MPI · survivorship · merge and unmerge
Charles Trinh — Master Patient Index | Las Vegas, NV · Las Vegas, Nevada · Henderson · Clark County, Nevada · North Las Vegas, Nevada

What an MPI is for
A master patient index exists to answer one question reliably: is this the same person as that record. Everything else it does, including demographics, identifiers and links, is in service of that question. An MPI that cannot answer it with evidence is a database of names, and a database of names is what produces duplicate charts, split histories and the wrong patient on the table.
Why matching is hard
Names change. Addresses change. Dates of birth are mistyped at registration by people under time pressure. Identifiers are issued by different systems with different formats and different scopes. The same person can legitimately appear with three spellings and two dates of birth, and two different people can share a name and a birth date in the same city. Deterministic matching on name plus date of birth fails in both directions.
Deterministic versus probabilistic
Deterministic matching compares exact values and links on agreement. It is fast, explainable and produces almost no false links, but it misses the majority of true duplicates. Probabilistic matching scores field agreement with weights learned from the population and links above a threshold. It finds far more duplicates and produces false links that must be reviewed. The right answer is usually both: deterministic for the automatic path, probabilistic to generate candidates for review.
The threshold is a clinical decision
Where you set the auto-link threshold determines how many false merges reach the chart. A false merge is not a data-quality inconvenience; it puts one patient allergy list, medications and history into another patient record. Set the threshold with clinical input, document the reasoning, and treat any change to it as a change requiring review.
Review queues are the product
The part of an MPI that actually works is the review queue: candidate pairs, the evidence for and against, and a decision recorded with a name and a date. An MPI without a staffed review queue either auto-links everything, which is unsafe, or links nothing, which is useless. Budget for the queue before you buy the engine.
Unmerge is not optional
Every MPI will produce a wrong merge. The question is whether you can reverse it completely, including downstream references in the EHR, the data warehouse and any analytics that already consumed the merged record. Design the unmerge path at the same time as the merge path, and test it before go-live, not after the first incident.
Identifiers and their limits
An enterprise identifier is only as good as its issuance discipline. If registration staff can create a new identifier when a search returns nothing, you will generate duplicates faster than matching can resolve them. The registration workflow is where duplicate rates are actually set, and it is usually outside the MPI project scope, which is why MPI projects underdeliver.
What to measure
Measure duplicate rate on a sampled chart review, not on the engine own report. Measure false-merge rate, review queue age, and the proportion of registrations that created a new record when a match existed. The last number is the one that tells you whether registration is helping or hurting.
The adversarial test
Take ten known duplicate pairs and ten known distinct pairs that look similar. Run them through the engine and inspect the scores. If the engine cannot separate them, no threshold will save it. Repeat after every tuning change, because tuning to reduce false merges reliably increases missed matches.
Interoperability dependency
Every external exchange depends on the MPI being right. A care-everywhere query that returns the wrong patient records is worse than one that returns nothing, because it is acted upon. Fix matching before expanding exchange, or you scale the error.
Sequencing
Start with registration discipline and a duplicate report, then deterministic matching, then a review queue, then probabilistic scoring. Each stage is useful on its own and each makes the next cheaper. Buying the engine first and building process around it is the standard failure.
The outcome
A working MPI produces a single chart per patient with a documented decision trail for every link, a review queue that stays under a week old, and an unmerge path that has been tested. That is achievable in months, not years, if the scope stays on matching rather than on replacing the EHR.
Registration is the real control point
Duplicate rates are set at the moment of registration, not in the matching engine. If the search returns nothing and the clerk can create a record anyway, duplicates accumulate faster than any engine can resolve them. Require a reason code to create a new record when a search returned candidates, and review those codes weekly. That single control does more for duplicate rate than any tuning of the matching algorithm.
Data quality inside the record
Matching is only as good as the fields it compares. A date of birth entered as today date, a name entered in the address field, a phone number shared by a household of six: each of these degrades scoring for every future comparison involving that record. Validate at entry, and treat obviously wrong values as defects to be corrected rather than as noise to be tolerated.
Consent and the MPI
Linking records is not the same as permitting their disclosure. An MPI that links two records has established identity, not authorisation. Keep the consent and disclosure logic in the exchange layer, and make sure the MPI cannot be used as a shortcut around it. Conflating the two is how an identity system becomes a privacy incident.
Vendor selection questions
Ask any MPI vendor how they represent a merge decision, whether the decision is reversible, what the unmerge does to downstream references, and how the review queue is staffed and aged. Ask for the false-merge rate on a reference customer with a similar population. A vendor who cannot answer the unmerge question has not built one.
The steady state
In steady state an MPI runs a daily candidate pass, a staffed review queue, a weekly duplicate sample, and a monthly threshold review. The work is routine and small. The failure mode is not a bad algorithm; it is a queue nobody staffs, which turns a working system into a growing backlog within a quarter.
Measuring the engine against reality
The engine reports its own confidence, and that number is not evidence. Evidence is a chart review: a sample of linked pairs and a sample of unlinked pairs, each adjudicated by a person who knows the patients. Report the engine accuracy against that review, not against its own score distribution. The two diverge, and the divergence is where the risk sits.
The cost of doing nothing
An unresolved duplicate is not a neutral state. It splits a medication list, hides an allergy, and produces two bills for one encounter. Quantify the cost in chart-review hours and in clinical near-misses before asking for budget, because the case for an MPI is made in clinical terms, not in data-quality terms.
The review queue is a staffing decision
Every probabilistic matcher produces a middle band, and the size of that band is set by the thresholds rather than by the data. A queue that is deep is a staffing problem; a queue that is old is a throughput problem; and a queue nobody owns is a duplicate record with a head start.
Staff it with someone who can see clinical context, give it a service level measured in days rather than weeks, and report its age weekly to the same person who owns the duplicate rate. The two numbers move together, and one without the other is a half-truth.
Review the threshold annually against the population. A matcher tuned on last year's demographics degrades quietly as the patient population shifts, and the degradation shows up as a duplicate rate nobody can explain.
Proof
- ONC (g)(10) self-test PASS, 317 of 317 subtests, April 2026 — public suite URL
- 67 public hostnames, 51 serving, one AWS account
- 104 clearinghouse enrolments across 53 jurisdictions
- USCDI v7 proposal — Patient Digital Identity Credential — listed at isp.healthit.gov
- SAM.gov UEI MVKQW9DKKGB8 · CAGE 20KM5 · NPI 1194645564
- AWS Business Associate Addendum active since 18 August 2026
Where this work is delivered
Engagements run from Las Vegas and reach across the valley: clients in Las Vegas, NV; Las Vegas, Nevada; Henderson; Clark County, Nevada; and North Las Vegas, Nevada — plus remote delivery nationwide.
How an engagement runs
1. Trace one real case
Follow a single patient or claim end to end with a timestamp at every hop, before any document is written.
2. Written findings
A findings list naming the defect, the system, the owner and the date.
3. Monthly review
Review the list monthly and check whether the dates held.
Frequently asked
What is this service?
This page describes charles trinh — master patient index — the scope, the failure modes it prevents and how the work is measured. It sits in the Identity and Consent category at trinh.io.
Where is the work delivered?
Engagements run from Las Vegas, Nevada across the valley — Las Vegas, Henderson, North Las Vegas, Spring Valley, Paradise, Winchester, Sunrise Manor, Enterprise, Summerlin, Clark County — and remotely anywhere in the United States. On-site work in the Las Vegas valley, remote delivery elsewhere.
What does it cost?
Standard rate $195 per hour, committed $165 per hour at twenty hours or more per week for three months or more; $1,600 standard day, $1,350 committed day. Scoped in writing before work starts, fixed fee where the deliverable is defined.
Who is Charles Trinh?
A technology architect in Las Vegas, Nevada, working in healthcare identity, consent and interoperability. Not a physician, not a medical practice, and not affiliated with any other Charles Trinh.
How does an engagement start?
With one trace of a real case and a written findings list naming the defect, the system, the owner and the date. The first call is twenty minutes and free: 702-559-3489.
Engagement
Scoped in writing before work starts; fixed fee where the deliverable is defined. First call is 20 minutes and free. 702-559-3489 · charles@trinh.io
Talk to Charles
- Phone: 702-559-3489
- Email: charles@trinh.io
- Based in Las Vegas, Nevada · working remotely across the United States
Record
Profile
Healthcare technology architect working where federal regulation becomes production infrastructure. Built and operates a FHIR R4 clinical platform — SMART App Launch authorization server, patient and provider portals, SQLite — on a hardened single-host AWS estate of 67 public hostnames. The platform returned a self-test PASS on the ONC (g)(10) Standardized API test suite in April 2026, 317 of 317 subtests, publicly verifiable. Designed a consent architecture in which a permissioned ledger records patient grant, refusal and revocation as immutable events while all clinical data stays off-chain and encrypted per patient — no PHI on the ledger by design. Filed a proposed USCDI v7 data element with ONC on verifiable patient identity. Translates the Federal Register, HHS/CMS guidance and HITRUST requirements into architecture, controls and audit evidence.
Selected engineering scale
- Self-managed estate of 67 public hostnames, 51 serving, on one AWS account.
- FHIR R4 API with a SMART App Launch authorization server; patient and provider portals; an AI service layer; Hyperledger Fabric peer, orderer and CA; IPFS; a 337-table SQLite clinical store.
- Two permissioned distributed ledgers in production, one operating a consent registry that records grant, refusal and revocation with no PHI on chain.
- X12 005010 pipelines across eight transaction sets.
- Designed, built, deployed and operated end to end by one person.
Credentials and standards work
- Inferno (g)(10) self-test PASS, 317 of 317 subtests, April 2026 — the ONC Standardized API test suite. Standards validated: US Core 6.1.0, USCDI v3, SMART App Launch 2.0.0, Bulk Data 2.0.0. Public suite URL.
- USCDI v7 public comment submission, April 2026 — proposed Patient Digital Identity Credential under the Provenance data class, listed publicly on the ONC Interoperability Standards Platform.
- SAM.gov UEI MVKQW9DKKGB8 · CAGE 20KM5 · NPI 1194645564. AWS Business Associate Addendum active 18 August 2026.
Technical environment
AWS us-east-1 (compute, storage, messaging, SMS, email, voice) · Linux and systemd hardening · nginx · Fabric and EVM ledgers · FHIR R4 · X12 005010 · HL7 · SQLite · self-hosted mail, drive, search and monitoring · Git-based change control and CI release gates.