MPI · matching · consent ledger · revocation
Charles Trinh — Patient Identity and Consent | Las Vegas, NV · Las Vegas, Nevada · Henderson · Clark County, Nevada · North Las Vegas, Nevada

The transport is solved. The matching is not.
For twenty years the interoperability problem was framed as a pipe problem: move the data from here to there in a standard format over a defined protocol. That problem is largely solved. FHIR R4 is the wire format, US Core defines the minimum content, SMART on FHIR defines the application surface, TEFCA defines the network.
What remains unsolved is the question underneath all of it: when a record arrives, whose record is it?
Why demographics are the wrong substrate
The industry answer has been to layer probabilistic matching on top of name, date of birth, sex and address. All four are mutable, all four are entered by humans under time pressure, and all four collide across a population measured in hundreds of millions.
The result is measurable in both directions. False negatives fragment a clinical history and cause a second workup. False positives put one patient's data into another patient's chart — the failure class no downstream security control repairs, because nothing was breached. The data is simply wrong in a place that looks authoritative.
Four disciplines that separate the good implementations
A deterministic tier that runs first and is trusted absolutely. Where a verified identifier exists — an enterprise MPI number, a payer member identifier tied to verified enrolment, a credential you issued — that identifier decides. A probabilistic score should never override it.
A probabilistic tier that is scored and auditable. Log the score with the attribute contributions that produced it. When a match is challenged, you need to say which attributes drove the decision and by how much. A black-box matcher cannot survive a patient complaint.
A human review queue with clinical context. The middle band needs someone who can see the chart, not a clerk comparing two forms, and it needs a service level. A review queue that sits for a week is a duplicate record with a head start.
An unmerge capability that works. Build it before merge. If a false positive is discovered after records are combined and the system cannot separate them without destroying the audit trail, a correctable error has become permanent.
Consent is a directive, not a checkbox
Most systems implement consent as a boolean on a patient record. That model cannot express "share with treating clinicians, withhold from the exchange, exclude substance use records, expires in twelve months, revocable at any time." Since it cannot express the patient's intent, organisations default to the strictest interpretation, and the exchange delivers less than the patient expected.
Real consent is period-bound, purpose-bound, category-scoped, recipient-scoped and revocable. Partial consent is the normal case. The FHIR Consent resource is a reasonable container for that structure, and it carries no universal enforcement semantics — two systems can both store a valid Consent and disagree completely about what it permits. Building the resource and validating the profile is not building consent. It is building a container for a disagreement.
Enforcement belongs in the data path
Before serialisation, with the decision logged. Filtering after the fact means the data already left the boundary and the log records a leak rather than a control.
And log the withholdings. When a patient or an examiner asks why a record did not travel, "we sent nothing" is indistinguishable from a bug. You need: these elements withheld, under this rule, by this decision, at this time. The withholding log is the difference between a control and an outage.
Segmentation is where it gets hard
Substance use disorder records under 42 CFR Part 2 sit inside the same chart as everything else. Withholding the whole record is clinically wrong; sharing the whole record is legally wrong. The only correct enforcement point is the segment, evaluated at read time, per requester, per purpose, per jurisdiction.
That requires the data model to carry the designation, the query layer to honour it, and the disclosure log to record what was withheld. Most systems do the first, partially do the second, and skip the third entirely.
Identity as the consent holder
The direction I have argued for, including in a proposed USCDI v7 data element, is that the patient's identity credential should carry the consent rather than a row in each provider's database. The reason is structural: consent stored per organisation must be re-obtained per organisation, and the patient's intent does not change when they change providers.
The cryptography is mature. The obstacle is that no payer or provider has an incentive to accept a consent it did not issue, and every one has a liability argument for declining. That is a governance problem wearing a technical costume, and it will be solved by whoever makes acceptance cheaper than re-collection.
What to build first
Enumerate and verify your deterministic identifiers. Instrument the matcher so every score is explained and logged. Build unmerge before merge. Model consent as directives with periods and categories. Design segment-level enforcement and log the withholdings. Then connect to the network.
Most programmes do this in reverse, because the network connection is the visible deliverable. The visible deliverable then propagates the invisible defect to every partner connected to it.
What I do
I built patient identity and consent into a clinical platform I operate — a 337-table clinical store behind a 51-hostname estate, with a Hyperledger Fabric and EVM ledger pair for consent and disclosure records. I work with payers, providers and vendors on matching design, consent directive architecture, and the segmentation enforcement Part 2 actually requires.
If your duplicate rate is a number you cannot explain, that is the first conversation, and it is measurable in a week.
Choosing a matching vendor without being sold a score
Every enterprise master patient index vendor will present an accuracy figure. The figure is not comparable across vendors, because it is generated against a test set the vendor chose, scored on a definition of match the vendor defined, and reported without the false-positive rate.
Ask instead for three things. The false-positive rate separately from the false-negative rate, because they have different clinical consequences and only one of them is reversible. The review queue volume the configuration produces per ten thousand records, because a matcher that is accurate and generates an unstaffable queue is not deployable. And the explanation output for a single match — the attributes, the weights, the score — because a matcher that cannot explain itself cannot survive a patient complaint.
The metrics that diagnose the problem
Duplicate rate per source system, not in aggregate. The problem is almost always concentrated in one interface or one registration site, and an aggregate number hides it.
Review queue depth and age. A queue that is deep is a staffing problem; a queue that is old is a throughput problem, and they need different fixes.
Merge and unmerge counts. A high unmerge count means the thresholds are too loose. A zero unmerge count over a year means nobody has found the false positives yet.
Queries issued against responses with clinical content. The single best indicator of whether the exchange is delivering anything, and the one that per-source reporting makes actionable.
Consent UX is a control, not a nicety
A consent screen that a patient cannot understand produces a consent that cannot be defended. Name the categories, name the recipients, name the duration, and make refusal as easy as acceptance.
Then record what was chosen, not merely that something was chosen. The dispute six months later is about scope, and a boolean cannot answer it.
Part 2: how to test that enforcement actually works
Create a test patient with a substance use disorder record, a general clinical record, and a consent directive permitting treatment disclosure while withholding the Part 2 content. Then request as a treating clinician entitled to everything else.
Exactly one category should be withheld, and the withholding should appear in the disclosure log with the rule, the decision-maker and the timestamp. If the response returns everything, enforcement is absent. If it returns nothing, enforcement is at record level rather than segment level. If it returns the right thing and the log is empty, you have a control with no evidence — which fails the first time anyone asks.
What a ledger buys, honestly
An append-only sequence of signed consent events makes history provable across parties that do not share a database. It does not compute the decision, and reading the chain on every request is a latency problem looking for a reason. Materialise current state into a fast read model and keep the chain as the authority for disputes.
What I do
I built patient identity and consent into a clinical platform I operate — a 337-table clinical store behind a 51-hostname estate, with a Fabric and EVM ledger pair for consent and disclosure. I work on matching design, consent directive architecture, and the segment-level enforcement Part 2 requires, and I start with the duplicate rate because it is the one number that can be established in a week.
The vendor question that ends the conversation
Ask any matching vendor to show you a single match decision, with the attributes, the weights and the score that produced it, for a record from your own data rather than theirs. A vendor who can will show you in five minutes. A vendor who cannot has built a black box, and a black box cannot survive the first patient who disputes a merged record.
Ask separately for the false-positive rate. Not accuracy, not match rate — the rate at which two different people were combined. It is the number that decides whether the system is safe to turn on, and it is the number most often absent from the deck.
Then ask what the review queue volume looks like per ten thousand records at the threshold they are recommending. If the answer is a figure, they have deployed this before. If it is a reassurance, they have not.
What the ninety-day sequence looks like
Days one to thirty: measure. Enumerate deterministic identifiers and establish the true duplicate rate per source system. Days thirty to sixty: instrument the matcher so every score is explained and logged, and staff the review queue for the middle band with someone who can see clinical context.
Days sixty to ninety: build the unmerge path, then model consent as directives with periods and categories, then test that a requester entitled to everything else has exactly one category withheld and the withholding is logged.
Most programmes invert this and start with the credential. The credential is the destination, and starting there produces a pilot that demonstrates beautifully and resolves nothing.
None of that requires a vendor. It requires a decision about which error the organisation is willing to have, recorded with its rationale, and a staffing commitment to the review band that decision produces. Most of the cost of a bad matching configuration is the review queue nobody funded, discovered a year later 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.
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
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 — patient identity and consent — 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.
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.