# MeSH Terms vs. Keywords in PubMed: How to Combine Them MeSH terms are the standardized vocabulary that indexers assign to every MEDLINE article, while keywords are the raw words you and the authors actually type. The single most useful rule of thumb is this: use MeSH terms for consistency and completeness, use keywords for currency and specificity, and in any search that matters (especially a systematic literature search), combine both with OR inside each concept block and AND between concept blocks. Understanding the difference between MeSH terms and keywords in PubMed is one of those skills that separates a quick keyword poke from a search you can defend in a thesis or a published review. Published reviews almost never rely on one or the other. A review of DOACs in valvular atrial fibrillation used "a combination of medical subject headings (MeSH) terms and relevant keywords" across PubMed, Embase, and Web of Science [1]. A cardiac rehabilitation meta-analysis searched PubMed, Embase, and CINAHL "using keywords and MeSH terms on heart failure and cardiac rehabilitation" [2]. This article explains what each tool does, where each one fails, and how to put them together in a search line you would be comfortable publishing. ## What MeSH Terms Are ### The controlled vocabulary idea MeSH stands for Medical Subject Headings. It is the National Library of Medicine's controlled vocabulary, a curated list of roughly tens of thousands of descriptors organized into a tree structure. Trained indexers at NLM read each article added to MEDLINE and assign the headings that describe its content, regardless of the words the authors happened to use. That assignment solves a real retrieval problem. One research group may write "heart attack," another "myocardial infarction," another "MI," and another "coronary thrombosis." A keyword search has to guess every variant. A MeSH search does not, because all of those articles carry the same heading. This is also why MeSH terms support bibliometric work such as comparing the subject profiles of journals, where analysts use the frequency of headings like surgery, diagnostic imaging, and methods to characterize what a spine journal publishes [3]. ### Explosion and the tree structure MeSH headings sit in a hierarchy. The heading Heart Failure sits beneath Cardiovascular Diseases, and narrower headings such as diastolic heart failure and systolic heart failure sit beneath it. By default, PubMed explodes a MeSH term, meaning a search for Heart Failure[Mesh] automatically retrieves articles tagged with any of its narrower headings. If you want only the exact heading without its children, you turn explosion off with the NOEXP tag. ### Subheadings and major topic MeSH headings can be qualified with subheadings that narrow the search to an aspect of the topic. Heart Failure/drug therapy retrieves articles about drug treatment of heart failure, and Heart Failure/therapy narrows to treatment in general. You can also restrict to major-topic articles with [MAJR], which retrieves only papers where the concept is the central subject rather than a passing mention. Both tools raise precision, which is valuable when an unqualified search returns thousands of tangential hits. ## What Keywords Are ### Free text and field tags Keywords, or free-text terms, are any words or phrases you search without a controlled heading. PubMed searches them against the fields you specify. The two most common tags are [tiab], which searches the title and abstract, and [tw], which searches text words in title, abstract, and other indexed text fields. Without any tag, an unqualified phrase goes through Automatic Term Mapping, which tries to interpret your input (more on that below). Free text has one enormous advantage: it exists immediately. An article appears in PubMed the day it enters the database, but MeSH indexing happens after publication. A brand-new paper on a fresh technique has keywords in its abstract today and may wait days or weeks for its headings. If you search only MeSH, you can miss the newest literature entirely. ### Synonyms, abbreviations, and author choices The other advantage of keywords is precision at the frontier of a field. New drug names, device brands, gene variants, and technique acronyms often exist as text long before any heading covers them. A review comparing eTEP and IPOM hernia repairs built its PubMed and Embase search from keywords such as "enhanced view extraperitoneal," "eTEP," "TEP," "laparoscopic retromuscular," and "IPOM," because those procedure names are exactly the kind of specific vocabulary that keyword lines capture well [4]. A globozoospermia review similarly combined "globozoospermia," "round-headed spermatozoa," "ICSI," and related terms with its MeSH lines [5]. The weakness of keywords is the mirror image of their strength. Authors choose their own words, and they do not coordinate. Miss one synonym, one British or American spelling, or one abbreviation, and you miss papers. That is why a keyword block for a concept tends to grow long, listing every variant you and your team can brainstorm. ## Automatic Term Mapping, the Bridge Between the Two When you type an untagged phrase such as "heart failure" into the PubMed search box, PubMed runs Automatic Term Mapping (ATM). It checks your phrase against its lists of MeSH headings, journal titles, and author names, and if it finds a match it translates the query. "Heart failure" gets translated to include Heart Failure[Mesh], plus the text words. If nothing matches, PubMed breaks the phrase into individual words and combines them with AND. This mapping is convenient for quick searches and genuinely helpful for beginners, which is why many casual users search PubMed for years without ever learning MeSH. But ATM is a translator, not a search strategist. It will not add the synonyms you failed to type, and it can map a phrase in a direction you did not intend. Automated query-generation research treats ATM as a post-processing step precisely because the raw output of term generation needs that translation layer to behave properly in PubMed [6]. The reliable habit is to click "Search details" under the search box and read exactly how PubMed translated your query before you trust it. ## Side-by-Side Comparison | Feature | MeSH Terms | Keywords (Free Text) | |--|--|--| | What it is | Controlled vocabulary assigned by NLM indexers | Words and phrases as written by authors | | Who controls it | National Library of Medicine | No one, author by author | | Synonym handling | One heading covers many author word choices | You must list every synonym yourself | | Availability for new articles | Added during post-publication indexing | Present the moment the record appears | | Precision | High, especially with subheadings and [MAJR] | Variable, depends on your terms | | Recall (completeness) | High within indexed records | High only if your synonym list is exhaustive | | Explosion | Automatic, includes narrower headings | None | | Subheadings | Available (for example /drug therapy) | Not applicable | | Spelling variants | Neutral, one heading per concept | Must include behavior and behaviour, pediatric and paediatric, and so on | | Typical failure mode | Misses the newest, not-yet-indexed papers | Misses papers whose authors chose different words | | Best used for | Concept completeness in indexed literature | New topics, specific techniques, abbreviations, recent records | ## How to Combine Them ### The concept-block pattern ![Flow of steps to split a question into concepts, fill each block with MeSH and keywords joined by OR, then join blocks with AND.](/article-images/illustrations/mesh-terms-vs-keywords-pubmed-concept-block-build.webp) The standard architecture of a serious PubMed search is called the concept-block pattern, and it is the structure behind virtually every published search strategy. 1. Break your research question into two to four core concepts. 2. For each concept, build one block containing every MeSH heading and every keyword variant for that concept, joined with OR. 3. Join the blocks with AND. 4. Add method filters (such as a randomized controlled trial filter) as a separate block if needed. The reason this works is that MeSH and keywords cover different gaps in the same concept. A MeSH heading catches the papers whose authors used unfamiliar wording. A keyword line catches the newest papers that carry no heading yet, plus author keywords that fall outside the indexer's choices. Removing either half removes a different set of relevant records, which is why reviews describing their methods so consistently report the combination. A scoping review of REMS programs searched PubMed, Embase, and Web of Science "using the relevant Medical Subject Headings terms and related keywords" [7]. A COVID-19 vaccination review in inflammatory bowel disease searched four databases "using MESH terms and the related keywords" [8]. A postpartum anemia meta-analysis used "keywords and MeSH terms such as anemia, low hemoglobin, postpartum, postnatal women, and Ethiopia" [9]. A zinc and abortion meta-analysis described its search as using both main and MeSH keywords targeting abortion, pregnancy loss, and zinc [10]. ### Worked example 1: cardiac rehabilitation in heart failure A concept-block version of the cardiac rehabilitation question, informed by the pairing of topics in a published meta-analysis [2], looks like this: ("Heart Failure"[Mesh] OR heart failure[tiab] OR "chronic heart failure"[tiab] OR "cardiac failure"[tiab]) AND ("Cardiac Rehabilitation"[Mesh] OR "cardiac rehabilitation"[tiab] OR "exercise training"[tiab] OR "exercise therapy"[tiab]) Reading it line by line, the first block asks for the concept heart failure, satisfied either by the MeSH heading or by any of the text variants. The second block does the same for the intervention. The AND at the end requires both concepts in the same article. If you wanted to restrict to major-topic indexing, you would swap [Mesh] for [MAJR] on one or both lines. If the drug-treatment angle mattered, you would add Heart Failure/drug therapy as a variant inside the first block. ### Worked example 2: a keyword-heavy topic Some topics lean heavily on keywords because the vocabulary is new, branded, or abbreviation-driven. The eTEP versus IPOM hernia repair question is a good model [4]. Here the strategy is still concept blocks, but the MeSH contribution shrinks and the keyword lines carry the load: ("eTEP"[tiab] OR "enhanced view totally extraperitoneal"[tiab] OR "totally extraperitoneal"[tiab] OR TEP[tiab] OR "laparoscopic retromuscular"[tiab]) AND ("IPOM"[tiab] OR "intraperitoneal onlay mesh"[tiab]) AND ("Ventral Hernia"[Mesh] OR "Hernia, Incisional"[Mesh] OR "ventral hernia"[tiab] OR "incisional hernia"[tiab]) Notice that the hernia concepts still get their MeSH lines. Even in a keyword-heavy search, the headings add safety, because indexers will have tagged relevant papers with the hernia headings even where the abstract avoids the exact technique name. ### Refinements that matter A few tags and habits change results dramatically: - Use [tiab] on keyword lines rather than leaving them untagged. Untagged terms trigger ATM, which can silently reinterpret your phrase. Tagging keeps the line literal. - Use double quotation marks around multiword phrases so PubMed treats them as phrases rather than loose word collections. - Remember that the asterisk for truncation (for example rehabilitat*) disables ATM for that term, so truncated terms should carry an explicit tag. - Check "Search details" after every exploratory search to see the translation PubMed actually performed. - Use the MeSH Database (linked from the PubMed homepage) to verify that a heading exists, to see its tree position, and to see which subheadings are common for it before you write the line. ### Document everything Whatever you build, record it verbatim. Reproducibility is a defining requirement of a systematic literature search, and bibliometric work depends on standardized, reproducible strategies built from synonyms, abbreviations, MeSH terms, and Boolean operators [11]. Full reviews report searches across many databases, not just PubMed. A palliative care review of Chinese societies searched eight sources including PubMed/MEDLINE, Scopus, Web of Science, APA PsycInfo, CINAHL, and three Chinese databases [12]. A protocol for a network meta-analysis listed MEDLINE, EMBASE, the Cochrane Library, PsycINFO, CINAHL, and ClinicalTrials.gov [13]. When you adapt your PubMed strategy for those databases, remember that each has its own vocabulary system. Embase uses Emtree headings, and CINAHL uses CINAHL Subject Headings, so a MeSH line pasted verbatim into Embase will not perform the same work. ## When to Use MeSH Terms vs. Keywords ![Path asking if the topic is new, abbreviation-driven, full of synonyms, or needs aspect precision, ending in using both tools together.](/article-images/illustrations/mesh-terms-vs-keywords-pubmed-mesh-or-keyword-lead.webp) You rarely choose between them for a serious search, but the balance shifts with your situation, and knowing when to use MeSH terms as the backbone versus when keywords lead will save hours. Lean on MeSH when: - Your topic is established, well-indexed, and stable in the vocabulary. - You need recall above all, as in a systematic review or meta-analysis. - Your concept suffers from synonym chaos (many author word choices for one idea). - You want aspect-level precision through subheadings or major-topic restriction. Lean on keywords when: - The topic is new, and articles have not been indexed yet. - Your concept is an abbreviation, device name, drug brand, or technique acronym with no heading (or a heading assigned inconsistently). - You need an exact phrase that the controlled vocabulary blurs. - You are searching very recent literature during a fast-moving event. A helpful mental model comes from how published reviews describe their own work. Reviews of established clinical questions, such as comparing 2D and 3D imaging systems in laparoscopic hysterectomy, report MeSH-plus-keyword strategies across PubMed/MEDLINE and Web of Science [14]. Reviews of rapidly evolving topics, such as the nonauditory effects of infrasound and low-frequency noise, also search eight databases but must contend with terminology that has shifted over time across exposure scenarios such as wind turbine noise and occupational noise [15]. In both cases the honest answer to "mesh terms or keywords pubmed?" is both, with the weighting determined by how mature the vocabulary for your topic is. For students learning how to write a literature review, the practical sequence is: draft your question, identify two to four concepts, harvest candidate MeSH headings from the MeSH Database and candidate keywords from the title and abstract fields of two or three papers you already know are relevant, then assemble the blocks. Test the search on known relevant papers. If a paper you want is missing, find out which block failed and add the term that would have caught it. ## Common Mistakes and Limitations These are the errors that most often damage a PubMed search in practice: - Trusting Automatic Term Mapping blindly. ATM is a convenience layer. Read "Search details" and confirm the translation, especially for ambiguous phrases that may match a journal title or author name. - Searching MeSH only and losing the newest literature. Articles appear in PubMed before indexing. Any time-sensitive search needs keyword lines against [tiab]. - Searching keywords only and losing synonym coverage. Author word choice varies enormously. One group writes "quality of life," another "health-related quality of life," another "HRQoL." The heading Quality of Life catches all of them. - Ignoring explosion or misjudging it. Explosion broadens MeSH searches automatically, which is usually what you want, but occasionally pulls in narrower headings you did not intend. Check the tree in the MeSH Database. - Writing keyword lists in only one spelling culture. Include American and British variants for clinical terms, since MEDLINE contains literature from many countries. - Bot Boolean precedence errors. PubMed applies Boolean operators in a defined order, and without explicit parentheses around each concept block your OR and AND combinations may not do what you think. Always parenthesize blocks. - Bloating the strategy with redundant keyword lines. ATM already adds the MeSH heading for common untagged phrases. Duplicating that work makes your recorded strategy harder to reproduce and audit. - Assuming every concept has a MeSH heading. Rare diseases, novel procedures, and very recent drug classes can lack headings, or have them applied inconsistently for years. The globozoospermia example, a rare genetic syndrome, required a keyword-rich strategy alongside its MeSH lines for exactly this reason [5]. - Copying a PubMed strategy into other databases unmodified. Emtree, CINAHL Subject Headings, and Scopus keyword structures differ from MeSH. Rewrite each block per database, as multi-database reviews routinely do [13]. - Failing to record the strategy with its date. Databases change daily. A systematic search is reproducible only if the exact query strings, filters, and run dates are archived [11]. One structural limitation deserves plain acknowledgment. For systematic reviews, methodological guidance such as PRISMA-shaped workflows expects searches across multiple databases with documented strategies, peer review of the search where feasible, and screening of titles and abstracts by two reviewers, which is the pattern visible in the methods of the reviews cited throughout this article [12] [13] [16] [17]. PubMed skill alone does not substitute for that structure. ## Frequently Asked Questions ### What is the difference between MeSH terms and keywords in PubMed? MeSH terms are standardized headings assigned to articles by trained indexers at the National Library of Medicine, so one heading unifies all author word choices for a concept. Keywords are free-text words you search against fields like the title and abstract, so they match exactly what authors wrote. MeSH gives consistency, keywords give currency and literal matching. ### Why do systematic reviews use both MeSH terms and keywords? Because each covers a different gap. MeSH lines catch relevant papers whose authors used unexpected wording, while keyword lines catch the newest records that have no headings yet and capture abbreviations and technique names the vocabulary lacks. Nearly every published search strategy in this area describes the combination, and professional guidance for systematic literature searches expects both. ### Do brand-new articles have MeSH terms? Not immediately. Articles enter PubMed right away, but indexing with MeSH headings happens after publication, so the newest records can only be found by keywords. Any search that must capture very recent papers needs keyword lines tagged to [tiab] or [tw]. ### How do I find the correct MeSH term for my topic? Open the MeSH Database from the PubMed homepage, search your concept, and inspect the heading, its tree position, and its common subheadings. A second trick is to find one known relevant article, click on its MeSH terms listed on the record, and borrow the headings the indexers actually used. ### What does the [tiab] tag do, and should I use it? The [tiab] tag restricts a term to the title and abstract fields and prevents Automatic Term Mapping from reinterpreting it. Using it on keyword lines keeps your search literal and predictable, which matters when you need to document and reproduce a strategy. Untagged terms are translated, sometimes in ways you did not intend. ### How do I know how PubMed translated my search? Run the search and look at "Search details," displayed beneath the search box. It shows the exact translation, including which MeSH headings ATM applied and how your terms were combined. Checking it after every exploratory search is the fastest way to catch mapping surprises before they cost you results. ## Sources 1. [Comparison of Effectiveness and Safety of Apixaban, Dabigatran, and Rivaroxaban in Patients With Valvular Atrial Fibrillation: A Network Meta-Analysis of Randomized-Control Trials and Observational Studies.](https://pubmed.ncbi.nlm.nih.gov/38707166/) 2. [Effects of Comprehensive Outpatient Cardiac Rehabilitation on Exercise Capacity, Functional Status, and Quality of Life in People With Heart Failure: A Systematic Review and Meta-Analysis.](https://pubmed.ncbi.nlm.nih.gov/37658777/) 3. [Comparing the similarity and differences in MeSH terms associated with spine-specific journals using the forest plot: A bibliometric analysis.](https://pubmed.ncbi.nlm.nih.gov/36343077/) 4. [The Comparison of eTEP and IPOM in Ventral and Incisional Hernia Repair: A Systematic Review and Meta-analysis.](https://pubmed.ncbi.nlm.nih.gov/35034068/) 5. [Globozoospermia: A Case Report and Systematic Review of Literature.](https://pubmed.ncbi.nlm.nih.gov/36047070/) 6. [Empowering open medium-sized generative language models for effective structured search in biomedical systematic reviews.](https://pubmed.ncbi.nlm.nih.gov/42107249/) 7. [Implementation and Use of Risk Evaluation and Mitigation Strategies Programs in Practice: A Scoping Review of the Literature.](https://pubmed.ncbi.nlm.nih.gov/36482261/) 8. [Comparison the effects and side effects of Covid-19 vaccination in patients with inflammatory bowel disease (IBD): a systematic scoping review.](https://pubmed.ncbi.nlm.nih.gov/35987619/) 9. [Prevalence and associated factors of anemia among postpartum mothers in public health facilities in Ethiopia, 2024: a systematic review and meta-analysis.](https://pubmed.ncbi.nlm.nih.gov/38678183/) 10. [Comparison of zinc levels in mothers with and without abortion: A systematic review and meta-analysiss.](https://pubmed.ncbi.nlm.nih.gov/38774104/) 11. [Research hotspots and evolution trends of human amniotic fluid stem cells: A cite space bibliometric analysis based on WoSCC and PubMed.](https://pubmed.ncbi.nlm.nih.gov/42888595/) 12. [Psychosocial and spiritual needs of children at the end of life and their families in Chinese societies: A systematic review.](https://pubmed.ncbi.nlm.nih.gov/42995460/) 13. [Palliative care educational intervention versus cognitive behavioral therapy for health-related quality of life in patients with incurable cancer: a protocol for a systematic review and network meta-analysis.](https://pubmed.ncbi.nlm.nih.gov/42989330/) 14. [The comparison of 2D and 3D systems in total laparoscopic hysterectomy: a systematic review and meta-analysis.](https://pubmed.ncbi.nlm.nih.gov/39180564/) 15. [Nonauditory Effects of Infrasound and Low-Frequency Noise in Humans: A Systematic Literature Review.](https://pubmed.ncbi.nlm.nih.gov/42976015/) 16. [Effectiveness of literature-based initiatives to enhance empathy in healthcare populations: A systematic review and meta-analysis.](https://pubmed.ncbi.nlm.nih.gov/42980381/) 17. [Comparative evaluation of shear bond strength of composite restorations and glass-incorporated restorations with calcium silicate-based cement liner: A systematic review and meta-analysis.](https://pubmed.ncbi.nlm.nih.gov/42936330/)