How to make a PRISMA flow diagram

Updated 2026-10-27

Making a PRISMA flow diagram is mostly bookkeeping: keep accurate counts at every stage of your review, then place them in the right boxes. Follow these five steps for the complete process, from choosing a template to checking your arithmetic before you export. Then check your figure against the ten mistakes peer reviewers find most often.

Open the free generator

Step 1: Choose the right template

Pick the PRISMA 2020 flow diagram that matches your review. Is it a new review or an update of an earlier one? Did you search databases and registers only, or also websites, organisations and citation searching? The four official templates cover each case.

Step 2: Count as you go

Record numbers while you work, not afterwards. Note how many records each database or register returned, how many duplicates or ineligible records you removed before screening, how many records you screened and excluded, and how many reports you sought, could not retrieve, assessed and excluded (with a reason for each exclusion).

Step 3: Fill in the boxes

Enter each number in its box. In the free PRISMA flow diagram generator, the diagram updates live as you type. Name your actual databases (for example PubMed, Embase, CENTRAL) rather than using one generic count, and list exclusion reasons with their counts.

Step 4: Check the arithmetic

Every box should equal the one before it minus what was removed. If you identified 1,100 records and removed 260 duplicates, you should screen 840. Mismatches are the most common reason reviewers query a flow diagram. See the worked example for a full check.

Step 5: Export and insert

Export the diagram as PNG for quick use, SVG for a sharp, editable vector, or PDF for print. Insert it into your manuscript as Figure 1. For Word, see the guide to a PRISMA flowchart template for Word.

Ten common mistakes to avoid

Most problems with a PRISMA flow diagram are small, but peer reviewers spot them quickly. Here are the ten that come up most, why each one matters and how to fix it before you submit.

1. The numbers don't add up

This is the error reviewers find first. Each box should equal the box above it minus whatever left the flow at that step. Records identified minus records removed gives records screened, and so on down the page. Fix: recalculate every step from your screening log, not from memory. The worked examples show each subtraction.

2. Mixing up records, reports and studies

A record is a title or abstract in a database. A report is a document you read in full. A study is the underlying investigation, which may have several reports. Using "studies" all the way down the diagram makes the counts impossible to follow. Fix: count records at the top, reports in the middle, and give both studies and reports in the final box.

3. One total for all databases

PRISMA 2020 asks for the number of records from each database or register searched, where feasible. A single combined figure hides where the evidence came from and makes the search harder to reproduce. Fix: name each source and give its own count, for example MEDLINE, Embase and CENTRAL on separate lines.

4. No reasons for full-text exclusions

"Reports excluded (n = 64)" with nothing else doesn't meet the guideline. Readers need to see why reports were rejected to judge whether the criteria were applied fairly. Fix: list each reason with its own count, and make sure the counts add up to the total.

5. Counting a report under two reasons

A report can fail on population and on outcome at once. If you count it under both, the reasons add up to more than the number of reports excluded. Fix: put your reasons in a fixed order and record only the first one that applies.

6. Leaving out reports that couldn't be retrieved

Some full texts can't be obtained: the journal is unavailable, the abstract was never published in full, or the authors don't reply. These aren't exclusions on eligibility, and PRISMA 2020 gives them their own box. Fix: report them under "reports not retrieved", even when the number is zero.

7. Using the 2009 layout

The older diagram with four phases and a meta-analysis box is still widely copied from earlier papers. Journals that follow PRISMA expect the 2020 version. Fix: rebuild the figure on the current layout. See PRISMA 2020 vs 2009 for how to convert it.

8. Choosing the wrong template

There are four PRISMA 2020 templates. If you searched websites or checked reference lists but used the databases-only version, those records have nowhere to go. If you updated an earlier review but used the new-review version, the previously included studies disappear. Fix: pick the template that matches your methods. For updates, see the updated review guide.

9. The diagram disagrees with the text

The number of included studies in the abstract, the results section, the tables and the diagram must all match. They often drift apart after a late change or a search update. Fix: after any change, check the figure against the abstract and the first paragraph of the results.

10. A figure that's hard to read

Screenshots, stretched images and tiny text in crowded boxes all make a correct diagram look careless. Fix: export a proper image file, resize it from the corner so it keeps its proportions, and shorten long labels. The Word guide covers inserting it cleanly.

Checklist before you submit

• You used the PRISMA 2020 template that matches your review: new or updated, with or without other methods.
• Every step reconciles with the step above it.
• Each database and register is named with its own count.
• Duplicates, automation-tool removals and other removals are reported separately, and boxes with a count of zero show n = 0.
• Reports not retrieved are shown.
• Every full-text exclusion reason has a count, and the counts add up to the total.
• The final box gives studies and reports.
• The included-studies number matches the abstract, results and tables.
• The figure has a caption and cites the PRISMA 2020 statement.

The generator shows a running total of the records identified across your databases and registers as you type. That gives you the starting number to check every later box against.

Quick answers

How long does it take to make a PRISMA flow diagram?

If your counts are ready, a few minutes with a generator. Keeping accurate counts during screening is what takes the time.

How do I complete a PRISMA flow diagram?

Work through the diagram top to bottom: total the records from each database, register and other method for Identification; subtract duplicates and automation-excluded records; then work through Screening by recording exclusions at title/abstract and at full text, with a reason for each; and finish with the studies and reports you kept for Included. See the arithmetic check above for how the boxes should line up.

How do I draw a PRISMA flow diagram by hand or in PowerPoint?

Reproduce the three bands — Identification, Screening and Included — and give every box a plain black outline, matching the official template's boxes for records identified, duplicates removed, records screened and excluded, reports sought, not retrieved and assessed, reports excluded with reasons, and studies included. It's usually faster to build it in a generator and export an SVG, which opens directly in PowerPoint, Illustrator or Inkscape for further editing.

Should I show a box when the count is zero?

Yes. Keep the box and report n = 0. It shows the step was considered, for example that no reports were unretrievable.

Can I change the wording of the boxes?

You can adapt labels to fit your review, such as naming your databases or using "sources of evidence" in a scoping review. Keep the structure and the order of the boxes.

Source of the flow diagram wording: Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. doi:10.1136/bmj.n71. Licensed CC BY 4.0. This site is independent and not affiliated with the PRISMA or EQUATOR Network groups.