A master patient index is one of those pieces of US healthcare infrastructure that everyone agrees is important and almost everyone underinvests in. Patient identifier matching is the foundation of every downstream workflow that touches more than one source system, which in 2026 is most of them. Picking the right MPI is a decision a US health network lives with for years, and the cost of picking poorly shows up as duplicate records, mis-routed lab results, and reconciliation work that grows faster than the volume it is supposed to reconcile.
This buyer's guide walks through what a master patient index has to do for a US digital health team in 2026, the capabilities to verify before signing, and how to read the trade-offs between cloud and on-premise picks. For the rest of the digital health series, the homepage covers the broader landscape.
What a Master Patient Index Has to Do
The narrow job of an MPI is taking patient records from many source systems and producing a single trusted view of which records belong to the same person. The longer job is doing that across the messy reality of US healthcare data: variant spellings, address changes, married-name updates, dual citizenship, twin births, and the cases where the same person has different birth dates recorded in different systems because of a data entry slip in 1987.
A serious US deployment also has to handle the regulatory and reporting requirements that MPI work touches. HIPAA-compliant audit trails, USCDI patient identifier conformance, state HIE participation requirements, and the increasingly common TEFCA Qualified Health Information Network workflows all assume a working MPI underneath.
Capabilities That Matter for US Health Systems
The capability set that matters most for US digital health is concrete.
- Strong patient matching algorithms that handle deterministic, probabilistic, and increasingly referential matching with comparable confidence.
- FHIR-native operation through Patient resources and standard interactions, since the rest of the stack is moving toward FHIR-first.
- Defensible audit trails for every merge and every match decision, since merges that turn out to be wrong have to be undone and explained.
- Address standardization and demographic normalization at ingestion time, since most matching failures trace back to inconsistent input.
An MPI that handles those four cleanly will serve a US health system for years. One that fails any of them produces reconciliation work that grows with the data volume.
Cloud or On-Premise
The choice between cloud EMPI and on-premise MPI follows roughly the shape of the wider infrastructure question, with US-specific wrinkles. Cloud-hosted picks like Verato Auto-Steward, NextGate's cloud product, and Smile Digital Health's hosted MPI handle the operational work, including matching algorithm tuning. On-premise picks like Mirth Match, the open-source OpenEMPI lineage, and the on-premise versions of NextGate keep the patient data on the network's own infrastructure.
The trade-off is real and the cloud EMPI vs on-prem MPI guide covers the specifics for US hospitals. Some networks end up running a hybrid where the matching engine is cloud-hosted but the source records stay on-premise; the pattern works in some regulatory contexts and not in others.
What to Verify Before Signing
Before signing anything, run four concrete tests.
- Run a deterministic match on a thousand-record sample with known duplicates and confirm the recall matches the published guidance.
- Run a probabilistic match on the same sample with degraded data quality, including missing date-of-birth and partial address fields, and verify the precision holds up.
- Walk through a merge-undo workflow with the audit trail visible and confirm the chain of custody is intact.
- Verify the patient match output lands in a FHIR Patient resource shape that the rest of the stack can consume directly.
Products that pass those four are products that will hold up over time. Products that fail any of them produce a long tail of reconciliation work.
Where to Go From Here
The natural next reads are the Top 5 MPI tools for state health information exchanges in 2026 and the Top 6 patient matching tools for newborn identity in US hospitals for two of the harder MPI deployments in US practice.
Sources
- Patient Identity and Patient Record Matching - HTML reference, ONC, 2025
- Interoperable Digital Identity and Patient Matching v2.0.0 - HTML IG, HL7 FHIR FAST Identity team, 2025
- Framework for Cross-Organizational Patient Identity Management - PDF, The Sequoia Project, 2018 (foundational)