Explainer

What is federated search?

One query, many databases, results merged in real time. Here's what that means in practice — and what it doesn't.

The one-sentence version

Federated search sends the same query, at the same time, to several independent databases that each hold their own records, and merges what comes back into one combined result list. Nothing is copied into a central index in advance — each source is queried live, and each record links back to where it came from.

Why this is different from "just searching PubMed"

A single-database search — PubMed, Google Scholar, a publisher's site — only shows you what that database happens to contain. PubMed indexes roughly 5% of the world's medical journals, weighted heavily toward North American and Western European publications. Google Scholar is broader but opaque: you can't audit what it searched, what it excluded, or in what order. Neither is wrong; both are partial.

Federated search is an answer to the partiality. Instead of picking one database and accepting its blind spots, you query several at once — including regional indexes the global tools don't reach — and you get back a merged list with the source of each record made explicit.

How it actually works

One query, many adapters. Your search terms are translated into the query syntax each database expects (MEDLINE field tags differ from CrossRef's, which differ from J-STAGE's).
Parallel requests. All databases are queried at the same time. Results stream back as each source responds — you don't wait for the slowest one to finish before you see anything.
Deduplication. The same paper can appear in PubMed, Europe PMC, CrossRef, and OpenAlex. A federated tool merges duplicate records using DOIs, PubMed IDs, and title+author matching, so you see one entry with multiple sources attached.
Provenance preserved. Every merged record carries a link back to each origin database. Nothing is laundered through an opaque "Symmathy ID" — the trail to the primary source stays intact.

What it is not

Not a meta-index. Federated search doesn't pre-crawl and store everything in one warehouse the way Google does. Each query is live, which means results reflect what each source contains right now — and which means slow or unreachable sources are visible as such, not silently dropped.
Not a ranking engine. Merging results is not the same as judging them. A federated tool can sort by date or relevance, but it shouldn't be telling you which paper is "best" — that's a job for the reviewer, with appraisal tools designed for it.
Not a replacement for systematic-review methodology. Federated search is a discovery step. PRISMA flow, eligibility screening, risk-of-bias assessment, and GRADE all still happen after the records are pulled.

Why it matters for medical evidence

The Cochrane Handbook (§4.3) already recommends searching beyond MEDLINE and Embase, including regional and non-English databases, to reduce language and geographic bias. The reason that step is so often skipped is operational: different interfaces, different query syntaxes, no shared export format, abstracts in languages the reviewer doesn't read. Federated search collapses that workload into one query, which makes the recommended methodology actually achievable inside a normal review timeline.

The practical consequence: for any disease with significant non-English-language literature — dengue, Chagas, leishmaniasis, schistosomiasis, Japanese encephalitis, snakebite, and many others — a federated search returns evidence a MEDLINE-only search wouldn't surface. Whether that evidence changes any specific review's conclusion is a question for the reviewer. The point of federated search is to let you answer it instead of assume it.

How Symmathy implements it

Symmathy runs one query in parallel across 12 databases: PubMed, Europe PMC (including bioRxiv and medRxiv), Cochrane, ClinicalTrials.gov, J-STAGE, LILACS, DOAJ, HAL, OpenAlex, CrossRef, ThaiJO, and KoreaMed. Non-English titles and abstracts are translated inline. Records are deduplicated across sources. Every result links back to its original database, and Symmathy generates a reproducible audit trail — sources queried, query terms, result counts — for the PRISMA flow diagram.

What Symmathy does not do: grade papers, rank by quality, recommend a "best" study, or offer clinical advice. The tool finds the evidence; interpretation stays with the reviewer.