CombineHealth Demonstrates New Benchmarks for Production-Scale Autonomous Medical Coding

CombineHealth Demonstrates New Benchmarks for Production-Scale Autonomous Medical Coding

Today's announcement reflects CombineHealth's vision for the next generation of autonomous medical coding. By reducing coding-related denials by 75% and increasing captured revenue by 4% while maintaining 98.4% coding accuracy, CombineHealth demonstrates that production AI can move beyond code assignment to improve reimbursement outcomes through integrated Clinical Documentation Integrity (CDI) and continuous payer-outcome learning.

"Healthcare organizations are no longer asking whether AI can assign the correct codes," says Sourabh Agrawal, CEO at CombineHealth. "They're asking whether it can operate reliably at production scale, adapt to changing payer behavior, strengthen documentation quality, and ultimately improve financial outcomes."

Production Evidence from a Hospital Deployment

Among the production deployments contributing to these benchmarks was a 400-bed Midwest hospital, where CombineHealth’s autonomous coding solution, Amy, identified a previously unquantified source of revenue leakage.

The hospital discovered that undercoding was preventing it from fully capturing reimbursement for care that had already been delivered.

Within the first three months, Amy reduced coding-related denials by 75% while improving captured revenue by 4% by intelligently orchestrating payer-specific coding policies, identifying undercoded encounters, and surfacing higher-specificity documentation opportunities that had previously gone undetected through manual review.

“The value wasn’t just better coding. We were surprised by how Amy identified 5× more CDI opportunities than our traditional workflow!” - HIM Director, 400-bed Midwest Hospital

Looking Beyond Coding Accuracy

Healthcare organizations have traditionally evaluated coding performance through a narrow lens: Were the correct codes assigned, were coding guidelines followed, and would the claim pass a coding audit?

But coding accuracy is only one part of whether a claim receives appropriate reimbursement.

CMS’s latest national review of Medicare Fee-for-Service payments illustrates the distinction. In FY2025, CMS estimated $28.83 billion in improper payments. Incorrect coding represented 11.1% of that amount, while insufficient or missing documentation represented a combined 65% and medical-necessity errors represented another 15.3%. 

Put differently, nearly nine out of ten improper-payment dollars in the review were attributed to something other than incorrect coding.

The implication is significant: a claim can be coded correctly and still be underpaid, denied, or found unsupported because the documentation does not substantiate the service, the care does not satisfy payer requirements, or another payment rule has not been met.

These findings reinforce CombineHealth’s view that the next generation of autonomous medical coding cannot be measured by coding accuracy alone. It must also be evaluated on whether it completely captures the care delivered, strengthens the supporting documentation, applies payer-specific requirements, and improves actual reimbursement outcomes.

Rather than asking only, “Did we assign the correct codes?” Amy reframes the question: “Did the codes completely and defensibly capture the care delivered, and is the claim positioned to receive appropriate payment without an avoidable denial?”

Amy was designed around that broader objective.

Before a claim is submitted, Amy identifies documentation gaps and inconsistencies, surfaces supported opportunities for more complete care capture, and validates whether the documentation and coding align with applicable coding guidelines and payer requirements.

How Amy Delivers These Outcomes

CombineHealth attributes these production outcomes to Amy's payer-aware coding architecture. Every billable CPT service is evaluated alongside its supporting ICD diagnoses to ensure appropriate clinical justification. Amy also validates modifier usage, verifies the appropriate provider to bill, and applies payer-specific coding and billing requirements before a claim is generated.

Amy’s architecture combines two capabilities: applying payer- and organization-specific requirements before claim generation and learning from downstream payer outcomes over time.

Applying Payer and Organization-Specific Rules

To code accurately, Amy incorporates and consistently applies current payer guidelines, including:

• Local Coverage Determinations (LCDs)
• National Coverage Determinations (NCDs)
• Specialty-specific coding guidelines
• Payer-specific billing rules

Beyond national standards, the platform is highly configurable to customer-specific coding preferences and organizational workflows.

Learning From Payer Outcomes

Amy is also payer-outcome-aware learning from real payer outcomes rather than relying solely on static coding guidelines. Claim outcomes, such as reimbursements, denials, underpayments, and payer edits, become feedback that helps Amy improve future coding decisions.

The Bigger Picture

CombineHealth believes autonomous medical coding is entering a new stage of maturity. Accuracy remains foundational, but production performance must also be judged by automation coverage, documentation quality, coding-related denials, supported revenue capture, and downstream payment outcomes.

The company’s latest coding benchmarks demonstrate that payer-aware autonomous coding is no longer just a future concept. It’s already being used in real production environments. As the system learns from real-world payer outcomes over time, better coding leads not only to cleaner claims but also to meaningful improvements across the revenue cycle.

About CombineHealth

CombineHealth is building the next generation of AI employees for healthcare revenue cycle operations. Its Revenue Cycle Intelligence platform helps hospitals, physician groups, and RCM teams automate medical coding, claims, denials, appeals, and accounts receivable workflows.

Its flagship autonomous medical coding solution, Amy, helps healthcare organizations automate more coding with confidence by combining advanced AI models, coding guidelines, specialty-specific rules, human review feedback, and real payer outcomes. Amy is designed not only to assign accurate codes, but also to improve documentation quality, reduce coding-related denials, identify missed reimbursement opportunities, and support cleaner downstream revenue cycle outcomes.