INTEGRATION

Charles Trinh — Clinical Workflow Automation | Las Vegas, NV · Las Vegas, Nevada · Henderson · Clark County, Nevada · North Las Vegas, Nevada

Diagram of a clinician workflow with an automated intake step and a human review step
Charles Trinh — Clinical Workflow Automation

Clinical automation fails when it speeds up the wrong step. I map the workflow a clinician actually runs, identify the step that is worth automating and design the integration that removes it without removing the clinician's judgement.

The workflow the clinician actually runs

Clinical workflow automation begins with the workflow as it is actually run, not the one written in the customer's process documentation. The documented workflow is the one the quality team signs; the lived workflow has workarounds, exception paths and small manual steps that make the clinician's day work. The small manual steps are often where the biggest time is being lost, and they are invisible to anyone reading the policy. I observe clinical sessions, one to three per role per shift, before designing any automation.

Observation is not a survey. I sit beside the clinician, with the clinician's consent, and write down every system interaction, every tab switch, every re-keying of data and every step the clinician swears at under their breath. The observation notes go into a task model: a list of tasks, their inputs, their outputs, their dependencies and their frequency. The task model is the customer's property; I hand it over at the end of the engagement. It is more useful than any vendor demo and it survives long after the automation.

The task model makes the automation decision possible. Not every task should be automated. Some tasks look automatable but are the clinician's judgement written in a few clicks. Some tasks look unautomatable but are a repeated data entry that the clinician does by hand. The observation tells them apart. The clinician's own explanation ("I do this so I can see the last three visits in one place") is the data point that a process document would not contain.

Choosing what to automate and what to leave alone

The automation worth building is the task that is deterministic, frequent and low-stakes. The triage nurse entering a temperature into a vitals form is deterministic (the temperature has one field), frequent (every patient) and low-stakes (a wrong entry is caught by the next step). The clinician deciding whether to admit a patient is none of those. The first is a candidate for automation; the second is a reason the clinician is there. I write the criteria into the customer's automation plan and use them to filter the task model.

The second filter is the clinician's trust in the automation. A deterministic data entry that the clinician checks anyway does not save time; it adds a check. The automation has to be trusted enough that the clinician does not re-do it, and trust comes from the automation showing its work. The automation's output includes the source of each value, the time it was captured and the system that provided it. The clinician sees the values in the form as if they had typed them, with a small provenance indicator,

The third filter is the exception rate. A task that automates 98 per cent of cases and fails 2 per cent of the time is a good task to automate; the exception is small and the clinician handles it as they would a manual step. A task that automates 80 per cent and fails 20 per cent is worse than manual because the clinician has to check every time. I refuse to scale a task above the agreed threshold (typically 10 per cent) without the clinical lead's sign-off. to scale a task whose exception rate sits above the customer's agreed threshold (typically 10 per cent) unless the customer's clinical lead signs the exceptions off explicitly.

Clinician-in-the-loop patterns and when to use each

The design patterns for clinical automation are three: manual-assisted, the manual step stays and the automation provides a suggestion; auto-filled with review, the automation fills and the clinician reviews; and fully automatic, the automation completes the task and the clinician can see record evidence. Most healthcare automation should be manual-assisted or auto-filled with review. Fully automatic is reserved for tasks where the automation's error has no clinical consequence, such as scheduling an appointment slot the patient requested.

Manual-assisted is the pattern the clinician will accept first because nothing changes in the step's output except the clinician has a suggestion. I build the suggestion layer with provenance and a confidence indicator. The clinician accepts or ignores. The automation records both, so the customer can see how often the suggestion was used and how accurate it was. A suggestion accepted 98 per cent of the time is a strong candidate for auto-fill; a suggestion accepted 40 per cent of the time is a candidate for refinement or removal.

Auto-filled with review is the pattern the customer will want next. The automation fills the form, the clinician sees the filled values and reviews them before the form is submitted. The review step is fast if the form's filled values are visibly distinct from the empty values, and the clinician has a one-click accept-all with the ability to correct individual values. The customer's quality team will ask what record exists to show the clinician reviewed, and the answer must be an audit trail that the clinician accepted the values at a specific time.

Integration and the connection to the EHR

The automation's integration to the EHR is where the design gets tested. The EHR may not have an API for the task, in which case the automation integrates through the same interface the clinician uses (a UI automation, a database update via a supported interface or a message). None of these is as clean as an API. UI automation is fragile to EHR upgrades. Direct database updates are outside the vendor's support. A message interface is robust but may not support the operation.

I pick the integration path per task and I document the cost of the choice. UI automation is the most fragile: it breaks on the next EHR upgrade, and the automation's owner must be part of the EHR upgrade regression testing. Database updates are the newest path to break, . I do not use direct database updates on an EHR where the vendor supports an API or a message interface for the same operation; I have used them only when the alternative is manual data entry and the customer's vendor has written consent.

The message interface is the choice I prefer when it exists. A HL7v2 Order message or a FHIR ServiceRequest create is a supported write path, and the automation can integrate with the same interface the EHR has already tested. The cost is that the message interface may not support every field the form has. This is not perfect, but it is robust and it survives upgrades. See the HL7 v2 interface work for the message design and the API gateway work for the API-side policy.

Measure the time and the error, not the activity

The measurement of an automation is time saved and errors avoided, not the number of automations. The time saved is measured from the task model: the baseline is the observed time in the pre-automation sessions, and the post-automation time is measured in a pilot of similar sessions. The errors avoided are measured by the exception rate and by the clinical incidents the automation eliminated. The two measurements are collected by the customer's quality team, not by the vendor, and the numbers are reported to the customer's operations leadership.

The time saved is often smaller than the estimate because the clinician still checks the automation. A task that took two minutes by hand and thirty seconds with review saves ninety seconds, not two. That is real across high volume and a rounding error across low volume. I report the actual number. Forty hours saved per quarter is a good result; a per-patient claim without the volume is marketing.

The error measurement is more important than the time. If the automation introduces a new error class, the time savings are moot. I monitor the automation's output for errors in the first quarter and stop the automation if the error rate exceeds the customer's clinical lead's threshold. The stop is a config change, not a code change, and the clinician returns to the manual step instantly. The ability to stop the automation is the reason the clinician trusts the automation.

Handing the automation back to the customer

The automation is a customer-owned asset when the engagement ends. The handover includes the task model, the automation's code or configuration, the runbook and the measurement dashboard. The runbook names the automation's owner on the customer's team, the conditions for pausing the automation, the procedure for the exception case and the escalation for the EHR upgrade impact. The owner is someone who runs the automation, not a project manager. I have seen automations fail because the owner left and the replacement did not know the automation existed.

The engagement's deliverable is the measured improvement, the ability to stop the automation and the ability to extend it. The customer's clinical leadership sees the numbers; the customer's engineering team sees the artefacts. The automation is a small system with an owner, a runbook and a budget. See the fractional architect service for the ongoing ownership model and the clinical workflow automation details in the engagement's own scope.

The automation I would ship is the one the clinicians asked for, built the way they expected, with a stop button they have used. Anything else is a vendor feature, not an improvement, and the clinicians will switch back to the manual step before the pilot ends.

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 — clinical workflow automation — the scope, the failure modes it prevents and how the work is measured. It sits in the AI and Automation 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

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.

TRINH — Charles Trinh, technology architect
TRINH

Charles Trinh

Fractional CTO/CIO and Technology Architect — Healthcare Data, Identity & Regulated Infrastructure · Las Vegas, NV

Las Vegas, Nevada · US citizen · Fractional and project engagements

(702) 559-3489 Email Charles Trinh