Full text: finding PDFs and assessing reports
Open-access PDFs are found for you; you upload the rest or mark a report as not retrieved. Then each report is assessed against every criterion, and every exclusion needs a reason.
Records included at title and abstract (and, by default, those everyone marked maybe) move on to full text: you read the whole report and decide again, more strictly.
Getting the PDFs
Find open-access PDFs looks for a legal copy of each report in:
- OpenAlex
- Semantic Scholar
- arXiv
- Unpaywall (when a contact e-mail is set in System)
- Europe PMC
It attaches the first valid PDF it finds. Paywalled papers are not downloaded. Optionally, also let the AI search the web for author copies and repositories (never shadow libraries).
For the rest:
- Upload the PDF you got through your library.
- Mark as not retrieved when you cannot get the report. That is a normal outcome: PRISMA 2020 counts "reports not retrieved" in the flow diagram.
Every attached PDF records where it came from, and the paper reports how the full texts were obtained.
Assessing a report
The full-text screen works like title and abstract screening, with the PDF beside the record:
- vote include or exclude, with your verdict on each numbered criterion;
- every exclusion needs one reason from the review's list of exclusion reasons; the PRISMA diagram counts the excluded reports per reason;
- two reviewers assess each report independently, and disagreements (including different primary reasons) become conflicts.
Send a point to the paper while you read: Note for the paper adds it to a manuscript section's author inputs, tagged with the study's citation key. See Writing the manuscript.
After full text
Studies included at full text are the ones your review synthesises. They go on to Data extraction, and they fill the "studies included in review" box of the flow diagram.