Join our Newsletter — 33% off our NHI Course
Home› Guides› Patient Identification and Identity Matching Guide: registration, MPI…
Guide Customer Identity

Patient Identification and Identity Matching Guide: registration, MPI matching and portal proofing

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 14 November 2026 14 min read
On this page

When a healthcare organisation links the wrong records or fails to link the right ones, clinicians act on another person's allergies, results and medicines, and portals show one patient's data to another. Patient identification is the customer identity problem of healthcare: the organisation must establish who a patient is at registration, find that person's existing record in a master patient index (MPI), keep the record clean over decades, and prove identity again when the patient or a proxy signs into a portal or app. This guide covers the provider and health information exchange side: demographic data capture, matching approaches and their errors, the FHIR and IHE interfaces that expose matching, identity proofing for patient-facing access, and how to measure and govern match quality. It is written for health IT, identity and data quality teams, and for security staff who own patient portal and API access. Workforce identity in hospitals is covered separately in the Healthcare Identity Security Guide, and the patient's own recovery steps in the Medical Identity Theft Guide.

Key takeaways

  • Two error types, two different harms. A duplicate (false negative) splits one patient's history across records; an overlay (false positive) merges two people into one record, which is the more dangerous error clinically and for privacy.
  • Data quality at registration beats a smarter algorithm. The Sequoia Project framework calls standardising phone and identifier formats the biggest immediate opportunity to improve match rates, and warns against matching on default or placeholder values.
  • A national identifier helps but does not remove matching. England uses the NHS number as the unique key through the Personal Demographics Service; the US has no national patient identifier, so demographic and referential matching carry the load.
  • Matching interfaces are an attack and privacy surface. A FHIR Patient $match endpoint that returns probable matches to any caller lets that caller discover who is a patient; restrict it and prefer certain-only results for automated linking.
  • Portal access needs proofing, not just a match. Finding a record from name and date of birth is not identity verification; bind portal and proxy accounts through verified evidence or an in-person check, as NIST SP 800-63A-4 describes.

Why patient matching is a security problem

Patient matching is usually treated as a data quality or interoperability issue, but each failure mode is also an identity failure with confidentiality and integrity consequences.

FailureWhat happensSecurity and privacy impactPatient safety impact
Duplicate (false negative)A second record is created for an existing patientAccess monitoring and consent flags are split across records; a restricted-record flag on one is missing on the otherIncomplete history: missed allergies, repeated tests
Overlay (false positive)Two people's data are merged into one recordUnauthorised disclosure to the other person, in clinic letters, portals and exchange; integrity of the legal record lostTreatment based on another person's data
Wrong-patient query resultAn exchange or $match query links an external record to the wrong local patientDisclosure across organisations; error propagates to partnersWrong results imported into the chart
Medical identity misuseA person registers using someone else's detailsFraudulent claims; the impostor's data enter the victim's recordVictim later treated on false history
Weak portal bindingA portal account is linked to a record without verifying the personAccount takeover or proxy abuse exposes full recordsMessages, results and prescriptions misdirected

The Sequoia Project's Framework for Cross-Organizational Patient Identity Management (2018, published as a proposed framework for comment) defines a false positive as records of two separate individuals merged as one, producing an overlay, and a false negative as a failure to detect a shared patient, producing a duplicate. Use those definitions consistently in policy and metrics so that teams count the same thing.

Capturing identity at registration

Most matching problems start at the front desk, the contact centre or a self-registration form. Fix capture first.

  • Required core traits: in the Sequoia case study, first name, last name, date of birth, sex, five-digit ZIP code and phone number together gave the strongest uniqueness, with middle name or initial a useful addition. Collect prior names and addresses, telephone type, email and address type as well, because they let the MPI link records across life events.
  • Standard formats: enforce one format for phone numbers and identifiers at entry, normalise addresses against the postal authority's standard, and validate fields (impossible dates, test names). In the US, the ONC Project US@ technical specification, released by ONC as Final Version 1.0 with a companion implementation guide developed with AHIMA, gives a cross-standard way to represent addresses for matching.
  • No placeholder values: "Baby Girl", 01/01/1900, 000-00-0000 and the hospital's own phone number create false matches. Mark unknown values as unknown and exclude defaults from matching and from exchange.
  • Search before create: registration screens must search the MPI with several combinations of traits before allowing a new record, and creating a new record when candidates exist should require a reason.
  • Fragile populations: newborns, multiple births, trauma and unidentified patients, and people without fixed address need specific workflows: temporary identities with a defined process to reconcile them, and follow-up to collect full demographics.
  • Photo and document checks: where a photo is captured into the record, or an ID document is inspected at check-in, record what was checked. Keep any photo access controlled and logged like other identity data.

NIST SP 800-53 PM-22 and SI-18 cover the quality of personally identifiable information, and SI-10 covers input validation; they are the natural control anchors for registration rules.

Matching approaches and identifiers

Organisations combine identifiers and algorithms. The choice depends on jurisdiction, volume and how many external partners exchange data.

ApproachHow it worksStrengthsLimits
National or regional identifierUnique number issued by a national service, for example the NHS number traced through the Personal Demographics ServiceOne key across organisations; central back office resolves duplicatesMust still trace patients without a number; local copies drift unless synchronised
Deterministic matchingExact or rule-based match on a defined set of fieldsPredictable, explainable, easy to auditMisses records with typos, name changes or moves (more duplicates)
Probabilistic matchingWeighted scores across fields, with thresholds for auto-link, review and rejectTolerates errors; tunableThresholds trade duplicates against overlays; needs tuning and review queues
Referential matchingCompares records with a reference dataset of historical identity dataLinks across address and name changesDepends on a third-party dataset: contract, data protection and quality terms matter
Supplemental identifiersAdds a verified identifier such as a driving licence number, or a biometricSequoia suggests a supplemental identifier for match rates above 95%Collection burden; biometric data needs specific legal handling

England and Wales: NHS England's Personal Demographics Service holds NHS number, name, address, date of birth, gender, contact details, registered GP and related people, with full audit of access and strong authentication (such as an NHS smartcard) required to update records. Its guidance states that the NHS number must be used to identify patients uniquely, under information standard ISB 0149, and that local copies of demographics should be synchronised with PDS regularly.

United States: without a national identifier, organisations rely on EMPI software, health information exchanges and frameworks such as the Sequoia work. ONC defines patient matching as linking multiple demographic fields such as name, birth date, phone number and address. For biometrics in the matching workflow, see the Biometric Verification vs Identification and the Biometric Privacy Laws Map.

Matching interfaces: FHIR $match, PDQm and exchange

Matching is no longer only an internal MPI function. Applications, partners and national networks query it through standard interfaces, which turns it into an access-controlled service.

  • FHIR Patient $match: the HL7 operation asks an MPI to find patients matching a full or partial Patient resource. It returns a searchset Bundle ordered from most to least likely, with a match-grade extension per result and a search score from 0 to 1. The onlyCertainMatches parameter, when true, requires the server to return only records it is certain belong to the same person. The specification does not define a matching algorithm or required fields, so check what your MPI needs.
  • IHE PDQm: the IHE Patient Demographics Query for Mobile profile (version 3.2.0, on FHIR R4) defines a RESTful interface to a demographics supplier, with the ITI-78 Mobile Patient Demographics Query and ITI-119 Patient Demographics Match transactions.
  • Cross-organisation exchange: the Sequoia framework advises running a trait analysis for each exchange partner, never exchanging default values, agreeing data standards, and setting network timeouts long enough that slow responses are not read as no match.

Security controls for matching endpoints:

  1. Authenticate every caller with a client identity (for SMART on FHIR backend services, signed client assertions rather than shared secrets) and scope it to the operations it needs. See the EHR Platform Security guide for platform-side settings.
  2. Use onlyCertainMatches=true for automated linking, and send probable results only to a human review queue inside the organisation.
  3. Limit the count of results and the attributes returned; a matching response should not double as a demographic lookup service.
  4. Rate-limit and log queries by client, and alert on high volumes of no-match or low-score queries, which indicate enumeration or a broken integration.
  5. Treat inactive and merged records carefully: the FHIR specification notes that a record inactive because of a merge may be excluded in favour of its merge target.

Proofing patients for portals, apps and proxies

Matching answers "which record is this?", not "is this person really the patient?". Patient-facing access needs both.

  • Self-enrolment: do not grant portal access because a user typed a name, date of birth and postcode that matched a record. Those values are often known to relatives, former partners and fraudsters. Use an activation code issued in person or by post to the address on the record, or remote identity proofing against evidence.
  • Assurance levels: NIST SP 800-63A-4 (which supersedes SP 800-63A) defines IAL1, IAL2 and IAL3, with identity resolution, validation of evidence and attributes, and verification that the applicant owns the evidence. It does not allow knowledge-based verification. IAL2 addresses scaled and targeted attacks, evidence theft and social engineering; IAL3 requires an attended session and a biometric. Map portal functions to a level: viewing results and messaging clinicians usually justifies more than a bill payment page. See the Identity Proofing and KYC Guide.
  • Name and address changes: NIST allows trusted referees to resolve mismatched core attributes, such as after a recent name or address change, by reviewing supporting evidence. Build that route instead of creating a new record.
  • Proxy access: parents, carers and attorneys need delegated access with its own proofing, documented authority and expiry, with special handling for adolescents' confidential care. See Delegated and Proxy Access in CIAM.
  • Record requests: under HIPAA, 45 CFR 164.514(h) requires a covered entity to verify the identity and authority of a person requesting protected health information if not already known. Apply the same verification to email, phone and contact centre requests, which are where social engineers go. See the Account Recovery and Help Desk Security Guide.

Choosing an approach

SituationRecommended emphasis
Single clinic or small practice on one EHRRegistration standards, search-before-create, monthly duplicate report, deterministic plus vendor probabilistic matching
Multi-hospital system with one EMPIProbabilistic matching with tuned thresholds, a staffed review queue, enterprise data stewardship and merge governance
Health information exchange or network participantPartner trait analysis, certain-only automated linking, restricted $match access, cross-partner overlay reporting
National identifier availableTrace and verify against the national service at every registration, synchronise local copies, route conflicts to the national back office
High-risk patient-facing access (results, messaging, prescriptions)IAL2-style proofing or in-person activation before binding the account to the record

Detection and measurement

The Sequoia framework recommends analysing duplicates, false positives, false negatives and overlays together, using a manually validated set of shared patients as the benchmark, and prefers precision over specificity as a metric. Useful indicators:

  • Duplicate creation rate: new records later merged as duplicates, per 1,000 registrations, by site and by registration channel.
  • Overlay count and time to correct: every overlay investigated as a privacy incident, with days from detection to correction.
  • Review queue age: probable matches waiting for a data steward, and the share resolved within the target time.
  • Precision on the benchmark set: the share of automated links confirmed correct on manual review.
  • Placeholder value rate: records containing default dates, names or numbers.
  • Portal binding without proofing: active portal and proxy accounts with no recorded verification method, which should trend to zero.

Log merges, unmerges and record links as security events: who performed them, which records, and on what evidence. An unmerge after an overlay should trigger a review of everything disclosed from the merged record, including portal views, letters and exchange queries. The EHR Access Monitoring and Patient Privacy Guide covers the access monitoring side.

Implementing an identity integrity programme

  1. Baseline: run a duplicate and overlay analysis on the MPI, list placeholder values, and build a validated benchmark set of record pairs.
  2. Govern: assign an identity integrity owner, data stewards, and a merge and unmerge policy that requires two-person review for clinical record merges.
  3. Fix capture: deploy registration standards, search-before-create, address normalisation and staff training, and pilot them at the sites with the highest duplicate rate first.
  4. Tune matching: adjust thresholds in a test copy of the MPI against the benchmark set, compare duplicates and overlays before and after, and keep the previous configuration ready to restore if overlays rise.
  5. Secure interfaces: inventory every system and partner calling $match, PDQ or equivalent, give each a scoped client identity, and turn on certain-only automated linking.
  6. Proof patient access: introduce verified activation for new portal and proxy accounts, then re-verify existing accounts with no recorded method, starting with proxies.
  7. Measure and report: publish the indicators above monthly to clinical safety, privacy and security governance.

Common failure patterns

  • Auto-merging on probable matches: a low auto-link threshold to clear the review backlog creates overlays that are far harder to undo than duplicates.
  • Portal activation on demographics alone: anyone who knows the patient's name, date of birth and postcode can open the record.
  • Default values in exchange: placeholder dates or phone numbers sent to partners create false matches in their systems.
  • Open matching endpoints: a $match or demographics query interface available to many integrations with broad scopes becomes a patient lookup service for any compromised client credential.
  • Overlays treated as data clean-up: correcting the record without assessing what was disclosed, and to whom, misses a reportable privacy incident.
  • Local copies never resynchronised: departmental systems that keep their own demographics drift from the master index and reintroduce duplicates.

Framework mapping

AreaNIST SP 800-53 Rev. 5NIST CSF 2.0Other
Identity proofing for patient accessIA-12, IA-12(2), IA-12(3)PR.AA-02NIST SP 800-63A-4
Identifier managementIA-4PR.AA-02NHS number, ISB 0149
Demographic data qualityPM-22, SI-18, SI-18(1), SI-10ID.AM-07Sequoia framework; Project US@
Matching interface accessAC-3, AU-6PR.DS-10HL7 FHIR $match; IHE PDQm
Disclosure verificationIA-12PR.AA-02HIPAA 45 CFR 164.514(h)

Practitioner checklist

  • Define duplicate and overlay consistently and report both monthly.
  • Enforce registration standards: required traits, standard formats, no placeholder values, search before create.
  • Run workflows for newborns, multiple births and unidentified patients with a reconciliation deadline.
  • Tune matching thresholds against a validated benchmark set, with a rollback configuration kept.
  • Require two-person review for clinical merges and treat every overlay as a potential privacy incident.
  • Give every system and partner calling matching interfaces its own scoped identity, rate limits and logs.
  • Use certain-only results for automated linking; route probable matches to stewards.
  • Proof portal and proxy users before binding accounts to records; never rely on demographics alone.
  • Verify identity and authority for every record request, including phone and email.
  • Synchronise local demographic copies with the master or national index on a schedule.

Standards and references

Related NHI Mgmt Group resources: Healthcare Identity Security Guide · EHR Platform Security · Medical Identity Theft Guide · Identity Proofing and KYC Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 14 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.