Varying Relationships Between Beneficiary Traits and Quality Measures Affect Comparisons in Medicare Advantage

• Activity Number: 529
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Laura Anne Hatfield* and Alan M. Zaslavsky
• Companies: Harvard Medical School and Harvard Medical School
• Keywords: Medicare ; mixed effects models ; quality scores ; multivariate modeling

Abstract

Casemix adjustment produces adjusted means for units as if they all enrolled an "average" population by regressing scores on individual traits. But what if the relationships between individuals' characteristics and survey responses vary across units? We study how differences in quality score adjustment coefficients across Medicare Advantage (MA) contracts change comparisons among individuals and contracts. We analyze responses to the MA implementation of the Consumer Assessments of Healthcare Providers and Systems (CAHPS) survey from 2010 to 2014. Using hierarchical models, we predict quality for individuals and contracts, adjusted for self-reported education, general health, and mental health. We allow the effects of these variables on quality measures to vary across contracts with a hierarchical model. We find that for average consumers, standard adjustment is sufficient to represent variation in contract quality standardized to a common population. For people with characteristics far from average, personalized reporting using their characteristics and contract-specific coefficients can substantially change the expected quality measures across contracts.