Winnow

The step before screening in a systematic review, in a few clicks.

  • Remove duplicatesCertain matches go; doubtful pairs are yours to decide
  • Exclude irrelevant studiesCase reports, reviews and more, with the evidence shown
  • Generate PRISMA diagramFlow diagram with counts filled in, in Word
  • Private by designRuns in your browser; files are never uploaded
ScopusPubMedCochrane .csv.nbib.ris Winnowmatch andcompare Unique records Doubtful pairs you decide

Sources (names are detected from the files; click a name to rename it. Amber means not recognised)

Exclude automatically applied after duplicates are resolved

A record is excluded only when its title, abstract or PubMed indexing states the study type explicitly. Nothing is deleted: excluded records stay in the Auto-excluded tab with the evidence, and you can restore any of them. Check them before you finalise.

How it works

1

Add your search results

Drop in the RIS, PubMed or CSV export from each database you searched.

2

Check the doubtful pairs

Winnow removes only certain duplicates. Uncertain pairs appear side by side with the differences highlighted, and you decide.

3

Exclude ineligible study types

Optional. Tick the study types your review excludes. Each excluded record shows the wording that triggered it, and you can restore any of them.

4

Export and report

Download a screening-ready RIS or CSV file, a log of every decision, and a PRISMA 2020 flow diagram in Word.

About Winnow

How matching works

Records are compared on DOI and PMID first, then on title, year and first author. Titles are compared without capitals, accents or punctuation, using character-pair similarity. Only certain matches are removed. Everything else that looks similar is flagged for you.

What to know

Winnow is being validated, so check its results before relying on them. It reads RIS, PubMed (.nbib), Web of Science (plain text, tab-delimited or RIS) and CSV files. Check the Removed tab and the flagged pairs before relying on a result. Trial registry entries are flagged when their title matches a publication. The optional exclusion of study types uses only explicit statements in the title, abstract or PubMed indexing, and every excluded record is listed with its evidence so you can restore it.

Privacy and use

Your files are read in your browser and never uploaded. The saved session stays in this browser only, so use Clear saved session when you finish a review. Winnow is released under the MIT licence.

Records are processed in your browser and are not uploaded. Matching order: DOI, PMID, exact normalised title with year and first author, then fuzzy title similarity (bigram Dice). Anything short of a certain match is flagged, never removed silently. Version 1.0.0.