Write your inclusion and exclusion criteria as explicit rules for population, intervention, outcome, and study type, then apply them in a fixed order and record a reason for every keep or drop decision. This produces a screening log that explains each decision and keeps the process consistent across records.
Define the four criteria
State each criterion as a test a record either passes or fails. For population, name the group the review targets. For intervention, name the exposure or treatment of interest. For outcome, name the result that must be reported. For study type, name the designs you will accept. Write the exclusion side as the mirror of each inclusion rule so the two cannot contradict each other.
Apply them in a fixed order
- Screen the title and abstract against population first; if the record clearly targets a different group, exclude it and log "population."
- Apply the intervention criterion to records that pass step 1; log "intervention" for failures.
- Apply the outcome criterion next; log "outcome" for records that do not report the required result.
- Apply the study-type criterion last; log "study type" for designs you do not accept.
- Mark records that cannot be judged from the title and abstract as "unclear" and carry them to full-text review instead of excluding them.
Each step ends with one of three outcomes: include, exclude with a reason, or unclear and pending full text. The reason field is what makes the log useful later.
Keep the log reportable
PRISMA 2020 provides a checklist and flow diagrams for transparent reporting of what studies were found and included, so structure your log so its counts can feed that flow. Cochrane guidance on searching for and selecting studies covers planning searches, documenting searches, and managing records, which is where the screening log sits in the wider review process. If you search PubMed, its field tags, filters, and saved searches can help you refine and re-run the same query, so the records you screen stay traceable to a documented search.
Before you start reading
Pilot the rule set on a small batch of records and check that each criterion produces a clear pass or fail. If a rule leaves many records "unclear," tighten its wording before screening the full set.