PRISMA 2020 flow diagram generator

Type your numbers and get the complete PRISMA 2020 diagram, with every box from the official template. The arithmetic is checked as you type, because numbers that do not reconcile are the most common reason a diagram gets sent back. Download as SVG for a manuscript or PNG for a slide. Nothing is uploaded and nothing is stored.

Your numbers

Identification

PRISMA asks you to report each database separately where you can, rather than one combined total.

Screening

The three greyed fields are calculated, so they always reconcile.

Reports excluded, with reasons

PRISMA expects a reason for every full-text exclusion, and the reasons should account for all of them.

Included

A study can have several reports, so these two numbers often differ.

Your diagram is below this form and updates as you type.

Other methods (optional)

Tired of counting by hand?

Verflux fills this diagram in from your actual screening decisions, keeps the numbers reconciled as you work, and lists your full-text exclusion reasons automatically.

Start a free project See what else it does

What goes in each box

The box names changed in 2020 and several of them are not self-explanatory. This is what each one is asking for.

Records identified from
Everything your searches returned, before you looked at any of it. PRISMA asks you to give each database and register separately where you can, because a single combined total cannot be checked or reproduced by anyone else. Registers such as ClinicalTrials.gov or the WHO ICTRP are listed apart from bibliographic databases.
Records removed before screening
Anything discarded mechanically, before a human judgement was made. Three separate counts: duplicates, records removed by automation tools, and records removed for other reasons. The 2009 diagram collapsed this into one deduplication figure; 2020 splits it because "an algorithm dropped 400 records" and "we removed 400 duplicates" are very different statements about a review.
Records screened
What a person actually read at title and abstract. This is identified minus removed before screening, so it is arithmetic rather than a judgement.
Records excluded
Rejected at title and abstract. No reasons are required here, unlike at full text. If automation tools did some of the excluding, say how many were machine and how many were human.
Reports sought for retrieval
New in 2020, and the box people most often leave out. These are the records that survived screening and whose full text you then tried to obtain. It is not the same as the number you read, because some full texts cannot be got hold of.
Reports not retrieved
Full texts you could not obtain at all: no institutional access, no reply from the author, a conference abstract with no paper behind it. Reporting this honestly is the point of the box, and a zero here on a large review invites questions.
Reports assessed for eligibility
The full texts you actually read. Sought minus not retrieved.
Reports excluded, with reasons
Every full-text exclusion needs a stated reason, and the reasons should account for all of them. Count each excluded report once, under a single reason, even when it fails several of your criteria. Reasons that sum to more than the number excluded is a common and immediately visible error.
Studies included in review
The research you are synthesising.
Reports of included studies
The documents describing that research, which is often a larger number. Eight included studies might be published across ten papers. PRISMA 2020 asks for both so that reports are not quietly lost between the assessed and included counts.

The four subtractions a reviewer will check

A diagram whose numbers contradict each other is the fastest way to lose a reader's confidence, and it is the single most common defect in submitted diagrams. Four identities have to hold:

screened = identified − removed before screening
sought for retrieval = screened − excluded at screening
assessed for eligibility = sought − not retrieved
studies included = assessed − excluded at full text

The generator above calculates all four for you, which is why those fields are greyed out: you cannot enter a number that contradicts the box above it. It also warns when your itemised exclusion reasons do not sum to the number excluded.

One thing no tool can check for you: the figures in the diagram must match the figures in your manuscript text. Reviewers notice a diagram that says 34 included and a results section that says 36.

What changed from PRISMA 2009

20092020
"Records" used throughout Records (database entries) distinguished from reports (documents)
One deduplication box Removed before screening split into duplicates, automation tools and other reasons
No retrieval step Adds reports sought for retrieval and reports not retrieved
Single "additional records from other sources" box, often filled in with only the records that were eventually included A full parallel pathway for other methods, expected to account for everything found that way
One included box Studies included and reports of those studies
One template Separate templates for new and updated reviews, with and without other sources

PRISMA 2009 is superseded for new reviews. If a journal endorses PRISMA, it expects the 2020 form.

Five mistakes that get diagrams sent back

  1. Numbers that do not subtract. The most common of all, and the easiest to catch before submission.
  2. Exclusion reasons that do not add up to the exclusions. Usually caused by counting a report under every criterion it failed instead of choosing one.
  3. Omitting reports sought for retrieval. Carried over from the 2009 layout out of habit. It is a required box in 2020.
  4. Under-reporting other sources. Under 2009 many authors put only the included studies in the "other sources" box. 2020 expects every record found that way to be accounted for, screened and reported like any other.
  5. A diagram that disagrees with the text. Treat the diagram and the manuscript as one document with two views of the same numbers.

Questions

What changed between the PRISMA 2009 and PRISMA 2020 flow diagram?

PRISMA 2020 distinguishes records (database entries) from reports (documents), adds boxes for reports sought for retrieval and reports not retrieved, requires the removed-before-screening figure to be broken down, makes the other-methods pathway explicit, and asks for exclusion reasons to be itemised. It also provides separate templates for new and updated reviews.

Which numbers in a PRISMA flow diagram have to add up?

Identified minus removed before screening equals screened. Screened minus excluded equals reports sought for retrieval. Sought minus not retrieved equals reports assessed for eligibility. Assessed minus excluded equals studies included.

What is the difference between studies included and reports of included studies?

A study is a piece of research; a report is a document describing it. One study can be published across several papers, so eight included studies might be described in ten reports. PRISMA 2020 asks for both numbers so reports are not lost between the assessed and included counts.

Do I need the other methods column?

Only if you found records outside database and register searching, such as citation chasing, websites, grey literature or contacting organisations. If every record came from a database or register, leave that column off.

Can I reuse and adapt the PRISMA flow diagram?

Yes. PRISMA 2020, including the flow diagram, is published under a Creative Commons CC BY 4.0 licence, so it can be used and adapted with attribution. Cite Page MJ et al. BMJ 2021;372:n71.

Is this PRISMA flow diagram generator free?

Yes, and there is no sign-up. It runs entirely in your browser, so nothing is uploaded or stored.

The numbers are easier when something else is counting them

Verflux tracks every record from import to inclusion, so this diagram fills itself in and stays reconciled while you work. Your full-text exclusion reasons appear in it automatically, and blinded dual screening keeps the counts defensible.

Start a free project See the rest