How CMS Picks Who Gets Examined
CMS doesn’t select RADV samples randomly. The agency uses a stratified sampling methodology that targets specific diagnosis categories based on their financial impact and historical error rates. Understanding how this selection works reveals which parts of a plan’s submissions face the highest probability of scrutiny and where pre-audit investment produces the greatest return.
CMS groups HCCs into risk strata based on their payment impact. High-impact categories (conditions that generate the largest per-member payment adjustments) are oversampled relative to their prevalence. This means acute conditions like stroke, myocardial infarction, and certain cancers appear in RADV samples at rates far higher than their share of the plan’s total submissions. Plans that concentrated coding effort in these high-value categories concentrated their audit exposure in the same categories.
The Stratification Logic Plans Misunderstand
Many plans assume RADV sampling is proportional: if 10% of their submitted codes are in diabetes categories, 10% of the audit sample will target diabetes. That’s not how stratified sampling works. CMS overweights categories where unsupported codes generate the largest overpayments. A plan with 5% of its codes in acute stroke categories may find 15% or 20% of its RADV sample targeting those codes specifically.
This creates a mathematical asymmetry. The categories a plan codes most aggressively for revenue purposes are the categories that appear most frequently in audit samples. The revenue-maximizing strategy and the audit-exposure-maximizing strategy point in the same direction. Plans that chased high-value codes disproportionately created audit profiles that test those exact codes disproportionately.
OIG’s March 2026 audits confirmed this pattern. Acute stroke and myocardial infarction categories, both high-value and high-audit-priority, showed 100% error rates across multiple organizations. The categories that generated the most revenue per code were also the categories with the worst documentation quality.
Using Stratification Knowledge Defensively
Plans can use CMS’s stratification logic to prioritize internal quality improvement. If CMS oversamples high-impact categories, those categories deserve the most rigorous internal validation. Run targeted MEAT assessments against your submitted codes in the diagnosis groups CMS is known to oversample. If your internal error rate in stroke, MI, or cancer categories exceeds 20%, those codes will likely fail at similar or higher rates in the actual audit.
This targeted approach produces more audit-relevant quality improvement per dollar than uniform quality reviews across all categories. A plan that reduces its error rate in the top 10 highest-impact categories from 40% to 10% will see a larger improvement in RADV outcomes than a plan that reduces its error rate across all categories from 25% to 20%, because the audit sample is weighted toward those high-impact categories.
The Selection Isn’t Random. Your Preparation Shouldn’t Be Either.
Plans preparing for radv audits should align their internal quality investment with CMS’s sampling methodology. The categories CMS oversamples are public knowledge based on payment impact data and OIG audit focus areas. Plans that apply their strongest validation resources to those categories are defending the exact codes auditors will examine most frequently. Plans that spread validation effort uniformly are underinvesting in the categories that carry the highest audit probability.