DATA

Charles Trinh — Population Health Analytics | Las Vegas, NV · Las Vegas, Nevada · Henderson · Clark County, Nevada · North Las Vegas, Nevada

Care management registry showing risk bands and open care gaps by primary care assignment
Charles Trinh — Population Health Analytics

Population health analytics only works when the denominator is honest, the attrition is visible, and the intervention it recommends is one a care manager can actually deliver this week.

The honest denominator problem

Population health reporting lives or dies on attribution. A panel of 40,000 attributed patients that does not actually count everyone who sees the organisation is a fiction, and almost every organisation runs at least one reporting artefact that uses a fiction. The most common cause is dual attribution: the same patient attributed to two primary care providers, to both of whom they are a non-returner. Both panels then overstate turnover and understate the size of the population they serve.

I start with a written attribution rule for each programme: who is included, for how long, under which encounter types, and how an overlap is resolved. The rule is decided by the clinical lead with the finance side, and written once. When the panel changes, the rule is the reference point, not the spreadsheet. This one document prevents most of the arguments about why the diabetic cohort is different in two reports.

The artefact is a patient-registry table with one row per patient per programme and one attribution row per month. Every downstream report joins that table. When someone asks why the panel is 41,300 rather than 39,880, the table shows the two attribution sources and the resolution applied, and the conversation ends. Without it, the conversation repeats quarterly and trust in all reports erodes with it.

Risk stratification that means something

Risk scores are useful when they rank accurately and are used within the period the score was calculated for. A model that predicts next year's cost is the wrong input for a programme that contacts patients this week about care gaps in the next ninety days. I separate retrospective risk, which drives contract financials, from prospective risk, which drives outreach lists, and I document which model produces which score, against which outcome, over which horizon.

Scores are calibrated against the actual outcomes in the local population, not the model vendor's reference set. Calibration plots are the review artefact, and a model whose top decile actually contains the top decile of events is the only one worth using for care management. A miscalibrated model floods the top of the list with healthy high-utilisation patients and leaves the modifiable risk outside the top cohort entirely.

Social risk factors belong in the model, but only where they are collected and reliable. I avoid imputing social segment from ZIP code because the assumption that everyone in a postcode has the same barrier produces systematic overreach in dense urban areas and underreach in rural ones. The quality measures work covers the clinical side of risk, which does not move as easily as financial risk when a population shifts.

Cohorts that care managers can act on

A cohort of 9,000 people with an open gap is not actionable. A cohort of 40 people who are due for a review, prefer a morning slot, have no recorded contact in three months, and are part of a programme whose funding depends on their follow-up is something a team can work this week. I build cohorts with care managers in the room, not from a desk in analytics, because the difference between those two cohorts is not visible in the data layout.

Every cohort has a defined exit condition as well as an entry condition. If a patient exits only when the gap closes and the gap is a permanent contraindication, they stay on the list for ever and the list stops being worked. The exit condition also tracks when a patient drops out, because attrition from a cohort is itself a finding about the programme and not only about the individual.

The list is delivered where the care manager works, which is usually a registry view in the EHR or a care management application rather than a separate analytics dashboard. Multiple lists in multiple tools are the same as no list. The artefact is the integrated view and the fact that the care manager does not need to open a second system to act on it.

Outcome measurement and the counterfactual

Programmes need a way to measure whether they worked that does not depend on the people who did not engage. A programme with ten per cent engagement and a large apparent improvement often has selection bias in the ten per cent. I build a matched or segmented comparison, using modelled comparison groups where randomisation is impossible, and I publish the comparison basis alongside the outcome. The reader can see what the number is being compared against and judge the claim accordingly.

For most programmes the comparison group is period-matched and risk-matched, from the same source population, over the same follow-up window. That is not the same as a randomised comparison and it does not prove causality. It does produce a number the clinical lead can defend and report to payers without overclaiming, and that is the practical requirement for a programme that is going to continue.

Attrition matters here. When patients leave the panel and the programme, the denominator for outcome changes. I report outcomes both for the full attributed cohort and for the actively managed subset. The two views answer different questions and publishing both prevents the most common form of outcome inflation, which is measuring against only the people who remain., and the exit condition is reviewed whenever the programme funding cycle changes.

Integrating claims, clinical, and social data

Most population health analytics needs claims, clinical and social data together, and the three rarely share identity cleanly. Member IDs and patient IDs differ and change. The integration starts with the master patient index, where the link between payer members and hospital patients is established, reviewed and versioned. When the MPI link is weak, every analytic downstream is weak in an invisible way.

I build source-specific views and join them on the MPI key, keeping the source-specific identifiers accessible for audit and for business rules that need them. Claims data needs its own latency contract because it arrives at month end and then restates. Population health reports carry the claims period explicitly, because a report that reads as current when claims are two months behind is a known trap.

Social data arrives in pockets and often from a separate application. The same person may have different housing status across systems. I never overwrite one with another; each field carries its source and its date. The registry picks the source according to a documented priority rule, and consumers see the rule when the value is queried., which prevents the trend line from bending every time a programme owner leaves.

Dashboard design for management

Management dashboards carry cohorts, gaps, engagement and outcomes in one page, with a visible line on what the data is current to. Anything that reads as current when it is not is a defect. I omit confidence bands from management views because they create decision paralysis, but they are available to clinical leads in the detailed view, and the toggle is visible.

Trends are shown on the attribution basis at the time of the period, not the current basis. When a panel changes in July, the January trend is still January's panel, and the reader can see when the rule changed. Recomputing retrospective panels undermines the point of a trend line: a series that is re-fitted each time the panel changes cannot show change, only today's snapshot in several shapes.

The dashboards are hosted on the hospital's platform and backed by a clinical data warehouse with the same access controls as the underlying data. A population health dashboard that exposes patient-level detail to a manager who should see aggregates is a privacy finding waiting for an audit, and I check access on every page as part of delivery rather than at the end as a review.

What handing it over looks like

Handover includes the attribution rules, the calibration review schedule, the cohort definitions with their exit criteria, the outcome models with their comparison basis, the intake form for a new programme, and a named owner for each. The new owner reads the pack and can run the next quarterly cycle without me. I test that by having them run one cycle while I observe and advise, before the engagement ends.

Programmes that fail on handover usually fail because the formulas are not written down. The formulas are written into a definitions file, versioned in the repository, with a test that runs on the last four cycles and recomputes the published numbers. If the recompute disagrees, the definitions file has changed or the data has, and either way the next owner sees it before the next cycle.

The healthcare data quality programme is a natural partner for this work, because a population health registry inherits the completeness of everything upstream. When the two run together, the registry owner is the first to know a source has degraded, and the fix usually lands before the cohort dissolves. Separate they tend to discover the same issue at the same review, several weeks later than necessary.

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 — population health analytics — the scope, the failure modes it prevents and how the work is measured. It sits in the Security and Compliance 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