A master patient index is the part of a FHIR stack that decides whether two records about the same person are recognised as the same person, or quietly drift into separate identities that nobody reconciles for years. The standard makes the surface look modest (a few Patient resource fields, a few search parameters), but the operational reality is wider: an MPI has to handle name variants, transposed dates of birth, NHS numbers that arrive late or wrong, address changes, and the long-tail edge cases that come with maternity, paediatrics, hospice care, and adult mental health. The choices on the market in 2026 are more mature than they were a few years ago, and picking the right fit pays back across the deployment's life.
This guide is the entry point to more healthcare interoperability notes, pulling together what a master patient index actually does, which capabilities matter most in 2026, and how UK and international teams should approach the choice.
What a Master Patient Index Actually Does
A master patient index is the service that takes patient identifiers from multiple source systems and resolves them into a single canonical patient identity. It maintains the mapping from each source identifier to the canonical one, runs matching logic to decide when two records refer to the same person, and exposes the canonical identity through standard FHIR Patient operations.
When this works, the rest of the stack stops worrying about identity. When it does not, every downstream component ends up implementing its own partial matching logic, which is exactly the failure mode an MPI was supposed to prevent.
Capabilities That Actually Matter in 2026
Three capabilities separate practical MPIs from prototypes:
- Configurable matching algorithms (deterministic, probabilistic, or hybrid), with clear behaviour around match confidence reporting.
- Robust manual stewardship tooling, because every realistic MPI workflow includes a queue of records the algorithm cannot resolve confidently.
- Clean FHIR API surface, with $match operations that behave predictably and Patient.link relationships that downstream systems can rely on.
A product that handles all three becomes a quiet background service. A product that misses any of them tends to surface as a recurring data quality issue elsewhere.
How UK Teams Should Frame the Choice
The British market has a few specific dynamics. NHS number is a strong identifier when it is present and correct, but the long tail of records without an NHS number (newborns before assignment, overseas visitors, urgent care attendances under stress) still needs reliable demographic matching. Integrated Care Systems running across multiple providers need the MPI to behave consistently across the providers' source systems. And the historic pattern of separate MPIs in primary care, secondary care, and community care creates a backlog of identity reconciliation that any new MPI inherits.
The deterministic vs probabilistic patient matching comparison walks through the matching style decision in detail, and the open-source vs commercial MPI comparison covers the operating model question.
Operational Realities Worth Planning For
Three operational details often surprise teams new to running an MPI in production. The initial historic load is bigger than expected and produces a long queue of stewardship cases that takes weeks to work through. Match thresholds need tuning against real data, not synthetic test sets, because real-world demographic noise is wider than test data captures. And the relationship with the source systems is a long-term one; an MPI is not a write-once integration but a continuing flow that needs monitoring and adjustment.
A team that plans for these realities tends to run a quiet MPI. A team that does not tends to run a noisy one and rediscover the same surprises every twelve months.
Where to Go Next
If the immediate decision is shortlisting products for an Integrated Care System or similar deployment, the top 5 MPI tools for integrated care networks in 2026 is the right starting point. For deployments where matching style is the deciding question, the deterministic vs probabilistic patient matching comparison is the next read. For deployments choosing between open-source and commercial operating models, the open-source vs commercial MPI for regional health networks covers the trade-offs.
A master patient index is one of those quiet decisions that compounds heavily across a deployment's life. Picking carefully now saves a lot of reconciliation work later.
Sources
- Interoperable Digital Identity and Patient Matching v2.0.0 - HTML IG, HL7 build.fhir.org
- MDM module overview - HTML docs, HAPI FHIR
- Scaling Patient Identity Solutions: the FAST Identity IG - HTML blog, HL7
