AI

Charles Trinh — Fraud, Waste and Abuse Analytics | Las Vegas, NV · Las Vegas, Nevada · Henderson · Clark County, Nevada · North Las Vegas, Nevada

An FWA alert board with per-rule precision, reviewer outcomes and trend
Charles Trinh — Fraud, Waste and Abuse Analytics

A rule that flags six hundred claims a month and is acted on twelve times is a rule that trains the team to ignore it. FWA analytics is a precision problem first and a detection problem second.

Precision before recall, in this domain

Fraud, waste and abuse analytics is usually framed as a detection problem, and that framing diverges from the reality of the teams that use it. A detection model with ninety percent recall and five percent precision will produce a queue that no investigation team can process, and the queue's growth will degrade the team's actual work more than its absence. The metric that matters is the number of paid investigations a rule produces per month and the confirmed-integrity rate among them. I measure both before any new rule goes live.

The distinction between fraud, waste and abuse is important here. The same rule can be accurate for one and worthless for another, and the response differs: fraud investigation, utilisation management, or provider education. Queues that blend the categories produce alerts that investigators cannot route. I keep the analytics strategy separated by category and the queue operational by category, so the reviewed outcome feeds the right corrective action.

Most of the practical value lies in waste, not fraud. Waste is more common, easier to confirm and often fixable by education or a rule change; catching it before it becomes fraud protects the money and the provider relationships together. I start by measuring waste classes in the claims data, because the measurement is quick and the remedy is usually non-adversarial.

Building a baseline of current performance

The first measurement is the existing queue. What rules run today, what is their monthly volume, what is the confirmed rate by rule and by category, and what fraction of the queue receives an investigation in the month it is raised. Most teams cannot answer the last question without a manual count. The answer typically shows that a handful of rules dominate the volume and a smaller handful dominate the confirmations, and those two sets are rarely the same.

The second measurement is the claims population against the benchmark. For each provider type, service category and geography, the rates that normalise before any anomaly is remarked on: the coding distribution, the procedure mix, the patient mix and the payer mix. A rate that looks anomalous against a general population may be entirely normal for the provider's specialty and location, and a model that ignores this context will alert on the specialty rather than on behaviour.

The third measurement is the investigating team's capacity. The number of alerts per month, the number that reach a decision, and the average hours per decision determine how many alerts per month the team can process. That is the constraint the analytics should design against. An analytics programme that expands the queue beyond this number has produced a report, not a workload.

Explainable flags and the reviewer's time

Every FWA flag should be reviewable in minutes, not hours. The reviewer should see the pattern that triggered it, the specific claims involved, the relevant benchmark or policy, and the provider context in one view. When I build flags I write the review workflow at the same time, and I measure the reviewer's average time with a short usability pass before the rule goes live. A flag that requires pulling claims one by one from a separate system will sit in the queue.

Explainable flags use explicit patterns more often than supervised models, because explicit patterns match how reviewers think and because reviewers can argue with them. A pattern such as “billing a high-level established visit for every encounter in a week” is a claim the reviewer can test by reading a few charts. A model that produces a “risk score of 0.71” from forty features does not. Where a supervised model beats patterns on confirmed-integrity rate, I use it with a pattern behind it. Where it does not, I set it aside.

Explainability has a limit in adversarial contexts. A provider who learns the rule will change behaviour around it, and certain rules are easily evaded. For the small category of alleged fraud, I do not rely on published patterns and I treat the specific review scope as confidential. For waste and abuse, transparent patterns are appropriate and often function as a compliance measure in their own right, because education is the corrective action.

Working with program integrity and payer rules

Program integrity work happens in more than one place: the payer's own SIU, the plan's provider network management, and the provider's own compliance programme. An analytics programme in any of these should align with the rules the payer applies, not invent its own. I work from the payer's published and internal policy on what constitutes a reportable issue and from the provider's contractual obligations, and the rule set is maintained against both. Where the rules conflict, I say so and the compliance team decides.

Some engagements support a provider's internal compliance programme, whose aim is to catch upstream problems before they are reported. There the analytics are a reframed version of the payer's: the same rules on the same claims, used for education and correction instead of investigation. The harness is shared; the legal and operational handling differs sharply, and the organisation's counsel owns that handling, not the analytics. In practice that means working from a compliance charter the organisation already has, not improvising one on the way in.

Claims enter through x12-837p-claims and x12-837i-claims, and revenue-cycle consumes the downstream output. Where a workload also touches enrolment, the identity layer in provider-directory-api matters for linking providers across payer identifiers. A surprising share of false positives originate in those links, so the identity work belongs in the analytics scope from the start rather than as a repair. Those links are also where duplicate provider records hide, and they affect the denominator of every rate it reports.

Reviewer feedback and rule lifecycle

FWA rules need a lifecycle, not a static list. A rule that worked well last year on a different provider mix may be producing noise this year. I assign each rule an owner, a review date and a minimum monthly confirmation rate. Rules that dip below the minimum go back to the analyst for retuning or are withdrawn. Rules that retire get archived with their history, and the archive is the reason analysts are willing to withdraw them, because they can be restored if the provider population changes back.

Reviewer feedback is recorded at the decision level with a structured reason. The reasons accumulate faster than analytics can retrain, so the feedback path also serves as a regular weekly analysis: which reasons are growing, which rules generate the most of them, and which downstream education may be relevant. Reviewer feedback in this domain is the most valuable data the programme has, and it is lost when the only field recorded is the decision.

The feedback loop needs governance too. A supervisor reviews the distribution of decision reasons each month and flags drift. A sudden shift in rejection reasons often indicates a reviewer learning curve, a training gap or a rule change that has not been communicated. None of those belong in the analytics pipeline; all of them are visible through it, and that visibility is worth building for on purpose.

Privacy, fairness and due process

FWA analysis uses claims and, in some investigations, chart data. That is PHI, and the analysis has to sit inside the same access rules as any clinical function. Access should be role-based, audited and limited to the minimum necessary. Analysts do not need full chart access to decide that a claim should be pulled for review. I design the pipeline so that access is graded: an analyst sees the claim summary and the flag, an investigator sees the chart, and the audit trail records each access separately.

Patterns can encode bias. A rule that flags providers serving a particular population because that population's coding patterns are different will drive reviews toward providers who serve that population. I measure alert rates by provider characteristics and by patient population where the data allows, and I say so when a rule's alert rate is disproportionately concentrated in one population. The rule is then either retuned or removed, and that decision is made with the compliance and legal teams.

There is a due process dimension that is easy to miss. A provider investigation is adverse to the provider, and providers have contractual rights to respond. The analytics programme should produce a case file that is sufficient for the provider to respond to, with the specific claims, the specific pattern and the specific evidence cited. That is not only fair; it is what makes a confirmed finding withstand legal challenge, and it is the part of the programme most likely to be under-built when the analytics team treats it as someone else's job.

What I hand over

The deliverable is a rules library with owners and history, a review workflow, a feedback loop and a monthly reporting pack that reports precision, volume, decision mix and confirmation rate by rule and category. The pack also reports the investigation team's capacity and the fraction of alerts that are actually investigated. Those numbers together are enough to tell the programme whether it is working, and they replace the scoreboard of flags raised that is the usual default.

The analytics are built so that a rule change is a configuration change and so that a new rule goes through the same review and calibration step before it reaches production. That is the difference between a programme that can improve and a programme that expands until the team stops reading. The discipline is not glamorous and it is what keeps the programme useful.

My background is engineering, not compliance or investigation. Policy judgments, legal handling and investigative decisions stay with the organisation's compliance and counsel teams and nobody else. I build the pipeline, the measurement and the review workflow, so those decisions rest on evidence rather than on summaries, and the evidence behind each decision is retrievable by the person who has to defend it at an external review.

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 — fraud, waste and abuse analytics — 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

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