Mispricing in the medicare advantage risk adjustment model

Area of Science:

• Health Economics
• Health Services Research
• Medical Informatics

Background:

• The Centers for Medicare and Medicaid Services (CMS) uses Hierarchical Condition Category (HCC) models to adjust payments for Medicare Advantage (MA) plans.
• These models aim to reflect enrollees' expected healthcare costs.
• Previous CMS models were simplified due to concerns about data submission burdens from health plans.

Purpose of the Study:

• To identify mispricing errors in CMS's HCC models for Medicare Advantage payments.
• To compare the accuracy of CMS models with more comprehensive Diagnostic Cost Group (DxCG) Medicare models.
• To evaluate the impact of model simplification on payment accuracy and predictive power.

Main Methods:

• Utilized Verisk Health's DxCG Medicare models, a refined version of the HCC framework with 189 clinical categories.
• Applied both CMS and DxCG models to Medicare's 2010-2011 fee-for-service 5% sample data.
• Analyzed discrepancies in predicted versus actual healthcare costs, particularly for patients with complex conditions.

Main Results:

• Identified two key mispricing errors in the CMS implementation: ignoring new enrollee data and using simplified models.
• CMS models recognized only 79 condition categories, excluding many diagnoses and merging heterogeneous conditions.
• In 2010, 13% of beneficiaries with higher-cost DxCG-recognized conditions had no CMS-recognized condition, leading to average underpayments of over one-third ($6628 in 2011).

Conclusions:

• The current CMS HCC model implementation results in significant mispricing and lower predictive accuracy compared to DxCG models.
• Simplified CMS models fail to adequately capture the costs associated with complex patient populations.
• CMS should consider adopting more refined, DxCG-like models, especially as MA plans now provide encounter data.