The Split That Explains the Error Rates
OIG’s March 2026 audits found error rates between 81% and 91% across three MA organizations. But that aggregate number obscures a split that matters for program design. Chronic condition codes (diabetes, CKD, COPD, heart failure) fail at meaningfully lower rates than acute or episodic condition codes (stroke, MI, cancer). The audits where acute categories hit 100% error rates were dragging the overall numbers up. Chronic categories, while still showing significant failures, didn’t collapse at the same rate.
This split makes clinical sense. Chronic conditions generate continuous documentation. A diabetic patient visits their physician quarterly. Each visit produces notes referencing A1C levels, medication adjustments, and treatment plans. A coder reviewing that patient’s chart for a retrospective HCC has multiple encounters to draw from, each containing MEAT evidence (Monitoring, Evaluation, Assessment, Treatment) because the condition is being actively managed across visits.
Acute conditions tell a different story. A stroke happens once. The acute documentation is thorough: emergency department notes, neurology consults, imaging reports, discharge summaries. But after the acute phase, documentation thins out. The patient may or may not have follow-up neurology visits. The condition may appear on problem lists indefinitely without any current management evidence. A coder three years later sees “history of stroke” and submits it, but the chart has nothing showing the condition is being actively addressed.
Why Programs Fail on Acute Conditions Specifically
Retrospective programs fail on acute conditions because the coding process treats all HCCs the same. The coder identifies the diagnosis. The system checks the HCC mapping. If the diagnosis appears in the chart and maps to an HCC, it gets recommended for submission. No separate validation layer asks: is this a carried-forward acute condition without current management documentation?
The financial incentive compounds the problem. Under V24, acute conditions like stroke and MI carried high RAF coefficients. Retrospective programs prioritized them on chase lists because each successful code generated substantial revenue. The programs were optimized to find these codes aggressively without corresponding optimization to validate whether the documentation could defend them.
V28 reduced some of these coefficients, but the validation gap remains. Programs still processing acute conditions through the same workflow as chronic conditions will keep producing the concentrated failures OIG documented, even if the per-code revenue has dropped.
Designing Category-Aware Validation
The fix is building diagnosis-category awareness into the coding workflow. When the system identifies a chronic condition, it applies standard MEAT validation: does the current documentation show active management? For chronic conditions with continuous care patterns, this evidence is typically present across multiple encounters.
When the system identifies a single-occurrence acute condition, it applies enhanced validation. It searches for current specialist follow-up, recent labs or imaging related to the condition, active medication management, and documented care plans. If the condition appears only in the problem list without any of these current management indicators, the system flags it as high-risk rather than recommending submission.
This tiered approach mirrors how OIG actually audits. Auditors apply heightened scrutiny to high-risk categories. Programs that apply the same heightened scrutiny internally catch the failures before submission rather than discovering them in audit findings.
Where the Biggest Quality Gains Are
The fastest path to lower RADV error rates isn’t improving chronic condition coding, which already has stronger documentation patterns. It’s fixing how programs handle acute conditions, where documentation decay is fastest and audit failure rates are highest. Retrospective hcc coding programs that build category-specific validation, applying enhanced evidence thresholds to acute and single-occurrence diagnoses, will see disproportionate quality improvement in the categories that currently produce the worst audit results. The chronic-acute split isn’t just an analytical observation. It’s a program design specification.
