Medicare Advantage: Changes Improved Accuracy of Risk Adjustment for Certain Beneficiaries

What GAO Found

The effect of CMS's revised community model on payment accuracy varied for the high-risk groups studied. Specifically, compared with the current community model, the revised community model slightly reduced the accuracy of MA payment adjustments for beneficiaries with multiple chronic conditions by $164, or about 1 percent of average actual expenditures. For beneficiaries with low income, the accuracy of the revised and the current community models was similar: estimates differed by $5, or less than 0.1 percent of average actual expenditures. For beneficiaries with dementia, the revised community model substantially improved the accuracy of MA payment adjustments by $2,674, or about 16 percent of average actual expenditures.

Compared with the general new enrollee model, the C-SNP new enrollee model substantially improved the accuracy of MA payment adjustments for new enrollees with C-SNP conditions, but considerable inaccuracy in the model's estimates remains for certain groups. The amount by which accuracy improved was similar across 14 severe or disabling chronic conditions: about $2,500. This reflects the design of the C-SNP new enrollee model, which increases expenditure estimates from the general new enrollee model by an amount that does not depend on beneficiaries' medical conditions. However, the C-SNP new enrollee model still underestimated expenditures for C-SNP-eligible new enrollees, on average, by about $1,500 and by more than $15,000 for beneficiaries who had certain conditions, such as end-stage liver disease or stroke. The C-SNP new enrollee model's results varied depending on the number of severe or disabling conditions the beneficiary had. Specifically, the model reduced the accuracy of estimated expenditures for new enrollees with only 1 severe or disabling condition by about 62 percent of average actual expenditures but improved the accuracy for those with 4 or more conditions by about 8 percent. However, the C-SNP new enrollee model still underestimated expenditures for beneficiaries with 4 or more conditions by over $20,000.

Accurate risk adjustment is particularly important for certain high-risk beneficiary groups that are more challenging and costly to treat and may benefit particularly from the coordination of care MA plans can provide. The decision to implement the revised community model that adjusts for dementia will depend on CMS's assessment of the advantages of more accurate payment adjustment for beneficiaries with dementia compared with the potential increase in the discretionary coding of dementia because of revised coding guidelines for Alzheimer's disease dementia published in April 2011. Additionally, while the introduction of the C-SNP new enrollee model improved the accuracy of payment adjustments for eligible new enrollees, on average, the model still considerably underestimated expenditures for certain groups, which could place plans that disproportionately enroll beneficiaries in these groups at a relative financial disadvantage.

In its comments on a draft of this report, CMS suggested that GAO assess the overall accuracy of the current risk adjustment model. GAO did not assess overall model accuracy because such an analysis was not within the scope of GAO's work and would have required additional data.

Why GAO Did This Study

The Centers for Medicare & Medicaid Services (CMS) pays plans in Medicare Advantage (MA)--the private plan alternative to Medicare fee-for-service (FFS)--an amount per beneficiary that is adjusted to reflect beneficiary health status. This adjustment, called risk adjustment, helps ensure that health plans have the same financial incentive to enroll and care for beneficiaries regardless of their health status. In 2010, CMS announced plans to revise the major medical conditions included in its principal risk-adjustment model--the community model--and add a model for new enrollees in chronic condition special needs plans (C-SNP), which target beneficiaries with certain severe or disabling chronic conditions. CMS began using the C-SNP new enrollee model in 2011, in place of the general new enrollee model, to adjust MA payments for new Medicare beneficiaries who enroll in a C-SNP. GAO was asked to examine the accuracy of these models for high-risk beneficiaries. Using data for a nationally representative sample of 2007 FFS beneficiaries, GAO computed the amount that expenditure estimates were above or below actual expenditures for 2007, the most recent data available at the time. GAO compared the accuracy of the current and revised community models for three high-risk groups: beneficiaries with multiple chronic conditions, with low income, and with dementia. GAO compared the accuracy of the general and C-SNP new enrollee models for new enrollees eligible to enroll in a C-SNP.

For more information, contact James C. Cosgrove at (202) 512-7114 or cosgrovej@gao.gov.