As health plans finalize their 2026 retrospective risk-adjustment strategies, a sobering reality emerges: The political and regulatory environment is shifting toward unprecedented scrutiny. Top health regulators in the Trump administration have pledged to closely examine upcoding practices, signaling that enforcement will remain a priority despite temporary pauses in specific audit methodologies.
Against this backdrop, the industry’s drive to capture every possible HCC isn’t just operationally exhausting, it’s strategically misaligned with where health care compliance is heading.
The New Enforcement Reality
While the Centers for Medicare & Medicaid Services (CMS) hasn’t yet achieved its ambitious goal of expanding to 2,000 medical coders, the agency continues building audit capacity through technology and staffing investments. This expansion signals intent: automated detection systems, pattern-recognition algorithms and data analytics will increasingly flag outliers with or without human reviewers.
The message from regulators is clear: Upcoding scrutiny will intensify. Plans that continue prioritizing volume over documentation quality are building tomorrow’s enforcement targets on today’s submissions.
The Evolution NLP Must Make
Natural language processing (NLP) has served the industry well, revolutionizing how we identify potential diagnoses in clinical documentation. NLP can scan thousands of charts, flag potential HCCs and surface coding opportunities at an unprecedented scale.
But the regulatory environment has evolved faster than technology. Today’s challenge isn’t finding potential codes, it’s proving them. NLP excels at the former but wasn’t designed for the latter.
Consider what happens when NLP identifies a potential diabetes diagnosis. It recognizes clinical terminology and flags the relevant ICD-10 code. What it typically can’t do is evaluate whether documentation satisfies MEAT criteria : monitoring, evaluation, assessment and treatment. While only one of these elements is required to support code selection, NLP struggles to verify that even this minimal threshold is met with sufficient clarity. It can’t determine if the documentation provides the definitive clinical evidence or assess whether that single supporting element would withstand audit scrutiny.
This gap between identification and validation creates the industry’s current crisis: millions of flagged codes requiring manual verification, overwhelming coding teams that must review each suggestion individually.
Why Your Current Technology Needs an Upgrade
The limits of current NLP implementations aren’t failures, they’re design boundaries. First-generation NLP was built to solve the discovery problem. It succeeded. Now we face a validation problem requiring different capabilities.
Modern risk adjustment demands technology that evaluates documentation completeness, traces diagnoses to specific clinical evidence, assesses programmatically whether MEAT criteria are satisfied, provides transparent reasoning for audit defense, and distinguishes between defensible and questionable documentation.
This isn’t about replacing NLP. It’s about augmenting it with explainable artificial intelligence (AI) that bridges the gap between finding and proving.
The Hidden Cost of Status Quo
Organizations continuing with volume-first strategies face mounting hidden costs. When technology identifies 5,000 potential codes requiring manual validation, coding teams become verification factories. Experienced coders spend their expertise on repetitive checking rather than strategic improvement. Every marginally documented piece of code submitted becomes a potential audit exposure tomorrow, creating “compliance debt” that compounds invisibly until an audit triggers repayment with interest.
The Path to Sustainable Excellence
Leading organizations are implementing “strategic restraint,” the discipline to pursue only codes with defensible documentation. This approach requires three fundamental shifts.
First, redefine success. Stop celebrating raw capture rates. Start measuring documentation-quality scores. Track not just codes found, but codes validated.
Second, upgrade your technology stack. Augment existing NLP with explainable AI that validates documentation completeness. Deploy systems providing audit-ready evidence trails for every recommendation.
Third, build quality systems. Create a feedback loop to improve provider documentation at the source. Implement continuous mock audits to identify gaps before submission.
Case Study Results
A large health plan recently transformed its retrospective program using these principles. Instead of processing every NLP-identified code, they implemented AI validation to assess documentation quality before human review.
Results after six months showed dramatic improvement. Chart review time decreased from over 40 minutes to under eight. Documentation completeness improved by 60 percent. Audit-readiness scores increased to more than 98 percent. Team overtime dropped 75 percent. And here’s the kicker: zero adverse audit findings.
Preparing for Enhanced Scrutiny
With regulators signaling increased scrutiny of upcoding, organizations must prepare for enhanced oversight. Every submitted code needs bulletproof supporting documentation. Avoid statistical outliers triggering algorithmic detection. Run your program as if every submission will be reviewed, because increasingly, through automated systems and expanded review capacity, they will be.
The 2026 Decision Point
The future belongs to organizations that understand that, in modern risk adjustment, how you capture matters more than how much. Quality beats quantity. Documentation beats discovery. Sustainability beats speed.
This requires courage to leave marginal codes unclaimed, invest in better technology rather than more reviewers and measure success differently from competitors. But this courage pays dividends. Organizations building quality-first retrospective programs don’t just survive audits, they welcome them.
______________________________________________
RAAPID transforms retrospective risk adjustment through explainable AI that validates documentation completeness beyond traditional NLP. Our Novel Clinical AI Platform ensures compliance with MEAT criteria, providing transparent audit trails for every diagnosis. We help health plans achieve more than 98 percent documentation defensibility while reducing chart review from over 40 minutes to under eight.
