Step 1
Decide what the search must support
A background scan and a systematic review need different methods. Write the purpose before choosing filters. These planning steps are Manusights guidance, not a certification checklist.
- Name the decision
- Record the population or setting, the phenomenon or intervention, and the decision the evidence will inform. Define exclusions before reading results so inconvenient findings are not quietly filtered out.
- Choose sources by coverage
- List which disciplines and evidence types each source covers. Note what you cannot access. One familiar database is not evidence that a cross-disciplinary question has been covered.
Step 2
Test the query, not just the result count
PubMed uses uppercase Boolean operators. Field tags, quoted phrases and wildcards can change automatic term mapping. Inspect Search Details and warnings rather than assuming the typed query is the executed query.
- Separate concepts from syntax
- Build alternatives within each concept first. Translate the structure into each platform's syntax; do not paste one database's field codes into another.
- Keep a small test set
- Choose known relevant papers for different parts of the question. Investigate misses: absent database coverage, a missing synonym, a restrictive field or a filter. Finding this set is a diagnostic, not proof of completeness.
Check the source
- NLM: PubMed User Guide · Searching for a phrase; Combining search terms; Search Details
Step 3
A useful miss is one you can explain
Fictional troubleshooting example: a question about shift work and sleep misses a known paper when the query requires the phrase 'night shift'. The paper's abstract uses 'rotating schedules'.
- Make one controlled change
- Add the alternative wording to that concept, rerun the search and record what changed. Check whether new irrelevant meanings dominate. Do not report an invented retrieval count.
- Keep the limitation
- If the paper is not in the database at all, editing synonyms cannot fix that gap. Record an additional source or citation-search step, with its own method and date.
Step 4
Preserve the search as an inspectable record
PRISMA-S provides reporting guidance for evidence-synthesis searches. Complete reporting helps others inspect a search; it does not establish that the methods were adequate.
- Save the exact run
- Keep the database and platform, executed query, filters and their rationale, run date, result count and exported file name. Document later amendments instead of replacing the original.
- Make the handoff explicit
- Give the next person the search record, source list, known misses and unresolved questions. For a formal review, plan method review and search updates within the review protocol rather than using an arbitrary number of papers as a stopping rule.
Check the source
- Rethlefsen et al.: PRISMA-S (2021) · Abstract; search strategies; dates of searches