A PRISMA flow diagram is not a design task. It is a reconciliation task. The diagram is only as trustworthy as the screening log beneath it, and the log is only as trustworthy as the search and selection process it records. If you are revising a manuscript, closing a project, handing off to a coauthor, or responding to a peer reviewer who asked where the numbers came from, the fastest path to a defensible diagram is to rebuild the count trail from your own records rather than to reconstruct it from memory.
This guide walks through that rebuild in order: diagnosing what state your records are in, fixing the artifacts you need, producing the counts, and knowing when the diagram is ready to submit. PRISMA 2020 supplies the reporting checklist and flow-diagram templates for original and updated systematic reviews, including transparent reporting of search, selection, and synthesis methods. That template is your target; your screening log is your source.
What does your screening log actually need to contain?
Before you draw anything, check whether your log can answer the questions the diagram will ask. A usable log records, for every record, where it came from, whether it was a duplicate, whether it was screened at title/abstract, whether it was assessed at full text, and why it was excluded at each stage.
Cochrane's guidance on searching and selecting studies is a useful audit standard here. It recommends planning searches from eligibility criteria, using multiple sources, combining free-text terms with subject headings, documenting the process, deduplicating records, screening in stages, linking multiple reports of one study, and rerunning searches before publication. Each of those recommendations maps onto a number or a box in the diagram.
Run a quick diagnostic against your own files:
- Source-level counts. Can you state how many records came from each database or source? If your exports were merged without source labels, you cannot populate the identification boxes.
- Duplicate handling. Do you know how many records were removed as duplicates, and by what rule? Cochrane's emphasis on deduplication means this should be a recorded step, not an assumption.
- Stage boundaries. Can you separate records screened at title/abstract from reports assessed at full text? Staged screening is what makes the two middle tiers of the diagram meaningful.
- Exclusion reasons. Are full-text exclusions coded with reasons, or only marked "no"? The diagram's exclusion box needs reasons, not just a total.
- Study-versus-report linkage. Do you know which reports belong to the same study? Cochrane recommends linking multiple reports of one study, and PRISMA templates distinguish records, reports, and studies.
If two or more of these fail, pause the diagram and repair the log first. A diagram built on a broken log will not survive a coauthor's question, let alone peer review.
Which artifacts should you assemble before counting?
Gather these in one working folder so the diagram and its audit trail stay together:
- The search record. The date each search was run, the sources searched, and the terms used. Cochrane's recommendation to rerun searches before publication means you may need more than one dated entry.
- Raw export files. The unedited result sets from each source, kept as exported.
- The deduplication record. Which records were removed and on what basis.
- The screening log itself. One row per record, with stage outcomes and exclusion reasons.
- The eligibility criteria. The written criteria your screening applied, since Cochrane ties search planning to eligibility criteria.
- The PRISMA 2020 template you are completing. Choose the original-review or updated-review version as appropriate.
This is also the point where a shared workspace helps. If your team is rebuilding counts across several people, keeping the search record, exports, and screening log in one project area prevents the version drift that produces mismatched numbers. WisPaper's Research Projects area is a place to create and hold that project context while you reconcile the trail.
How do you turn log rows into the diagram's counts?
Work bottom-up from the log, not top-down from the template. The template tells you which boxes exist; the log tells you what goes in them.
Step 1: Count identifications by source
Group your raw exports by source and count records per source. Keep the per-source numbers visible even if the template shows a combined total, because reviewers and coauthors often ask for the breakdown.
Step 2: Count duplicates removed
Apply your recorded deduplication rule and count removals. If the rule was applied by a tool, confirm the count against the log rather than trusting a summary figure.
Step 3: Count records screened and records excluded at title/abstract
Total screened minus exclusions at this stage gives you the reports sought for retrieval. If these two numbers do not reconcile, the log has a gap.
Step 4: Count reports assessed for eligibility and exclusions with reasons
This is where coded exclusion reasons pay off. Tally each reason category so the exclusion box can be populated with reasons rather than a bare number.
Step 5: Count studies included, and reconcile reports to studies
Because multiple reports can describe one study, the included-studies count may differ from the included-reports count. Record both and note the linkage.
Step 6: Reconcile every tier
For each transition, confirm that the incoming count equals the outgoing count plus exclusions. Any tier that does not balance points to a specific log defect you can fix.
What does a worked example look like?
Suppose a team is completing a review and has kept a screening log with source labels, duplicate flags, stage outcomes, and coded full-text exclusion reasons. Their reconciliation proceeds like this:
- They count raw records per source from the untouched exports and record the per-source totals.
- They apply the recorded deduplication rule and count removals, then subtract to get the records screened.
- They tally title/abstract exclusions from the log and subtract to get reports sought for retrieval.
- They tally full-text assessments and group the coded exclusion reasons into categories for the exclusion box.
- They check the included reports against the study linkage record, since Cochrane's guidance on linking multiple reports of one study means the two counts can legitimately differ.
- They confirm each tier balances before transferring numbers to the PRISMA 2020 template.
The team never estimates a number. Every figure traces to a row, a flag, or a coded reason in the log, which is exactly the audit trail a coauthor or reviewer can follow.
Where does literature search fit into the trail?
If you are starting a new review rather than closing one, the diagram's top boxes depend on a search you can document. Cochrane's recommendation to plan searches from eligibility criteria and to combine free-text terms with subject headings gives you the structure for that record.
When you run searches, keep the query and the date with the export. WisPaper's scholarly search interface is where you can find academic papers, and the Scholar Search entry point explains the paper-finding capability for readers new to the tool. Save the papers you retrieve into a library so the identification stage has a stable record: My Library lets you save or upload papers and organize them. That library then becomes the working set your screening log describes.
For longer reviews where search, screening, and synthesis run over weeks, WisPaper's Scholar Agent area supports literature review and related research tasks. Use it to keep the review work moving while your screening log remains the authoritative count record.
How should you handle citation records during screening?
Screening often surfaces citation problems that affect the reference list rather than the flow diagram, but they surface at the same moment and are worth resolving while the trail is open. If your BibTeX contains entries you cannot verify, TrueCite checks BibTeX against real academic databases to flag hallucinated references. Catching those during screening prevents a reference-list correction after the diagram is finalized.
If you need to interrogate the papers already in your library, Library QA lets you ask questions based on papers uploaded to My Library or added from search results. That is useful when a full-text exclusion decision needs a second look at what a paper actually reports.
How do you keep the method trail auditable for coauthors and reviewers?
An auditable trail has three properties: it is dated, it is versioned, and it is traceable from any diagram number back to a source row.
- Dated. Record when each search ran and when screening stages closed. Cochrane's recommendation to rerun searches before publication means the trail should show the final search date relative to submission.
- Versioned. Keep the screening log and the diagram in the same project folder so a coauthor opening the project sees matching versions.
- Traceable. For each number in the diagram, note the log column or export file it came from. A one-page count reconciliation sheet is enough.
NIH's Scientific Rigor and Reproducibility guidance frames transparent methods, validation, and reproducibility as core parts of rigorous research practice. A count reconciliation sheet is a small, concrete way to meet that standard for the selection portion of your review.
What checklist should you run before finalizing the diagram?
Use this as a pre-submission gate. Each item should be answerable with a yes and a pointer to the artifact.
| Check | Evidence to point to |
|---|---|
| Per-source identification counts recorded | Raw exports, one per source |
| Duplicates removed and rule documented | Deduplication record |
| Records screened and title/abstract exclusions reconcile | Screening log stage columns |
| Reports assessed for eligibility counted | Screening log full-text column |
| Full-text exclusions coded with reasons | Reason codes in log |
| Included reports reconciled to included studies | Study linkage record |
| Every tier balances (in = out + excluded) | Count reconciliation sheet |
| Search dates recorded, including any rerun | Search record |
| Template version matches review type | PRISMA 2020 template |
| Diagram and log stored together | Project folder |
When is the diagram ready?
The diagram is ready when four conditions hold at once:
- Every tier balances. No transition has an unexplained gap.
- Every exclusion has a reason. The full-text exclusion box is populated from coded reasons, not a residual total.
- Reports and studies are reconciled. The included counts reflect the linkage between multiple reports and single studies.
- The trail is stored with the diagram. A coauthor or reviewer can move from any number to its source without asking you.
If any condition fails, the fix is in the log, not in the diagram. Correct the log, rerun the reconciliation, and update the diagram from the corrected counts.




