Systematic Review Search Strategy and Study Identification

Study identification begins with a chain of decisions

By Dr. Esmaeel Saeedy Robat, Founder of MetaSyn Academy · Published meta-analyst (Nature Human Behaviour, 2026) · Evidence checked July 2026

A systematic-review search is not a one-time database task. It is the controlled process that turns eligibility criteria into searchable concepts, selects information sources, develops and tests terms, creates platform-specific strategies, adds supplementary discovery methods, records peer review, executes dated searches, and preserves the evidence needed for reporting and updating.^1,2

Every later review stage depends on the candidate evidence set created here. Screening cannot recover a study that was never retrieved. At the same time, indiscriminate retrieval can impose a screening burden that consumes time without improving coverage. The methodological aim is a sensitive, transparent, question-appropriate search whose decisions can be assessed and whose execution can be rerun to the extent each source permits.¹

From eligibility criteria to an auditable identification record

The lifecycle begins with the review question, but the eligibility criteria provide the operational boundary. They define the populations, phenomena, interventions or exposures, evidence types, settings, dates, languages, and report forms that may matter. Question frameworks help organize those elements; they do not require every element to become a search concept. Some concepts are poorly represented in searchable metadata and can reduce recall when added. Their inclusion or omission should be tested and documented rather than assumed.¹

Source selection follows the evidence pathway. A database, register, repository, regulator, organization, website, citation route, or contact method belongs because it can plausibly contain eligible evidence. The Cochrane Handbook sets source expectations for Cochrane intervention reviews, while JBI and Campbell provide guidance for their own review traditions. The transferable principle is to justify coverage in relation to the question, not to copy a universal source list.^1,4,5

Conduct, peer review, and reporting solve different problems

Search quality becomes confused when three distinct functions are collapsed. Conduct guidance informs planning and execution. PRESS structures peer review of an electronic strategy. PRISMA-S sets expectations for transparent reporting of the search.^2,6,7 Good reporting cannot repair weak conduct; a PRESS review does not certify completeness; and a carefully designed strategy still needs complete reporting.

Known-item and relative-recall testing belong to the diagnostic layer. They can reveal missing language, indexing problems, overly restrictive concepts, or translation errors. Their conclusions depend on the benchmark chosen: retrieving all benchmark records does not prove that all unknown eligible records were found.^3,8

Keep the logic stable and the execution source-specific

A master strategy preserves concepts and relationships; it is not a universal executable string. Each database–platform pair needs its own version because field tags, subject headings, phrase behavior, proximity, truncation, mapping, limits, and export functions differ. Database and platform must be recorded separately. Automated translation can assist, but the output remains a draft until it is checked against current official help and the executed interpretation is inspected.¹

Grey-literature and supplementary methods need a parallel but source-sensitive record. A stable database query may be exactly reproducible; a ranked web search may not be. Preserve the precise query or browse route, date, searcher, filters, screen depth, stopping rule, yield, captured records, storage location, and limitations. Transparent uncertainty is more rigorous than treating a dynamic source as if it were stable.⁵

The weakest point can occur before, inside, or after the query

A search can fail before syntax is written because the scope is ambiguous, evidence types have not been mapped, or important source families are absent. It can fail inside the query because a concept is over-constrained, terminology is incomplete, controlled vocabulary is misapplied, parentheses change precedence, or a platform ignores an unsupported behavior. It can fail after the query because the wrong version is executed, limits are applied without justification, records are not preserved, peer-review changes are lost, or updates cannot be distinguished from the original search.¹

These failures require different responses. Adding synonyms cannot repair a missing source. Searching another database cannot repair a logic error copied into every translation. Polished methods prose cannot recover an undocumented execution. The record architecture in H02 is designed to locate the problem at its level and preserve the evidence used to correct it.¹

Precision is not the same as quality

A shorter result set is convenient, but it is not automatically better. Precision has to be weighed against the risk of excluding eligible evidence and the team’s screening capacity. Restrictions that remove records date, language, document type, age, human, study design, or publication format, need a protocol-based rationale or a tested fit-for-purpose method. A filter’s reported performance belongs to the context in which it was developed and evaluated; it should not be treated as a permanent property across databases, platforms, topics, and time.^1,3

Reproducibility is a property of the record and the source

Some methods can be reproduced closely when exact strategies, dates, platforms, and versions are retained. Others, such as ranked web searching or browsing a changing site, can only be reproduced partially. The responsible response is to increase transparency: define the route and boundary, preserve captured evidence, and disclose instability. Reproducibility should not be treated as an all-or-nothing label when the underlying source is dynamic.⁵

Specialist review remains an escalation pathway

Complex review questions, unfamiliar databases, validated-filter decisions, difficult translations, and high-stakes searches may require an information specialist. The Academy tools support structured preparation and auditability, which can make specialist review more efficient. They do not simulate professional judgment or convert a self-check into formal PRESS review.^2,6