The PRISMA 2020 flow diagram, explained
What the PRISMA flow diagram is, what each box means, and how to fill it in without losing track of your numbers.
If you're writing a systematic review, a reviewer will eventually ask for your PRISMA flow diagram. It's the chart that shows how you got from thousands of search results down to the handful of studies you actually included, with the numbers dropping at each stage and a reason given for every exclusion.
PRISMA stands for Preferred Reporting Items for Systematic reviews and Meta-Analyses. It's not a method, it's a reporting standard: a checklist and a diagram that make your review transparent enough for someone else to follow and trust. The current version is PRISMA 2020.
What the diagram tracks
Read it top to bottom. You start with a big pile of records and end with the studies in your review. At each step, some records leave, and the diagram records how many and why. The whole point is that the numbers add up: anyone should be able to trace a single record's journey through your process.
The four stages
- Identification: how many records you found, broken down by source, plus how many duplicates you removed before screening.
- Screening: how many titles and abstracts you reviewed, and how many you excluded at this quick first pass.
- Eligibility: how many full-text articles you assessed, and how many you excluded after reading them in full, with reasons.
- Included: how many studies made it into your final synthesis, and how many into a meta-analysis if you ran one.
What changed in PRISMA 2020
The 2020 update replaced the 2009 version and reflects how people search now. It separates records found in databases from those found other ways, such as citation chasing or contacting authors. It also makes room to report records removed by automation tools before a human ever screened them. If you're following an older template you found online, check the date, because reviewers will notice the wrong one.
Filling it in without losing the thread
The diagram is easy in principle and fiddly in practice, because the numbers have to reconcile. Records screened minus records excluded must equal records assessed for eligibility, and so on down the chart. One miscount early and nothing balances. The trick is to log the count at the moment you make each decision, not reconstruct it from memory at the end.
- Record your raw counts per source before you merge anything.
- Log duplicates removed as a separate number, not folded into screening.
- Keep a running tally of exclusion reasons at the full-text stage.
- Check that each stage's maths reconciles with the one above it.
- Keep the figures somewhere they update as you screen, rather than counting twice.
This is exactly the kind of bookkeeping software should handle. When your screening and the diagram share the same underlying counts, the PRISMA figure stays correct on its own, and the number you report is the number you can defend.
Generate a PRISMA diagram from your review.
Try Deep Review →