How to build publication-ready PRISMA 2020 flow diagrams — from Word templates to AI-automated generation — and why getting this right is the difference between a desk rejection and a published paper.
If you’re conducting a systematic review or meta-analysis, you’ve almost certainly encountered the PRISMA flow diagram. It’s the standardized flowchart that documents every step of your study selection process — from the initial database search through screening, eligibility assessment, and final inclusion.
For many researchers, creating this diagram is one of the most frustrating parts of the review process. You’ve spent months searching databases, screening thousands of titles, and reading hundreds of full texts. Now you need to translate all of that work into a single, precisely formatted figure — with exact counts at every stage, exclusion reasons itemized, and a layout that satisfies both the PRISMA guidelines and your target journal’s formatting requirements.
This guide covers everything you need to know: what a PRISMA flow diagram is, why it matters, how the 2020 update changed the standard, step-by-step instructions for creating one, the tools available to help, and how CoChat’s new PRISMA flow diagram feature eliminates the manual busywork entirely.
Already have your counts? Use our free PRISMA flow diagram generator: type the numbers and it draws the PRISMA 2020 diagram and does the arithmetic. Download PNG or SVG, no sign-up.
What Is a PRISMA Flow Diagram?
A PRISMA flow diagram is a standardized visual representation that documents the flow of information through the different phases of a systematic review. It maps out the number of records identified, included, and excluded — and the reasons for each exclusion.
PRISMA stands for Preferred Reporting Items for Systematic Reviews and Meta-Analyses. It’s a reporting guideline — not a methodology for conducting reviews, but a standard for reporting them. The flow diagram is one component of the broader PRISMA framework, which also includes a 27-item reporting checklist and an abstract checklist.
The PRISMA statement was originally published in 2009 and quickly became the gold standard for systematic review reporting. It has been cited over 60,000 times, endorsed by nearly 200 journals and systematic review organizations, and adopted across disciplines from medicine and public health to education, engineering, and the social sciences.
Why Does It Exist?
Systematic reviews are only useful if readers can assess their trustworthiness. The flow diagram exists to make the study selection process transparent and reproducible. Without it, a reader has no way of knowing:
- How many studies were initially identified
- How many were removed as duplicates
- How many were excluded at title/abstract screening versus full-text review
- What specific reasons led to each exclusion
- Whether the final included set is the result of a rigorous, unbiased process — or a cherry-picked selection
This transparency isn’t academic vanity. Systematic reviews inform clinical guidelines, policy decisions, and funding allocations. A poorly reported review can mislead treatment decisions for millions of patients. The flow diagram is a safeguard against that.
What Changed in PRISMA 2020?
In 2021, the PRISMA statement received its first major update since the original 2009 version. The PRISMA 2020 statement (Page et al., 2021) reflected a decade of advances in systematic review methodology, terminology, and technology.
The flow diagram changed in several significant ways:
1. Three Phases Instead of Four
The original 2009 flow diagram used four labeled phases: Identification, Screening, Eligibility, and Included. PRISMA 2020 simplifies this to three: Identification, Screening, and Included. The old “Eligibility” step is now folded into the Screening phase as the “reports assessed for eligibility” box.
2. Separate Tracking for Databases and Registers
The Identification phase now explicitly separates records from databases (like PubMed, Embase, Scopus) and registers (like ClinicalTrials.gov, CENTRAL, ICTRP). The 2009 version lumped these together, which obscured important differences in how records were sourced.
3. Records Removed Before Screening
PRISMA 2020 adds a new box for records removed before screening — including duplicates, records removed by automation tools, and records marked as ineligible by other means. This acknowledges that modern review workflows often use software to strip duplicates and flag obviously irrelevant records before human reviewers ever see them.
4. Four Template Variants
Instead of a single template, PRISMA 2020 provides four flow diagram variants:
| Template | Use Case |
|---|---|
| New review, databases only | Original systematic reviews that searched databases and registers exclusively |
| New review, databases + other sources | Original reviews that also used citation searching, websites, organizations, or other grey literature methods |
| Updated review, databases only | Updates to previously published reviews using databases and registers only |
| Updated review, databases + other sources | Review updates that also included non-database sources |
Choosing the right template is the first decision you’ll make. If you searched only PubMed, Embase, and Cochrane CENTRAL, you need the “databases only” version. If you also hand-searched reference lists, contacted experts, or searched Google Scholar, you need the “databases + other sources” variant, which adds a parallel “Other methods” arm to the diagram.
5. Reports vs. Studies Distinction
PRISMA 2020 distinguishes between records (individual database entries), reports (published documents), and studies (the actual research). A single study may produce multiple reports across different journals. The flow diagram now tracks this distinction explicitly in the lower phases.
Why PRISMA Compliance Matters More Than Ever
The numbers tell the story. As of 2019, approximately 80 systematic reviews were published per day — a 20-fold increase from the year 2000. That volume has only accelerated since.
At the same time, compliance with PRISMA remains poor. A 2024 meta-epidemiological study found that among systematic reviews published after the PRISMA 2020 statement, only 30% even cited the updated guideline. Of those that did, the average adherence to the expanded checklist was just 42.64% — and not a single review achieved full compliance. The weakest areas were the methods and results sections, with adherence as low as 25.55% and 40.57% respectively.
This matters because journals are increasingly enforcing PRISMA compliance as a condition of publication. The PRISMA statement has been endorsed by nearly 200 journals, and many top-tier venues — including The BMJ, PLOS Medicine, The Lancet, and Systematic Reviews — require PRISMA checklists and flow diagrams at submission. A missing or non-compliant flow diagram can trigger a desk rejection before peer review even begins.
For researchers, this creates a paradox: the flow diagram is simultaneously one of the most important components of a systematic review submission and one of the most tedious to produce correctly. Getting the counts wrong, choosing the wrong template variant, or omitting required exclusion reasons can delay publication by months.
How to Create a PRISMA Flow Diagram: Step by Step
Whether you’re using a Word template, an online generator, or an automated tool, the process follows the same logical sequence. Here’s how to do it right.
Step 1: Choose the Correct Template
Before anything else, determine which of the four PRISMA 2020 templates applies to your review:
- Did you search only databases and registers? → Use the “databases only” template
- Did you also search websites, reference lists, or contact experts? → Use the “databases + other sources” template
- Is this an update of a previously published review? → Use the “updated” variant of whichever template above applies
Getting this wrong is a common mistake. If you used Google Scholar or hand-searched reference lists, you need the “other sources” template — even if those methods only contributed a handful of records.
Step 2: Record Identification Numbers
For each database you searched, record the number of results returned. Enter the total across all databases (including duplicates) in the first box of the diagram.
Best practice: Note individual database counts as annotations. For example:
Databases (n = 2,847) PubMed (n = 1,203), Embase (n = 1,044), Scopus (n = 600)
If you searched trial registers separately, record those counts in the “Registers” box.
Step 3: Document Records Removed Before Screening
This is the new PRISMA 2020 step. Calculate and record:
- Duplicate records removed — using your citation manager (EndNote, Zotero, Mendeley) or deduplication software
- Records removed by automation tools — if you used machine learning or text mining to flag obviously irrelevant records
- Records marked as ineligible by other reasons — any other pre-screening removals
Step 4: Calculate Records Screened
Subtract the records removed in Step 3 from your total identified records. This gives you the number of records that went into title/abstract screening.
Step 5: Record Title/Abstract Screening Exclusions
After screening titles and abstracts, record the number excluded. You can optionally add exclusion reasons at this stage, but they’re not required until full-text screening.
Step 6: Calculate Reports Sought for Retrieval
Subtract the excluded records (Step 5) from the screened total (Step 4). This is the number of full-text articles you attempted to obtain.
Step 7: Record Reports Not Retrieved
Document any full-text articles you couldn’t obtain despite interlibrary loan requests and other retrieval efforts. Be specific about why they were unavailable.
Step 8: Record Full-Text Eligibility Assessment
Subtract reports not retrieved (Step 7) from reports sought (Step 6). This is the number of articles whose full text you actually read and assessed against your inclusion/exclusion criteria.
Step 9: Document Full-Text Exclusion Reasons
This is the most important exclusion box in the diagram. For every article excluded at full-text review, categorize the reason. Common categories include:
- Wrong study design
- Wrong patient population
- Wrong intervention or exposure
- Wrong comparator
- Wrong outcome
- Duplicate report of an included study
- Insufficient data
Each excluded article should be counted once, even if it meets multiple exclusion criteria. Use a hierarchical approach: assign the first applicable reason in your pre-defined exclusion hierarchy.
Step 10: Calculate Final Included Studies
Subtract full-text exclusions (Step 9) from full-text assessments (Step 8). Report both the number of studies included and the number of reports (since one study may have multiple publications).
If your review includes a quantitative synthesis (meta-analysis), add a sub-box showing how many of the included studies contributed data to the meta-analysis.
Step 11: Validate Your Numbers
Before finalizing, check the arithmetic:
- Records screened = Records identified − Records removed before screening
- Reports sought = Records screened − Records excluded at title/abstract
- Reports assessed = Reports sought − Reports not retrieved
- Studies included = Reports assessed − Reports excluded at full text
If the numbers don’t add up, you have an error somewhere in your documentation. Fix it now — reviewers and editors will catch it.
Our free PRISMA diagram generator does this arithmetic for you: it fills in the derived boxes from your counts and flags any box where the numbers can’t add up.
Current Tools for Creating PRISMA Flow Diagrams
Until recently, researchers had limited options for producing PRISMA flow diagrams. Here’s what’s currently available:
Word and PowerPoint Templates
The official PRISMA website provides downloadable Word templates for all four diagram variants. These are the most commonly used option, but they come with significant limitations:
- Manual data entry into text boxes
- Formatting headaches when numbers overflow box boundaries
- No validation of count consistency
- Difficult to update when screening numbers change during the review process
- Poor export quality for journal submissions
The PRISMA2020 R Package and Shiny App
The most widely cited software solution is the PRISMA2020 R package by Haddaway, McGuinness, and Pritchard (2022), which includes a web-based Shiny app hosted at eshackathon.org. It generates interactive HTML diagrams with clickable links and tooltips.
Strengths: Publication-quality output, interactivity, R integration for reproducible workflows.
Limitations: Requires R installation (or reliance on the hosted Shiny app, which can have cold-start delays), depends on Graphviz/DiagrammeR for layout, and the output is difficult to embed outside of R Markdown environments.
Python Packages
Several Python packages have emerged, including prisma-flow (a lightweight SVG generator) and prisma-flow-diagram (which uses matplotlib). These are useful for researchers working in Python-based analysis pipelines but require programming knowledge and produce static images.
Online Generators
A growing number of web-based tools offer point-and-click PRISMA diagram creation:
- SciFig — AI-assisted generator with multiple publication styles and editable SVG/PPTX export
- prismaflowdiagramgenerator.com — Free tool with live preview and all four templates; PNG export (free version watermarked)
- ConceptViz — Offers both precise mode (count entry) and AI illustration mode for presentations
- Lumina — Simplified generator built into a broader systematic review platform
- CoChat’s PRISMA generator — Free, no sign-up; live preview as you type, flags counts that can’t add up, PNG and SVG export without a watermark
These tools, our free generator included, solve the formatting problem but share a fundamental limitation: they are all disconnected from the actual search and screening process. You complete your review in one set of tools (databases, citation managers, screening software), then manually re-enter all your numbers into a separate diagram tool. Every number is a manual transcription — and every transcription is a chance for error.
Creating a PRISMA Flow Diagram in CoChat
This is where CoChat takes a fundamentally different approach. Instead of treating the PRISMA flow diagram as a separate deliverable you produce after your review, CoChat integrates it directly into the research workflow — and auto-populates it from the searches you’ve already run.
How It Works
CoChat’s PRISMA flow diagram feature is built as an editable artifact — just like the literature review tables, flashcard decks, and other research tools already available in the platform. It consists of two components:
1. The PRISMA Flow Diagram (prisma_flow)
An interactive, editable flow diagram that covers all four PRISMA 2020 template variants:
- New review, databases only
- New review, databases + other sources
- Updated review, databases only
- Updated review, databases + other sources
The diagram renders as a clean, deterministic SVG — no layout engine needed, no external dependencies. You can edit every box directly through a form-based interface with live preview. When you’re done, export as SVG, PNG, or print-ready PDF.
2. The PRISMA Reporting Checklist (prisma_checklist)
The full canonical 27-item (42-row) PRISMA 2020 reporting checklist, auto-seeded with the official item text for every sub-item (10a/b, 13a–f, 16a/b, 20a–d, 23a–d, 24a–c). Mark each item as Reported, Not Reported, or N/A, with fields for the location where each item is reported and additional notes.
A progress meter tracks your overall compliance. Filters let you focus on unreported items. Collapsible per-item guidance explains what each checklist item requires. Export as CSV or Markdown for journal submission.
The Auto-Fill: Why This Changes Everything
Here’s the feature that no other PRISMA tool offers: automatic population of the Identification band from your actual searches.
When you use CoChat’s multi-database paper search to run your systematic review searches — across PubMed, Semantic Scholar, CrossRef, arXiv, Europe PMC, and ClinicalTrials.gov — CoChat captures the raw per-database counts before deduplication and truncation.
A per-chat search ledger accumulates these counts across every search you run. When you create a PRISMA flow diagram, CoChat reads this ledger and auto-seeds the Identification phase:
- Database sources — with per-database counts (PubMed: 340, Semantic Scholar: 892, CrossRef: 1,203 …)
- Duplicate records removed — calculated automatically from the deduplication step
You don’t re-enter a single number. The top two boxes of your flow diagram — the ones that require the most tedious bookkeeping — are already filled in, with exact counts that match your actual search history.
The Screening and Included phases remain for you to fill in, because those represent human judgment decisions that the search tool can’t observe. But the identification phase — which is entirely mechanical — is handled for you. And any explicit counts you enter will override the auto-seeded values, so you always have full control.
A Workflow That Eliminates Transcription Error
Here’s what the complete workflow looks like in CoChat:
1. Run your searches Use CoChat’s paper search to query your target databases. The search ledger records raw counts in the background.
2. Build your literature review table As results come in, organize them in a literature review table with AI-powered citation verification against CrossRef and Semantic Scholar.
3. Screen and assess Work through your records, noting exclusions and reasons. The literature review table tracks your screening decisions.
4. Generate the PRISMA flow diagram Ask CoChat to create a PRISMA flow diagram. The Identification band auto-populates from your search ledger. Fill in the Screening and Included counts from your review notes.
5. Complete the reporting checklist Generate the PRISMA 2020 checklist. Mark each item as reported, with page/section locations. The progress meter shows your compliance at a glance.
6. Export everything The flow diagram exports as Figure 1 (SVG/PNG). The checklist exports as the required PRISMA supplement (CSV/Markdown). Both can be shared as read-only links with download capability intact.
Soft Validation, Not Hard Blocks
One design choice worth noting: CoChat’s PRISMA flow diagram includes soft arithmetic warnings — it’ll flag if your numbers don’t add up (e.g., if records screened exceeds records identified minus duplicates removed). But it won’t block you from saving.
This is intentional. The PRISMA 2020 guidelines themselves permit independent reporting of some phases, and real-world data sometimes has legitimate inconsistencies (for example, when some records are identified from “other methods” sources that don’t go through the standard deduplication pipeline). A warning tells you to double-check; a hard block would prevent valid diagrams from being created.
What Good PRISMA Compliance Looks Like
Now that you understand the mechanics, let’s look at what separates adequate PRISMA reporting from excellent PRISMA reporting.
The Minimum Standard
At a bare minimum, your PRISMA flow diagram should:
- Use the correct 2020 template variant for your review
- Include exact counts in every box — no empty boxes, no approximations
- List specific exclusion reasons at the full-text eligibility stage with counts per reason
- Distinguish between studies and reports in the Included phase
- Match the numbers reported in your manuscript text
What Elevates a Flow Diagram
The best systematic reviews go further:
- Annotate database counts individually — not just the total, but per-source breakdowns
- Document automation decisions — if you used machine learning to remove irrelevant records, state the tool and the count
- Link exclusion reasons to eligibility criteria — your exclusion categories should map directly to your PICO/PECO criteria
- Include the checklist supplement — the flow diagram is Item 16b of the PRISMA checklist; submitting the complete checklist demonstrates thoroughness
- Version the diagram — if you update your searches during revisions, regenerate the diagram with updated counts and note the search dates
Common Mistakes to Avoid
- Using the 2009 template — many journals now specifically require PRISMA 2020; submitting the old version signals carelessness
- Miscounting duplicates — citation managers sometimes under-count or over-count duplicates depending on settings; cross-check with a second tool
- Excluding a study for multiple reasons — each excluded article should appear only once in your exclusion reasons, under the first applicable category
- Forgetting the “other methods” arm — if you reference-chased or contacted experts, those records should appear in the parallel arm, not the main database arm
- Omitting the records removed before screening box — this is new in 2020 and is frequently left empty or missing entirely
Who Needs PRISMA Flow Diagrams?
PRISMA was originally developed for health sciences, but its adoption has expanded dramatically:
Core Users
- Medical and clinical researchers conducting systematic reviews of interventions, diagnostics, or prognosis
- Public health professionals synthesizing evidence for policy recommendations
- Cochrane review authors (PRISMA compliance is essentially mandatory)
Growing Adoption
- Social sciences — psychology, education, sociology, economics
- Environmental sciences — conservation biology, climate research, toxicology
- Engineering and computer science — software engineering, HCI, AI/ML systematic mapping studies
- Graduate students — many thesis committees now require PRISMA-compliant reporting for literature review chapters
If you’re writing any kind of systematic review, scoping review, or evidence synthesis — in any discipline — PRISMA is either required or strongly recommended by your target venue.
The Future of Systematic Review Reporting
The trajectory is clear: systematic reviews are growing in volume and importance, reporting standards are tightening, and the tools for producing compliant reviews are moving from manual templates toward integrated, automated platforms.
PRISMA 2020 already acknowledges the role of automation in its expanded “records removed before screening” box — a nod to the text mining and machine learning tools increasingly used in modern reviews. The next frontier is closing the gap between the doing of systematic review work and the reporting of it.
That’s the fundamental shift that CoChat’s PRISMA feature represents. Instead of doing your searches in PubMed, managing your records in EndNote, screening in Covidence, and then manually reconstructing the whole process in a Word template or Shiny app — you work within a platform that tracks the process as it happens and generates the reporting artifacts automatically.
The flow diagram stops being a tedious after-the-fact chore and becomes a live byproduct of doing the work.
Get Started
If you’re ready to try it:
- Start a systematic review in CoChat — run your searches using multi-database paper search
- Ask for a PRISMA flow diagram — CoChat will generate one with the Identification phase auto-populated
- Complete the Screening and Included phases — fill in your human-judgment counts
- Generate the reporting checklist — track your compliance across all 27 PRISMA items
- Export and submit — SVG/PNG for the flow diagram, CSV/Markdown for the checklist
No R installation. No Graphviz dependencies. No manual number transcription. Just the PRISMA reporting your review deserves, built directly into the research workflow.
CoChat’s PRISMA flow diagram and reporting checklist tools are available now to all users. Visit cochat.ai to start your next systematic review.
