Most systematic reviews use 3–5 databases; at minimum, pair two core databases with 1–3 subject sources and registries for broad, low-bias coverage.
Quick Answer: Database Counts That Work
There is no magic number that fits every review. The right count depends on the field, question type, and how well the topic is indexed. A safe working range is three to five databases, with a floor of two large multidisciplinary or field-leading sources. Health topics often need more, since studies scatter across MEDLINE, Embase, CENTRAL, and subject databases. Outside biomedicine, a core pair such as Scopus and Web of Science, backed by a subject database, often lands near the sweet spot.
Beyond the database list, quality comes from search craft, peer review of strategies, and transparent reporting. That mix raises recall while keeping noise under control.
Field-To-Database Map (Start Here)
| Field | Core Databases To Include | Notes |
|---|---|---|
| Clinical Interventions | MEDLINE, Embase, CENTRAL | Add trial registries; add CINAHL for nursing topics. |
| Public Health | MEDLINE, Embase | Add Global Health or Sociological Abstracts as the topic demands. |
| Nursing & Allied Health | MEDLINE, CINAHL | Add Embase when drug or device issues appear. |
| Behavioral Sciences | MEDLINE, PsycINFO | Add Embase if pharma content is central. |
| Education | ERIC, Scopus or Web of Science | Pair ERIC with one large index to widen coverage. |
| Social Sciences | Scopus or Web of Science | Add Sociological Abstracts or Social Services Abstracts. |
| Engineering | Scopus or Web of Science | Add Compendex and Inspec for depth. |
| Business & Management | Scopus or Web of Science | Layer ABI/INFORM or Business Source. |
| Food & Farming | CAB Abstracts | Pair with MEDLINE or Web of Science; add FSTA when food science is central. |
| Multidisciplinary Topics | Scopus, Web of Science | Add one subject database aligned to the question. |
Why More Than One Database Matters
Databases index different journals, conference sets, and language zones. Even when coverage overlaps, indexing terms, update cycles, and record quality differ. A single source can miss trials or turn up thin detail that blocks screening. Using several sources reduces the odds of missing eligible studies and balances out quirks in any one index.
Health reviews show this gap. MEDLINE and Embase each bring distinct records. CENTRAL adds controlled trial records and hand-searching content that do not always appear in other places. In many non-medical areas, Scopus and Web of Science capture distinct slices of the literature. Pairing them surfaces more citations than either alone.
How To Choose Databases For Your Topic
Start With The Core Pair
Pick two large, complementary sources before adding niche tools. In biomedicine, that usually means MEDLINE and Embase or MEDLINE and CENTRAL for trial-heavy questions. In other fields, Scopus and Web of Science make a reliable pair.
Add Subject Databases
Add one to three subject indexes that match your concepts. Examples include CINAHL for nursing, PsycINFO for mental health, ERIC for education, Compendex for engineering, and CAB Abstracts for farming. These sources often hold journals that general indices downplay.
Include Trial Registries And Grey Sources
Balance the database plan with trial registries and grey sources. ClinicalTrials.gov and the WHO ICTRP help catch unpublished or ongoing trials. Theses repositories and preprint servers can reveal emerging work. Be selective and match these sources to your question so screening stays manageable.
How Many Databases For A Systematic Review: Practical Ranges
Use the ranges below to plan effort and coverage. These are field-tested patterns, not hard rules.
Two Databases
Use this when the topic is narrow and well indexed, the review is a rapid effort, or you have strong registry coverage. Pair a large general index with a subject source. Risk: distinct studies may sit outside the pair, so describe limits clearly and add a citation chase from included papers.
Three To Five Databases
This is the sweet spot for most teams. Start with a core pair, then add one to three subject sources that align with your concepts. Add registries and citation chasing to round out retrieval. Time cost rises, but recall rises too.
Six Or More Databases
Reserve this for broad, policy-relevant questions, multi-disciplinary topics, or when early scoping shows poor overlap. Plan time for de-duplication and tight screening workflows. Gains in recall can be real, but noise grows fast if strategies are not tuned.
Search Craft That Lifts Recall
Write Precise Concepts
Break the question into main concepts, map controlled terms and keywords for each, and combine with tested operators. Keep proximity operators consistent across platforms where possible.
Peer Review Your Strategy
Ask a librarian to run a PRESS-style review on the main strategies. Small tweaks in terms, spellings, and proximity can change yield in a big way. Record the final strings and platform details for full transparency.
Document Everything
Note the date searched, platform, limits, and record counts by source. Keep a log of changes. Transparent methods raise trust and make updates smooth.
What The Guidance Says
Leading handbooks stress multiple databases and clear reporting. The Cochrane search chapter sets out core sources for trials and calls for several databases plus trial registries. PRISMA-S checklist spells out how to report every source, strategy, and record count so others can follow the path. Linking strategies and counts in an appendix makes replication and audits easier.
When Fewer Databases Can Be Enough
Some RCT-focused topics with tight indexing may show little change in the final conclusion once you add more databases. That does not mean a single source is safe. A small, well chosen set can work if you combine a solid core pair, strong registry searches, and forward-backward citation chasing from included studies. Spell out the trade-offs and keep the door open for an update when new studies surface.
Database Count: Fit And Trade-Offs
| Database Count | Best Use Case | Main Risks |
|---|---|---|
| 2 | Narrow scope, rapid work, or well indexed topics with strong registry coverage. | Missed distinct studies; thin coverage of peripheral journals. |
| 3–5 | Standard reviews that need balance between recall and workload. | Higher screening load if strategies are loose. |
| 6+ | Broad, cross-disciplinary, or policy-sensitive questions. | De-duplication burden; rising noise without tight logic. |
Practical Steps To Build Your List
Scope The Terrain Fast
Run a quick seed search in one large index to see where studies cluster. Scan journal titles and indexing terms to reveal candidate databases. Check where sentinel trials are indexed.
Draft Your Core Set
Lock in a core pair that complements each other. Add one to three subject databases that match the concepts. Add trial registries that fit the question. Keep the list short enough to manage well.
Test And Refine
Run pilots, check recall with known sentinel papers, and tighten terms that drive noise. When a database yields only duplicates, drop it and explain why. Plan time and document choices clearly throughout.
De-Duplication And Record Management
Plan the workflow before you search. Export with rich fields, keep database tags, and pick one tool for de-duplication. Run exact matches first, then fuzzy passes. Keep a log of record counts removed at each step so your PRISMA flow is clear.
When export formats differ, capture both RIS and CSV if allowed. Preserve persistent IDs, accession numbers, and source names. These markers help you trace a record back to its origin and spot gaps later.
If several team members screen titles, set clear rules for tie-breaks and reasons for exclusion. Agree on a short list of tags that everyone uses the same way. Consistency saves time when you draft the methods section.
Handling Grey Literature
Grey sources can fill gaps. Use them when publication bias is a concern or when studies tend to sit outside journals. Pick a short, targeted list: priority repositories, policy sites, or conference sets. Record URLs, dates, and terms so others can repeat the path.
Be realistic about volume. Wide web searches can flood the screen with noise. Set caps on pages or facets, or use known lists of organizations and funders to keep the task bounded.
When you find items that overlap with journal records, keep one version only and note the reason. When grey items inform context rather than outcomes, store them in a separate folder so they do not slip into meta-analysis steps.
Field Examples That Show The Range
Drug Safety In Adults
Core: MEDLINE and Embase. Add CENTRAL if randomized trials are in scope. Add a regulator site search for signals and trial registries for ongoing work.
Rural Water Access
Core: Scopus and Web of Science. Add CAB Abstracts for farming links and Social Services Abstracts for policy angles. Scan NGO portals for grey material.
Quality Checks Before You Stop Searching
Ask: does the core pair span the main journals and languages for this topic? Do subject databases add distinct yield? Do registry searches turn up trials that the databases miss? Can a citation chase from two or three sentinel papers pull in fresh records? If the answers look weak, extend the search set and retest.
Run a short audit on the top ten included studies. Note where each record came from, the indexing terms that fired, and which sources missed them. Use that snapshot to decide if one more subject database is worth the time.
Cost And Access Notes
Some databases sit behind paywalls. If access is limited, lean on PubMed for MEDLINE content and public trial registries, plus one broad index your library provides. When a subject database is missing, use citation chasing and hand searches to plug gaps.
Final Take
There is no fixed count that works for every question. A practical answer is this: plan for three to five databases for most reviews, with a floor of two large, complementary sources. Add one to three subject databases and registries that match the topic. Keep searches transparent, peer reviewed, and well documented. With that plan you raise recall, keep noise in check, and give readers a clear map of how the evidence base was found.