A PRISMA flow diagram is the figure that shows how many records your searches found, how many you removed or excluded at each stage, and how many studies survived into the synthesis. It is written for anyone running an evidence synthesis who has finished screening and now has to turn a spreadsheet of decisions into a defensible figure that a journal editor, a peer reviewer, and a future replicator will all accept. PRISMA stands for Preferred Reporting Items for Systematic Reviews and Meta-Analyses, a reporting guideline first published in 2009 to reduce incomplete and inconsistent reporting of reviews [1].
What you need before you start is straightforward but often underestimated. You need a per-database count of records retrieved with the date each search was run, a duplicate count from your reference manager or screening tool, title and abstract exclusion totals, a list of full texts you could not obtain, and a tally of full text exclusions grouped by reason. You also need to know whether your review counts studies, reports, or both, because the 2020 guideline asks you to distinguish them [2]. If any of those figures are missing, you cannot complete the diagram without guessing, and a guessed number is the single most common source of a figure whose arithmetic does not reconcile.
The rest of this article walks through what each box means, how to report the assessment methods that sit alongside the figure, how to assemble the PRISMA flow diagram step by step, where the official explanatory guidance lives, and which template or software route suits your project. Open source tooling now exists specifically to build compliant diagrams and to make them interactive for digital publication [3], which changes the practical calculation for larger reviews.
Key takeaways
- A PRISMA flow diagram records how many records entered your review, how many left at each stage, and why, so a reader can reconstruct your selection path from the figure alone.
- The 2020 version separates records found in databases and registers from records found through websites, citation searching, and contact with authors, and adds a box for records removed before screening.
- Every number must reconcile: records screened minus records excluded should equal reports sought for retrieval, and exclusion reasons at full text must be itemized rather than summed.
- Records are database entries; reports are documents. One study can produce several reports, which is why the final box counts studies, not papers.
- Diagrams can be drawn from a downloadable template, produced by an open R package and web app, or generated from live screening data in review software.
What Is a PRISMA Flow Diagram and What Does It Show?
A PRISMA flow diagram is a four stage map of study selection: identification, screening, eligibility assessment, and inclusion. It shows the total records retrieved from each source type, the records discarded before a human read anything, the records screened on title and abstract, the reports sought and actually retrieved, the reports assessed in full, and the studies finally included. Each stage carries a number, and each drop between stages must be explained by a labeled exclusion box.
What the figure shows that prose cannot is proportion. A reader can see instantly that 4,812 records collapsed to 61 full texts and then to 9 included studies, which tells them something about the precision of your search and the tightness of your eligibility criteria. The 2009 statement introduced the diagram as part of a minimum reporting set precisely because narrative descriptions of selection were routinely vague [1], and the 2020 update rebuilt it to handle sources that databases do not index, such as trial registers, organizational websites, and citation chasing [2].
The diagram is not an audit certificate. It reports what you did; it does not demonstrate that your searches were sensitive or your criteria sensible. Using a PRISMA flow diagram generator, whether the open R package and web app described by Haddaway and colleagues [3] or a feature inside review software, removes drawing effort and arithmetic slips, but the underlying counts still have to be right. Garbage numbers render beautifully.
How to Report Risk of Bias Assessment Methods and Assessors
Risk of bias reporting belongs in your methods text, not in the figure, and the 2020 guideline is specific about what that text must contain. Name the tool and version you used, state how many people assessed each study independently, explain whether assessors worked with the studies blinded or unblinded, describe how disagreements were resolved, and say whether any automation tools contributed to the process [2]. If different tools applied to different designs, for example one instrument for randomized trials and another for cohort studies, report the mapping rather than a single blanket statement.
Two details are commonly omitted. First, report the domain level judgments, not only an overall rating, because readers cannot interpret a summary label without knowing which domain drove it. Second, state who the assessors were in terms of role and training, since a review assessed by two experienced methodologists is not equivalent to one assessed by a single student. The original statement placed quality assessment alongside selection reporting for exactly this reason [1].
Keep this separate from the PRISMA flow diagram. Studies excluded on methodological grounds after eligibility was confirmed should be reported in the results text with their bias ratings, not folded into a selection exclusion box. If you are working from a PRISMA flow diagram template, resist the temptation to add a bias box to it; interactive diagram tools can link out to appraisal detail instead [3].
How to Build a PRISMA Flow Diagram Step by Step
Building the figure works best as a fixed sequence, completed in order, with each number written down before the next stage begins.
- Record the hits from every database and register separately, with the platform, date run, and result count. Record non database sources, such as websites, citation searching, and author contact, in their own column.
- Deduplicate, and log how many records automated deduplication removed, how many an automation tool marked ineligible, and how many you removed for other stated reasons.
- Screen titles and abstracts against your eligibility criteria and record the number excluded as one figure.
- List the reports you sought for retrieval, then log how many you could not obtain and why.
- Assess the retrieved reports in full and record exclusions grouped by reason, each with its own count.
- Report the number of studies included, and separately the number of reports those studies generated [2].
Then reconcile. Records screened minus records excluded must equal reports sought; reports sought minus reports not retrieved must equal reports assessed. Check both arithmetic chains before drawing anything, because correcting a figure later means correcting the abstract and results text too. If you start from a PRISMA flow diagram template, fill the numbers in a spreadsheet first and paste them in once. The original statement set this transparency expectation [1], and diagram software can rebuild the PRISMA flow diagram automatically whenever a count changes [3].
Where to Find the PRISMA 2020 Explanation and Elaboration Paper
The explanation and elaboration paper is published in the BMJ as Page et al., PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews, BMJ 2021;372:n160 [4]. It is open access, and it is the document to consult when the checklist wording alone does not tell you what a journal expects. For each item it gives the rationale, the essential elements to report, and a published exemplar showing acceptable wording, which is far more useful than the summary statement when you are arguing a reporting decision with a reviewer.
Read it alongside the main 2020 statement in the BMJ [2] and the 2009 original, which remains useful for understanding why particular items exist [1]. Three parts repay close reading: the guidance on distinguishing records, reports, and studies; the treatment of sources that are not bibliographic databases; and the discussion of when a study selection diagram needs an additional column or a second panel, for example in a review that updates a previous one.
Practical tip: extract the exemplar wording for the items you find hardest and keep it beside your draft. It is faster than reinventing phrasing, and it gives you a citable precedent. If you build your study selection diagram with the open R package, the field labels follow the same guidance, which reduces the risk of mislabeling a box [3].
Every Box in the Flow Diagram and Which Numbers Go Where
Each box in a PRISMA flow chart takes one specific number, and mixing them up is the fastest way to produce a study selection diagram that does not reconcile. The table below maps every box to its source and the check that confirms it.
| Box | Number it takes | Where the number comes from | Reconciliation check |
|---|---|---|---|
| Records identified from databases and registers | Raw hits per source, then total | Search logs with dates and platforms | Per source counts sum to the total |
| Records identified from other methods | Hits from websites, citation searching, author contact | Separate log, kept apart from database counts | Reported in its own column |
| Records removed before screening | Duplicates, automation exclusions, other removals | Reference manager and screening tool logs | Three subtotals sum to the box |
| Records screened | Records seen at title and abstract | Identified minus removed before screening | Matches screening tool total |
| Records excluded | Single figure, no reasons required | Screening decisions | Screened minus excluded equals reports sought |
| Reports sought for retrieval | Full texts requested | Retrieval list | Equals screened minus excluded |
| Reports not retrieved | Full texts unobtainable | Retrieval log with reasons | Sought minus not retrieved equals assessed |
| Reports assessed for eligibility | Full texts read in full | Eligibility decisions | Equals assessed exclusions plus included reports |
| Reports excluded | Itemized count per reason | Eligibility form | Reason counts sum to exclusions |
| Studies included; reports of included studies | Two distinct figures | Final included set | Reports are equal to or greater than studies |
The 2020 guideline requires exclusion reasons at the eligibility stage but not at title and abstract screening [2], a distinction the 2009 version handled less explicitly [1]. Diagram generators enforce these box definitions, which is part of their value [3].
Reporting Your Review With the PRISMA 2020 Abstract Checklist
The abstract checklist is a 12 item companion to the main checklist, and it governs what your abstract must state even though no figure appears there [2]. It asks for the objectives, eligibility criteria, information sources with the last search date, risk of bias methods, synthesis methods, the number of studies and participants included, the main results with effect estimates and precision, limitations of the evidence, an interpretation, funding, and the registration number.
The number you create PRISMA flow diagram figures around must match the abstract exactly. If the diagram says 14 included studies and the abstract says 15 because a companion report slipped through, reviewers will notice, and the mismatch casts doubt on the rest of the reporting. Reconcile the abstract, the results text, the PRISMA flow chart, and the characteristics table as one final step, not as four separate edits.
Two practical points the checklist implies but does not spell out. Report the last search date in the abstract, not the submission date, since the gap between them tells readers how current the evidence is. And give participant totals alongside study totals, because 14 studies with 320 participants supports a very different conclusion from 14 studies with 90,000 [1]. Tools that regenerate the figure from live counts help keep the two in step [3].
Why Screening Transparency and Reproducibility Depend on the Flow Diagram
Reproducibility depends on the figure because the numbers in it are the only public trace of thousands of private decisions. A reader who has your search strings, your eligibility criteria, and your stage by stage counts can rerun the search, repeat the screen, and compare their yield with yours. Without the counts, they can only compare final reference lists, which hides where two teams diverged. The 2009 statement framed complete reporting of selection as a minimum condition for a review to be appraisable at all [1].
The 2020 guideline sharpened this by asking authors to state how many reviewers screened each record, whether they worked independently, and what automation tools assisted them [2]. The explanation and elaboration paper gives exemplars for that wording and explains why vague phrasing such as two reviewers screened the results fails the test [4]. When you create PRISMA flow diagram output that ties to a dated audit trail of screening decisions, you also give yourself protection: if a reviewer queries an excluded study, you can show when and why it was excluded.
Interactive versions extend this further by letting readers click through from a box to the underlying record list [3]. A PRISMA 2020 flow diagram published that way turns a static count into something a reader can verify rather than trust. That is the direction reporting is moving, and it is worth considering before you finalize a flat image.
Templates, Tools, and Workflows for Filling In Your Diagram
Three routes exist for producing a PRISMA 2020 flow diagram, and the right choice depends on review size and how often your numbers will change. The official downloadable templates, available in editable word processor and PDF formats, cost nothing and work well for a small review with stable counts. Their weakness is that every number is typed by hand, so a late correction means re-editing boxes and re-checking arithmetic that no tool is validating for you.
The second route is a dedicated PRISMA flow diagram generator. The open source R package and web application described by Haddaway and colleagues takes a structured data file of counts and renders a compliant figure, with an interactive variant for digital publication [3]. It enforces the box structure defined in the 2020 statement [2] and the explanatory guidance [4], which catches label errors before submission, and it makes regeneration trivial when counts change.
The third route is review software that holds your screening decisions and derives the figure from them, which is how Systematicly handles it. The trade-off is that the figure is only as good as the data hygiene in the platform, and you still need to verify the counts yourself.
Whichever route you take, keep the raw counts in a version controlled spreadsheet with search dates. That file, not the image, is your real record, and it is what you will need if you update the review later. Our systematic review software overview and PRISMA reporting guide cover the wider toolchain.
What Is a PRISMA Diagram?
A PRISMA diagram is the figure that documents study selection in a review reported under the PRISMA guideline, and in practice the term is used interchangeably with flow diagram or flow chart. It sits in the results section, almost always as Figure 1, and it is usually the first thing an editor looks at when assessing whether a submission is reportable.
Its practical role is threefold. It proves the selection process happened in a defined order rather than being assembled opportunistically. It gives peer reviewers a fast way to test internal consistency against the abstract and the methods text. And it becomes the baseline for any later update, since an updated review reports new records against the original counts. Treat it as a data product with a traceable source, not as artwork, and it will survive review. For the wider methodology context, see our Complete Guide to Systematic Reviews: Methodology, Phases, and Reporting and Meta-Analysis Guide: Methods, Forest Plots, and Heterogeneity guides.
Tips and Common Mistakes When Filling In the Boxes
The recurring errors are arithmetic, not artistry. Counts that do not reconcile between screened, sought, and assessed; duplicates removed twice, once in the reference manager and again in the screening tool; full text exclusions reported as a single lump rather than itemized by reason; and a final box that counts papers when the guideline asks for studies with their associated reports listed separately [2]. Reviews that update an earlier synthesis often omit the prior included set entirely, leaving readers unable to see what is new. Practical habits help more than software does. Freeze your search date before drafting, keep per-database counts in a spreadsheet that the figure is copied from, have a second person re-add the chain, and write the exclusion reason labels from your protocol rather than inventing them after the fact. If a number changes, change it in the spreadsheet, the figure, the abstract, and the results text in one sitting. The 2009 statement made complete selection reporting a condition of appraisability [1], and the explanation and elaboration paper supplies exemplar wording for the cases most authors get wrong [4].
How Journals and Peer Reviewers Read the Figure
Editors and reviewers use the figure as a fast integrity check rather than a decorative element. The usual sequence is to test whether the boxes add up, compare the included count against the abstract and the characteristics table, look for itemized exclusion reasons at full text, and check that non-database sources such as registers, websites, and citation chasing are reported in their own path as the 2020 statement requires [2]. Where those checks fail, the query that follows is normally about the underlying records, not the layout, so keep your dated screening log retrievable. Claims that a given tool guarantees compliance should be treated with care: software can enforce the box structure and the arithmetic, which is genuinely useful [3], but no tool can judge whether your eligibility criteria were applied consistently or your search was sensitive enough. That judgment stays with the review team, and the figure only reports it [1].
Frequently asked questions
What is a PRISMA flow diagram?
A PRISMA flow diagram is a figure that tracks records through four stages of an evidence synthesis: identification from databases, registers, and other sources; screening on title and abstract; eligibility assessment on full text; and final inclusion. Each stage reports how many records entered, how many left, and, at the full text stage, why they left [2]. A flowchart is worth using rather than prose because it makes the selection path checkable at a glance and lets a reader judge whether your search was broad and your criteria narrow, or the reverse [1].
What are the four steps of a PRISMA?
The four steps are identification, screening, eligibility, and inclusion. Identification captures raw search yield by source and the records removed before any human reading. Screening covers title and abstract decisions and the single exclusion count. Eligibility covers full text assessment, including reports you could not retrieve and exclusions itemized by reason. Inclusion reports the number of studies and, separately, the number of reports those studies produced [2]. The explanation and elaboration paper gives published exemplars for each step [4].
What does PRISMA stand for?
PRISMA stands for Preferred Reporting Items for Systematic Reviews and Meta-Analyses. It began as QUOROM in 1996, was rebuilt and renamed with the 2009 statement, which set out a minimum reporting set of items and the flow diagram [1], and was updated in 2020 to reflect changes in search methods, automation, and reporting practice, including separate handling of non database sources [2].
How to do a PRISMA flow diagram in Word?
To build one in a word processor, download the official editable template rather than drawing shapes yourself. Type your reconciled counts into the existing text boxes, keep the box structure unchanged, and export as a high resolution image or PDF for submission. Check the arithmetic in a spreadsheet before you type, because word processor shapes do no calculation and will happily display a figure that does not add up.
Does Microsoft Word have a flowchart template?
Yes. Mainstream word processors include flowchart shape libraries and pre-built diagram galleries, and they are fine for generic process charts. They are not PRISMA layouts, though, so the box names, the split between database and non database sources, and the exclusion reason list all have to be added by hand. Starting from the official PRISMA template avoids that work and the labeling errors that come with it.
How do I turn a Word document into a flow chart?
There is no reliable automatic conversion. Pull your stage counts out of the document into a small table, then either type them into the official template or feed the table to a generator. The best app to create a flow chart for a large or updatable review is a data driven one, such as the open source R package and web application, because it rebuilds the figure from a structured input file and supports interactive versions [3].
References
- Moher D, Liberati A, Tetzlaff J, Altman DG; The PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med. 2009;6(7):e1000097. ↩
- Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. ↩
- Haddaway NR, Page MJ, Pritchard CC, McGuinness LA. PRISMA2020: an R package and Shiny app for producing PRISMA 2020-compliant flow diagrams, with interactivity for optimised digital transparency and open synthesis. Campbell Syst Rev. 2022;18(2):e1230. ↩
- Page MJ, Moher D, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews. BMJ. 2021;372:n160. ↩