The 2020 Playbook Hit a Wall
Risk adjustment programs designed before 2024 were built around a set of assumptions that no longer hold. The payment model rewarded volume. Enforcement was sporadic and penalties were manageable. Add-only retrospective review was standard practice. The AI available at the time was limited but sufficient for code identification. Plans optimized for these conditions, and their technology reflected it.
Every one of those assumptions changed. CMS-HCC V28 restructured the payment model to penalize coding intensity (100% effective January 1, 2026). The DOJ collected over $670 million from Kaiser and Aetna in settlements targeting inflated risk adjustment submissions. OIG’s February 2026 guidance named add-only chart reviews as a specifically identified high-risk practice. And AI capabilities evolved to the point where explainable, evidence-based validation is now achievable at scale.
Programs built for the old conditions don’t just underperform in the new environment. They create active regulatory exposure. The practices they were designed to execute, high-volume add-only coding without structured documentation validation, are now the exact practices that generate enforcement attention.
What Changed and Why Retrofitting Doesn’t Work
Some vendors have tried to adapt older systems by layering compliance features on top of revenue-optimized architectures. Add a MEAT validation check here. Include a deletion flag there. Generate an evidence report as a secondary output. The problem is that systems designed around volume as the core function treat validation as a peripheral task. The architecture optimizes for throughput, and compliance features compete with that optimization rather than reinforcing it.
A system built from the ground up for defensibility works differently. Documentation validation isn’t a post-processing step. It’s the primary function that every other feature serves. The AI doesn’t identify codes and then check if they’re supported. It evaluates evidence first and recommends codes only where the documentation meets the standard. That’s a fundamentally different design philosophy, and it produces fundamentally different audit outcomes.
The distinction shows up under scrutiny. When CMS audits a code, they follow the evidence trail backward from submission to documentation. A system that built the trail as the primary output produces clean, auditable records. A system that generated the trail as an afterthought produces records that look like what they are: compliance bolted onto a revenue engine.
The Criteria That Matter Now
Plans evaluating technology in 2026 should test for four capabilities. First, native MEAT validation: does the system evaluate clinical evidence against Monitoring, Evaluation, Assessment, and Treatment criteria as part of its core workflow, not as a secondary report? Second, two-way coding: does it identify both codes to add and codes to remove? Third, explainable AI: can the system show its reasoning for every recommendation in a format that auditors can follow? Fourth, audit simulation: does it score defensibility before submission and flag weak documentation proactively?
Any system that fails on even one of these tests was designed for an environment that ended in 2025. The regulatory changes aren’t cyclical. They’re structural. V28 isn’t a transitional model. The DOJ settlements set permanent precedent. OIG guidance doesn’t expire.
Replacing, Not Patching
Plans still running technology built for the pre-2024 environment face a choice: patch or replace. Patching means adding compliance layers to a system that was architected for revenue optimization. Replacing means selecting a Risk Adjustment Solution designed from the start for defensibility, two-way coding, evidence validation, and audit readiness. The patching approach saves procurement cycles. The replacement approach saves settlement checks. The math favors replacing, and the enforcement trajectory makes the case stronger every quarter.
Also Read-How South Carolina Businesses Can Navigate Group Health Insurance for Remote Teams
