Master patient index (MPI)
How a health system knows that records from a dozen places are the same person — identity matching, the EMPI, and why a wrong match can be dangerous.
In one line
A master patient index (MPI) is the system that resolves identity: it decides that "Priya Sharma" registered at the clinic and "P. Sharma" admitted to the hospital are the same person — and links their records under one identifier. Across organisations it's an EMPI (enterprise MPI).
How matching works
There's rarely one perfect ID, so MPIs compare demographic traits — name, date of birth, sex, address, phone, national ID — with:
- Deterministic rules (exact matches on key fields), and
- Probabilistic scoring (weighted similarity that tolerates typos, nicknames and changed addresses), with a manual-review queue for the uncertain middle.
A national health ID like India's ABHA makes this far easier by giving everyone a stable anchor — but matching is still needed where the ID is absent.
Why it's safety-critical
- A false negative (failing to link) splits a patient's history — a clinician misses an allergy or a prior result.
- A false positive (wrongly merging two people) is worse — one patient's data contaminates another's chart, a genuine patient-safety event.
Identity is therefore a core data-quality discipline, not a registration afterthought.
Where it connects
Every EHR, HIE and FHIR exchange
depends on it: a FHIR Patient.identifier only means something if the MPI behind it is
sound. Get identity wrong and every downstream record, claim and analysis inherits the
error.
Identifiers help, but don't replace matching
A strong national ID like India's ABHA (or an Aadhaar-linked anchor) dramatically improves matching — but doesn't eliminate it: people present without their ID, IDs get mistyped, and legacy records predate the scheme. So even with a national ID, an MPI still runs deterministic + probabilistic matching with a human review queue for the uncertain middle.
Key takeaways
- An MPI/EMPI resolves identity — deciding that records from many places are one person.
- Matching combines deterministic rules + probabilistic scoring, with a human review queue.
- Errors are safety-critical: a false negative splits a history; a false positive contaminates a chart.
- Identity is a core data-quality discipline; every EHR/HIE/FHIR exchange inherits its accuracy.
Check your recall
0 of 2 recalledActive recall beats re-reading — try to answer, then reveal.
What does a master patient index (MPI) do?
Why is patient matching safety-critical?