Rural-Urban Variation in CMS Hierarchical Condition Categories (HCC) Risk Scores

Lead researcher:
Contact:
Project funded:
September 2017
Project completed:
September 2020
Topic:

CMS uses hierarchical condition categories (HCC) risk scores to risk adjust cost and quality measures in several physician payment programs. Previous research has shown that rural Medicare beneficiaries have lower average HCC risk scores than urban beneficiaries. In addition, Table 34 of the proposed rule for the 2018 Quality Payment Program shows that the smallest physician practices have the lowest average HCC risk scores, and that these scores increase with practice size. Small practices are more likely to be located in rural areas; e.g., Liaw et al. (2016) find that 16-20 percent of practices with fewer than six physicians were located in rural areas, compared to nine percent of practices with at least 20 providers. Thus, the lower HCC risk scores among rural beneficiaries and small practices suggest it is likely that physicians practicing in rural areas will have lower average HCC risk scores than those practicing in urban areas. The goals of this study were to: (1) examine the difference in HCC risk scores by rurality of the provider, and (2) understand the sources of HCC risk score variation. Since HCC risk scores will affect provider reimbursement rates, understanding whether there is a bias against rural providers in the HCC risk model will enable federal and state policymakers to best design alternative payment models.

Using 2015 data from the Medicare Physician and Other Supplier Public Use File and the Chronic Conditions Warehouse, this study tested the following two hypotheses: (1) Controlling for provider characteristics likely to be associated with HCC risk score (e.g., practice size, provider type, provider specialty), a provider's location in a rural area will be independently associated with lower average HCC risk scores; (2) Among rural residents, rural beneficiaries treated in rural areas will have lower HCC risk scores than rural beneficiaries treated in urban areas. The study design was observational and analyses included both bivariate and multivariate (regression) analysis. We produced two findings briefs summarizing the distribution of HCC risk scores by rurality. One focused on the physician/practice level, while the other focused on the individual/community level.


Publications