237 – SPEED: Missing Survey Data: Analysis, Imputation, Design, and Prevention
Nonresponse Bias Analysis for the Medicare Current Beneficiary Survey
Ying Li
NORC at the University of Chicago
Whitney Murphy
NORC at the University of Chicago
Kirk Wolter
NORC at the University of Chicago, Chicago
Survey nonresponse occurs when data are not collected for an eligible sampled individual. If non-respondents differ from respondents in meaningful ways, then nonresponse bias may occur. This paper focuses on evaluating the presence and extent of unit nonresponse bias in the Medicare Current Beneficiary Survey (MCBS). The MCBS is a continuous, multipurpose survey of a nationally representative sample of the Medicare population, conducted by the Centers for Medicare & Medicaid Services through a contract with NORC at the University of Chicago. Unlike most nonresponse bias analyses that only have limited types of data available to conduct comparisons between respondents and non-respondents, the MCBS offers a variety of measures, such as demographic characteristics, self-reported attributes, Medicare administrative Fee-For-Service payment information, and administrative chronic conditions indicators to do analysis. We applied Rao-Scott chi-square test, adjusted logistic regression model, and generalized linear model with contrast analysis to identify statistically significant differences. Using these additional measures, we found that only limited attributes would indicate bias.