Beyond Diabetes: Uncovering High-Value Cardiovascular HCCs with AI

Cardiovascular diseases are among the most complex and highest-value conditions in risk adjustment. Correctly capturing the specificity of diagnoses like heart failure, stroke, and vascular disease is critical for accurate RAF scores, but it's also an area rife with common coding errors. The subtle details that differentiate a low-value code from a high-value one are often buried deep within unstructured clinical notes.

This is where a clinically-intelligent AI becomes an invaluable partner for a coding team. Let's explore a common scenario to see how AI helps coders capture a more accurate and complete picture of a patient's cardiovascular health.

The Challenge: Coding Congestive Heart Failure (CHF)

A coder reviews a chart for a 78-year-old patient. The primary care physician's problem list simply states "I50.9 - Heart Failure, unspecified." This is a valid code, but it's the least specific and has a lower value in most risk adjustment models. A manual reviewer, pressed for time, might stop there.

However, buried on page 12 of a faxed cardiologist report is the sentence: "Echocardiogram shows a left ventricular ejection fraction of 35%, consistent with systolic dysfunction." On page 4 of the PCP's notes, a medication change to include Entresto is documented.

A human coder could find this, but it would require reading every word of every document. Here’s how our AI assists:

1. It Identifies the General Diagnosis: The AI first flags the "Heart Failure" diagnosis on the problem list.

2. It Scans for Clinical Specificity: The AI analyzes all unstructured data, including the cardiologist's report. It recognizes "ejection fraction of 35%" as a key clinical indicator for Systolic Heart Failure.

3. It Connects Treatment to Diagnosis: The AI identifies "Entresto" as a medication primarily used to treat Systolic Heart Failure, providing MEAT evidence.

4. It Presents a High-Confidence Suggestion: The platform presents a suggestion to the coder:

   Suggested Rationale for Review: Patient has a documented ejection fraction of 35% and is on Entresto. Consider updating diagnosis to I50.2 - Systolic (congestive) heart failure.

This AI-surfaced insight allows the coder to instantly validate the evidence and capture a more specific, accurate, and higher-value HCC.

From Reactive Stroke Care to Proactive Sequelae Capture

Another common challenge is capturing the late effects (sequelae) of a cerebrovascular accident (CVA), or stroke. A patient's chart may list a "history of CVA," which is not a risk-adjustable diagnosis. However, the patient may have ongoing, codable deficits like hemiplegia or aphasia.

Our AI can scan therapy notes, nursing assessments, and physician narratives for terms related to these deficits. By flagging these ongoing conditions, it prompts the coder to review for a payable sequelae diagnosis, ensuring the patient's current health status is accurately reflected. This is especially critical under the CMS-HCC V28 model, which has specific logic for these types of conditions.

By empowering coders with AI that can connect these disparate dots, you enable your team to move beyond basic code capture. They become clinical data investigators, ensuring that the complexity and severity of your patients' cardiovascular health are documented accurately and completely.