How to Complete a PRISMA 2020 Flow Diagram

Start with the correct purpose

PRISMA 2020 serves as a transparent reporting guideline. Its flow diagram visually maps the pathway from initial identification to final study inclusion. It does not replace a protocol, explicit eligibility rules, dual-screening procedures, or an auditable selection ledger. Those methods must be planned and executed separately.^1,2,10

A reconstructive diagram answers key audit questions: How many records entered the review? What occurred during screening? How many full-text reports were sought and evaluated? Why were reports excluded after appraisal? How do final reports map to included studies? The diagram cannot verify whether search strategies captured all eligible literature or whether individual inclusion calls were scientifically correct. Its evidentiary value relies entirely on the quality of underlying selection logs.

Complete the diagram only after reconciling the primary screening ledger, full-text retrieval logs, exclusion accounts, and report-to-study maps. Populating blank boxes prematurely encourages reverse-engineering—authors manipulate numbers to force balance rather than tracking what occurred for each record. The correct sequence requires event logging, unit reconciliation, narrative explanation of legitimate discrepancies, and final diagram generation.

Select the appropriate template family

Official PRISMA 2020 flow-diagram templates are distributed under CC BY 4.0.³ Template choice depends on two parameters: review status (new vs. updated) and identification route (databases/registers only vs. addition of other methods). “Other methods” encompass website searches, organizational contacts, citation tracking, and expert correspondence that do not yield standard database exports.

Avoid choosing simpler templates merely because their arithmetic is easier. If secondary identification methods contributed candidate papers, incorporating the parallel branch ensures transparent reporting. Similarly, updated reviews must separate baseline studies from newly identified records. The chosen template must align with your methods section and source inventory.

Maintain clear distinctions between records, reports, and studies

Report counts do not always equal study counts. Maintain a report-to-study linkage ledger prior to populating final inclusion boxes.

PRISMA 2020 explicitly defines records, reports, and studies because analysis units shift as full-text data becomes available.² Cochrane guidelines similarly emphasize that studies—not individual papers—are the primary unit of synthesis, requiring multi-publication trials to be collated prior to data extraction.^4,5 Database records and registry citations represent duplicates when they refer to identical papers. However, a protocol, conference abstract, and full journal paper from a single trial represent distinct reports of one study rather than duplicate records to discard.

Populate diagram stages sequentially

1. Identification: Log search hits by database before and after deduplication. Track automation removals and pre-screening removals separately.
2. Screening: Record total unique records screened alongside title/abstract exclusions.
3. Retrieval: Enter reports sought for full-text evaluation, separating unretrieved papers from retrieved full texts.
4. Eligibility: Assign mutually exclusive primary exclusion reasons for full-text exclusions, ensuring their sum equals total excluded reports.
5. Inclusion: Record included studies and included reports separately. Do not force these numbers to match.

Populate every stage directly from primary decision logs. Search logs supply identification counts; deduplication records supply pre-screening removals; screening ledgers supply retrieval totals; full-text exclusion logs supply primary reason categories; and study-report maps supply included study and report totals. Documenting data provenance prevents platform exports from altering counts.

Enter zero explicitly when an event was evaluated but yielded no hits; do not leave blank fields. Blank entries signify incomplete screening, whereas explicit zeros provide audit data about your search pathway.

Document non-database identification methods transparently

When incorporating website searches, organizational reports, citation tracking, or expert correspondence, select templates featuring the “other methods” branch. Track each source independently. Hit counts from non-database methods may not equal reports sought for retrieval because screening or deduplication occurs prior to full-text requests.

Secondary identification methods yield records, reports, or study leads depending on execution. Citation tracking identifies citation records; organizational website browsing yields reports; direct correspondence reveals ongoing trials. Document the observed unit rather than reclassifying all hits as database records. Explain custom transitions in figure notes.

Separate baseline cohorts in updated systematic reviews

Select an updated-review template and report baseline studies separately from new search hits. Ensure total included studies equal baseline inclusions plus newly identified studies, documenting any study reclassifications.

An updated review differs from a new review with expanded inclusion totals. Baseline study cohorts may require re-evaluation if companion papers surface, eligibility criteria undergo amendment, or missing data becomes available. Document reclassifications explicitly to clarify how new searches impact baseline findings.

Distinguish unretrieved reports from full-text exclusions

Unretrieved full texts were never evaluated against eligibility criteria and belong in the “reports not retrieved” box. Full-text exclusion reasons apply exclusively to retrieved reports evaluated against criteria.

Log retrieval efforts in detail: interlibrary loan requests, publisher contacts, request dates, and author responses. If unretrieved papers surface later, update their status and proceed to full-text appraisal. Assigning speculative exclusion reasons to unretrieved papers violates transparency guidelines.

Understand arithmetic validation boundaries

Establish arithmetic validation checks prior to generating flow records. For database searches, identified hits minus removals must equal screened records; screened records minus exclusions equal reports sought; reports sought minus unretrieved papers equal assessed reports; and primary exclusion reasons must sum to full-text exclusions. Discrepancies warrant ledger audits.

Included study totals rarely equal simple subtraction remainders of assessed reports minus excluded reports. Subtractions occur at the report level, whereas inclusion reporting captures both reports and studies. Study-report linkage ledgers bridge this gap, explaining why report totals naturally exceed study totals.

Audit software exports before finalizing flow diagrams

Screening platforms handle duplicates, conflicts, linked reports, and workflow transitions differently. Document export dates, applied filters, and software versions. Verify platform exports against primary decision logs before transferring counts.

Open-source tools like the PRISMA2020 R package and Shiny app streamline diagram generation.⁶ However, digital tool usage does not replace verifying template selection, data labels, and imported counts. Archive generation scripts and source files to ensure reproducibility. Avoid labeling diagrams as “PRISMA compliant” without verifying underlying data accuracy.

Apply review-specific reporting extensions

PRISMA 2020 provides core reporting guidelines, but specialized synthesis types require extensions. Scoping reviews should follow PRISMA-ScR guidelines, and diagnostic test accuracy syntheses should incorporate PRISMA-DTA standards.^7,8 JBI evidence syntheses should align with JBI manual standards rather than assuming intervention review layouts apply universally.⁹ State chosen reporting frameworks explicitly.

Flow diagram adaptations should preserve core functions: demonstrating how records entered, moved through screening, and reached final inclusion. Avoid adding decorative complexity or omitting required count categories.

Define primary exclusion hierarchies before aggregating counts

PRISMA 2020 requires reporting full-text exclusions alongside primary exclusion reasons. Construct reason categories from operational eligibility criteria rather than generic templates. Categories must be mutually exclusive, consistently applied, and sufficiently specific to explain evidence boundaries.

Because papers may fail multiple criteria, establish a primary exclusion hierarchy in your protocol. Assign a single primary reason for reporting while logging all secondary failed criteria in decision ledgers. For example, a team may prioritize “ineligible study design” as primary even if population criteria also failed. PRISMA requires transparent exclusion reporting but leaves hierarchy definitions to review protocols.²

Following conflict resolution, verify that total excluded full texts match the sum of primary exclusion reasons. Resolve discrepancies at the case level without deleting secondary audit evidence.

Reconcile changes of unit using a practical example

Consider a review where database searches yield 1,420 records. Deduplication removes 220 records, automation filters remove 10 records, and manual checks remove 5 records, leaving 1,185 records for screening. Title/abstract screening excludes 1,000 records, leaving 185 reports sought. Seven reports cannot be retrieved, leaving 178 full texts assessed. If 142 reports are excluded after full-text evaluation, the report-level remainder is 36 eligible reports.

The review may ultimately report 31 included studies if five trials contributed two companion reports each. Report 36 as included reports and 31 as included studies. The study-report map explains this difference cleanly. Forcing included reports to equal 31 distorts document counts and discards five eligible papers from the audit trail.

If exclusion reasons total 144 while excluded reports equal 142, do not alter the report count. Audit the decision log. Two reports were likely double-counted across criteria or unresolved conflicts were logged prematurely. Resolving case-level entries restores both audit trail integrity and diagram accuracy.

Report automated screening rules transparently

PRISMA 2020 includes dedicated boxes for records marked ineligible by automation tools prior to screening.² Use this category only when software automatically excluded records without human review. Document software name, version, tasks, input data, threshold rules, calibration procedures, human verification checks, and restored records. Using machine learning to prioritize screening order differs from automated exclusion.

Automation also supports duplicate detection, language filtering, report clustering, or active-learning stopping rules. Detail these technical procedures in methods sections. A populated automation box does not guarantee rule sensitivity or safety.

If no records underwent automated exclusion, report zero explicitly rather than describing software that merely supported human screening. This maintains clear distinctions between automated filtering and workflow optimization.

Freeze a fully explainable selection account

Generate final flow diagrams from a dated selection dataset after resolving conflicts, retrieval statuses, exclusion reasons, and study-report links. Have a second reviewer trace diagram counts back to primary ledgers and verify alignment with methods sections, exclusion tables, and included-study lists.

A robust flow diagram addresses three core requirements: selecting official template structures, reporting data at correct analysis units, and explaining legitimate count discrepancies. Preserve these standards across your manuscript. The diagram serves as a concise, reproducible summary of an auditable selection method.

Methodological Scope & Limitations

This guide details core PRISMA 2020 flow structures for new and updated systematic reviews. Specialized synthesis types may require extension checklists. Always follow applicable extension guidelines, protocol specifications, and journal instructions.

Frequently Asked Questions