Should you build or buy Enterprise Master Patient Index (EMPI)?

An Enterprise Master Patient Index resolves duplicate and fragmented patient records across a health system's or exchange's source systems into one golden identifier, exposed through HL7 or FHIR. Buyers are health system IT teams and health information exchanges that need one accurate patient identity behind clinical, billing, and analytics systems.

Buy it. Probabilistic matching is open, but referential identity data and clinical-grade merge governance are what vendors sell, and no health system is documented running its own EMPI in production.

Build it, buy it, or bridge?

⚒ Build it
✓ Buy it
➔ Bridge
Cost shape
Splink is free, but no health system runs it as a production EMPI in place of a vendor.
Rhapsody lists roughly $115,000 a year for up to 500,000 identities on AWS Marketplace.
Buy referential matching for sparse records; keep stewardship rules and merge governance in-house.
Time to value
NHS England is still developing its Splink-based linkage service; it isn't a live EMPI yet.
Vendor platforms come with matching, stewardship, and referential resolution already built and supported.
Deploy the vendor platform first; tune matching thresholds and stewardship rules as you go.
Differentiation captured
None. Every health system needs the same demographic-matching and duplicate-resolution workflow.
Vendor owns referential identity data; you own stewardship rules and how you tune matching thresholds.
Accurate identity is a step above pure hygiene, but no competitor is beaten on it.
AI feasibility today
AI-assisted matching can improve tools like Splink, but no U.S. health system runs that build in production.
AI hasn't displaced referential data as the vendor's core moat; it's layered on top as added features.
Use AI-assisted probabilistic matching to estimate your own duplicate rate before shopping vendors.
Who it fits
Nobody documented in production. Even NHS England's Splink service is still in development.
Any health system or HIE that needs to resolve identity across source systems reliably.
Health systems piloting internal matching before committing to a vendor's referential data license.

When building makes sense

Probabilistic record linkage itself is mature and open. Splink is widely used open-source software for exactly this kind of matching problem, and NHS England is developing a national linkage service built on it. That's real evidence the underlying technique works. But no U.S. health system or health information exchange is documented running Splink, OpenEMPI, or any custom matcher as its production enterprise master patient index, and NHS England's service is still in development rather than live. One instructive data point: Manifest MedEx, a health information exchange, added Verato's referential matching on top of its own EMPI specifically to resolve duplicate records that system had flagged but couldn't clear on its own. The reason isn't that matching logic is hard to write. It's that resolving sparse, stale, or inconsistent demographic records reliably requires referential identity data that no health system holds on its own. Build only as a pilot to understand your own duplicate rate before evaluating vendors.

When buying makes sense

Buying is the clear call because the hard part isn't the matching algorithm, it's the reference data behind it. A 2022 JAMIA study, whose funder holds a stake in referential-matching vendor Verato, found referential matching's sensitivity 30 points higher than probabilistic matching (93.5% versus 63.7%), and Verato counts health information exchanges like Manifest MedEx among its customers. Rhapsody's EMPI lists roughly $115,000 a year for up to 500,000 identities on AWS Marketplace, which sounds significant until you compare it to standing up threshold governance, manual review queues, merge and unmerge controls, and a referential data license on your own. Every health system runs the identical problem, matching demographics across source systems and exposing one golden identifier via HL7 or FHIR, so there's no competitive edge in owning the plumbing. Run a blinded bake-off against your own adjudicated records before signing, and keep the stewardship rules around the platform as your actual area of ownership.

The desk read

Build-versus-buy analysis for Enterprise Master Patient Index (EMPI) is being written. In the meantime, the framework that drives every B4 call is on the B4 Index page.

Representative vendors 4medica EMPIExperian Health Universal Identity Manager4medicaInterSystemsLexisNexis Risk SolutionsRhapsody NextGateVeratowhich differentiate mainly on the depthsource of their referential identity data

Frequently asked

Is open-source patient matching software like Splink good enough to replace a vendor EMPI?

The matching technique is genuinely mature and NHS England is building a national service on Splink, but no U.S. health system runs it as a production EMPI today. The gap isn't the algorithm, it's referential identity data: a large external pool of identity records that resolves sparse or stale demographics, which vendors compile and internal teams don't have access to.

How accurate is vendor patient matching compared to building it ourselves?

A 2022 JAMIA study, funded in part by a foundation with a stake in Verato, found referential matching's sensitivity 30 points higher than probabilistic matching (93.5% versus 63.7%), a gap attributed to matching against a large external identity database rather than only your own records. Internal probabilistic matching compares your records only to each other, so it can't close that specific gap.

What does an EMPI actually cost?

Rhapsody's EMPI lists roughly $115,000 a year for up to 500,000 identities on AWS Marketplace, including implementation services, while other vendors like Verato price by organization rather than publishing a rate card. Compare quotes against what a self-run matching program would cost in threshold governance, manual review, and referential data licensing before deciding.

What are the main EMPI vendors?

Representative vendors include 4medica EMPI, Experian Health Universal Identity Manager, 4medica, InterSystems, LexisNexis Risk Solutions, Rhapsody NextGate, Verato, which differentiate mainly on the depth, source of their referential identity data. B4 Pro includes the category score and the full vendor list.

The B4 Index scores every software category on two axes, strategic differentiation and AI feasibility, to classify it Build, Buy, Bridge, or Beware. See the full methodology.