Estimates Adjustment Strategies for the 1999 Health Outcome Survey

• Activity Number: 400
• Type: Contributed
• Date/Time: Thursday, August 15, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Section on Health Policy Statistics*
• Author(s): Yi-Ting Hwang*+ and Arlene Bierman+ and Samuel Haffer and Lap-Ming Wun
• Affiliation(s): National Taipei University and Agency for Healthcare Research and Quality and Centers for Medicare & Medicaid Service and Agency for Health Care Policy and Research
• Keywords: HOS ; nonresponse ; selection bias ; weight adjustment

Abstract

The Medicare Health Outcomes Survey (HOS), a new quality measure for Medicare+Choice Plans, is designed to measure casemix-adjusted two-year change in physical and mental function among plan enrollees. Using data from the 1999 Cohort II baseline survey, we developed survey weights accounting for selection and nonresponse bias, and compared point estimates and standard errors from individual multivariate analyses done with and without sampling weights. This survey randomly sampled 1000 enrollees from each Medicare+Choice in the US (278,334 eligible beneficiaries from 338 plans with a response rate of 66%). The characteristics of nonrespondents were characterized and contrasted to those who responded to the survey. Response rates differed significantly by gender, race, Medicaid status, and region, which were used to construct the adjustment for nonrespondents. The weights also included adjustment for selection bias, the reciprocal of the probability of a subject being selected. We then implemented the weights into the individual level analysis and found estimates with the weights were similar to those without the weights but with higher standard errors.