US Core Patient profile support is table-stakes for MPI implementations in 2026, but the profile's constraints have implications for matching that aren't obvious until production. The US Core Patient profile mandates specific data elements — race, ethnicity, birth date, identifier — that shape which matching signals are reliable.
Required elements set the deterministic-match floor. The profile requires at least one of identifier, name, or telecom. Identifier is the highest-signal match field (SSN, MRN, driver's license); name is next; telecom is weakest for identity. MPI systems that don't privilege identifier over other fields underperform when identifiers are present.
Race and ethnicity as blocking factors, not matching factors. US Core mandates the OMB race and ethnicity extension, but using race in match scoring is regulatory territory (HIPAA, ONC guidance). The right use: block matching to focus computation on likely-match subgroups, don't score on it.
Address as multi-field signal. US Core Patient allows multiple address entries with period.start/end. MPI systems that treat address as a single point-in-time miss the reality that patients have historical addresses that overlap with siblings, roommates, or previous tenants. Match on address + period intersection, not just current address.
Birth date fuzzy matching. US Core requires birthDate in full ISO format, but 2-3% of records have transposed month/day or single-digit typos. MPI systems that don't do fuzzy birth-date matching miss ~1-2% of true matches. Implement Damerau-Levenshtein on the birth-date string, not just exact match.
Deceased-status handling. US Core allows deceasedBoolean or deceasedDateTime. MPI systems that block matching on deceased patients (skipping them from candidate lists) speed up queries and prevent identity mistakes from resurrecting deceased profiles.
Implementation trade-off table
| Signal | Match weight | Regulatory sensitivity | Common data quality |
|---|---|---|---|
| Identifier (SSN/MRN) | Very high | HIPAA restrictions | Present ~90% |
| Name (given + family) | High | Low | Present ~99% |
| Birth date | High | Low | Present ~99% |
| Address (current + hist) | Medium | Low | Present ~80% |
| Race / ethnicity | Blocking only | Regulatory | Present ~60% |
US Core Patient conformance is necessary but not sufficient for effective MPI. The matching sophistication that ships with the profile is deliberately minimal — production systems add the calibration and fuzzy matching that make MPI actually work.
