The Era of 'Coding Intensity' Is Ending. The Era of 'Defensible Accuracy' Is Here.

By Chetan Parikh, Founder and CEO, RAAPID

For more than a decade, Medicare Advantage plans operated under one unspoken rule: more codes meant more revenue. The strategy was simple. Cast a wide net. Capture every possible HCC. Let the volume carry the bottom line.

That strategy is now a liability.

The Centers for Medicare & Medicaid Services has made its position clear. The era of "coding intensity," where aggressive capture trumped documentation quality, is over. What CMS demands now is something fundamentally different: defensible coding supported by clinical evidence.

This isn't a temporary adjustment. It's a permanent recalibration of the compliance and financial reality for every Medicare Advantage organization (MAO).

Enforcement Is Accelerating

In January, the Department of Justice secured its largest Medicare Advantage False Claims Act settlement ever: $556 million from Kaiser Permanente affiliates. The allegations centered on a multi-year effort to retroactively add a diagnosis to medical records, sometimes more than a year after the patient encounter. DOJ alleged the health plan used data-mining tools to identify diagnoses that hadn't been submitted to CMS, then systematically pushed physicians to add them.

The same month, a Congressional investigation concluded that at certain MAOs, the process of identifying and coding conditions has become "a business in itself," separate from the clinical care being delivered. That was never the intent of risk adjustment.

These aren't isolated events. DOJ has relaunched its joint False Claims Act Working Group with HHS, with Medicare Advantage enforcement as a top priority.

When Diagnosis Capture Becomes Disconnected From Care

The problem isn't risk adjustment itself. Accurate risk adjustment is essential. Plans need fair compensation for the true disease burden of their members, and CMS needs accurate data to manage the program effectively.

The problem is when the objective becomes volume rather than accuracy. When health plans build entire infrastructures dedicated to finding codes rather than validating them. When providers face pressure to add diagnoses without corresponding changes to care plans. When tools are deployed to maximize capture rather than ensure defensible coding.

The investigation documented exactly this pattern: in-home assessment teams, chart review operations, and pay-for-coding programs designed to identify "untapped" diagnoses. Proprietary diagnostic criteria that differed from clinical standards. Automated systems are built to surface "risk score opportunities" rather than verify documentation.

The V28 Model Was a Direct Response

CMS didn't remove over 2,000 diagnosis codes from the V28 risk adjustment model by accident. Codes for conditions like secondary immunodeficiency, secondary hyperaldosteronism, protein-calorie malnutrition, and asymptomatic peripheral artery disease were excluded specifically because of how inconsistently they were being applied.

Congressional investigators saw this firsthand. Some plans had trained providers to diagnose secondary immunodeficiency in any patient taking steroids, regardless of clinical presentation. Others used probability-based approaches to diagnose heart failure based on symptoms alone, without confirmatory testing.

When MAOs create their own diagnostic standards that diverge from clinical guidelines, it becomes impossible to determine whether the coding is appropriate.

The Real Problem With Capture-First Tools

The investigation revealed automation embedded throughout diagnosis capture workflows: identifying which enrollees had the greatest revenue potential, determining which charts to prioritize, and auto-populating diagnoses based on logic rules triggered by checkboxes.

These are tools optimized for capture, not defensible coding.

When systems are designed to find every possible code, they will find every possible code, whether the clinical documentation truly supports it or not. What's needed instead is Neuro-Symbolic AI that can reason through clinical documentation and provide transparent justification.

At RAAPID, we built our Novel clinical platform for retrospective risk adjustment, powered by Neuro-Symbolic AI, with a fundamentally different objective: validation and justification, not inflation. Our Neuro-Symbolic AI combines deep learning with clinical reasoning and knowledge graphs. It doesn't just identify codes. It maps every suggested HCC to specific MEAT evidence in the clinical note, then surfaces that evidence for your coders to confirm, making every code defensible.

The difference matters when an auditor pulls the chart. Plans using opaque, capture-first tools will struggle to explain why a code was submitted. Plans using Neuro-Symbolic AI to validate documentation will have the receipt ready.

A New Standard for Risk Adjustment

MAOs that will thrive in this new environment aren't the ones that capture the most code. They're the ones that capture the right codes and can prove it.

This requires a different approach. Not "how many codes can we find?" but "which codes can we defend?" The workflow must shift from reactive chart-chasing to proactive validation.

The regulatory environment will only become more demanding. The plans that position themselves for defensible coding today will be the ones that protect their revenue and their reputation tomorrow.

The era of coding intensity is ending. The era of defensible accuracy powered by Neuro-Symbolic AI is here.