03 · What You Need to Know
The Number of Databases Matters Less Than What They Collectively Cover
Databases are not interchangeable containers of the same literature
Bibliographic databases differ in what they index and how they describe it. Their coverage can vary by discipline, journal, country, language, publication type, historical period, and indexing system.
This means that searching one database does not simply retrieve a smaller version of what you would find by searching several. Another database may contain records that the first one does not contain at all.
It may also index the same article differently, making it retrievable through different subject headings or search terms.
| Database characteristic |
Why it matters |
| Disciplinary coverage |
A specialist database may index journals and research communities poorly represented in a broad multidisciplinary source. |
| Journal coverage |
Not every database indexes the same journals, even within the same discipline. |
| Geographic and language coverage |
Some sources provide better access to literature from particular regions or publication languages. |
| Document types |
Coverage of conference proceedings, books, reports, theses, and other materials varies. |
| Controlled vocabulary |
Databases may use different indexing systems that provide different routes to the same concept. |
| Historical coverage |
The earliest searchable year and completeness of older records differ among databases. |
Database selection is therefore a coverage decision rather than an exercise in collecting database names.
There is no universal “three-database rule”
You may encounter recommendations that a systematic review should search at least two, three, or some other fixed number of databases. Such numbers can be useful as rough warnings against relying on a single source, but they should not be treated as universal methodological laws.
JBI guidance takes a question-driven approach. Its current manual states that databases and sources should be appropriate for the review question and that systematic searching should use the widest reasonable collection of relevant information sources. JBI systematic reviews commonly use multiple bibliographic databases alongside other appropriate sources such as trial registers and grey literature.
The key word is appropriate.
Three poorly chosen databases do not become comprehensive merely because the count reached three. Two highly complementary databases may cover a particular narrow literature better than several sources with nearly identical scope, although formal evidence synthesis may still require additional sources under the relevant methodology.
Your research question should determine the database mix
Consider three hypothetical questions:
- the effectiveness of a medication for a clinical condition;
- teachers' experiences of generative AI in higher education;
- the relationship between social media use and adolescent mental health.
The most appropriate databases are unlikely to be identical.
The clinical intervention question may require major biomedical databases and trial sources. The higher-education question may require education, multidisciplinary, and perhaps technology-related databases. The adolescent mental-health question crosses health, psychology, and social-science literatures.
A useful selection process therefore begins by asking:
Which research communities could plausibly have produced evidence relevant to this question, and where is their work indexed?
If the question crosses disciplinary boundaries, your database selection should usually cross them too.
General databases and specialist databases serve different purposes
Broad multidisciplinary citation databases can provide extensive coverage across fields. Specialist databases concentrate more heavily on particular disciplines and may use subject-specific indexing.
Multidisciplinary database
Covers research across many fields and can be particularly useful for interdisciplinary questions and citation searching.
Subject-specific database
Concentrates on a disciplinary literature and may provide journals, indexing, or terminology not represented as well in broader sources.
For example, CINAHL is especially relevant to nursing and allied health questions, PsycINFO to psychology and behavioral science, and ERIC to education. MEDLINE provides extensive biomedical coverage. Embase provides substantial biomedical and pharmacological coverage that overlaps with, but is not identical to, MEDLINE.
These examples should not be interpreted as a universal shopping list. A specialist database is useful when its subject coverage intersects meaningfully with your question.
One database can retrieve much of the literature and still miss something important
Overlap among major databases can be substantial, which can create a false sense of security.
You may search a second database and find that most of its relevant records also appeared in the first. That does not mean the second source was useless. The few unique records may include an important eligible study.
Research examining database combinations illustrates this point. In a prospective study of 58 published systematic reviews, Bramer and colleagues found that 16% of the included references retrieved through database searching were found in only one database. Embase contributed the largest number of unique references in that sample.
The study also found that specialist databases such as CINAHL and PsycINFO added unique included references particularly when review topics matched their disciplinary focus.
These findings support a broader principle rather than a universal database recipe: overlap does not eliminate the possibility of unique retrieval, and specialist sources become especially valuable when the question enters their domain.
Evidence about an “optimal combination” is context-specific
Bramer and colleagues also examined combinations of databases in biomedical systematic reviews. In their sample, Embase, MEDLINE, Web of Science Core Collection, and Google Scholar together achieved an overall recall of 98.3% of the included references found through database searching. The authors recommended that combination for biomedical systematic reviews, with specialist databases added when appropriate.
That finding is useful, but it should not be generalized into a rule for every discipline.
The study was based largely on health-related systematic reviews performed within one institutional search environment. A review in education, engineering, communication, economics, or the humanities may have a very different distribution of relevant literature.
Watch Out
Do not turn a database combination validated for one body of literature into a universal minimum for all research. Database performance depends on the topic, discipline, eligible study types, publication patterns, and the search strategies used within each source.
For randomized trials, methodological guidance may identify core sources
Some review types have more specific source expectations.
For Cochrane intervention reviews, current guidance identifies CENTRAL, MEDLINE, and Embase, when Embase is available, as core bibliographic sources for identifying reports of randomized trials, alongside appropriate trial registers and other sources depending on the review.
That recommendation follows the particular evidence needs of intervention reviews. It should not be transplanted mechanically to a qualitative education review or a historical literature review.
The review methodology and eligible evidence types matter.
Scoping reviews may need a wider disciplinary reach
A scoping review often asks how a broad topic has been studied, what concepts appear in the literature, or what types of evidence exist. The evidence may therefore be distributed across several disciplines and publication forms.
JBI guidance states that scoping review searches should ideally be as comprehensive as possible within the constraints of time and resources and should seek both published and unpublished sources of evidence where appropriate.
This may justify searching several complementary databases, but again the objective is not a particular count. It is adequate representation of the evidence landscape defined by the review question.
Database selection should consider the type of evidence you need
Two reviews of the same general topic may need different databases because their eligible evidence differs.
Suppose one review asks whether an educational technology improves examination performance. Another asks how students experience using that technology.
The first may prioritize databases likely to index intervention and evaluation studies. The second may benefit from sources with stronger qualitative, education, psychology, or social-science coverage.
The search question is not merely “Which databases contain papers about this topic?” It is “Which databases are most likely to contain the kinds of studies eligible for this review?”
Database selection also interacts with language and geography
If your research question concerns a particular region, population, or international phenomenon, major English-oriented databases may not capture all relevant evidence.
Regional databases, national indexes, local repositories, and non-English sources may contribute literature that is poorly represented elsewhere.
This is why avoiding a formal language restriction in the search does not by itself guarantee multilingual or geographically comprehensive retrieval. Your information sources must actually index the literature you hope to find.
Historical coverage matters too
A database can be topically appropriate yet unsuitable for part of your review's time period.
Suppose relevant research began in the 1960s, but one of your chosen sources provides reliable coverage only from the 1990s. Searching that database from inception does not recover the missing decades.
You may need another bibliographic source, historical index, archive, or complementary search method.
The question of how far back the literature search needs to go should therefore inform database selection rather than being decided after the sources have already been chosen.
Databases are only one category of information source
A comprehensive literature search may require more than bibliographic databases.
Depending on the review, relevant sources can include:
- clinical trial or study registers;
- dissertation and thesis repositories;
- conference proceedings;
- government and organizational websites;
- regulatory sources;
- institutional repositories;
- reference lists of relevant studies and reviews;
- forward citation searching;
- targeted journal searching;
- contact with authors or experts.
JBI explicitly frames systematic searching in terms of information sources, not bibliographic databases alone. Its guidance for effectiveness reviews includes multiple relevant bibliographic databases as well as trial registers, grey-literature sources, relevant journals, websites, and other appropriate approaches.
This is particularly important after deciding not to restrict the search automatically to peer-reviewed literature. Searching six journal databases will not retrieve a dissertation that none of them indexes.
Google Scholar is not simply another bibliographic database
Researchers frequently ask whether Google Scholar counts as one of the databases in a systematic search.
It can be useful for locating scholarly material, citation searching, identifying additional reports, and finding material not easily located elsewhere. However, its search interface, ranking, result limits, indexing transparency, and reproducibility differ from conventional bibliographic databases.
Its role should therefore be specified rather than treated as interchangeable with a database such as MEDLINE or PsycINFO.
Bramer and colleagues found that Google Scholar contributed useful retrieval in their biomedical review sample, but their search method examined a limited number of highly ranked results. That is a specific search procedure, not evidence that a generic Google Scholar search can replace carefully selected bibliographic databases.
More databases create real costs
Adding another database is not free methodologically or practically.
Each source may require you to:
- translate the search syntax;
- map controlled vocabulary;
- adapt field codes and proximity operators;
- run and document another strategy;
- export another set of records;
- deduplicate overlapping results;
- screen additional unique and duplicate records;
- maintain another search for future updates.
Bramer and colleagues observed this trade-off directly. Additional databases increased recall but also increased the number of records that needed to be examined.
That does not argue for minimizing databases. It argues for selecting them deliberately.
Search quality inside each database still matters
Searching eight databases poorly is not necessarily more comprehensive than searching several well-chosen databases with carefully adapted strategies.
Database interfaces differ. Controlled vocabularies differ. Phrase syntax, truncation, proximity searching, and field codes differ. Copying the same query verbatim into every database can therefore produce very different retrieval behavior.
JBI requires database-specific searches in its systematic search process and emphasizes documenting the strategy used for each source. Search strategies need to be adapted rather than merely pasted.
A database can contribute unique literature only if your query is capable of retrieving it.
Do not choose databases only because you already know how to use them
Convenience inevitably influences research practice. Researchers have institutional subscriptions, familiar interfaces, and preferred platforms.
But familiarity is not evidence of coverage.
If your topic involves education and psychology, using only two multidisciplinary databases because they are convenient may leave specialist literatures underrepresented. If your topic concerns nursing, omitting a nursing-focused database deserves consideration. If the review is international, regional sources may matter.
A defensible methods section should be able to explain why the selected sources were relevant to the question.
Nor should you add databases merely because reviewers expect a long list
The opposite behavior is equally unhelpful.
Researchers sometimes accumulate databases because a longer methods section looks more comprehensive. Several sources may index essentially the same core literature, while a specialist database that would contribute genuinely different coverage remains absent.
Search breadth should be conceptual, disciplinary, geographical, methodological, and temporal. A database count captures none of those dimensions very well.
This leads directly to the related question of whether searching more databases automatically makes a literature search better. It does not, although stopping too early can certainly leave relevant evidence undiscovered.
Think in terms of diminishing returns, but do not use that as a casual stopping rule
As additional databases are searched, the proportion of duplicate records often increases. Eventually, another source may contribute very few unique relevant records.
This resembles diminishing returns.
However, you usually do not know in advance whether the next database contains the one study that matters. Nor is the absence of many unique records sufficient evidence that the source was unnecessary, particularly in a systematic review where missing an eligible study can matter more than screening duplicates.
Use expected coverage, methodological guidance, subject expertise, and evidence about database performance to choose sources prospectively rather than stopping opportunistically when another database seems inconvenient.
A librarian or information specialist can materially improve database selection
Choosing databases is partly a subject decision and partly an information-retrieval decision.
Researchers know the intellectual boundaries of the question. Information specialists know how bibliographic sources differ in coverage, indexing, platforms, syntax, and practical search behavior.
JBI recommends seeking advice from an expert research librarian when developing systematic review search strategies where possible. This collaboration can help identify specialist databases that researchers might overlook and avoid redundant sources that add little useful coverage.
For complex systematic reviews, asking for this input before finalizing the protocol is usually more useful than asking after the screening team has already processed 15,000 records. Librarians, like reviewers, appreciate being invited before the methodological fire starts.
Your protocol should identify the databases and sources you plan to search
For protocol-based evidence synthesis, database selection should be planned rather than reconstructed after the review.
JBI expects protocols to specify the information sources that will be used, including bibliographic databases and other relevant sources. The search platform should also be identified where appropriate because the same database may be available through different interfaces with different syntax and functionality.
The completed review should then report what was actually searched, the strategies used, and the dates of searching.
This documentation is part of the scientific validity of the search, not clerical decoration.
