Sample Selection for Medicare Risk Adjustment Due to Systematically Missing Data.

Area of Science:

• Health Economics
• Health Services Research
• Biostatistics

Background:

• Medicare Advantage (MA) risk adjustment aims to account for health status when paying plans.
• Nonrepresentative sampling in MA risk adjustment can lead to payment inaccuracies.
• Assessing sampling methods is crucial for equitable Medicare payments.

Purpose of the Study:

• To evaluate the impact of propensity-score matching on Medicare Advantage risk adjustment.
• To determine if matching improves the performance of risk adjustment formulas.
• To assess the effect of sampling methods on MA plan payments.

Main Methods:

• Utilized Medicare enrollment and claims data (2008-2011).
• Created risk adjustment predictor variables from Part A, B, DME, and HHA claims.
• Employed propensity-score matching to create a TM sample resembling MA enrollees.

Main Results:

• Propensity-score matching improved balance on observable characteristics.
• Risk adjustment formula performance metrics (R², predictive ratios) were similar between matched and random TM samples.
• Fitting formulas on matched versus random samples showed minimal differences in MA plan payments.

Conclusions:

• Current methods for fitting MA risk adjustment formulas show little difference in payments whether using a matched or random sample.
• Potential improvements may arise with more reliable MA encounter data and additional variables for risk adjustment.
• The matching method shows promise for future risk adjustment refinements.