Embase and MEDLINE: plan for overlap, then measure your actual retrieval

1. Separate the collection from the search interface

Elsevier documents Embase as a biomedical database with Emtree indexing and several content types, including journal articles and conference abstracts. MEDLINE uses MeSH. A database’s scope, the platform through which you access it, and the particular query you execute are separate parts of the search record.

State the database, interface, selected collections, date range and access route. Do not write “searched all medical literature” because a platform contains several resources. If an institution provides a restricted collection or a different interface, record the actual arrangement. Product documentation can describe a capability without demonstrating that your team has access to it or has used it successfully.

Reference: Elsevier: Embase content coverage · NLM: About MEDLINE

2. Read overlap studies within their actual scope

Bramer and colleagues’ 2017 prospective exploratory study examined searches for 58 published systematic reviews. Among 1,746 included references found through database searches, 132 were retrieved uniquely from Embase. This concerns the searches and reference set in that study; it is not a current percentage of Embase content absent from MEDLINE.

The authors describe an experienced information specialist, topic-dependent results and limits to generalisability. They did not test whether missing references changed review conclusions. The useful distinction is between a database containing a record and a particular strategy retrieving it. Do not turn that historical result into a promise that Embase will add 132 relevant papers, improve a treatment decision, or outperform another source for your next question.

Reference: Bramer et al. (2017): prospective database-search study

3. Draft the search before counting results

Imagine a team planning a review of how hospital staff implement medication-reconciliation workflows for adult inpatients. This is a fictional search-planning question, not evidence about whether an intervention works. The team first defines the setting, eligible report types and what counts as implementation evidence. It asks an information specialist to check whether additional subject sources are needed.

For each concept, examine the relevant controlled vocabulary and add justified free-text variants. Translate field codes, phrase handling and other operators for the selected interface rather than pasting an untested query. Elsevier documents advanced searching and export tools, but the exact proposed strategy must still be executed and inspected. Save unsuccessful or revised queries with a reason for the change.

Reference: Elsevier: Embase features and access

On small screens, scroll the table horizontally, or focus it and use the arrow keys.

4. Use a small invented export to understand the arithmetic

Consider two fictional exports, each already deduplicated internally: 120 records from search A and 90 from search B. A verified cross-export comparison finds 50 records present in both. The combined set contains 120 + 90 − 50 = 160 distinct records. Search B therefore adds 40 records beyond A in this exercise.

Forty additional records are not forty eligible studies. Suppose title-and-abstract screening sends six of those forty to full-text assessment, excludes thirty-two under the stated criteria and leaves two unresolved. Six + thirty-two + two = forty. Full-text eligibility remains pending for the six; the two unresolved records must not be recorded as exclusions merely to close the tally.

5. Keep reports, studies and access problems separate

A conference abstract and a journal article may report the same study without being duplicate bibliographic records. Preserve their relationship and assess what each version supplies. Do not remove a report only because its author and topic resemble another record; compare identifiers, participants, setting, dates and the details available in the sources.

Likewise, an inaccessible full text is an access problem, not an automatic ineligibility reason. Record the permitted route attempted and the unresolved status. Any decision to proceed without it belongs in the review’s documented method. A citation database listing does not supply permission to redistribute a PDF or prove that your institution subscribes to the underlying journal.

6. Report execution and limitations without a sales claim

Write the database and platform names, search dates, strategy versions, export details and the deduplication procedure actually used. Distinguish records retrieved, records screened, reports assessed and studies included. If a planned database could not be searched, state that limitation and its reason rather than describing the plan as completed.

A useful pilot compares the sources for the team’s defined task, with eligibility decisions checked and unanswered questions retained. It is not a universal ranking of products. Vendor coverage pages can help scope the plan, while current licensed access and an executed search establish what the team could retrieve. Neither replaces methodological review of the search or critical appraisal of the studies found.

Questions for a practical final check

• Question and eligible report types defined.
• Database and actual interface named.
• Vocabulary and syntax checked separately.
• Original exports and overlap decisions retained.
• Additional records not called additional eligible studies.
• Unknown eligibility and unavailable access remain explicit.