Every leader responsible for the revenue cycle is acutely aware that denial rates have been rising across the healthcare industry. This trend is partly driven by the use of AI-based systems by payers, as well as the increasing intricacy of CMS guidelines. Additionally, most revenue cycle professionals recognize that AI coding tools have been developed with the intention of addressing these challenges.
These tools are indeed showing marked improvements:
- Documentation reviews are being completed more rapidly.
- Code suggestions provided by the systems are growing more precise.
- Audit flags are now able to identify discrepancies and issues that human coders might overlook.
Despite these advancements, denials continue to occur—even on claims that have been coded accurately. This is not a matter of random chance. Instead, it highlights a fundamental, structural issue for which traditional coding tools were never designed as a solution.
The Actual Capabilities of Coding Tools:
The majority of AI coding solutions are engineered to tackle a specific challenge: they analyze clinical documentation and suggest appropriate codes. They are adept at flagging inconsistencies or gaps between what the provider has documented and what coding requirements dictate. Some further include compliance checks that are cross-referenced with established audit patterns.
This functionality is genuinely helpful. A coder working with a well-trained AI model can typically generate correct codes more efficiently and accurately than a coder without such assistance.
However, it is crucial to recognize that accurate coding and a payable claim are two distinct outcomes. The space between these two endpoints is precisely where a substantial amount of avoidable denial volume—especially in high-complexity specialties—originates.
Identifying the Core Issue:
For instance, consider a typical wound care encounter. A provider may:
- Perform sharp selective debridement
- Apply a skin substitute
- Use chemical cauterization to address a granulation tissue complication
A coder, upon reviewing the clinical note, may assign these codes:
- 97597
- 15271
- 17250
All three codes accurately reflect the documented procedures. Nevertheless, the claim will still face denial. The reason is that CPT 17250 is bundled with 97597 under the Correct Coding Initiative (CCI) edits. Submitting both codes for the same wound in a single encounter constitutes a bundling violation. While the codes are correct, the way the claim was constructed is not.
This is just one scenario from a single specialty. When you expand this issue across:
- Modifier application requirements
- Payer-specific coverage conditions
- LCD (Local Coverage Determination) and NCD (National Coverage Determination) medical necessity rules
- Surface area measurement logic
- Accurate assignment of HCPCS product codes
- Authorization matching
—you begin to grasp the extensive scope of the real problem. None of these are strictly coding problems, yet each directly affects whether a claim gets paid.
Why Coding Tools Fall Short
AI coding tools are purpose-built for a particular workflow: clinical notes are ingested, code suggestions are generated, and the model is trained on a mix of documentation patterns, coding guidelines, and audit logic.
What these tools do not account for is the inter-relationship between codes once assigned:
- Can two correct codes be billed together for the same encounter?
- Is the modifier required by one payer different from the one another payer accepts?
- Does the diagnosis tied to a procedure align with an active LCD in the claim’s specific jurisdiction?
This type of logic is not embedded within the clinical documentation or coding guidelines. It exists within payer policies, coverage determinations, and proprietary claim-level edit libraries—all realms beyond the coding workflow. Therefore, a tool designed merely to read notes and suggest codes cannot enforce those claim-level rules.
Critical Steps Before Submission:
Claim payability is not determined at the coding stage, but during claim construction. This is the point in the revenue cycle where all prior data and codes are assembled into a structured claim and validated against complex adjudication rules.
Effective validation must:
- Check for bundling edits tailored to the specific procedures performed
- Match required modifiers to the payer’s unique requirements, not just general CPT guidance
- Confirm diagnosis codes for each procedure are on the covered lists in the relevant LCD
- Ensure every product billed via HCPCS code has the correct unit of measure attached
Specialties such as wound care, pain management, and DME billing can involve a single encounter with six to eight interrelated codes. The outcome of claim adjudication depends on the relationships between these codes, not on each code individually.
The Resulting Structural Gap:
Healthcare organizations are making considerable investments in AI coding technologies. These tools are legitimate, are evolving, and deliver real returns in documentation efficiency. However, what remains largely unaddressed is the post-coding layer, where the majority of preventable denials actually arise: the claim construction and payer intelligence layer. This is where a correctly coded claim is transformed into a transaction that either passes adjudication or lands in a denial work queue.
This gap is invisible in reports measuring coding accuracy. Instead, it is revealed by metrics such as first-pass resolution rates, denial reason codes, and the amount of rework Medical billing teams must do even on claims that were coded correctly.
The industry, as it stands, is measuring the wrong things and mistakenly believing the problem is resolved.
The Future for Revenue Cycle Tool
Solving this challenge requires more than just enhanced coding models. It demands claim-level intelligence that can predict what a payer will do with a particular set of codes before submission. This means:
- Specialty-specific logic developed with a deep understanding of the entire revenue cycle.
- Payer-aware validation applied at the claim level (not just the code level).
- The ability to catch bundling violations, missing modifiers, and LCD mismatches prior to claim submission—thus preventing denials rather than managing them after they occur.
Providers who implement these capabilities first will not simply reduce denials—they will eliminate entire categories of rework that cannot be fully mitigated by post-denial management efforts.
The next generation of revenue cycle tools will not be judged by their ability to process denials, but by how effectively they prevent denials from arising in the first place.
Revenue integrity is not an issue to be addressed after billing. It never has been.
Source: Healthcare IT Today