Chapter 36 — Key Takeaways

The design

Risk-adjusted payment pays for a member, not for a visit. A plan receives a fixed amount per member per month, adjusted by a score built from the diagnoses documented and reported for that member last year. No claim pays for a diagnosis; the diagnoses set the pot before any claim exists.

The failure mode that makes this chapter hard: an incomplete record produces NO artifact. No denial. No variance. No work queue entry. No report. The claim pays exactly as expected.

Why risk adjustment exists

A fixed per-member payment pays an organization to avoid sick people. Adjusting the payment by expected cost removes the incentive. Three properties of the models:

• They predict cost, not sickness — fit statistically against spending
• They are relative — 1.0 is an average member, normalized
• Evidence from last year, payment for next year — which generates the annual reset

Under-reporting and over-reporting are the same error with opposite signs. Downcoding is not the safe option (Ch. 5 §5.8): it describes a population as healthier than it is, distorts the quality data computed from the same codes, and leaves the practice unable to answer questions about its own patients.

HCCs and the RAF

Most ICD-10-CM codes map to nothing. A risk model selects the conditions that predict cost. Where it does distinguish severities, it does so through the code's specificity.

More than one model, and they legitimately disagree: CMS-HCC (Medicare Advantage) · CMS-HCC ESRD · RxHCC (Part D) · HHS-HCC (ACA individual and small group) · state Medicaid models.

   RAF = DEMOGRAPHIC COMPONENT      (age/sex, dual status, institutional,
                                     originally disabled - the chart cannot
                                     touch this)
       + DISEASE COEFFICIENTS       (one per surviving condition category
                                     - THIS is what the chart controls)
       + INTERACTIONS               (modeled combinations; never codable)
       then, at the PLAN level:
       x normalization factor
       x statutory coding intensity adjustment (Medicare Advantage)

   PAYMENT = RAF x benchmark, per member per month

Worked example [constructed — verify current coefficients and the model version at CMS]: 0.346 + 0.328 + 0.302 + 0.127 + 0.331 + 0.121 = 1.555; × \$1,000.00 PMPM = \$1,555.00/month = \$18,660.00/year.

Hierarchies, trumping, interactions

The asymmetry is deliberate. Over-reporting gains nothing; under-reporting costs everything. That converts "code accurately" from a moral instruction into the obviously correct strategy.

The annual reset

Every condition returns to zero on January 1. Collection year N pays every month of year N+1. Nothing carries forward — not the amputation, not the transplant, not the paraplegia.

Recaptures: a documented, face-to-face encounter with a diagnosing provider who assessed the condition. Does not recapture: a problem list · a lab result · a refill · a copied-forward note · a claim for a service like a venipuncture · an unsigned note.

The vehicle: the annual comprehensive visit, often anchored on G0439 (subsequent annual wellness visit) — but a preventive visit is not automatically a problem-oriented one.

And the reset runs both ways: a resolved condition must not be carried forward. Chapter 10's Case Study 2 — the patient with cancer on paper for five consecutive years — is the warning.

Chronic ≠ weighted. M17.11 is chronic, permanent, disabling, and not among the categories the model weights. The model selects for cost prediction, not for importance.

MEAT

Monitor · Evaluate · Assess/Address · Treat — any ONE suffices.

MEAT is a mnemonic, not a regulation. The binding authority is Section IV of the Official Guidelines plus the documentation standards and the face-to-face requirement. (TAMPER is the common alternative and encodes the same rules.)

The six failure modes:

1. The problem list is not documentation
2. The history is not the assessment (Account 10-4471's gap 1: the conservative-therapy statement sits in the HPI)
3. "Stable" means at treatment goal (Ch. 15 §15.5) — fully reportable; and a stable patient is not a low-risk member
4. A medication list is not treatment documentation
5. Copy-forward is a liability, not six years of MEAT
6. Specificity is documentation, and the coder may not supply it by inference (Ch. 4 §4.7)

Chart review — the compliance line

The bright line: recognizing that a chart does not support a condition is the skill. Making a chart appear to support one is not, and is not taught here.

How it is caught: data analysis with no chart at all (a diagnosis appearing from one source and nowhere else) · RADV — CMS's January 2023 final rule extrapolates from payment year 2018 with no fee-for-service adjuster · extrapolation (Ch. 37 §37.6) · the OIG Work Plan and audit reports · the False Claims Act with a relator (Ch. 5 §5.3).

Discovering an unsupported submission starts the sixty-day clock (Ch. 31 §31.9).

The number that makes a review program credible is its DELETION COUNT.

Quality measures

CPT Category II codes report measures and are never paid. 0001F — heart failure assessed (process). 3044F — most recent A1c below 7.0% (outcome). Exclusion modifiers 1P (medical) / 2P (patient) / 3P (system) / 8P (not otherwise specified) record why a standard was legitimately not met.

MIPS (MACRA 2015): four categories — quality, cost, improvement activities, promoting interoperability — one final score against a threshold, a Part B adjustment two years later, a low-volume exemption, and MIPS Value Pathways. Every number in it changes annually.

Value-based contracts

Shared savings = a share of (benchmark − actual), gated by quality. Worked [constructed]: 5,000 × \$12,000.00 = \$60,000,000.00 benchmark; actual \$57,600,000.00; savings \$2,400,000.00 = 4.0%; minimum savings rate 2.0% met; × 50% sharing × 90% quality = \$1,080,000.00.

The four terms that decide it: attribution (which patients are yours, prospective or retrospective) · the benchmark (against what, and is it risk-adjusted) · the quality gate · the data feed. A payment model a practice cannot measure is one it should not sign.

And the interaction people expect is backwards: in a capitated arrangement, complete coding raises revenue. In a shared-savings arrangement it usually raises the benchmark too, which is why programs cap risk score growth. None of it changes the code.

The chapter's answer

Q2 is CLOSED. Under MEAT, four of the five conditions on Account 10-4471's March 14 note pass on all four criteria. Chronic kidney disease, stage 3a fails on all four — one appearance, on a problem list, addressed by nobody. The metformin dose says the clinician thought about the kidneys; the note does not, and no coder may say it for them.

The March 14 code is not wrong. It is incomplete as a description of the patient, and the difference between those two statements is worth money.

Third instance of the book's model finding: Ch. 14 §14.4 (the decision to inject, unwritten) · Ch. 33 §33.10 ("hypoxic" for acute respiratory failure, \$1,867.44) · here. Not wrong. Unwritten.

One member, one year [constructed]: 0.752 − 0.428 = 0.324 → \$324.00 PMPM → **\$3,888.00**. And the money goes to the plan; what the practice receives is a contract term.

Monday morning

You should be able to: explain to a physician why a fixed per-member payment must be risk-adjusted; build a RAF from its components and check it; apply a hierarchy and say why stacking gains nothing; list what does and does not recapture a chronic condition; apply MEAT to an assessment condition by condition and defend the result in writing; ask a chart-review vendor the three questions in §36.8; tell a process measure from an outcome measure and assign the Category II code that reports each; read a value-based contract for attribution, benchmark, quality gate, and data; and explain — without cynicism and without overstating — how a code that was right can leave a patient described incorrectly for a year.