could not obtain, whatever the reason.
6. Reports assessed for eligibility (full-text review), and reports excluded, with reasons. Give a count for each reason; the counts must sum to the box total.
7. Included: studies included in review and reports of included studies.

If you also searched other sources (citation searching, websites, organisations, contacting authors), use the template with a second column. That column starts at records identified from those sources and runs through reports sought, not retrieved, assessed and excluded. Both columns feed the single included box.

A worked PRISMA flow diagram example

The numbers below are synthetic, for an invented systematic review of randomized trials; they illustrate the arithmetic only.

Included: 11 studies, reported in 14 reports. The 14 reports include 3 secondary publications (a follow-up and two subgroup papers) of trials already counted, which is why reports exceed studies.

Check the arithmetic the way a reviewer will: 3,326 minus 1,018 is 2,308; 2,308 minus 2,201 is 107; 107 minus 4 is 103; 103 minus 90 is 13; and 13 plus 1 is 14 reports. If any subtraction fails, your log is incomplete. Keep the per-database search counts and full search strategies in a supplement, as PRISMA 2020 and PRISMA-S ask.

What is the difference between the new and updated review templates?

PRISMA 2020 provides four flow diagram templates on prisma-statement.org.

In the updated-review templates, the previous version's studies sit in a separate box, new studies are counted through the usual stages, and the bottom box gives the total included across both. The grey boxes in the templates are completed only if applicable.

What are the 27 items in the PRISMA 2020 checklist?

PRISMA 2020 has 27 items in seven sections, several split into sub-items, and a separate 12-item checklist for abstracts. Below is what each section covers, in our own words; download the official checklist from prisma-statement.org for submission.

Compared with 2009, PRISMA 2020 added the abstract checklist, items on certainty of evidence, competing interests and data availability, and asks for full search strategies for every database, register and website searched.

Which PRISMA extension do I need? PRISMA-ScR, PRISMA-S, PRISMA-P and others

prisma-statement.org lists 17 official extensions. The ones physician-researchers meet most often:

As of October 2026, prisma-statement.org lists updates of PRISMA-ScR, PRISMA-P and the network meta-analysis extension as under development, along with new extensions (for example, rapid reviews and AI interventions). Check the site before submission in case a new version has been published. If your review is a scoping review rather than a systematic review, use PRISMA-ScR and its checklist, and adapt the flow diagram to the scoping review.

What tools can I use to draw a PRISMA flow diagram?

1. Official Word templates from prisma-statement.org: the safest choice, with the correct 2020 wording.
2. The PRISMA2020 Shiny app (linked from the PRISMA site) and the PRISMA2020 R package by Haddaway and colleagues, which produce 2020-compliant diagrams from a simple data sheet, including interactive versions.
3. Your screening software's counts. Platforms that manage deduplication and screening can report the numbers you need; check them against your own log before transcribing.
4. General drawing tools (PowerPoint, draw.io) if you need a custom layout, keeping the PRISMA box labels.

CliniPaper can check that every box adds up, that the included count matches your results tables and text, and that all 27 checklist items are addressed. You remain responsible for the review, and any AI use in preparing the manuscript should be disclosed per your journal's policy; see our AI disclosure statement examples.

Common mistakes reviewers flag

• Using the PRISMA 2009 diagram (or its wording) in a 2020-era review.
• Counting only deduplicated records in the identification box instead of the raw total per source.
• Reports and studies conflated, so the included box shows one number when multiple reports per study exist.
• Exclusion reasons that do not sum to the reports excluded box, or vague reasons such as "not relevant".
• "Reports not retrieved" left out, hiding full texts you could not obtain.
• Only included items counted in the other-sources column, instead of everything that column identified.
• Search strategies missing for some databases, registers or websites (item 7).
• Included count differs from the characteristics table or the forest plot.

For the trial equivalent of this diagram, see the CONSORT 2025 checklist and flow diagram; for the references in your review, see Vancouver referencing with examples.