IAMAI V28 Deterministic Zero-PHI Value Ledger

What it solves

Every risk adjustment tool on the market predicts. It scores a chart, guesses at the likely V28 codes and hands back a probability. That leaves two open wounds at once: revenue you legitimately earned but never captured, and codes that cannot be defended when a CMS RADV audit asks the model to explain itself. And because the model is probabilistic, the same chart run twice can answer differently.

IAMAI V28 does not predict. It applies deterministic rules to the record and reveals the complete, unalterable picture — the same chart returns the same result today, next quarter and five years from now. You can feed it raw clinician notes or charts that have already been audited and signed off, and it acts as a final sovereign filter that closes whatever gap the earlier pass glossed over. Nothing is inferred, nothing is hallucinated, and no PHI leaves your control.

The problem nobody wants to name

Risk adjustment has quietly become an act of faith. A chart goes into a model, a score comes out, and somewhere inside that transaction the organisation accepts that a machine has made a judgement it cannot fully reconstruct. When the numbers look good, nobody asks. When CMS asks, everybody does.

The uncomfortable part is that probabilistic tools fail in both directions at once. They under-capture, quietly leaving legitimate V28 revenue uncollected because the model was not confident enough to surface it. And they over-reach, attaching codes on inference that will not survive a RADV review. The same technology creates the revenue leak and the audit exposure, and it does so invisibly.

A different premise entirely

IAMAI V28 begins from the opposite premise: that a risk adjustment report should be a mathematical fact, not an opinion with a confidence interval attached. The engine applies deterministic rules to what is actually documented in the record. Where the documentation supports a V28 condition, the condition is captured and the rule that captured it is recorded alongside it. Where it does not, nothing is invented.

This is what makes the output sovereign. It is not a suggestion that a coder must adjudicate, and it is not a black box that a compliance officer must defend on trust. It is a ledger — reproducible, explainable and permanent. Run the same chart in five years and the answer will be identical, because there is no model drift, no retraining, and no randomness in the path from documentation to code.

It runs after everyone else

The most valuable place to sit in this workflow is last. Most organisations have already invested in a coding vendor or an AI abstraction layer, and replacing it is a twelve-month political exercise nobody wants. IAMAI V28 is not asking for that seat. It is the final filter that runs on the output of whatever you already use.

You send charts that have already been audited, verified and signed. The engine treats them as unfinished, because probabilistically they are. It closes the residual gap — the conditions that were documented, were legitimately payable, and were simply not confident enough for the earlier model to claim. That gap is the money that is currently walking out of the building every reporting cycle.

Raw notes work the same way. Inject an unprocessed clinician narrative and the ledger returns a fully loaded, V28-coded, zero-PHI report without a human abstraction step standing in the middle of it.

Why the price is one dollar

A second opinion is only useful if it is cheap enough to take on every report rather than on the ones you already suspect. So verification costs one dollar per report, flat, with no platform fee, no minimum volume and no annual commitment in front of it.

If the engine finds nothing, you have spent a dollar to produce documentary proof that your existing process was clean — which has its own value the next time an auditor asks how you validate your vendor. If it finds recovered yield, the fee is ten percent of the newly recovered money only. Never on what you had already captured. Never in advance.

The order of that sentence is the whole commercial philosophy. We save first. We are paid out of the saving, afterwards, and only if there is one.

Zero PHI, by construction

The ledger is engineered so that protected health information is not required to leave your environment for the deterministic pass to run, and is not carried into the output artefact. Compliance teams get a report they can circulate, archive and hand to an auditor without a redaction exercise attached to it.

That is not a privacy feature bolted to the side of a product. It is a consequence of the same design decision that removes the guessing: if the engine only works from documented fact and rule, it does not need to hold the patient in order to hold the answer.

What a healthy ending looks like

The goal of this product is not to sell software into healthcare. It is to let a reporting cycle actually close. A final report where every payable code is captured, every captured code is supported, and the whole thing is deterministic — unchanging, reproducible and defensible for as long as it needs to be retained.

That is a healthy way to end an audit. Everything else is a position you are holding until someone checks.

What it does

100% deterministic calculation

Statistical probability is converted into audit-proof mathematical fact. Every code carries the rule that produced it, so the output can be re-derived by anyone, at any time, with identical results.

Auditing the audited

Designed to run after your existing vendor, not instead of it. Feed in previously audited and verified charts and the engine surfaces the residual gap the probabilistic pass could not close.

Raw note ingestion

Inject unstructured clinician notes and receive a fully loaded, V28-coded, zero-PHI report — no pre-cleaning, no template, no manual abstraction step in between.

Zero-typing selection grid

Clinicians capture structured, MEAT-backed V28 records at the point of care by selection alone. No free-text dictation to be re-interpreted downstream, so the documentation is defensible the moment it is created.

Full studio reports

Compliance teams generate complete, CMS-shaped auditable reports on demand, with the deterministic trail attached to every captured condition.

RADV-ready evidence trail

Because nothing is probabilistic, there is no model behaviour to explain to an auditor. Each line is traceable to documentation and to the rule that qualified it.

What you receive

The V28 ledger

One line per captured condition, each carrying the documentation reference and the deterministic rule that qualified it. Re-derivable by anyone, at any time, to the same result.

The gap statement

What your previous pass captured, what this pass captured, and the difference between them stated as recovered yield — the number the ten percent is calculated from.

The audit trail

Because no step is probabilistic, there is no model behaviour to explain. The trail an auditor receives is a chain of documentation and rule, not an explanation of a prediction.

A zero-PHI artefact

A report compliance can circulate, archive and hand over without running a redaction exercise first.

What it costs

The commercial model is deliberately inverted. You do not buy a licence, a seat count or an annual platform commitment before you have seen a single result. You pay one dollar to verify a report.

For that dollar you receive a full deterministic second opinion on a chart you have already paid someone else to audit. If our rules find nothing your previous pass missed, the story ends there and it has cost you a dollar to prove your existing process is sound — which is itself a defensible audit artefact.

If the engine does find V28 yield that was left on the table, you pay ten percent — strictly on the newly recovered money, never on revenue you had already captured. We are paid out of what we find for you, after we find it. We save first and charge second.