Build the log by recording each database, query string, filters, date run, and hit count, then map retained records to inclusion decisions in a PRISMA-style flow so the analyst can rerun and verify each step.
1. Fix the search plan before running anything. Cochrane guidance covers planning searches, selecting sources, and developing search strategies. Write down which databases you will search and the inclusion criteria that will decide which records stay. This gives the analyst the target the log is measured against.
2. Capture one log row per search execution. For each database, record the exact query string, every filter applied, the date the search ran, and the number of hits returned. PubMed supports field tags, phrase searching, filters, and Advanced Search, so record the field tags and filter settings exactly as used. Europe PMC supports keyword and fielded searches across literature, grants, preprints, citations, and references, so note which content types your query covered.
3. Record supplementary sources separately. Cochrane guidance includes searching reference lists and other sources alongside database searches. Log each supplementary source, how you searched it, and what it returned, so the analyst can see where records came from.
-
Map records to decisions in a PRISMA-style flow. PRISMA 2020 provides flow diagrams for reporting what studies were found and included. Carry the hit counts from your log into the flow: records identified, records screened, records excluded with reasons, and records retained. Each retained record should trace back to the log row that found it.
-
Package the log with the flow counts and a rerun note. The handoff should let the analyst repeat each query on the stated date-sensitive settings and compare new hit counts against your recorded ones. Where a rerun returns a different count, mark the discrepancy as unresolved and check the query string, filters, and database content coverage before drawing conclusions.
What the handoff should contain
- A search audit log with one row per database or source: query string, filters, date run, hit count.
- A PRISMA-style flow with counts at each stage and reasons for exclusion.
- A record-to-log mapping so each retained study links to its source search.
- A short note on rerun steps and any counts that changed.