The PRISMA flow diagram, explained
The diagram is not decoration. It is an audit trail, and reviewers check that it adds up.
What is a PRISMA flow diagram?
It is a four-stage diagram reporting what happened to every record in a systematic review: how many were identified, how many remained after duplicates were removed, how many were screened and excluded at each stage, and how many studies were finally included. Every number must reconcile with the ones above it, and full-text exclusions must state reasons.
What the diagram is for
A systematic review claims to have found and fairly assessed the relevant evidence. The flow diagram is the evidence for that claim: it accounts for every record from the moment it was retrieved to the moment it was included or dropped, and it makes selective reporting visible.
PRISMA 2020 replaced the 2009 version and changed the structure in one significant way: records found through databases and registers are now reported separately from records found by other methods — citation chasing, hand-searching, contacting authors, grey literature. The two columns run in parallel down the diagram.
The four stages
| Stage | What it counts | Where the number comes from |
|---|---|---|
| Identification | Records retrieved, per database, before any filtering | The result count logged when each search was run |
| Screening | Records after duplicates removed; records screened; records excluded on title and abstract | Your reference manager after deduplication, and the screening log |
| Eligibility | Reports sought for retrieval, reports not retrieved, reports assessed in full text, and full-text exclusions with reasons | The full-text pass, with a recorded reason per exclusion |
| Included | Studies included in the review, and reports of those studies | The final set |
Studies and reports are counted separately in PRISMA 2020, and the distinction is real: one multi-arm trial may be described across four papers. The review includes one study and four reports.
The numbers have to reconcile
This is what reviewers check first, and it is the most common reason a diagram is sent back.
- Records identified, minus duplicates removed, equals records screened.
- Records screened, minus records excluded at title and abstract, equals reports sought for retrieval.
- Reports sought, minus reports not retrieved, equals reports assessed for eligibility.
- Reports assessed, minus the sum of the reasoned exclusions, equals reports of included studies.
- The reasons listed under full-text exclusions must sum exactly to the total excluded at that stage.
If a line does not balance, the cause is usually a record silently dropped between stages — most often a duplicate removed late, or a paper set aside as awaiting retrieval and never resolved. Find it rather than adjusting a number, because the discrepancy is a real gap in the audit trail.
Common mistakes
| Mistake | Why it fails |
|---|---|
| Reporting one identification total across all databases | PRISMA 2020 asks for a per-database breakdown; a single total hides which source contributed what. |
| Giving reasons for title and abstract exclusions | Reasons are required only at full-text stage. Recording them earlier is not wrong, but the box should carry the count. |
| Omitting reasons at full-text stage | This is a hard requirement, and the reasons must sum to the exclusion total. |
| Folding citation-chasing finds into the database column | The 2020 diagram separates them for a reason: it shows how much the supplementary methods contributed. |
| Counting reports as studies | Inflates the apparent evidence base when several papers report one study. |
| No box for reports not retrieved | Papers you could not obtain are a limitation and must be visible rather than absorbed into exclusions. |
Fill it in as you go
Every number in the diagram is trivial to record at the moment it happens and painful to reconstruct afterwards. Log the per-database result count when you run each search, the post-deduplication count when you deduplicate, and the exclusion reason at the moment you exclude.
Reviews that build the diagram at write-up almost always find a discrepancy they cannot explain, and the only honest fix is rescreening.
Questions
- What changed between PRISMA 2009 and PRISMA 2020?
- The 2020 diagram separates records found via databases and registers from records found by other methods, adds explicit boxes for reports sought and reports not retrieved, and distinguishes studies from reports of studies. It also comes in a version for reviews updating a previous review.
- Do I need a PRISMA diagram for a scoping review?
- Yes. PRISMA-ScR is the scoping review extension and includes a flow diagram. Scoping reviews report the selection process the same way, even though they do not assess risk of bias or synthesise effects.
- Where do I get a PRISMA template?
- The PRISMA statement website publishes the 2020 flow diagram templates as free Word and PDF files, in database-only and combined-source variants. Several open tools also generate the diagram from a screening log.
- How do I count duplicates found across databases?
- Record the raw per-database totals at identification, then a single duplicates removed figure before screening. Duplicates are removed once, at that point, not netted out of individual database counts.
Related guides
- Building a systematic review search strategy — How do you build a systematic review search strategy?
- Screening papers for relevance — How do you screen papers for relevance quickly?
- Citation chasing and snowballing — What is snowballing in a literature review?
- Searching grey literature — What is grey literature and how do you search it?
Last updated 2026-08-25. Part of the litscout literature search guides.