Systematic Review Literature Management:Keep the Decisions, Not Just the PDFs

In a systematic review, several pieces of information accumulate around every paper: which search it came from, whether it's a duplicate, the title/abstract screening decision, the full-text decision, the reason for exclusion, and — if included — the extracted data and risk-of-bias assessment.

A well-managed review is one where all of that history stays attached to the paper. Even if every PDF is accounted for, the review has failed as a record the moment that history breaks.

In practice, no single tool handles everything. A three-layer division of labor works better.

There's a layer for references and citations (Zotero, EndNote, and the like), a layer for screening decisions (a dedicated review tool or a spreadsheet), and a layer for reading and decision notes — where you open the PDF, check the content, and record why you decided what you did. That third layer is the one most people neglect.

Confusion starts once the same paper exists in multiple versions — different filenames, duplicate downloads, records pulled from different databases.

At the deduplication stage, give every paper a single unique ID and link everything that follows — decisions, reasons, extractions — to that ID. It sounds like a small detail, but if this step slips, your PRISMA numbers will stop adding up too.

The reasons you exclude papers at the full-text stage will be needed later, both in the manuscript and in the PRISMA flow diagram.

A paper you drop because it "just doesn't feel right" will come back to haunt you months later when you have to justify the call. The moment you exclude a paper, write one short line that maps to your criteria — "population outside age range," "outcome definition doesn't match." That habit will save you during write-up.

Paperfy isn't a reference manager or a screening tool. It lives in the third layer — reading and decision notes.

Upload the papers that reach full-text review, and each one gets a single page with its AI summary, figures, tables, and original text. Edit the summary to add your inclusion or exclusion reasoning and what you checked, and the decision stays attached to the PDF — no more "why did I keep this?"

A systematic review typically takes months to a year, usually running alongside everything else you do. Your management system has to survive being paused and picked back up.

The hard part of resuming is remembering what was decided and where you left off. If every paper carries its own decision notes, getting back up to speed takes minutes instead of hours.

Finally, something obvious but important: don't keep your screening and extraction tables in a single file on one personal machine.

Back them up to a shared drive and version them according to your team's rules. The record of those hundreds of hours of work is a research output in its own right.