AI
Charles Trinh — Prior Authorization Automation | Las Vegas, NV · Las Vegas, Nevada · Henderson · Clark County, Nevada · North Las Vegas, Nevada

A prior authorization is a conversation between systems that disagree about what the patient's situation is. Automating it means building the disagreement handler and the audit trail, not just the submitter.
The real work sits after submission
Submitting a X12 278 request is the easy part of prior
authorization. The difficult part is what happens when the payer answers pending, and the
payer will answer pending for a large fraction of realistic requests. A pending response
is not a failure of automation; it is the payer's process telling you that it wants
something specific, and the job is to read that request precisely and answer it without
losing the case in a queue.
The failure mode I see most is a script that submits and then treats anything other than
an authorisation as an error to retry. Retrying a pending case creates duplicates, and
duplicates at the payer side are a path to denials. I build the submitter with an explicit
state machine for each 278 response code, and I log every transition with the
payer's own reason text, because that text moves once they refine a policy and yours does
not.
Measurement comes before build. On a sample of a couple hundred recent authorisations I compute the pended rate, the reasons driving it, the time in each state, and the fraction that required a phone call. That distribution tells us which exception reasons justify automation and which should be routed to a human queue by default, and it is the only honest basis for deciding where the automation should stop.
Reading a payer's rules without guessing
Payer requirements come from several places that do not agree with each other. The
published policy sits in a coverage bulletin, the portal form asks different questions,
the companion guide specifies the segments we must send, and the call centre says something
else again. I read all four for each of the eight X12 005010
transaction sets we touch and record where they conflict, because a conversation with the
payer that cites the conflict is more persuasive than one that cites the bulletin.
Some of the payer's logic can only be read from its own response behaviour. I run controlled probes across cases that differ in one field and record how the response changes. This is measured behaviour, not a guess at intent, and it gives the team a table of observed rules with a date and a response behind each. When a payer revises a policy the probes catch the change before the denials do.
The answer to a payer is a specific set of observations, not a narrative. Where the requirement is an imaging indication, a trial-of-therapy record or a specific diagnosis, I assemble the observations from the clinical store as structured values with a reference range and a source. Sending a paragraph of pasted notes is the practice that pushes requests into a human reviewer, and it is the practice the automation removes.
Exception routing with a human at the seam
The submitted fraction is the wrong denominator; the decision-without-retyping fraction is the right one. A system that submits 95 percent of requests but sends 60 percent of pending cases to a manual queue has automated the bureaucratic step and left the costly one alone. The report claiming a 95 percent automation rate has simply picked the flattering number and should be ignored.
The human review point has to be visible in the case record. When a case is routed to a
reviewer, the record says which rule routed it, what evidence was insufficient, and what
the clinician or authorisation specialist did next. That is the human-in-the-loop-review
design pattern applied to a clinical workflow, and it turns the review queue into a source
of measured data instead of a backlog of guilt.
Reviewers also need a way to disagree with the automation that does not require a ticket. A small override control that records a reason serves as a compliance control and an evaluation feed at once. Reasons accumulate, and the automation extends one reason class at a time, once that class is common enough to justify the engineering work. Overrides are also the most honest source of evidence that the automation is improving.
Turnaround and the states that cost time
Patients wait for prior authorisations in days, and the number worth managing is the median time from clinical decision to payer decision, not the median time from submission to decision. A case can sit in the internal queue for two days before it is submitted and look fast in the payer's reporting. I instrument the clock across all of it, including the time a case waits on a clinical document, and I report the slowest state weekly because that is the one worth fixing.
Adding a human review point changes the turnaround and must be priced honestly. A queue with a two-hour service-level target and a single reviewer is a different system from one with a four-hour target and coverage across shifts. When I compute the projected turnaround I use the observed review capacity, not a theoretical one, and I say when a proposed review volume exceeds the capacity the team currently has.
Denials and resubmissions are part of the turnaround picture and should share the same case record. When a resubmission repeats a request that was denied, the record shows the denial reason and the change made. If the same denial reason recurs across cases, that is a data problem in submission and the fix is upstream, in what the automation reads from the clinical store.
Payer portal versus X12 versus fax
Not every payer accepts a 278, and the mix varies by plan and by state.
Some accept it for certain service categories only, some disagree about whether a pended
response can be followed by an interactive attachment, and some require the portal for
everything. I measure the channel mix for your top payers and build the relevant channels
in order of case volume, starting with X12 and adding a portal automation with
its own state machine where the volume justifies it.
Browser-driven portal automation carries hazards that X12 does not. The
portal changes without notice, the login requires the staff member's credentials, and a
script that clicks through a payer portal can submit something the staff member did not
intend. I keep portal automation to reading status and attaching documents where the
portal permits it, require an explicit approval for any mutating click, and record every
action with a screenshot and a timestamp for the compliance record.
A practical mix is 278 for submission, portal for status where the payer
answers faster there than in 278, and fax as a last resort with the case
record linking the fax image to the case. None of this should look like one system to the
authorisation team; it should look like one queue with the channel as an attribute, and the
attribution evidence lives with the case.
Evidence and attribution in the case record
Every payer interaction leaves an artefact: the request, the response, the attachments, the phone call notes and the decisions. I store all of them against the case with a sha256 and a received timestamp, and I make the record retrievable by the patient and by the auditor. For the payer-to-payer and appeal workflows a case that cannot produce its evidence efficiently is a case that will lose, regardless of the clinical merit behind it.
The attributable record also matters for the audit trail around a determination that looked automated. When a payer's decision arrives quickly, the case record should show what the automation sent and what the human decided, separated. That separation is how an organisation answers whether an automated system made a clinical decision without a person reviewing it, and it is increasingly what a regulator or an appeal asks for.
I have engineered across the eight X12 005010 transaction sets,
which is how the 278 work connects to claims and eligibility. Where the
authorisation queue also generates denials, the remedy usually reaches into
revenue-cycle analytics, and the case record described here is the input that
this downstream work consumes. The connection runs both ways: denial reasons feed back into
the authorisation rules.
What starts the engagement
I start with a read of the current authorisation workload: case volumes by payer, service type and channel; current turnaround by state; the fraction of pending cases handled by phone; and the denial rate by reason. It takes a week of data and a conversation with the authorisation team, and it produces a map of where automation pays and where it would only move labour around. Some of the best outcomes are a rule removed from a payer, not a script added to ours.
The deliverable is the case state machine, the channel adapters, the review queue and the weekly report, in your repository, running from your pipeline. The state machine and the review queue interface are reused for other payer workflows, so the second engagement is usually smaller than the first. That reuse is intentional, and I build it up front rather than retrofitting it.
Engagements are quoted after the read, based on the channels and the exception rules found, not before. Prior authorisation work is one of those areas where a number quoted before the data often doubles once the payer mix is known. My estimates after the read have usually landed within a couple of working days of eventual effort. That is a track record on a small set of engagements, not a guarantee.
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 — prior authorization automation — the scope, the failure modes it prevents and how the work is measured. It sits in the X12 and EDI 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.