Complexity and risk adjustment are two strategies employed to understand chronic disease and healthcare cost within patient populations. There is a general assumption that the data applied to risk adjustment models, such as the clinical risk groups (CRG) is sufficient to infer patient complexity. Our aim in this study was to compare the calculated complexity of a patient population using the 3 M™ CRG software to complexity data extracted from community-based primary care electronic medical records (EMR). We found that the distribution of the 3 M™ CRG health status was significantly different from the primary care EMR health status distribution, and that the number and type of chronic conditions identified differed between the two methods. We calculated a new variable that combined the information from the 3 M™ CRG software with the primary care EMR data. The distribution of the Combined-CRG distribution was significantly different from the 3 M™ CRG software; specifically, we saw many patients originally classified as being healthy or having minor chronic condition(s) re-categorized into having significant chronic condition(s). The CRG health statuses alone may be sufficient to predict future health expenditures, but caution is warranted if CRGs are to be used to infer complexity of the patient population.