Deduplicate PubMed and Embase results: a worked example
Follow a downloadable PubMed and Embase overlap example: four records, one duplicate, three unique citations, with PRISMA counting and matching limits.
Quick answer
Keep PubMed and Embase exports separately, import them into one review, and check duplicate matches before screening. In our synthetic example two records from each source produce four identified records, one duplicate and three unique records for screening.
Keep identification counts separate
Record the number exported from each database before deduplication. Our PubMed-format file has two synthetic records; the Embase-style RIS also has two. One citation has the same DOI in both files, but a different source identifier.
The resulting example counts are PubMed 2 + Embase 2 = 4 identified; duplicates removed 1; unique records 3. These are record counts, not final included-study counts. Keep subsequent screening and full-text exclusions separate.
What matches and what needs checking
Lumina first compares normalized DOIs. It also uses normalized title similarity and an author/year/title fallback when eligible. Records with distinct nonempty DOIs are not merged solely because titles are similar.
Fuzzy matching can still misclassify records, especially short or generic titles. Different reports of one study are not automatically duplicate publications. Compare identifiers, authors, years and content and follow your protocol.
Run the example yourself
Download the two exports and the benchmark results. The reproducibility command parses both files and calls the production duplicate-matching function in memory; it does not create a review or change your database.
Our fixture has one deliberately obvious duplicate and three unique records. It measures that example only; it does not establish real-world deduplication accuracy or a recall guarantee.
Carry the counts into PRISMA
Use four for database records identified, one for duplicate records removed and three for records entering screening in this example. Add actual screening outcomes only after reviewers have made decisions. Use the free PRISMA tool to draw the flow.
| Stage | Records | Meaning |
|---|---|---|
| PubMed-format input | 2 | Before duplicate removal |
| Embase-style input | 2 | Before duplicate removal |
| Identified total | 4 | Sum of source records |
| Duplicates removed | 1 | One matching normalized DOI |
| Unique for screening | 3 | 4 minus 1; not included studies |
Download and inspect the examples
These CC0 examples are synthetic. Their identifiers do not refer to real publications. Use them to check an import, not as research evidence.
Read the benchmark method and limitations →Sources and editorial review
Maintained by the Lumina Editorial Team. Product mappings and examples were checked against Lumina’s implementation on October 7, 2026. External interfaces can change; consult the primary documentation below. This is an independent product guide, without a claim of independent scientific review.
Suggest a correctionInspect the screening workflow
Explore the public demo with sample papers. AI recommendations remain advisory; reviewers make the screening decisions.