How do I reconstruct and document a search strategy that was not recorded?

You can reconstruct the strategy by rebuilding the query from your saved searches and notes, recording the databases, dates, search terms, and filters you used, then rerunning the search to confirm it returns the same result count. The Cochrane Handbook describes documenting searches as part of planning and reporting how studies were found, so the goal is a written record that another reader could follow and repeat.

Rebuild the query

1. Gather your saved searches, browser history, and notes. PubMed supports saved searches, field tags, phrase searching, and filters, so check whether any of these were stored in your account or notes. Expected result: a starting list of terms and any filters you applied.
2. Write down each database you searched and the date you ran it. The Cochrane Handbook treats selecting sources and documenting searches as separate planning steps, so record them separately. Expected result: a source list with dates.
3. Reconstruct the exact terms, including field tags and phrases, and the filters you used. PubMed's Advanced Search and field tags let you specify where terms were matched. Expected result: a query string you can copy and rerun.
4. Rerun the query in each database and compare the result count with what you remember or recorded. Expected result: a matching count confirms the rebuilt query; a different count means a term, filter, or date is still missing, so adjust and rerun.

Document the strategy

Write a short methods note that lists, for each database: the database name, the date searched, the full query string, and any filters or limits. Add the result count for each search. PRISMA 2020 provides a checklist and flow diagrams for reporting what methods were used and what studies were found, which gives you a structure for presenting the record in your paper.

If a result count still does not match after rerunning, treat the discrepancy as unresolved and check the next variable: the date range, a filter, or a term that may have been entered differently. Record the version that reproduces the count you report.