Search Strategy Systematic Review: Build Yours

A search strategy systematic review is the structured, reproducible plan that defines exactly which databases you will search, which terms you will use, and how those terms are combined with Boolean operators to retrieve every relevant study on your research question. It is the single most consequential methodological decision in a systematic review, because every downstream step, from screening to meta-analysis, depends on the completeness and precision of your search. When your search is final, screen the records it returns after removing duplicates for a clean PRISMA count.

A poorly constructed search produces a biased review. An overly narrow search misses relevant studies. An overly broad search buries your team in thousands of irrelevant records. The search strategy is what separates a rigorous systematic review from a narrative literature review, and it is the component most scrutinized by peer reviewers, journal editors, and Cochrane assessors. Teams that want the search and screening handled to that standard often rely on a professional literature review writing service. In our systematic reviews, we build search strategies across 3-5 databases with librarian validation and PRESS peer review before executing a single search.

This guide walks you through the complete process of building a comprehensive, reproducible search strategy, from defining your PICO framework through MeSH and Emtree subject headings, Boolean logic, database selection, syntax translation, PRESS validation, and PRISMA 2020 documentation.

Why Search Strategy Matters

The search strategy is the foundation upon which every systematic review stands. The Cochrane Handbook for Systematic Reviews of Interventions, Chapter 4 (Lefebvre et al., 2023), dedicates an entire chapter to searching because the validity of a systematic review is directly proportional to the comprehensiveness of its search. A systematic review that misses relevant studies produces biased estimates of treatment effects, overstates certainty, and may lead to clinical decisions that harm patients.

Consider the consequences. If your search strategy fails to retrieve three randomized controlled trials that show no treatment effect, your meta-analysis will overestimate the intervention's benefit. If your search misses studies from non-English databases or grey literature sources, your review is vulnerable to publication bias. These are not theoretical risks, they are documented in methodological research comparing systematic reviews that used comprehensive versus restricted search strategies.

Sensitivity refers to the proportion of all relevant studies that your search retrieves. A highly sensitive search captures nearly every relevant study but also returns many irrelevant records. Specificity refers to the proportion of retrieved records that are actually relevant. The tension between sensitivity and specificity is the central challenge of search strategy design.

For systematic reviews, sensitivity takes priority over specificity. It is better to screen 5,000 records and find 50 relevant studies than to screen 500 records and miss 20 relevant studies. The cost of over-inclusion is time spent screening. The cost of under-inclusion is a biased review. Cochrane guidance is explicit: err on the side of sensitivity (Lefebvre et al., 2023).

Reproducibility is the second pillar. Every search strategy must be documented with enough detail that another researcher could execute the identical search and retrieve the identical set of records. This requirement is codified in PRISMA 2020, which mandates that authors present the full search strategy for at least one database in a supplementary appendix (Page et al., 2021). Without reproducibility, the systematic review cannot be updated, replicated, or audited.

Building from PICO, Six Steps to a Complete Search

Every search strategy begins with a well-defined research question, and the PICO framework is the standard tool for structuring that question. PICO stands for Population, Intervention, Comparison, and Outcome. Each PICO element becomes a concept block in your search strategy, and the way you translate PICO into search terms determines whether your strategy is comprehensive or incomplete.

As you work through the six steps below, you can turn a PICO question into a database-ready Boolean string and keep every concept block in one place.

Step 1: Define your PICO elements clearly. Write out each element in plain language before touching a database. For example: Population = adults with type 2 diabetes; Intervention = metformin; Comparison = sulfonylureas; Outcome = glycated hemoglobin (HbA1c). The clearer your PICO, the easier the translation. Use our interactive pico framework builder to structure your question before building your search.

Step 2: Generate synonyms and related terms for each PICO element. For "type 2 diabetes," your synonym list might include: type 2 diabetes mellitus, T2DM, non-insulin-dependent diabetes, NIDDM, adult-onset diabetes. For "metformin," include brand names: Glucophage, Fortamet, Glumetza. Do not rely on a single term for any concept, the vocabulary of medicine is broad, and authors use different terminology across disciplines, countries, and time periods.

Step 3: Identify MeSH terms and Emtree terms for each concept. Search the MeSH Browser (meshb.nlm.nih.gov) for your population and intervention terms. Note the preferred MeSH heading, its tree number, and any relevant narrower terms. If you are searching Embase, repeat this process using the Emtree thesaurus. Subject headings are the backbone of a systematic search because they standardize terminology, a study about "sugar diabetes" and a study about "type 2 diabetes mellitus" are both indexed under the same MeSH heading.

Step 4: Combine synonyms within each concept block using OR. Within the population block, combine all your synonyms and MeSH terms with OR. Within the intervention block, do the same. OR is the Boolean operator that broadens your search, it tells the database to retrieve records containing any of the listed terms.

Step 5: Combine concept blocks using AND. Once each PICO element has its own block of OR-combined terms, join the blocks with AND. AND is the Boolean operator that narrows your search, it tells the database to retrieve only records that contain at least one term from every block. A typical systematic review search combines two to four concept blocks with AND.

Step 6: Add filters and limits only when justified. Language filters, date limits, and study design filters should be applied only with explicit methodological justification. The Cochrane Handbook warns against routine use of language restrictions because they introduce bias (Lefebvre et al., 2023). If you restrict to English-language publications, document why and acknowledge the limitation.

This six-step process produces a search strategy that is comprehensive, transparent, and reproducible. For a detailed walkthrough of the broader systematic review process, see our guide on how to write a systematic review step by step.

MeSH Terms, Emtree, and Subject Headings

Subject heading systems are controlled vocabularies maintained by database producers to standardize the indexing of biomedical literature. Understanding how they work is essential for building a search strategy that captures every relevant study, not just those that happen to use your preferred terminology in their title or abstract.

MeSH (Medical Subject Headings) is the controlled vocabulary used by the National Library of Medicine to index articles in PubMed and MEDLINE. Every article in MEDLINE is reviewed by trained indexers who assign MeSH terms based on the content of the article, regardless of the terminology the authors used. This means that a search using the MeSH term "Diabetes Mellitus, Type 2" will retrieve articles about T2DM, NIDDM, adult-onset diabetes, and any other synonym, as long as the article has been indexed.

Emtree is the controlled vocabulary used by Elsevier to index articles in Embase. Emtree is larger than MeSH, containing over 90,000 preferred terms compared to MeSH's approximately 30,000. Emtree includes more drug terms, device terms, and disease terms than MeSH, making it particularly valuable for pharmacological and device-related systematic reviews.

CINAHL Subject Headings are used in the Cumulative Index to Nursing and Allied Health Literature. They are based on MeSH but include additional terms specific to nursing, physiotherapy, occupational therapy, and other allied health disciplines. If your systematic review includes nursing or rehabilitation outcomes, CINAHL subject headings capture studies that MeSH may miss.

The critical principle is this: always combine subject headings with free-text searching. Subject headings only work for articles that have been indexed, and indexing takes time, recently published articles may not have MeSH terms assigned for weeks or months after publication. Free-text searching (searching title, abstract, and keyword fields) catches these unindexed articles as well as articles where the indexer chose a different MeSH term than you expected.

Within each concept block, combine your subject headings and free-text terms with OR. This is the standard approach recommended by the Cochrane Handbook and taught in every information literacy course for systematic reviewers. For a deeper understanding of Boolean logic in literature searching, see our Boolean search strategy guide.

Database Selection and Syntax, A Comparison

Choosing which databases to search is a methodological decision that directly affects the comprehensiveness of your review. The Cochrane Handbook recommends searching MEDLINE and at least one other database, but the optimal set depends on your research question, discipline, and the types of studies you need to find.

PubMed is the most widely used database for biomedical systematic reviews and the logical starting point for most searches. It is free, its MeSH system is well-documented, and its search interface provides transparent automatic term mapping that shows you exactly how your query was interpreted. PubMed search strategy development benefits from the MeSH Browser, which lets you explore the hierarchical structure of subject headings and identify narrower terms you might otherwise miss.

Embase is essential for any systematic review in pharmacology, toxicology, or device-related topics. Embase indexes approximately 2,900 journals not covered by MEDLINE, with particularly strong coverage of European and non-English-language publications. Its Emtree thesaurus is more granular than MeSH for drug and device terms. Drug reviews also draw on regulatory and trial-registry sources, which our pharmacy and pharmacology review guide lists.

CINAHL adds value when your review involves nursing, allied health, or patient-reported outcomes. CINAHL indexes journals and conference proceedings not found in MEDLINE or Embase.

The Cochrane Library contains CENTRAL (the Cochrane Central Register of Controlled Trials), which aggregates trial records from multiple sources including MEDLINE, Embase, and hand-searched journals. Searching CENTRAL is particularly efficient for identifying randomized controlled trials.

Web of Science provides citation indexing, which enables forward citation searching, a strategy for finding studies that cited a known relevant article. This is a valuable supplement to traditional keyword searching and is recommended by the Cochrane Handbook as an additional search method.

For most biomedical systematic reviews, a minimum set of PubMed, Embase, and one discipline-specific database provides adequate coverage. Our dental systematic review guide shows how that choice plays out for oral health topics. Adding the Cochrane Library and Web of Science increases comprehensiveness further.

Want a comprehensive, peer-reviewer-approved search strategy? Our information specialists build and validate multi-database searches with full PRISMA documentation and sensitivity analysis. request an expert opinion on your project, or explore our systematic review search strategy support.

Translating Across Databases

One of the most common errors in systematic review search strategies is copying a PubMed search and pasting it directly into Embase or CINAHL. Each database uses different syntax, different field tags, different subject heading systems, and sometimes different truncation or proximity operators. A search that works perfectly in PubMed may retrieve zero results or wildly different results in another database if it is not properly translated.

A MeSH to Emtree syntax converter handles the mechanical part of this rewrite, though every translated line still needs a human check before you run it.

Subject heading translation is the most important step. Replace every MeSH term with its Emtree equivalent when moving to Embase. Use the Emtree browser to find the corresponding preferred term, it is not always a direct one-to-one match. Some MeSH terms map to multiple Emtree terms, and some Emtree terms have no direct MeSH equivalent. Similarly, when translating to CINAHL, replace MeSH terms with CINAHL Subject Headings.

Field tag translation requires changing the syntax that tells the database where to search. In PubMed, "[MeSH Terms]" searches the MeSH field. In Embase on Ovid, the equivalent is ".sh." or "/exp" on the Embase.com interface. In CINAHL on EBSCO, it is "(MH "term")". These differences seem minor but cause search failures when overlooked.

Truncation and wildcards are generally consistent across databases (the asterisk * is nearly universal), but proximity operators vary significantly. PubMed does not support proximity searching at all. Embase uses NEAR/n and NEXT/n. CINAHL on EBSCO uses Nn and Wn. Web of Science uses NEAR/n. If your PubMed search relies on phrase searching, you may want to use proximity operators in databases that support them to capture variations in word order.

Boolean operator handling is consistent across all major databases, AND, OR, and NOT work the same way, but the order of operations and nesting with parentheses can produce different results if not carefully managed. Always use explicit parentheses to control the logic of your search, and never rely on a database's default operator precedence.

A practical approach is to build and validate your search in PubMed first, then create a translation table that maps every PubMed element to its equivalent in each target database. This systematic translation process ensures nothing is lost in conversion and makes your search documentation complete for PRISMA 2020 reporting.

Our free free search translation tool automates much of this process, converting PubMed syntax into Embase, CINAHL, and Cochrane Library formats with the appropriate field tags and subject headings.

Validating Your Search, The PRESS Checklist

A search strategy should never be executed without validation. Just as a manuscript undergoes peer review before publication, a search strategy should undergo peer review before execution. The PRESS (Peer Review of Electronic Search Strategies) checklist, developed by McGowan et al. (2016), is the gold standard tool for this purpose.

PRESS provides a structured framework for evaluating six domains of search quality:

1. Translation of the research question. Does the search strategy accurately reflect the research question? Are all PICO elements represented? Has any concept been omitted or inadequately represented?

2. Boolean and proximity operators. Are AND, OR, and NOT used correctly? Is the logic sound? Are parentheses used appropriately to group terms within concept blocks?

3. Subject headings. Are the correct MeSH terms, Emtree terms, or other subject headings used? Are explosion (broader searching) and subheadings applied appropriately? Are any relevant subject headings missing?

4. Text word searching (free text). Are sufficient synonyms, spelling variations, abbreviations, and acronyms included? Is truncation applied correctly to capture word variants without introducing irrelevant terms?

5. Spelling, syntax, and line numbers. Are there typographical errors? Are field tags correct for the database being searched? Do line number references in combined searches point to the correct lines?

6. Limits and filters. Are any limits or filters applied? If so, are they justified? Could they introduce bias by excluding relevant studies?

The PRESS process involves having a second researcher, ideally an information specialist or librarian, review your search strategy against these six domains and provide written feedback. Studies have shown that PRESS peer review identifies errors in the majority of search strategies submitted for review, including errors that would have resulted in missed relevant studies (McGowan et al., 2016).

In practice, PRESS is the step most often skipped, because it requires a second person with search expertise that many teams do not have in-house. If no librarian or second information specialist is available to you, our information specialists peer-review the electronic searches against all six domains and return the strategy with the documentation PRISMA-S expects.

Beyond PRESS, you should validate your search empirically by testing it against a set of known relevant studies. Before finalizing your strategy, identify five to ten studies that you know are relevant to your research question, studies found during scoping searches, cited in previous reviews, or recommended by subject experts. Run your search and verify that it retrieves all of these studies. If any known relevant study is missing, analyze why and adjust your strategy accordingly.

This validation step is the closest thing to a quality assurance test for your search. If your strategy cannot find studies you already know exist, it is almost certainly missing other relevant studies you do not yet know about.

Common Mistakes in Systematic Review Search Strategies

Even experienced researchers make errors in search strategy construction. Understanding the most frequent mistakes helps you avoid them and builds a stronger, more defensible search.

Using only free-text terms without subject headings. Researchers who are unfamiliar with MeSH or Emtree often search only title and abstract fields with their own terminology. This misses studies indexed under different terms and fails to exploit the standardized vocabulary that makes database searching powerful. Always combine subject headings with free-text terms.

Failing to explode subject headings. In PubMed, searching a MeSH term without explosion retrieves only studies indexed under that exact heading. Exploding the term retrieves studies indexed under that heading and all narrower headings in the MeSH hierarchy. For most systematic review searches, explosion is appropriate and significantly increases sensitivity.

Copying searches between databases without translation. As discussed above, each database has unique syntax, field tags, and subject heading systems. Pasting a PubMed search into Embase without replacing MeSH terms with Emtree equivalents and adjusting field tags will produce an incomplete and unreliable search.

Using NOT to narrow results. The NOT operator excludes records containing a specified term, but it can inadvertently remove relevant studies. A study about "diabetes NOT gestational" will exclude any study that mentions gestational diabetes anywhere in its record, even if the study is primarily about type 2 diabetes and only mentions gestational diabetes in passing. The Cochrane Handbook recommends avoiding NOT unless absolutely necessary and carefully evaluating its impact on sensitivity.

Applying unjustified date or language limits. Restricting to English-language publications or to studies published after a certain date introduces bias unless there is a specific methodological justification. Peer reviewers routinely flag unjustified limits as a weakness.

Inadequate documentation. Many researchers record their search terms but not the complete search string with Boolean operators, field tags, and line numbers. PRISMA 2020 requires the full, executable search strategy, not a summary or paraphrase. Without complete documentation, the search cannot be reproduced, and the review's validity is undermined.

Neglecting grey literature. Published studies represent a biased sample of all research conducted, because studies with positive results are more likely to be published. Searching grey literature sources, clinical trial registries (ClinicalTrials.gov, WHO ICTRP), conference proceedings, theses and dissertations, and organizational reports, helps mitigate publication bias. The Cochrane Handbook recommends grey literature searching as a standard component of comprehensive systematic review searches.

Not involving a librarian or information specialist. The Cochrane Handbook explicitly recommends that systematic review teams include or consult an information specialist for search strategy development. Librarians are trained in database syntax, subject heading systems, search validation, and documentation standards. Studies comparing librarian-developed and researcher-developed searches consistently find that librarian searches are more comprehensive and contain fewer errors.

Documenting Your Search for PRISMA 2020

PRISMA 2020 (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) requires transparent reporting of search strategies as a core component of systematic review methodology (Page et al., 2021). Proper documentation serves three purposes: it enables reproducibility, it demonstrates rigor, and it satisfies journal and Cochrane submission requirements.

For each database searched, you must document:

The complete search string, every term, field tag, Boolean operator, and line number, exactly as executed. Do not paraphrase, summarize, or simplify. The reader should be able to copy your search string, paste it into the database, and retrieve the same results (subject to database updates since your search date).

The database name and interface, specify not just "MEDLINE" but "MEDLINE via Ovid" or "PubMed via pubmed.ncbi.nlm.nih.gov." The same database can produce different results depending on the interface used, because interfaces may implement search features differently.

The date the search was executed, record the exact date for each database. If you run updated searches before manuscript submission, document those dates and results separately.

The number of records retrieved, report the raw count before deduplication. This number, combined with the search string, allows readers to assess the scope and sensitivity of your search.

PRISMA 2020 recommends presenting the full search strategy for at least one database in the main manuscript or supplementary materials. Many journals now require full search strategies for all databases. The PRISMA-S extension (Rethlefsen et al., 2021) provides additional guidance specific to search reporting, including recommendations for documenting grey literature searches, hand-searching, and citation searching.

Maintaining a search log throughout the process, recording every iteration, modification, and rationale, makes PRISMA documentation straightforward. Without contemporaneous records, reconstructing your search history months later for manuscript preparation is error-prone and time-consuming.

For additional guidance on extracting and organizing the data your search retrieves, see our data extraction for systematic reviews guide. And for an overview of what to expect when working with a professional team on your review, read our guide on what to expect from a systematic review writing service.

Putting It All Together

Building a comprehensive, reproducible search strategy is one of the most technically demanding tasks in systematic review methodology. It requires familiarity with controlled vocabularies, Boolean logic, database-specific syntax, and validation frameworks, skills that take years to develop and that most researchers use infrequently enough that each new review requires relearning.

The process follows a clear sequence: define your PICO, generate comprehensive synonym lists, identify subject headings in each database's controlled vocabulary, combine terms with Boolean operators, translate the strategy across all target databases, validate with PRESS peer review and known-study testing, and document everything for PRISMA 2020 compliance.

The investment in a rigorous search strategy pays dividends throughout the review. A comprehensive search reduces the risk of bias in your findings. A well-documented search satisfies peer reviewers and journal requirements. A validated search gives you confidence that your results are based on the best available evidence.

Whether you build your search strategy independently, with the help of a librarian, or with professional support, the principles remain the same: be comprehensive, be systematic, be transparent, and be reproducible. Those four qualities are what distinguish a credible systematic review from a selective literature summary, and they all begin with the search strategy.

A comprehensive search extends beyond indexed databases. Our guide to grey literature sources for systematic reviews covers where and how to find unpublished studies.

After running your search, the screening phase requires dual review. Learn why you need a second reviewer for screening and how to manage disagreements.

To handle large search yields efficiently, consider using AI tools that accelerate screening and extraction.

After building your PubMed strategy, you need to adapt it for other databases. Our guide on translating searches between PubMed, Embase, and Cochrane covers syntax differences.

The search can also be commissioned on its own, separately from a full review. A search designed and documented by an information specialist covers the database-specific syntax, the PRESS peer review, and the complete search log.