SSO · provisioning · least privilege
Charles Trinh — Identity and Access Management | Las Vegas, NV · Las Vegas, Nevada · Henderson · Clark County, Nevada · North Las Vegas, Nevada

Three populations, one model
Enterprise identity answers who you are and what you may do. Healthcare identity has to answer that for workforce, patients, and trading partners at once, across systems never designed to share an authorisation model. The three populations have incompatible lifecycles, and a model that works for one will fail the other two.
Workforce joins and leaves on an HR schedule, moves between roles on a manager's decision, and is expected to lose access the moment employment ends. Patients appear, vanish for years, and return with new coverage and a new name spelling. Trading partners are organisations with delegated humans inside them, and the delegation changes with no notice to you at all.
The failure to separate them produces the two outcomes that matter: an access model nobody can explain to an auditor, and a portal that locks patients out of their own data. Both are expensive, and only one of them is loud.
HR is the authoritative source, not the directory
Directory-first provisioning is the most common structural mistake in healthcare identity. The directory becomes the place where access is granted, and the HR system becomes a thing that is consulted occasionally, so a person's entitlements drift from their actual role with nothing to pull them back.
Make HR authoritative for employment state and role, and derive access from that. The derivation should be a function: role in, entitlements out, computed fresh rather than accumulated by whoever filed the last request. A derived model can be reviewed against its source; an accumulated one can only be reviewed against itself.
The test is simple and rarely passed. Pick an employee, ask what they can reach, and reconstruct the chain from HR record to each entitlement. If any link in that chain was granted by a person who has since left, the model is not derived. It is drift with documentation.
Movers are where access accumulates
Joiner automation is easy and most organisations have it. Leaver automation is easy and most have it too, sometimes with a delay measured in days. Movers are the population nobody wrote a rule for, and they are where excess access quietly accrues for years.
Someone changes teams in March and keeps the group membership from their previous role because no process revoked it. Six months later they can see a patient population they have no clinical relationship with. No review flagged it, because the review compared their current role to their current role and found them consistent.
The fix is re-derivation rather than delta. On any role change, recompute the full entitlement set from the new role and revoke whatever is not in it. That is more disruptive than additive provisioning, which is precisely why it is correct behaviour and precisely why nobody has implemented it.
Privileged access and the standing grant
Standing production access for engineers is normal in small organisations and indefensible at any scale where an auditor is involved. The defensible model is time-bound elevation with a recorded reason, an approver where the risk warrants one, and automatic expiry that does not depend on a human remembering.
The friction is real and it is small. The finding it removes is large and it recurs, because standing access is the first thing a reviewer asks about and the easiest thing for a team to have never reconsidered since the company was eight people and everyone knew each other.
Break-glass access should exist and should be loud. An organisation with no emergency path invents one informally, and the informal path is the one that never appears in the log, because it was built by exactly the person who needed it and had no time to file a ticket.
Authentication strength is the highest-value control
Multi-factor authentication everywhere, and phishing-resistant authenticators for administrative roles. This is the control that prevents the incident rather than detecting it afterwards, and it is also the one most often implemented as SMS, which is a control in name only.
The distinction matters in healthcare because the credential in question reaches clinical data at population scale. A compromised workforce account is not one record; it is every record that role was entitled to, queried at machine speed before anyone notices an unusual pattern in a system where unusual patterns are normal.
Pair readiness for this with the operational reality: clinicians work on shared workstations and move between them mid-shift. An authentication design that assumes a personal device produces workarounds, and workarounds in a shared clinical environment are how sessions stay open on a screen nobody is watching.
Patients are not workforce
Patients need different proofing, different recovery, and a different consent model. Treating them as another directory population produces a portal that locks people out of their own data, which under information-blocking rules is no longer just a support cost.
Proofing should match the risk of what the portal exposes. Viewing a record and downloading it are different actions, and a single level of assurance applied to both either over-burdens the common case or under-protects the sensitive one. Grade the actions, then grade the requirement.
Consent belongs in this model too. A patient grant is not an entitlement in the workforce sense; it is a directive with a period, a scope, and a revocation. Modelling it as an entitlement row is the mistake that makes revocation impossible to honour immediately.
Recovery is the attack surface
Account takeover happens at recovery, not at login. A recovery process that verifies less than the login process is the weakest link in the chain, and it is the link users are pushed toward under time pressure when the normal path is unavailable.
Design recovery with the same rigour as authentication. Who can reset, what evidence is required, what is logged, and what stops a help-desk agent from being socially engineered by someone holding a name, a date of birth and a former address — all of which are in the record the attacker is trying to reach.
For privileged accounts, recovery should require a second person. For patient accounts, recovery should never be faster than enrolment, because the asymmetry between the two is exactly what an attacker exploits.
Trading partners and delegated authority
Organization-level identity with delegated human identity inside it, and the delegation effective-dated. A partner administrator who can add users is a privileged role, and the model should say so explicitly rather than implying it through a UI that happens to permit it.
Machine-to-machine access uses mutual TLS or signed assertions. Delegated human access uses OAuth with scopes that reach the data path. Neither substitutes for the other, and mixing them produces system accounts with human names, which is how an audit trail becomes useless.
Maintain a register: which partners, which users, which entitlement, granted when, reviewed when. That register is the artefact that answers the question after an incident, and it is the one nobody builds until the day it is requested.
Policy in one place, evaluated in the data path
Roles encode too much and attributes encode too little, so use both. Role for the coarse grant, attribute for the constraint: this clinician, this panel, this facility, this purpose, this window of time. One expression, evaluated at the point of data access.
A policy expressed as an if-statement inside a route handler cannot be audited, cannot be reviewed as a whole, and cannot be tested independently of the endpoint it guards. The first auditor who asks for the access model receives a code review instead of an answer, and that is a finding regardless of whether the access was correct.
Express the policy as data, log the decision with the inputs that produced it, and test the policy by itself. An access decision you cannot replay is an access decision you cannot defend six months later when the logs are all you have.
The audit trail must name the human
Who, what, when, from where, under which entitlement, and on whose authority. For clinical data the last item is the one that matters: not which application made the request, but which person was behind it and what relationship entitled them to the record they opened.
An audit trail without the human in it answers a security question and not a clinical one. When a patient asks who looked at their record, an application name is not a response, and a session identifier is worse.
This is where the fhirUser claim earns its place. It connects a token to a practitioner the record recognises, and omitting it means the entire access log describes software rather than people.
Three tests that prove it works
Take a person who changed roles mid-quarter and confirm their access reflects the change on the day it happened. Most organisations can answer the join question and the leaver question. Almost none can answer the mover question, and that is where the excess access lives.
Terminate a test account and confirm every session dies within the defined window rather than at token expiry. Then revoke a partner delegation and confirm their users fail on the next request rather than at the next review cycle, because a revocation that takes effect at review time is not a revocation.
Those three tests take an afternoon and cover the failures that actually produce incidents. Every other identity control is a proxy for them, and a proxy that has never been run against a real change is a belief rather than a control.
What I build
I have run identity, access and consent for a clinical platform on a single AWS account under an active business associate agreement: 51 serving hostnames, a 337-table clinical store, workforce populations, patient populations, and partner access scoped to contract with effective dates.
I do not sell a directory product. I fix the derivation model, put the policy in one place, make the data path enforce it, and instrument the three tests above so the answer exists before it is asked for.
If you cannot answer the mover question today, that is the first thing to fix, and it is measurable in a week rather than a quarter.
What good looks like in ninety days
Days one to thirty: make HR authoritative and prove derivation works for one department. Pick the team with the most role churn, recompute entitlements from HR, and diff the result against what people currently hold. That diff is your finding, and it costs nothing beyond access to two systems you already run.
Days thirty to sixty: kill standing production access and replace it with time-bound elevation carrying a recorded reason. Expect two weeks of complaints and then no further complaints, because the friction is one form and the alternative was never as convenient as people believed. Log every elevation from the first day, including the approvals that were granted in under a minute.
Days sixty to ninety: run the mover test on last quarter's actual transitions. Every person who changed role, checked against their access the week after the change. Anything they retained that the new role does not require is either a derivation defect or an undocumented exception, and both need a decision rather than a clean-up.
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 — identity and access management — 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.