Should You Use Built-In “Humans,” “Age,” or “Study Type” Filters?

Can You Trust the Filters Already Provided by the Database?

After running a database search, you may see convenient options such as "Humans," "Adult," "Child," "Randomized Controlled Trial," "Clinical Trial," or other study and population categories. If one of those labels matches your eligibility criteria, clicking it seems much easier than constructing another search block.

The filter itself may work exactly as designed. The more difficult question is whether what it was designed to retrieve is identical to what you mean.

Built-in filters often depend on indexing terms, publication types, database-specific search expressions, or other metadata. Those mechanisms can be useful, but they do not necessarily reproduce the judgments you would make after reading an article. Understanding what lies behind the checkbox is therefore part of search-strategy design.

Use Built-In Filters Selectively, Not Automatically

What Actually Happens When You Click a Database Filter?

The label you see is not necessarily the mechanism being used

A database interface may present a filter using an intuitive label such as "Humans" or "Young Adult." Behind that label is a technical retrieval rule.

PubMed provides a particularly transparent example because its documentation shows the search expressions behind many filters. The Humans filter corresponds to humans[mh]. Its age filters likewise use Medical Subject Headings (MeSH), such as "young adult"[mh] for ages 19 to 24 and aged[mh] for ages 65 and older.

This means the database is not reading every retrieved article and independently deciding whether its participants are human or whether their ages satisfy your protocol. It is retrieving records according to the metadata and indexing represented by those search expressions.

Built-in age filters depend on MEDLINE indexing in PubMed

PubMed explicitly warns that its age filters restrict results to citations assigned the corresponding MeSH age terms during MEDLINE indexing. As a result, relevant records can be excluded when those MeSH terms are absent. PubMed specifically identifies records that are not indexed for MEDLINE, including some preprints and Online Ahead of Print citations, as examples.

This is an important limitation because your eligibility criterion may concern the actual participants, whereas the filter concerns how the bibliographic record has been indexed.

Suppose your review includes adults aged 18 to 25. PubMed's documented categories include Young Adult at 19 to 24 years and broader adult categories beginning at 19 years. None is an exact representation of 18 to 25.

A precise age criterion in your protocol therefore does not guarantee a corresponding database filter. This is one reason you may sometimes need to avoid a population filter and assess the characteristic during screening.

The Humans filter can have the same indexing problem

In PubMed, the Humans filter is implemented through the Humans MeSH heading.

That sounds straightforward until a relevant human study lacks the indexing needed to satisfy the filter. The broader principle is that a study characteristic and its bibliographic representation are not the same thing.

If your search absolutely requires high sensitivity, particularly for a systematic review, relying on an indexing-dependent restriction should therefore be a deliberate decision rather than an automatic cleanup step.

Study Type can mean several different things

Study-type filters require particular care because the label "study type" can refer to publication types, methodological filters, database-generated subsets, or combinations of several search mechanisms.

PubMed's article-type filters commonly use publication types. For example, its documentation shows that many article types correspond to the article type name searched with the publication-type field tag. The Systematic Review filter is an exception because it uses a search strategy that includes, but is not limited to, the systematic-review publication type.

This distinction matters. A database filter labelled "Systematic Review" is not necessarily equivalent to searching:

"systematic review"[pt]

Nor should you assume that a filter for one evidence design uses the same logic as another.

Publication type and methodological design are related but not identical

A bibliographic publication-type label is metadata assigned to a record. A methodological design is a characteristic of the underlying study.

Often they align. Sometimes they do not.

An eligible randomized trial may not yet have the publication-type indexing expected by a filter. A paper may use terminology associated with a design without satisfying your methodological definition. A systematic review may be represented differently depending on the database and indexing state.

The two can correspond closely without being interchangeable.

A tested methodological filter may be more sophisticated than a simple checkbox

Methodological search filters can combine text words, indexing terms, publication types, and Boolean logic to identify particular evidence types. Their performance can then be evaluated against a reference set.

Research on filters for randomized controlled trials, for example, has compared multiple MEDLINE strategies against large gold-standard sets and reported sensitivity approaching 99% for the most sensitive filters.

Search-filter performance is not uniform across evidence types, however. A Cochrane review of filters for identifying systematic reviews found filters with different combinations of sensitivity, specificity, and precision, with particularly variable performance reported for Embase.

So "use a filter" is not a complete methodological instruction. You need to know which filter, designed for what purpose, in which database, and with what retrieval performance.

Different evidence types can require very different caution

Methodological filtering is particularly risky when the underlying design is difficult to identify from titles, abstracts, and indexing.

Diagnostic test accuracy research offers a useful example. An evaluation of 23 MEDLINE filters found sensitivities ranging from 20.6% to 86.9%, leading the authors to conclude that the tested filters did not provide an adequate sensitivity-precision trade-off for identifying studies for systematic reviews.

Another evaluation found that adding methodological filters to subject searches for diagnostic accuracy studies reduced sensitivity from 91% for the subject searches to between 43% and 87% for filtered searches.

These findings should not be generalized mechanically to every filter or evidence type. They demonstrate why study design may sometimes be better left out of the search strategy.

Filters involve sensitivity and precision trade-offs

A filter can remove thousands of irrelevant records and therefore improve precision. That may substantially reduce screening workload.

But if some relevant records disappear too, sensitivity decreases.

The appropriate trade-off depends on the search purpose. A researcher seeking a handful of representative articles may tolerate losses that would be difficult to justify in a systematic review intended to identify all eligible evidence.

A database filter can be accurate yet still inappropriate for your question

Suppose a database accurately applies its Young Adult category to records indexed as involving people aged 19 to 24. Your review includes people aged 18 to 25.

Nothing has malfunctioned. The database has correctly executed its own rule. The problem is that its category does not match yours.

This distinction prevents a common misunderstanding: filter problems are not necessarily database errors. They often arise because researchers ask a database category to stand in for a methodological criterion that was defined differently.

Convenience should not substitute for verification

Built-in filters are attractive because they are immediate. You can reduce 12,000 results to 2,000 with a click. That reduction may be completely justified, but the number itself provides no evidence that the right 10,000 records disappeared.

Before retaining a consequential filter, compare the results with and without it. Check known eligible studies. Examine records that disappear. If the filter has published validation evidence, consider whether that evidence applies to your database, evidence type, and purpose.

This follows the broader principle that database filters can accidentally remove relevant studies.

When “Young Adult” Does Not Mean Your Young-Adult Criterion

The problem was not that the filter was incorrectly labelled or malfunctioning. Its documented definition simply answered a slightly different question from the one in the review protocol.

Common Misconceptions About Built-In Filters

Look Behind the Checkbox Before You Use It

Before applying a built-in filter, find its database documentation. Determine whether it relies on controlled vocabulary, publication types, text searching, a predefined query, or another metadata field.

Then compare that mechanism with your actual eligibility rule.

For reproducible evidence synthesis, record which filters were used. A checkbox is still a search decision. Someone attempting to reproduce your search needs to know about it just as much as they need your keywords and Boolean operators.

Before Clicking Humans, Age, or Study Type

Questions About Humans, Age, and Study Type Filters

Use the Filter's Actual Definition, Not Just Its Label

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.