Building a Literature Search Strategy: Databases, Filters, and Grey Literature

A literature search strategy determines the entire evidence base your systematic review can draw from -- studies your search doesn't find cannot be included, however rigorous the rest of your methodology is. Building one properly means treating it as a genuinely deliberate design process, not copying a single search string across every database and hoping for the best.

Choosing your databases

Database selection should be driven by where your topic's relevant literature actually concentrates, not a fixed, generic list applied to every review regardless of subject area. Medical and clinical topics typically need PubMed/MEDLINE, Embase, and CENTRAL at minimum, with CINAHL added for nursing and allied health topics, and PsycINFO for anything touching psychological or behavioral outcomes. Education, social science, or policy-focused reviews often need a different core set entirely -- ERIC, Scopus, or Web of Science depending on the field. Justify your specific database choices in your protocol, rather than defaulting to whatever list a previous review happened to use.

Controlled vocabulary versus free text

Most major databases use their own controlled vocabulary system -- MeSH terms in MEDLINE, Emtree in Embase -- and a comprehensive search strategy uses both the relevant controlled vocabulary terms and free-text keyword synonyms in combination, since controlled vocabulary alone can miss recently published studies not yet indexed with the current term, while free text alone can miss studies using different terminology than you anticipated.

Building the search string

A well-built search string organizes your PICO or PICOS components into separate concept blocks, each containing the relevant controlled vocabulary and free-text synonyms combined with OR, with the separate blocks then combined with AND. This structure makes your search both more comprehensive within each concept and easier to audit and adjust than a single long string of mixed terms.

Applying filters thoughtfully

Study design filters, language filters, and date filters can all narrow your search, but each one is a deliberate trade-off that should be justified and disclosed, not applied by default. A study design filter restricting to RCTs, for instance, is appropriate if your eligibility criteria genuinely require RCT evidence, but inappropriate if you're planning to include observational evidence too.

Searching beyond bibliographic databases

Grey literature -- conference abstracts, theses, government and institutional reports, and preprints -- often contains relevant studies that never appear in standard bibliographic databases, and searching sources like OpenGrey, ProQuest Dissertations, or relevant organizational websites is expected for a comprehensive search, particularly on topics where publication bias toward positive findings is a known concern.

Trial registry searches

For intervention reviews, searching clinical trial registries like ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform can identify completed but unpublished trials, or ongoing trials relevant to your review's currency, and comparing registered outcomes against what gets published is also a useful check for outcome reporting bias.

Hand-searching and citation chasing

Manually checking the reference lists of included studies and closely related reviews, and using forward citation searching to find studies that have cited your key included papers, catches relevant studies that database searches sometimes miss, particularly in fields with inconsistent indexing.

Testing your search before finalizing it

Before running your full search, test it against a small set of studies you already know should be included -- if your search string fails to retrieve a study you know is relevant, that's a signal your terms need adjustment before you commit to the full search across all databases.

Documenting for reproducibility

PRISMA 2020, and the more detailed PRISMA-S extension, expect your final search strategy reported with enough specificity that another researcher could reproduce it exactly -- the full search string for each database, the date each was last searched, and any filters or limits applied. A search strategy that can't be reproduced from your methods section isn't finished yet, regardless of how comprehensive the underlying search actually was.

Working with an information specialist

Medical librarians and information specialists bring genuine expertise in database-specific syntax, controlled vocabulary structures, and search filter design that most subject-matter researchers haven't developed, and involving one early in a review, ideally during protocol development rather than after a first search attempt has already returned unsatisfying results, tends to produce a measurably more comprehensive and better-documented search strategy. Many institutions have librarians specifically trained in systematic review support, and this is often an underused resource relative to how much it can improve search quality.

Search strategy peer review

PRESS, the Peer Review of Electronic Search Strategies checklist, provides a structured way for a second, independent searcher to review your search strategy before you run it, checking for missed synonyms, incorrect Boolean logic, and inappropriate filters. This step, run before your full search rather than after, catches errors while they're still cheap to fix, rather than after screening has already begun against a flawed search.

Updating your search close to submission

For reviews with a long gap between initial search and final submission, rerunning your search shortly before submission to capture newly published relevant studies is increasingly expected practice, particularly in fast-moving fields. This update search should be documented with its own date and, if it identifies additional eligible studies, they should be incorporated and clearly noted as part of an updated search rather than silently folded into your original results without explanation. Treating an update search as a normal, expected part of a longer review timeline, rather than an awkward exception, keeps your final evidence base genuinely current at the point of submission. Journals and committees increasingly expect this kind of update search as standard practice for any review with a long gap between initial search and final submission, and building it into your project timeline from the outset avoids the scramble of an unplanned late-stage search update. This is a small planning habit with an outsized payoff: reviews that build a scheduled update search into their original timeline rarely feel the last-minute pressure that reviews without this planning step commonly experience just before submission. A small amount of upfront planning here reliably pays for itself many times over by the time your manuscript is actually ready to go out.