5 Common HCC Coding Errors and How AI Helps Solve Them
Hierarchical Condition Category (HCC) coding is a complex process where even minor errors can have a major financial impact. As the healthcare landscape shifts towards value-based care, the accuracy of your risk adjustment coding has never been more critical. Here are five of the most common errors we see and how AI technology can provide a powerful safety net for your coding team.
1. Missing Unspecified Codes
One of the most frequent errors is failing to code for conditions that are documented in clinical notes but not captured in structured EHR fields. A physician might detail a patient's 'Type 2 diabetes with nephropathy' in their notes, but if only 'diabetes' is checked in the EHR, the more specific and higher-value HCC is missed.
AI-Assisted Solution: Our platform uses Natural Language Processing (NLP) to scan physician notes and identify potentially more specific diagnoses. It then flags the corresponding HCC (e.g., HCC 18 for Diabetes with Chronic Complications) for the coder to review and validate. We break down a real-world example of the AI supporting this process to find diabetic nephropathy in another post.
2. Lack of MEAT Criteria Documentation
Every coded HCC must be supported by documentation showing it was Monitored, Evaluated, Assessed, or Treated (MEAT) during a face-to-face encounter. Simply listing a diagnosis is not enough for it to be billable.
AI-Assisted Solution: MedChartScan helps coders by highlighting the exact sentence where a physician may have prescribed a medication (Treated) or adjusted a care plan (Assessed). This allows the coder to quickly validate the evidence and create an instant, audit-ready trail for every human-approved code.
3. Incorrectly Linking Manifestations
Many conditions are only risk-adjustable when linked to a specific manifestation. For example, 'diabetic neuropathy' is a payable diagnosis, but coding 'diabetes' and 'neuropathy' separately may not be.
AI-Assisted Solution: Our clinically-trained AI is designed to identify when two separate entries in a chart could represent a single, linked manifestation. It then recommends the correct, combined HCC code for the coder's expert review. You can see a detailed breakdown of this exact type of AI-supported review process here.
4. Using Outdated or Deleted Codes
With the transition to models like CMS-HCC V28, thousands of diagnosis codes have been removed from the payment model. Using a code that was valid last year might result in a zero-value submission this year.
AI-Assisted Solution: The MedChartScan platform is continuously updated with the latest CMS guidelines. It acts as a safety net, automatically cross-referencing all potential codes against the current V28 model and flagging outdated codes to help prevent submission errors.
5. Failure to Recapture Chronic Conditions Annually
Many chronic HCCs must be documented and coded every single year to contribute to a patient's RAF score. It's a common oversight to miss recapturing a condition that was documented in the previous year.
AI-Assisted Solution: Our platform can review a patient's entire historical record. If a chronic condition like COPD (HCC 111) was coded last year but isn't mentioned in the current year's encounters, MedChartScan can flag it as a potential recapture opportunity for the clinical documentation improvement (CDI) team to investigate. This is the first step in shifting from reactive recapture to a proactive CDI strategy.