Scientific Literature Search: From Question to Audit Trail
A dependency-aware scientific literature search workflow with a worked query, stopping rules, and recovery for low recall or low precision.
Next step
Choose the next useful decision step first.
Use the guide or checklist that matches this page's intent before you ask for a manuscript-level diagnostic.
Quick answer: Run a scientific literature search by fixing the question and eligibility criteria first, translating concepts into controlled terms and synonyms, testing the query in the right databases, recording every search, deduplicating results, and using citation chasing to find studies the database query missed.
Evidence basis: We reviewed the PubMed user guide, Cochrane Handbook chapter on searching, and PRISMA-S reporting guideline on September 8, 2026. Those sources describe established search and reporting practices. The decision workflow below is Manusights guidance and should be adapted to the review type.
Official sources define search and reporting expectations. Manusights judgment does not certify completeness; it tests whether the manuscript's evidence trail is reproducible and proportionate to its claims.
Search dependency | Required input | Output before moving on | Failure signal |
|---|---|---|---|
Question | Population or system, concept, relationship, and decision | A bounded review question | Search terms are being chosen before scope is fixed |
Eligibility | Study designs, dates, languages, settings, and evidence types | Written inclusion and exclusion rules | Criteria change after seeing results |
Vocabulary | Controlled terms, synonyms, abbreviations, spelling, and known phrases | Concept blocks with provenance | One preferred phrase carries the search |
Databases | Subject coverage, indexing method, and access | Named sources matched to the question | Convenience determines database choice |
Audit trail | Full strings, platforms, dates, limits, counts, and exports | Reproducible search log | The team can only describe the search from memory |
Define the review before writing a query
A scoping review, systematic review, rapid review, and background search have different recall and documentation needs. State the purpose, intended decisions, eligible evidence, and stopping rule. Do not call a search systematic because it used many keywords.
For a manuscript introduction, the goal may be to map the closest evidence and live disagreement. For a systematic review, the goal is a reproducible attempt to identify all eligible studies. The second requires protocol discipline, multiple databases, documented searches, deduplication, screening records, and an update plan.
Build concept blocks
Start with two or three load-bearing concepts. For each, list controlled vocabulary from the database, author language from known relevant papers, acronyms, spelling variants, older terms, and narrower or broader terms. Combine synonyms with OR and concepts with AND.
Do not add an outcome block automatically. Outcomes are often described inconsistently and can reduce recall. Add one only after testing against known relevant records and documenting what it excludes.
Worked PubMed example
Question: among adults after stroke, do home-based telerehabilitation programs improve functional recovery compared with usual care?
Concept blocks:
- Stroke: controlled heading plus stroke, poststroke, cerebrovascular accident.
- Telerehabilitation: controlled heading where available plus telerehab, remote rehabilitation, video rehabilitation, home-based digital therapy.
- Trial design: add a validated study-design filter only if the eligibility criteria require it.
Test the query against five known eligible papers. If one is missing, inspect its title, abstract, indexing, and publication type to learn which concept or field restriction failed. That is a better repair than adding dozens of untested synonyms.
Search across complementary sources
Use databases because they index different journals, document types, and vocabularies. Add backward citation searching from included studies and strong reviews, forward citation searching, author searches, trial or preprint registries where relevant, and targeted grey-literature sources.
Record the platform as well as the database. A PubMed query is not identical to MEDLINE searched through another interface. Dates, field tags, automatic term mapping, and export behavior affect reproducibility.
Deduplicate and preserve provenance
Keep stable identifiers such as DOI, PMID, accession number, and database record ID. Deduplicate in stages: exact identifiers, exact title-year combinations, then careful fuzzy matching. Preserve a record of merges rather than deleting without trace.
Every included study should retain where it was found. This reveals which sources added unique evidence and makes later updates more efficient.
Failure states and recovery
Symptom | Likely diagnosis | Recovery |
|---|---|---|
Known relevant paper is missing | Vocabulary, indexing, or field restriction failed | Inspect the record and repair the affected concept block |
Screening burden is unmanageable | One concept is too broad or an eligibility boundary is absent | Add tested precision without hiding relevant study types |
Results are dominated by another field | Ambiguous terms lack context | Add a field-specific concept or controlled heading |
Team cannot reproduce the count | Platform, date, limits, or full string is missing | Re-run and store the exact search record |
Search becomes stale before submission | No update rule was set | Re-run the saved strategy and document the update date |
Completion checklist
- Name the review type, question, eligibility criteria, and stopping rule.
- Build concept blocks from controlled vocabulary and observed author language.
- Test recall against known relevant papers.
- Match databases and supplementary search routes to the question.
- Save full queries, platforms, dates, limits, and result counts.
- Deduplicate while preserving source provenance.
- Record screening decisions and reasons.
- Schedule an update before submission when needed.
The literature search strategy example owns the worked reporting artifact. After the evidence is assembled, use the literature review guide to synthesize rather than catalogue it, and the peer-review verification guide before assigning source labels.
Readiness check
Run the scan while the topic is in front of you.
See score, top issues, and journal-fit signals before you submit.
Submit if / think twice if
Submit the search methods when another researcher could reproduce the route and understand every scope decision. Think twice if one database carries a cross-disciplinary question, if a known eligible paper is still missing, or if the final manuscript cites studies that cannot be traced to a documented search or supplementary route.
Manusights connects the search record to the manuscript. We trace major claims in the introduction and discussion to included evidence, test whether exclusions distort the stated boundary, and flag conclusions that are broader than the searched population, setting, intervention, or date range. This is a claim audit, not a guarantee of exhaustive retrieval.
Frequently asked questions
Define the review question, eligibility criteria, evidence types, and stopping rule before choosing databases or writing search terms.
Use the sources needed for the question and review type. Systematic work usually needs multiple complementary databases plus documented supplementary searching.
Save the database and platform, complete query, field tags, limits, date, result count, export, deduplication method, and update record.
Set the update rule in advance and rerun before submission when the field, review standard, or journal requires current coverage.
Sources
- 1. PubMed user guide, U.S. National Library of Medicine.
- 2. Searching for and selecting studies, Cochrane Handbook.
- 3. PRISMA-S reporting guideline, Systematic Reviews.
Before you upload
Choose the next useful decision step first.
Move from this article into the next decision-support step. The scan works best once the journal and submission plan are clearer.
Use the scan once the manuscript and target journal are concrete enough to evaluate.
Private API processing. Your manuscript is not used to train models.
Put the guidance to work
Choose the practical next step for your research.
Start with the question in front of you, or use the guided learning path when you want a clearer view of the full publishing process.