NLP to Support Hierarchical Condition Categories (HCC) Coding

Hierarchical Condition Categories (HCC) coding was implemented by the Center for Medicare and Medicaid Services (CMS) in 2004. It’s designed to estimate the possible healthcare costs of Medicare enrollees in the next coming year. It is used in Medicare, Medicaid, and commercial healthcare plans to design the health mix of their member enrollments and the reimbursements they can expect from CMS. Each HCC code has a specific weight and score. The patient’s risk adjustment factor (RAF) can be determined by the HCC code in combination with the patient demographics, such as age and gender. Risk scores are associated with the estimated payments. Figure 1 to 3 shows three scenarios with different patients; however, all are the same age and sex but with different HCC codes and RAF scores.


To accurately assign the HCC code to the patient ICD-10-CM diagnosis codes and different granularities of patient context/conditions will be considered. Some of the conditions are only available in the unstructured EHR notes. With our clinical information extraction pipeline, patients’ major diagnoses, demographics, and other chronic conditions can be recognized and extracted to support HCC coding. In Figure 3, our NLP engine can recognize chronic conditions in addition to type 2 diabetes so that there will be a more accurate risk and payment estimation, and the patient will be mapped to scenario 3.


Figure 1: Discharge summary from patient #1

Figure 2: Discharge summary from patient #2

Figure 3: Discharge summary from patient #3


To learn more about Melax Tech's clinical information extraction pipeline and how it can support your organization's HCC coding needs, request a demo today!