The flow diagram is the most-scrutinised figure in any systematic review: it declares, in four stages, exactly what happened to every record you found. Reviewers check its arithmetic before they read your findings — a diagram whose numbers don't add up discredits the whole review. Here's how to build the PRISMA 2020 version correctly.
The four stages and their boxes
1. Identification
Records found, per source: each database with its count (Scopus n = 412, Web of Science n = 356, PubMed n = 289…), plus registers if searched. PRISMA 2020 adds a parallel column for other methods (citation searching, hand-searching, websites) — keep it separate all the way down. Then the removals before screening: duplicates, plus any records removed by automation tools.
2. Screening (title/abstract)
Records screened → records excluded. This is the fast pass on titles and abstracts; you do not itemise exclusion reasons here. Screening in pairs with a conflict-resolution rule is the methodological standard — state it in the text.
3. Eligibility (full text)
Reports sought for retrieval → reports not retrieved (paywalled/unavailable — be honest) → reports assessed for eligibility → reports excluded, with reasons and counts (wrong population n = 14, wrong outcome n = 9, not empirical n = 7…). These reasons come straight from your protocol's eligibility criteria — reviewers cross-check them.
4. Included
Studies included in the review, and (where applicable) studies included in the meta-analysis — the two numbers often differ, since not every included study has poolable data.
The arithmetic rules that catch people
- Every stage must reconcile: screened = identified − removed-before-screening; sought = screened − excluded; and so on down. Sum your per-reason exclusions and check they equal the box total.
- Count records (search hits) at the top but reports and studies at the bottom — one study can have multiple reports (a thesis + a paper), and PRISMA 2020 names the boxes accordingly.
- The other-methods column merges only at the included stage — don't fold it into the database column early.
The diagram is unbuildable retrospectively from memory. Log counts at every decision point from day one — reference-manager folders or a screening tool give you the audit trail, and a simple spreadsheet of exclusion reasons at full-text stage writes the diagram for you. The official PRISMA flow-diagram templates (Word/R Shiny) are free at prisma-statement.org — use them rather than redrawing boxes.
Beyond the diagram
The diagram is 1 of 27 items on the PRISMA 2020 checklist — you'll also need the full search strings per database (appendix), the protocol/registration statement, and the risk-of-bias assessment. The complete workflow, from protocol to synthesis, is in our PRISMA systematic review guide and the what is PRISMA explainer.
A systematic review stands or falls on process discipline months before writing begins. Our systematic literature review mentoring and PRISMA support set up the protocol, screening workflow and audit trail so the diagram assembles itself.
Frequently asked
Is the PRISMA flow diagram mandatory for a systematic review?+
Effectively yes — journals and examiners treat it as non-negotiable evidence of a transparent process, and PRISMA 2020 is the current version to follow. Even scoping reviews use an adapted version (PRISMA-ScR).
Do I report exclusion reasons at the screening stage?+
No — itemised reasons with counts belong at the full-text eligibility stage only. At title/abstract screening you report a single excluded total; listing reasons there is a common formatting error reviewers flag.
What's the difference between records, reports and studies in PRISMA 2020?+
Records are search hits (one per database result); reports are retrieved documents (a paper, thesis or preprint); studies are the underlying investigations. Multiple records can point at one report, and multiple reports at one study — PRISMA 2020's box labels track this deliberately.
phdguide's mentors are senior academics, former supervisors, statisticians and publication specialists with 25+ years' average experience guiding MBA, MPhil and PhD scholars from topic to viva.
Ethical, compliant guidance: We provide academic support, mentoring, analysis, editing and structuring — not authorship. Your work stays compliant with university policies.