01 · The Question
If Your Study Has a Defined Population, Should You Filter the Search to That Population?
Your eligibility criteria may specify children, adolescents, older adults, university students, healthcare professionals, people from a particular setting, or another clearly defined population. Restricting the database search to that group can seem like an obvious way to eliminate irrelevant results.
The difficulty is that databases can retrieve a population only through information represented in searchable text, indexing, or filter metadata. A study can contain exactly the participants you need without advertising that characteristic in the way your population filter expects.
This creates an important distinction: a population may be precisely defined in your research question while remaining surprisingly difficult to identify reliably at the database-search stage.
03 · What You Need to Know
Why Population Restrictions Can Be More Fragile Than They Look
A population criterion and a population filter are not the same thing
Suppose your review includes only adolescents aged 12 to 18. That age range defines which studies are eligible. It does not necessarily follow that the database should be instructed to return only records classified within an adolescent age category.
The eligibility criterion operates when you decide whether a study belongs in the review. The search restriction operates earlier: it decides which records you are allowed to inspect at all.
If the database fails to recognize an eligible study as belonging to your target population, filtering can remove the record before screening. This follows the broader principle that not every concept in the research question must become a mandatory search condition.
Authors may describe populations differently from your protocol
Population terminology rarely follows one tidy vocabulary. Children may be described as pediatric patients, school-aged participants, pupils, minors, infants, neonates, adolescents, or by specific ages. Older adults may be described using terms such as elderly, geriatric, seniors, ageing populations, retirement-age adults, or simply by an age distribution reported in the methods.
Other populations can be even more context-dependent. "University students" might appear as undergraduates, college students, tertiary students, medical students, nursing students, preservice teachers, or participants recruited from a university without a generic student label in the title.
Adding alternative expressions with OR can improve coverage. Still, no vocabulary list can guarantee retrieval when the relevant characteristic is absent from the searchable fields.
Age is especially difficult to translate into database categories
Age appears objective because it is numerical, but age filtering creates several retrieval problems. Database age categories may overlap, their boundaries may not match your eligibility criteria, and relevant studies can include mixed-age samples.
Empirical work on age-specific MEDLINE searching has documented difficulties caused by inconsistent indexing, overlapping age categories, and the distribution of relevant literature across journals. This is why an age restriction should not be assumed to map neatly onto the population definition in a protocol.
Imagine that your review includes participants aged 18 to 25. A study of university students has a mean participant age of 21.2 but includes several 17-year-olds and 26-year-olds. Another article reports only "first-year undergraduate students" in its abstract and provides the age distribution in the full text. A rigid database age filter may behave differently from the eligibility decision you would make after reading those studies.
Population information may be missing from searchable metadata
Authors cannot put every study characteristic in a title or abstract. A record may emphasize the intervention, condition, or principal finding while providing detailed participant information only in the methods section or a table.
This problem is particularly relevant when population membership is inferred from several characteristics rather than represented by one obvious term.
A study can therefore satisfy your population criterion while failing the database's textual or indexing representation of that criterion. Whether searching the full text rather than bibliographic metadata would help is a separate retrieval decision, since full-text searching introduces its own precision problems.
Indexing is useful but should not be treated as infallible
Controlled vocabulary and demographic indexing can improve population retrieval. However, indexing practices differ between databases, and population characteristics are not always represented consistently.
Cochrane's guidance on equity and specific populations specifically warns that standard database filters can pose risks because many population-related characteristics are not consistently indexed. It also notes that pediatric studies may be poorly indexed and that authors of child-specific research may fail to use pediatric terminology explicitly even within the article.
For pediatric searching, Cochrane consequently recommends using a pediatric search filter that combines subject headings, age limits where available, and free-text terms rather than relying on indexing or age limits alone.
This distinction matters. "Avoid population filters" does not mean "never search the population." It means that an apparently convenient restriction may be insufficient when the population requires richer representation.
Broad population concepts can still be useful search blocks
Some questions cannot be retrieved meaningfully without a population concept. If your intervention or phenomenon occurs across many populations, population terminology may be necessary to distinguish the literature you need.
The choice is therefore not simply between filtering and ignoring the population. You may have several options:
| Approach |
Potential advantage |
Potential limitation |
| Built-in population or age filter |
Fast and easy to apply |
Depends on database categories and metadata that may not capture every eligible study |
| Free-text population block |
Can represent terminology used by authors |
May miss unexpected wording or records that omit population terminology |
| Controlled vocabulary plus free text |
Combines indexing with author terminology |
Still depends on the quality and coverage of both approaches |
| Tested population filter |
May provide documented retrieval performance |
Performance may depend on database, population, topic, and context |
| No population restriction |
Reduces the risk of exclusion caused by population terminology |
May produce a much larger screening burden |
Mixed populations create another problem
Relevant evidence may be embedded in studies containing participants both inside and outside your target population.
Suppose your review concerns adults aged 65 and older. A study enrolls participants aged 55 to 80 and reports results separately for those aged 65 and above. Depending on your protocol, the older subgroup may provide eligible evidence. Yet a database-level age restriction might not classify the record in the way your review requires.
The same issue arises with studies spanning school levels, professional groups, geographic populations, disease severities, or other mixed samples.
Database filters make decisions at the record level. Eligibility decisions may depend on information at the subgroup or analysis level. Those are not always equivalent.
Very narrow populations deserve particular caution
The more specific the population, the more tempting it becomes to encode every characteristic into the search. A review might concern female nursing students aged 18 to 24 in rural universities, for example.
Requiring separate search blocks for sex, profession, student status, age, geographic setting, and institutional setting would create multiple opportunities for an eligible record to fail retrieval. Every additional mandatory concept can narrow the search further.
Some of those characteristics may be better applied during screening, especially when they are reported reliably only in the full article.
The cost of avoiding a population filter is usually more screening
Removing a population restriction can increase retrieval dramatically. That does not automatically mean removing it was wise. Search strategy design involves balancing the risk of missed evidence against the resources required to screen irrelevant records.
If a tested population strategy reduces 20,000 records to 4,000 while retaining the studies that should be found, it may be highly useful. If a convenient age limit reduces 4,000 records to 600 but quietly removes eligible studies, the apparent efficiency is misleading.
This is fundamentally a sensitivity-versus-precision decision.
Population filters should be tested rather than trusted by default
Compare the search with and without the population restriction. Check whether known eligible studies remain. Inspect a sample of records lost when the filter is applied.
For consequential evidence syntheses, a population filter should ideally have some evidence supporting its performance in the relevant database and context. PRESS treats limits and filters as a specific domain requiring peer review, reflecting the fact that seemingly minor restrictions can materially change retrieval.
Watch Out
A population filter can make the results look impressively clean while hiding the records it removed. Do not judge it solely by the reduction in screening workload.
06 · What This Means for You
Use Population Restrictions Only When Their Retrieval Behavior Is Defensible
Begin with the population definition in your protocol, then examine how relevant studies represent that population in the databases you plan to search. Look beyond the exact terminology in the research question.
If the population is difficult to identify reliably, consider whether a broader search followed by screening offers a safer trade-off.
A simple decision framework
If the population is distinctive and consistently represented
A population search block may improve precision without substantial retrieval loss.
If age or population metadata are incomplete or poorly aligned with your eligibility definition
Avoid relying solely on a built-in population limit.
If a tested population filter exists for the relevant database
Examine its construction and reported performance before deciding whether it suits your search.
If eligible studies contain mixed populations
Be cautious about record-level restrictions that may remove studies containing an eligible subgroup.
If applying the filter removes known relevant studies
Investigate why and reconsider the restriction rather than accepting the smaller result set.
For systematic reviews, record the exact population terms, filters, and limits used. If you intentionally leave a population characteristic for screening, document that decision as part of search development rather than allowing it to become an undocumented convenience.
