Understanding CMS hierarchical condition categories for risk adjustment

Each calendar year, the CMS risk adjustment model resets: diagnoses and codes do not carry forward automatically. For an ambulatory practice, that annual cycle creates both the opportunity to recapture supported conditions and the risk of a partial picture when documentation is missed.

HCCs and RAF scores

The HCC model groups related ICD-10 diagnoses into hierarchical condition categories. These categories organize documented disease severity for risk adjustment rather than treating every diagnosis as an independent measure.

A RAF score, or risk adjustment factor, combines beneficiary demographics with HCC categories supported by diagnoses from the prior year. The resulting score helps represent expected resource needs for a Medicare Advantage member.

HCCs are hierarchical: when related conditions fall in one hierarchy, only the highest-weighted supported condition counts.

Annual recapture requirement

Codes do not carry forward from one year to the next. Each condition must be re-documented in the current year's chart and supported by the clinician's assessment and plan.

The annual wellness visit and chronic disease visits are primary opportunities to review active conditions, update specificity, and connect diagnoses to the current assessment/plan.

A problem list may retain a condition while the current assessment/plan leaves it unaddressed. In a 4-provider internal medicine practice where 30% of the panel is Medicare Advantage, that disconnect across several hundred patients can produce only a partial RAF score, even when CKD stage 4, COPD, CHF with preserved ejection fraction, or other conditions remain clinically relevant.

V28 model update

V28 is the new model, phased in beginning 2024. It replaced V24, the prior model in use since 2020, with revised categories and weights.

V28 makes diagnostic detail more consequential for some conditions. For depression, F33.1 (major depressive disorder, recurrent, moderate) provides more specific information than F32.9 (unspecified depressive disorder), while unspecified codes can lose value under V28 when the chart supports a defined subtype.

Under V28, ICD-10 specificity and documentation detail must move together.

Common under-coded conditions

Several chronic categories are under-coded in Medicare Advantage charts.

For diabetes, HCC 19 (type 2 diabetes, no complications) differs from HCC 18 (type 2 diabetes with peripheral neuropathy). The complication must be explicitly linked to diabetes in the documentation.

CKD staging matters: N18.4 (CKD stage 4), N18.31 (CKD stage 3a), and N18.32 (CKD stage 3b) carry different weights. The stage belongs in the assessment.

Heart failure specificity includes systolic, diastolic, or combined disease, as well as acuity. Those distinctions drive the supported ICD-10 code and HCC category.

COPD coding depends on current severity and exacerbation status. J44.1 (COPD with acute exacerbation) and J44.0 (COPD without exacerbation) require documentation of the current state.

Upcoding boundary

Accurate specificity records the clinical reality the clinician has evaluated and documented. Upcoding occurs when a code claims greater severity, a complication, or acuity that the chart does not support.

The appropriate boundary is evidence: document and code the condition at the highest supported level, but do not manufacture severity to increase a RAF score. A more specific code is appropriate only when the completed chart establishes that specificity.

Automated HCC capture

Automated systems can identify where note content supports a more specific ICD-10 code than the one applied. For example, a chart coded E11.9 (type 2 diabetes without complications) may support E11.40 (type 2 diabetes with diabetic neuropathy, unspecified) when the assessment documents the linkage and gabapentin appears in the neuropathy context.

Arintra reads the complete chart and proposes the most specific code that the documentation supports. Arintra assigns CPT and ICD-10 codes from completed clinical charts, without adding clinical content.

The value is catching routine gaps across hundreds of encounters each year, where familiar nonspecific codes are often selected by default.