GUIDE 2 OF 5
Efficiency in evidence: title and abstract screening
Screening is usually the bottleneck. Spreadsheets hide disagreements and scramble PRISMA counts. Dual independent decisions, a conflict log, and a clean library are the workflow — not speed for its own sake.
The job at this stage is not to judge quality. It is to ask whether the record could meet eligibility. Risk of bias, sample size, and journal prestige are not exclusion reasons at title/abstract.
Preparing the library: deduplication and data cleanliness
Accurate screening starts with a defensible starting count. EvidenceFlow imports from PubMed search and from RIS, BibTeX, and CSV uploads (EndNote, Zotero, and similar). Automatic deduplication by DOI, PMID, and title/author similarity is what keeps the PRISMA “identified” and “after duplicates” boxes honest.
OpenAlex and Crossref appear later in the pipeline as full-text retrieval sources, not as the citation import itself. Do not mix those two jobs or the identification numbers will not match the methods text.
The gold standard: blind dual screening
Two reviewers, working the same records without seeing each other’s decisions, reduce the chance that one person’s reading of PICO decides the corpus. Allowed decisions: include, exclude (with a reason code), or unsure. Unsure always goes to full text — never to exclude.
Practice the same codes on the free screening worksheet, then run them on an imported set in a project.
Navigating conflicts and consensus
Do not re-screen the whole library together. Meet on disagreements only. A third reviewer, the review manager, or a recorded consensus — pick one in the protocol and stick to it. Editors ask for that log when dual review is claimed.
Start a free project to use the screening interface, then continue with Guide 3: data extraction.
FAQ
Can AI make the include/exclude call?
No. A relevance score can reorder the queue. The decision stays with human reviewers. EvidenceFlow core screening does not require paid AI.
What if the abstract is missing?
Treat it as unsure and retrieve the record. Do not silently drop missing abstracts.
Should I exclude on title alone?
Only when eligibility is impossible (wrong species, clearly the wrong intervention). Vague titles go to abstract or full text.
Core import, screening, extraction, meta-analysis, and PRISMA reporting are free. AI relevance scoring and extraction auto-fill are the paid upgrade.