01 · The Question
Does Every Part of Your Research Question Belong in the Database Search?
Your research question may specify a population, intervention or exposure, comparator, outcome, setting, age range, study design, and other conditions. When constructing the search, it seems logical to translate every one of those elements into its own block of keywords and subject headings.
After all, if a concept matters to the question, shouldn't it appear in the search?
Not necessarily. A research question defines what you want to investigate. A database search has a narrower technical job: retrieving records that could plausibly provide the evidence you need. Some characteristics that are essential for determining eligibility are not expressed consistently enough in searchable titles, abstracts, keywords, or indexing to function safely as mandatory search concepts.
03 · What You Need to Know
Research Questions and Search Strings Do Different Jobs
The research question defines the evidence you want
A well-framed research question helps define the scope of an investigation. Depending on the methodology, a framework such as PICO may distinguish the population or problem, intervention, comparator, and outcome. Other question frameworks may emphasize exposures, phenomena, contexts, settings, or study types.
These elements can help establish eligibility criteria and clarify what evidence would answer the question. Question frameworks can also help identify candidate concepts for a systematic search.
Candidate is the important word. Moving from a conceptual question to a database query requires another judgment: which of those concepts should actually constrain retrieval?
The search string defines what the database is allowed to retrieve
Most Boolean strategies organize related terms into concept blocks. Alternative expressions for the same concept are commonly joined with OR, while distinct concepts are commonly intersected with AND. Search-strategy development literature describes this progression from research question to searchable concepts, terms, Boolean combinations, testing, and refinement.
Suppose a question can be represented by four candidate concepts:
Population AND Intervention AND Comparator AND Outcome
In a Boolean database search, a record generally needs to satisfy every concept connected with AND. If the outcome is absent from the searchable metadata, that record may not be retrieved even if the full article measures exactly the outcome required by your protocol.
This is why understanding what AND and OR actually do to retrieval matters when translating a question into a search.
Eligibility concepts are not automatically retrieval concepts
This distinction is fundamental.
Eligibility concept
A characteristic used to decide whether a study belongs in the review or answers the research question.
Search concept
A characteristic represented in the database query because requiring it helps retrieve the intended literature effectively.
A characteristic can be essential for eligibility without being safe to require during retrieval.
Imagine that your inclusion criteria require participants aged 18 to 25. A relevant article may identify participants as "university students" in its title and abstract while reporting a mean age of 21.4 years only in the methods section. If your database query requires age terminology that does not occur in the searchable record, you may never retrieve the article to screen it.
The article meets the eligibility criterion. Its metadata simply does not advertise that fact.
Some concepts are poorly represented in searchable metadata
Bibliographic databases usually provide searchable combinations of titles, abstracts, author keywords, indexing terms, and other metadata, although available fields differ among databases. Authors do not necessarily mention every study characteristic in those fields.
KU Leuven's information-retrieval guidance makes this point explicitly: because literature searching often depends on titles, abstracts, and keywords, researchers need to consider how relevant authors will describe the topic in those metadata fields, and some concepts from the research question may consequently be unsuitable for the search strategy.
Several types of concepts deserve particular scrutiny.
Outcomes can be difficult retrieval concepts
An outcome can be central to a research question yet expressed in many ways. Authors may report the construct, an instrument name, a broader outcome family, a narrower measure, or no outcome terminology in the title and abstract at all.
For example, a review may concern "academic achievement," while eligible studies describe examination scores, course grades, learning performance, achievement tests, or particular disciplinary assessments. Making one outcome vocabulary mandatory can therefore exclude eligible records.
That does not mean outcomes should never be searched. It means their retrieval value must be assessed. The question of when an outcome should be left out of a search strategy depends partly on how consistently and distinctively that outcome is represented in the literature.
Comparators often do not need their own search block
A comparator may be indispensable to the research question while contributing little to retrieval. An intervention study may compare a new instructional approach with usual teaching, another intervention, no intervention, or a control condition. The comparator may be described inconsistently or only within the full text.
Requiring comparator terminology can therefore add another failure point without necessarily identifying many useful records that the population and intervention concepts would not already retrieve.
This principle is not a universal prohibition. Some questions involve a comparator that is distinctive and searchable enough to contribute meaningfully. The decision should follow the literature and retrieval behavior rather than the letters in a question framework.
Study design may be essential for inclusion but risky to search
Researchers may also be tempted to add terms such as "qualitative," "cross-sectional," "randomized," "experimental," or "mixed methods" because the protocol specifies a study design.
Sometimes validated or well-established methodological filters are appropriate. In other situations, design terminology may be inconsistently reported, indexed, or understood. A relevant study may satisfy your methodological criterion without using the terminology you anticipated in its searchable record.
Whether to leave study design out of the search strategy should therefore be considered separately rather than treated as an automatic consequence of the eligibility criteria.
Population concepts can also create unexpected exclusions
Population searching often appears straightforward, but populations can be represented indirectly. University students might be described through educational level, institutional setting, degree program, year of study, or simply as participants recruited from a university. Age groups can be particularly difficult when age information appears only in tables or methods sections.
A population concept that is highly distinctive may be indispensable. A generic or inconsistently reported population restriction may add little except lost sensitivity.
This becomes especially important when considering whether population filters should be avoided.
Question frameworks are thinking tools, not database commands
A framework such as PICO helps researchers formulate and structure certain types of questions. It should not be interpreted as an instruction that P, I, C, and O must each become a mandatory Boolean block.
Search-method guidance similarly distinguishes the concepts contained in a question from those selected for searching. Educational guidance on systematic review searching notes that keeping the number of concepts relatively low can broaden retrieval and that some question elements, including outcomes, may not always be suitable search concepts.
PRESS makes the same issue testable during peer review by asking whether too many or too few question elements have been included and whether the resulting concepts are too narrow or too broad.
Omitting a concept from the search does not mean ignoring it
This is perhaps the most important distinction.
If you omit an outcome from the database query, the outcome can remain an eligibility criterion. You retrieve candidate records using more searchable concepts and then determine during title, abstract, or full-text screening whether each study reports the required outcome.
Likewise, an age range, setting, comparator, or design requirement can remain fully operational during study selection even when it is not encoded into the retrieval query.
You are moving the decision from retrieval to screening, not abandoning it.
There is a cost to leaving concepts out
Broadening retrieval usually increases screening burden. A two-concept search may return thousands of records, whereas adding a third concept might reduce that number substantially.
That reduction can be useful if the third concept removes mostly irrelevant records. It is harmful if it also removes too many eligible studies.
| Approach |
Potential advantage |
Potential risk |
| Include another concept with AND |
Can improve precision and reduce screening |
Can exclude relevant records lacking searchable terminology for that concept |
| Leave the concept for screening |
Can protect sensitivity when metadata are inconsistent |
May substantially increase the number of records requiring screening |
| Use a tested filter |
May improve precision efficiently for some characteristics |
Performance depends on the filter, database, evidence type, and context |
There is therefore no universal rule that fewer search blocks are always better. The objective is to use enough concepts to distinguish the target literature without imposing unnecessary retrieval conditions.
04 · A Practical Example
Turning a Four-Part Question Into a Two-Concept Search
Hypothetical Example
Does generative AI improve critical thinking compared with conventional instruction among undergraduate students?
The question contains at least four obvious elements: undergraduate students, generative AI, conventional instruction, and critical thinking. A researcher initially assumes that all four should appear in the database strategy.
Candidate concept 1: Undergraduate students This population may be searchable through terms such as undergraduate*, universit*, college student*, and relevant database indexing.
Candidate concept 2: Generative AI This is central to distinguishing the target literature and may be represented through terms such as "generative artificial intelligence," "generative AI," ChatGPT, "large language model*" and appropriate controlled vocabulary where available.
Candidate concept 3: Conventional instruction Eligible papers may describe the comparator as traditional teaching, usual instruction, standard practice, control, lecture-based instruction, or simply describe the comparison within the methods. Making this block mandatory removes some otherwise relevant records during testing.
Candidate concept 4: Critical thinking Some relevant papers use "critical thinking," but others describe reasoning, argument evaluation, analytical reasoning, or particular instruments. Requiring the outcome also removes known relevant records.
Search decision The researcher tests a strategy centered on the undergraduate/higher-education population and generative-AI concept, while retaining comparator and critical-thinking requirements for eligibility screening.
The resulting search does not reproduce the research question word for word. It does something more useful: it translates the question into concepts that the database can plausibly retrieve, while preserving the complete question for study selection and analysis.
06 · What This Means for You
Choose Search Concepts by Retrieval Value, Not by Framework Completeness
Begin with the complete research question and identify its conceptual elements. Then treat those elements as candidates rather than mandatory search blocks.
For each candidate, ask how relevant articles are likely to express it in searchable metadata. Examine known papers. Look at titles, abstracts, author keywords, and database subject headings. Test the search with and without the concept and inspect what changes.
A simple decision framework
If the concept is central, distinctive, and consistently represented
It is a strong candidate for inclusion as a search concept.
If the concept is essential for eligibility but inconsistently reported
Consider retrieving without requiring it and applying the criterion during screening.
If adding the concept removes known relevant studies
Investigate whether the terminology is incomplete or whether the concept itself should remain outside the retrieval query.
If omitting the concept produces overwhelming irrelevant retrieval
Explore whether a broader or better-designed version of the concept can improve precision without unacceptable loss of sensitivity.
If you are unsure whether the strategy has become too restrictive
Test its retrieval rather than assuming conceptual completeness equals search quality.
For systematic reviews, document these decisions. If an outcome, comparator, population characteristic, or study-design criterion was intentionally omitted from the query, recording why can make the strategy easier to defend, peer review, reproduce, and update.
