Written and maintained by CASRAI Editorial Board
Last updated
PubMed Clinical Queries is a set of five pre-built search filters — Therapy, Diagnosis, Etiology, Prognosis, and Clinical Prediction Guides — layered on top of a normal PubMed search, each run at either a Broad or a Narrow scope. It exists to solve one specific problem: a plain-text PubMed search for a clinical question returns a mix of case reports, animal studies, editorials, and the occasional well-designed clinical study, and sorting that mix by hand is slow. Clinical Queries filters that mix down automatically using methodologic terms tuned to the study design a clinician actually needs, not the subject matter.
This page covers what each category and scope setting actually does, where the filters come from, what they will not tell you, and — the question that matters most if you are working from a research office or writing a protocol — when Clinical Queries is the right tool versus when you need to build and document your own filtered search instead.
What Clinical Queries actually is
Clinical Queries lives at pubmed.ncbi.nlm.nih.gov/clinical and is also reachable from the links on the PubMed homepage. It is not a separate database — it searches the same MEDLINE/PubMed index as a normal search box, but it automatically appends a validated methodologic filter to whatever subject terms you enter, restricting results to records that match a particular study design and evidence purpose.
The National Library of Medicine attributes the underlying filters to research by Haynes RB and colleagues, whose group at McMaster University spent years testing candidate search filters against hand-searched reference sets of methodologically strong clinical studies to see which combinations of subject headings, publication types, and text words best separated strong studies from everything else in MEDLINE. That validation work is why Clinical Queries filters are a meaningfully different tool from an ad hoc AND search of your own field-tagged terms: each one was tested for how well it actually performs at retrieving (or excluding) the type of study it targets, not just assembled by intuition.
One limit worth stating up front: the PubMed interface does not publish the literal search strings each filter runs. You choose a category and a scope, PubMed applies the corresponding filter behind the scenes, and you see the retrieved records — not the query. That matters for the “build your own” question below.
The five clinical study categories
Clinical Queries organizes filters under “Clinical Study Categories.” Pick the one that matches the kind of clinical question you are asking, not the disease or population — the disease/population terms are what you type into the search box; the category is the filter layered on top:
- Therapy — studies evaluating the effect of an intervention (a drug, procedure, or other treatment) on outcomes. Use this for “does treatment X help condition Y” questions.
- Diagnosis — studies evaluating how well a test or sign discriminates between people who do and do not have a target condition. Use this for “how accurate is test X for detecting Y” questions.
- Etiology — studies evaluating whether an exposure is associated with, or causes, an outcome. Use this for “does exposure X increase the risk of Y” questions, including harms of an intervention.
- Prognosis — studies following patients over time to identify factors that predict a future outcome. Use this for “what happens to patients with condition X” or “which factors predict outcome Y” questions.
- Clinical Prediction Guides — studies deriving or validating a clinical prediction rule: a formal combination of findings used to estimate the probability of a diagnosis or outcome. Use this for questions about a named rule or score, or about building one.
PubMed also runs a separate, narrower COVID-19 category (General, Mechanism, Transmission, Diagnosis, Treatment, Prevention, Case Report, Forecasting, and Long COVID) on the same Clinical Queries page. It works the same way — pick a sub-category, get a filtered set — but it is a distinct set of filters from the five clinical study categories above, built for pandemic-era retrieval rather than the general Haynes filter set.
Broad vs. Narrow scope: the sensitivity/specificity trade-off
Every one of the five clinical study categories can be run at one of two scopes, and this choice matters as much as the category itself:
- Broad (sensitive) scope casts a wider net. It is built to catch as many genuinely relevant studies as possible, at the cost of also pulling in more studies that turn out not to be relevant once you screen them. Use Broad when missing a relevant study would be worse than spending extra time screening false positives — the classic case is any search that is meant to be comprehensive.
- Narrow (specific) scope is the opposite trade: fewer results, but a higher proportion of them are actually on-target. Use Narrow when you want a short, high-precision list fast — the classic case is a point-of-care question where a clinician needs a handful of directly relevant studies in the next few minutes, not an exhaustive set to screen later.
Neither setting is “better” in the abstract; they answer different questions. A search built for speed at the point of care should default to Narrow. A search built to make sure nothing was missed should default to Broad — and, as the next section covers, even Broad is not the same thing as a search built to be comprehensive and reproducible.
What Clinical Queries is not: the Systematic Reviews filter is separate
It is easy to assume “Broad scope” on Clinical Queries is PubMed’s tool for finding systematic reviews. It is not. PubMed has a distinct, separate mechanism for that: the Systematic Review article-type filter in the sidebar of a normal search, or appending AND systematic[sb] to a query in the main search box. That filter is not one of the five Clinical Study Categories and does not have a Broad/Narrow toggle — it is a different tool answering a different question (“is this itself a systematic review,” rather than “is this a strong primary study of a given design”).
If what you actually need is to locate existing systematic reviews on a topic, see CASRAI’s guide to the PICO framework for formulating a review question and the Cochrane Handbook entry, rather than reaching for Clinical Queries’ Broad scope and assuming it is doing that job.
When to use Clinical Queries vs. build your own filtered search
This is the decision that actually matters once you know what the five categories and two scopes do. Clinical Queries and a custom filtered search solve different problems, and the difference comes down to two things Clinical Queries does not give you: the literal search string, and a citable methods statement.
| Situation | Use Clinical Queries | Build your own filter |
|---|---|---|
| Point-of-care question, need an answer in minutes | Yes — this is exactly what it is built for | No — too slow for the purpose |
| Quick check of “has anyone studied this” during protocol planning | Yes, as a first pass | Not yet — save the real build for the registered protocol |
| Systematic review, meta-analysis, or any search that must be reported in a PRISMA-S / methods section | No | Yes — a peer reviewer or a registry like PROSPERO needs to see the literal strategy you ran, and Clinical Queries does not expose one |
| Search needs to be reproduced exactly by someone else later, or re-run identically across multiple databases | No | Yes — a documented strategy (built from MeSH terms, free text, and, where appropriate, a published validated hedge) can be copied, translated to Embase or CINAHL, and re-run |
| You need to justify, in writing, why certain study types were included or excluded | No — the filter logic is not published | Yes — every inclusion/exclusion term in a custom strategy is visible and defensible |
In practice, a lot of real searches use both, in sequence: Clinical Queries to get oriented fast and sanity-check that relevant literature exists, then a fully documented custom strategy — built with the same MeSH-plus-free-text approach covered in CASRAI’s guide to MeSH terms in PubMed and, for multi-database work, translating a PubMed strategy to Embase — for the search that actually goes in the manuscript or the protocol. Clinical Queries is a fast filter for judgment calls made by one person in real time; a documented strategy is evidence, built to survive someone else checking it.
Where to find Clinical Queries
Go directly to pubmed.ncbi.nlm.nih.gov/clinical, or from the PubMed homepage look for the “Clinical Queries” link under the search box. Enter your subject terms (the disease, drug, exposure, or test — not the study design), choose a Clinical Study Category, choose Broad or Narrow, and run the search. The results list works exactly like a normal PubMed results page — you can still apply additional filters (date range, article type, species) on top of it.
Frequently asked questions
What is PubMed Clinical Queries?
A tool at pubmed.ncbi.nlm.nih.gov/clinical that applies one of five pre-built methodologic search filters — Therapy, Diagnosis, Etiology, Prognosis, or Clinical Prediction Guides — on top of your own subject search, at either a Broad (sensitive) or Narrow (specific) scope, to surface studies of a particular design more quickly than a plain-text search would.
What’s the difference between Broad and Narrow scope?
Broad scope is tuned for sensitivity: it retrieves more studies, including more that turn out not to be relevant, so you are less likely to miss something. Narrow scope is tuned for specificity: fewer results, but a higher share of them are on-target. Use Narrow for a fast, focused answer; use Broad when completeness matters more than speed.
Who created the Clinical Queries filters?
The National Library of Medicine attributes them to research by Haynes RB and colleagues at McMaster University, who validated candidate search filters against hand-searched reference sets to determine which term combinations best retrieved methodologically strong studies of each design.
Can I see the exact search string Clinical Queries runs?
No. The PubMed interface applies the filter behind the scenes and shows you the results, not the underlying query syntax. That is the main reason Clinical Queries is not a substitute for a documented search strategy in a systematic review or protocol — there is nothing to cite or reproduce.
Is Clinical Queries the same as PubMed’s Systematic Review filter?
No. The Systematic Review filter (accessible as an article-type sidebar filter or via systematic[sb]) is a separate PubMed feature for finding systematic reviews themselves. It is not one of the five Clinical Study Categories and has no Broad/Narrow toggle.
Should I use Clinical Queries for a systematic review search?
Use it, if at all, only as an early orientation step. The actual search that goes into a systematic review protocol or manuscript needs to be a documented, reproducible strategy built from visible MeSH terms and free text, because reviewers, registries, and anyone attempting to replicate the review need to see exactly what was run — which Clinical Queries does not expose.
Sources: PubMed Clinical Queries tool and PubMed User Guide, National Library of Medicine (pubmed.ncbi.nlm.nih.gov/clinical, pubmed.ncbi.nlm.nih.gov/help), consulted 29 August 2026.
