name: PubMed
description: Search and evaluate biomedical literature with effective queries, filters, and critical appraisal.
metadata: {"clawdbot":{"emoji":"π¬","os":["linux","darwin","win32"]}}
PubMed Research Rules
Query Construction
Use MeSH terms for precise searching β controlled vocabulary ensures you find related concepts regardless of wording
Boolean operators must be uppercase: AND, OR, NOT β lowercase is ignored
Phrase searching with quotes: "heart failure" not heart failure β unquoted searches terms separately
Field tags narrow searches: [Title], [Author], [MeSH Terms] β example: aspirin[Title] AND prevention[MeSH]
Truncation with asterisk: therap* finds therapy, therapies, therapeutic
Essential Filters
Article type matters: Clinical Trial, Systematic Review, Meta-Analysis β filter by study design
Publication date for recent evidence β older studies may be superseded
Free full text filter if access is limited β but don't ignore paywalled high-quality studies
Humans filter excludes animal studies β relevant for clinical questions
Language filter if translation isn't feasible
Study Hierarchy
Systematic reviews and meta-analyses synthesize multiple studies β start here for established topics
Randomized controlled trials (RCTs) are gold standard for interventions β but not all questions are answerable by RCT
Cohort studies for long-term outcomes and rare exposures
Case-control for rare diseases
Case reports are lowest evidence β interesting but not generalizable
Guidelines synthesize evidence into recommendations β check who wrote them and when
Critical Appraisal
Sample size matters β small studies may show effects that don't replicate
Check confidence intervals, not just p-values β narrow CI with meaningful effect size beats p<0.05
Funding source and conflicts of interest affect interpretation β industry-funded studies favor sponsors
Primary vs secondary outcomes β cherry-picking significant secondary outcomes is common
Intention-to-treat vs per-protocol analysis β ITT is more conservative and realistic
Common Traps
Abstract conclusions may oversell results β read methods and results sections
Single studies rarely settle questions β look for replication and systematic reviews
Statistical significance isn't clinical significance β 1% improvement may not matter to patients
Retracted papers still appear in searches β check Retraction Watch for controversial papers
Predatory journals publish low-quality research β verify journal reputation
Preprints haven't been peer-reviewed β useful for speed but not vetted
