PRISMA Systematic Review Template

Turn each part of your review question (population, intervention, comparator, outcome and study design) into a rule a screener can apply from a title and abstract alone. The US Institute of Medicine's 2011 standards for systematic reviews say to include or exclude studies based on the protocol's prespecified criteria¹. Write those rules into your protocol before anyone screens.

Then fix a short, ordered list of exclusion reasons: wrong population, wrong intervention, wrong comparator, wrong outcome, wrong study design. Screeners record the first reason that applies, so two people excluding the same paper give it the same reason. If you need more than about 10 reasons, some are one reason under two names.

Pilot the criteria before the full screen: both screeners decide the same 50 or so records, then compare. The same standards call for training screeners with written documentation, then testing and retesting them. Rewrite any rule the two read differently, and keep the record they argued over beside it as a worked example.

Combine your search exports and remove duplicates before anyone screens. PRISMA 2020 treats records that refer to the same report² as duplicates, such as one journal article indexed in both MEDLINE and Embase. A trial's registry entry and the paper reporting its results are two reports of one study, so keep both and link them later.

Match from the strongest rule to the weakest: an exact DOI, then title with year and first author, then near matches a person confirms. Keep a register of how many records each rule removed for each pair of databases. Take your duplicate count from it, not from a reference manager's summary screen.

Before screening starts, check the first subtraction. Removals before screening are three boxes in the PRISMA 2020 diagram: duplicates, records marked as ineligible by automation tools³ and records removed for other reasons. Records identified minus all three should equal the rows in your screening sheet; if not, find out why while the export is fresh.

Two screeners should decide every record independently, without seeing each other's calls. The Institute of Medicine standards require two or more members of the review team, working independently¹, and its committee concluded that reviews may not be reliable without two screeners. The report cites a 2002 study in which two reviewers found up to 32 percent more eligible trials than one¹.

Send disagreements to a third reviewer instead of letting the more senior screener decide. When a record is borderline, let it through to full text. PRISMA 2020 asks you to report how many reviewers screened each record³, whether they worked independently and how you resolved disagreements.

Measure agreement with Cohen's kappa, which discounts the agreement you would expect by chance. A 2012 paper on the statistic in Biochemia Medica argues that any kappa below 0.60 indicates inadequate agreement⁴. In screening, low agreement often means a criterion can be read two ways, so reword it and rescreen the records it touched.

Retrieve every report that passed title and abstract screening. A report you couldn't obtain is not one you assessed: the PRISMA 2020 diagram gives reports not retrieved their own box. Two screeners then apply the full criteria. PRISMA asks for the primary reasons for exclusion³ at this stage, so give each report you drop one reason from your ordered list.

If a study looks eligible but doesn't report your outcome in a usable form, write to the authors and log the date. PRISMA asks you to say whether such a study was excluded because the outcome wasn't measured or because it wasn't reported³.

Keep reports and studies apart. One trial can produce a protocol, a main results paper and a follow-up paper: three reports, one study. The Institute of Medicine standards say to link publications from the same study¹ so its data are not counted twice. PRISMA asks for the number of studies and reports included³.

As you go, list the near misses. PRISMA 2020 asks you to cite studies that might appear to meet the inclusion criteria² but were excluded, and to explain why.

Draw the diagram last, and count every box from your log. PRISMA publishes flow diagram templates⁵ for new and updated reviews. Four of their boxes are subtractions, and each one has to add up.

Add up the full-text exclusion reasons too: a missing reason is easy to overlook, because the total still looks right. If any subtraction fails, find the records responsible and fix the log; never nudge a box until the arithmetic works.

Then write the study selection paragraph from the same counts, so the text and diagram can't drift apart. Item 16a of PRISMA 2020 asks you to describe the results of the search and selection process², from records identified to studies included.

Once the included set is settled, the literature review synthesis tool helps you write up the findings. Before you submit, a reference and citation check tests each citation against its source and flags retracted papers.

If peer reviewers question your flow, the response to reviewers template ties each answer to a change. The template on this page keeps your log, counts and diagram together, and River can check each subtraction for you.