The short answer
PRISMA means Preferred Reporting Items for Systematic Reviews and Meta-Analyses. It is reporting guidance that helps authors explain why a review was undertaken, how it was conducted and what it found. PRISMA 2020 provides the main 27-item checklist. Use it with applicable companion guidance to check the report; completing a checklist does not establish that the review methods or conclusions are sound.
Key takeaways
Choose guidance for the document you are preparing: a review, protocol, scoping review or search report.
Map each reporting requirement to a project record, a manuscript location and any action still needed.
Reconcile the checklist, flow diagram, results and supplements against the same final version.
What do the PRISMA guidelines cover?
A useful review report lets a reader follow the question through the search, selection, appraisal, synthesis and conclusions. The PRISMA 2020 statement, published in 2021, replaced the 2009 statement. Its name refers to the version, not the publication year.
Methods determine how the review is conducted; project records preserve its decisions and evidence; reporting explains what happened. PRISMA addresses that last responsibility. Use a methods handbook to plan the research, and the systematic-review explainer if you are still deciding which kind of review you need.
For example, “two reviewers screened the records” leaves important questions unanswered. Did they decide independently? How were disagreements resolved? Adding those details makes the process assessable, but does not show whether the eligibility criteria were appropriate. A report or software product should therefore not be described as “PRISMA certified.”
Which PRISMA document do you need?
Start with the document you are preparing and the review type. A protocol describes intended methods; a completed review describes what actually happened. Scoping reviews have a different purpose from reviews of intervention effects. Read the scope of each guideline and check any relevant journal or funder requirements.
These documents can work together. PRISMA-S adds detail about searches; it does not replace the main reporting guidance for the review. Its 16 search-reporting items were published in 2021. Use the official links below to obtain the full documents rather than treating this selection table as a substitute.
| Reporting task | Starting document | Use alongside | Check first |
|---|---|---|---|
Completed systematic review | PRISMA 2020 | Abstract guidance and applicable extensions | The scope of any question- or design-specific extension |
Systematic review protocol | PRISMA-P | Relevant guidance for planned search and review methods | Planned methods, registration and amendments |
Scoping review | PRISMA-ScR | Search-reporting guidance where applicable | The evidence-mapping purpose and reporting scope |
Literature search | PRISMA-S | The main guidance appropriate to the review | Sources, full strategies, dates and search processes |
Map the checklist to your manuscript and evidence
Gather the protocol and amendments, search logs, selection decisions, data definitions, appraisals and analysis records before entering page numbers in a checklist. Those records let you distinguish something that was done but not reported from something that was never done.
Build a working map with five fields: reporting topic, project record, manuscript location, current status and next action. Give an exact subsection, table or appendix name rather than simply “Methods.” The fictional examples below show how the same map can handle a completed explanation, a writing gap and an omitted procedure. They are teaching examples, not official checklist wording or evidence that these procedures occurred in your review.
In the first example, the protocol specified 12-month outcomes but the available studies reported six-month outcomes. Linking the original plan to a dated amendment makes the change visible; whether the change is appropriate still needs a methodological judgment. Likewise, an omitted assessment should be reported honestly, not relabeled “not applicable.” Use that label only when the requirement genuinely does not apply.
Collect
Bring together the records behind the report.
Map
Locate the explanation for each applicable requirement.
Resolve
Add missing reporting and identify limitations that remain.
Recheck
Have a colleague follow the map in the final submission version.
| Reporting topic | Project record | Manuscript location | Status | Next action |
|---|---|---|---|---|
Change from 12 to 6 months | Protocol v1 and dated amendment log | Methods: protocol amendments; Appendix A | Mapped | Check that outcome labels and results also use 6 months |
Complete search strategies | Search log and source exports | Methods currently says only “we searched the literature” | Done; not yet described | Create Appendix B and link it from the search methods |
A relevant assessment was omitted | Project record confirms it was not performed | Not yet stated in Methods or limitations | Not performed; needs explanation | Describe the omission and its consequences rather than marking it inapplicable |
Build a reporting package that can be followed
The manuscript carries the main explanation; its supplements make the detail inspectable. Name the files and tell readers where to find them. A practical package brings together the manuscript, completed checklist, search appendix, selection records and relevant analysis files. Where material cannot be shared, explain the access restriction and what information is available instead.
The examples below show how vague statements can point to something a reader can actually inspect. The subsection and appendix names are fictional. Use locations and descriptions that match your own work, and check them again after editing. Compare the abstract with the full text too: its study counts, estimates and account of uncertainty should agree with the underlying results.
| Vague statement | Record needed | A specific reporting location |
|---|---|---|
“We searched the literature.” | Source names, dates and full search strategies | Methods: information sources and search; full strategies in Appendix B |
“Disagreements were resolved.” | Reviewer assignments and conflict-resolution records | Methods: study selection; describe independent decisions and adjudication |
“AI helped extract data.” | Assisted fields, source locations and correction records | Methods: data collection; task and verification log in Appendix C |
How to choose and read a PRISMA flow diagram
The reporting package also needs a clear account of how the search led to the included evidence. A PRISMA flow diagram follows that path, showing the numbers identified, screened, retrieved and included, with exclusions along the way. Read its units carefully: a record is a database or website entry, a report is a document describing a study, and a study is the underlying investigation. One study may have a registry entry, a conference abstract and a journal article, so reports and studies need separate counts.
Choose among the four official PRISMA 2020 templates using two questions: is this a new review or an update of an existing review, and did you search only databases and study registers or use other sources as well? The table below links to the corresponding Word templates. The diagram summarizes selection; the methods and search appendix still need to explain how each source was searched.
Grey literature is not a separate template category. It refers to material such as dissertations, conference abstracts and reports outside traditional commercial publishing, and it can be found in databases as well as on websites. The additional branch describes other ways of finding evidence, such as consulting organisations or following reference lists, which can also identify journal articles. Classify results by how they were identified and tracked, rather than treating the two branches as published versus unpublished literature. The PRISMA flow-diagram FAQ explains source tracking; the Cochrane search guidance discusses grey literature sources.
| Review type | Databases and registers only | Also using other sources |
|---|---|---|
New systematic review | ||
Update of an existing review |
Two worked examples for a new review
These fictional examples concern the same review topic under two different search scenarios. Switch examples to see what changes when other sources are added. All counts are invented for teaching; the diagrams are adaptations of the official templates, not results from an actual review.
Scenario A searches databases and registers only. Of 500 records, 100 duplicates are removed before screening. The process ends with 20 reports describing 18 studies. No automation or other removals before screening are assumed in this example.
- Databases (n = 480)
- Registers (n = 20)
- Duplicate records (n = 100)
- Ineligible population (n = 15)
- Ineligible study design (n = 12)
- Ineligible intervention (n = 8)
Scenario B keeps the same database/register results and adds 12 records from websites, organisations and citation searching. For this example, each leads to a different report sought for retrieval, and the reports and studies do not overlap with the database branch. Four additional reports describe three additional studies, giving 24 reports and 21 studies overall. Real reviews must resolve overlap between sources before combining counts.
- Databases (n = 480)
- Registers (n = 20)
- Duplicate records (n = 100)
- Ineligible population (n = 15)
- Ineligible study design (n = 12)
- Ineligible intervention (n = 8)
- Websites (n = 5)
- Organisations (n = 3)
- Citation searching (n = 4)
- Ineligible population (n = 3)
- Ineligible study design (n = 2)
- Ineligible intervention (n = 1)
Scroll horizontally to follow both branches, or enlarge the diagram.
PRISMA 2020 source templatesLicense CC BY 4.0
How to check the numbers
- Follow each branch separately. In A, 500 − 100 = 400 records screened, and 400 − 340 = 60 reports sought. Of these, 5 cannot be retrieved, leaving 55 for eligibility assessment. Those 5 belong in “not retrieved”, not among the reports excluded after assessment.
- Give each excluded report one primary reason for this count. In the database branch, 15 + 12 + 8 = 35 exclusions, leaving 55 − 35 = 20 included reports. In B’s other-sources branch, 12 − 2 = 10 reports assessed and 3 + 2 + 1 = 6 excluded, leaving 4 included reports.
- Count reports and studies separately. A’s 20 reports describe 18 studies; B adds 4 reports from 3 distinct additional studies. Several reports from one study should be linked to that study, not automatically discarded as duplicate reports.
- Track other sources from identification onward, including reports that were not retrieved or were excluded. Do not enter only the final included reports in that branch. Reconcile duplicate records across sources with the search log and document any adaptations to the flow. The simplified examples assume one report sought per record reaching retrieval.
If you are updating an existing review
Use the updated-review template to distinguish evidence carried forward from the previous version from evidence identified in the update, and reconcile both with the new total. Rerunning a search before submitting a first review does not by itself make it an updated review: combine those search runs transparently and retain their dates and counts in the search log. Check the final counts against the search log and results tables, then describe any assistance used along that selection path.
How should AI assistance be reported?
Describe what the AI did and how its output was checked. At the source check on 18 October 2026, the official PRISMA extension directory still listed the partial update on AI tools in the review process as under development. The log suggested here is an editorial aid, not an adopted PRISMA AI checklist; check the directory and your journal policy when preparing a submission.
For each assisted task, retain the tool and available version information, date, inputs, output, source checks, corrections and responsible reviewer. Explain what happened to records or values the tool did not resolve. For screening prioritization, for example, readers need to know how the rest of the queue was handled, not just what the tool surfaced.
A fictional report might state: “The tool suggested study-design labels from abstracts. A reviewer checked every label against the source, and uncertain records remained eligible for full-text assessment.” Use such wording only when it describes the recorded procedure. Keep restricted inputs out of public logs. If you are deciding how to evaluate assistance before using it, continue to AI tools for systematic review.
A final audit before submission
After the last substantive edit, ask someone who did not write the report to follow its evidence map. A failed lookup is useful feedback: it tells you which location, explanation or attachment needs attention before submission.
Keep the completed checklist with the exact submitted version. Obtain the original documents and any available PRISMA translations from the official site; this guide is an independent teaching explanation, not an official translation.
Every applicable checklist entry points to a specific, current passage or attachment.
The flow diagram, abstract and results tables agree on report and study counts.
Changes from the protocol have a dated record and an explanation.
Supplement links work, filenames match and access restrictions are stated.
The conclusions reflect the reported uncertainty and limitations.
Common questions
Is PRISMA a method for conducting a systematic review?
PRISMA guides reporting. Use an appropriate methods framework to conduct the review, then use PRISMA to explain what was done. A complete checklist does not by itself establish methodological quality.
Which checklist should I start with?
Start with the document and review type: PRISMA 2020 for a completed systematic review, PRISMA-P for a protocol and PRISMA-ScR for a scoping review. Check their scopes and applicable extensions; PRISMA-S can supplement the search-reporting component.
Is PRISMA 2020 still the main statement?
At the October 2026 source check, the official site still presented PRISMA 2020 as the main statement. Check the official directory for relevant extensions and updates when preparing your report.
Can I mark a procedure that was not done as not applicable?
Only if it genuinely does not apply. If a relevant procedure was omitted, describe that fact and its effect on the review rather than concealing the omission with an inapplicable label.
Do I need a meta-analysis to use PRISMA?
No. Explain how findings were organized and synthesized when estimates were not pooled. The report still needs a clear account of the review methods and results.
Sources & further reading
Follow the original sources for their complete methods, scope, and latest updates.
- PRISMA 2020 statementPRISMA
- PRISMA-P: reporting review protocolsPRISMA
- PRISMA-ScR: reporting scoping reviewsPRISMA
- PRISMA-S: reporting literature searchesPRISMA
- PRISMA 2020 flow-diagram templatesPRISMA
- PRISMA extensions and development statusPRISMA
- PRISMA translationsPRISMA
- PRISMA 2020 and PRISMA-S: common questions on tracking records and the flow diagramRethlefsen & Page, 2022
- Cochrane Handbook, Chapter 4: Searching for and selecting studiesCochrane