Multiple Imputation of Missing Income Data in the National Health Interview Survey

• Activity Number: 150
• Type: Topic Contributed
• Date/Time: Monday, August 9, 2004 : 2:00 PM to 3:50 PM
• Sponsor: General Methodology
• Author(s): Nathaniel Schenker and Trivellore Ragunathan and Pei-Lu Chiu and Diane Makuc*+ and Guangyu Zhang and Alan Cohen
• Companies: National Center for Health Statistics and University of Michigan and National Center for Health Statistics and National Center for Health Statistics and University of Michigan and National Center for Health Statistics
• Keywords: multiple imputation ; surveys ; item nonresponse

Abstract

The National Health Interview Survey (NHIS), a household survey of the U.S. civilian noninstitutionalized population, is a major data source for studies of health status and health care access and utilization. Although item nonresponse in the NHIS is generally low, it is high for annual family income, a key variable in many analyses. In 2001, for example, 29% of families did not report family income. Starting with the 1997 NHIS, multiple imputation of unknown family income and unknown personal earnings is being carried out. Five sets of imputed values are being created for each year using sequential regression multivariate imputation. The application to the NHIS is discussed in documentation for public use files containing the 1997-2001 imputed income values. This presentation will describe the patterns of missing income data and the multiple imputation of income in the 1997-2001 NHIS. Analyses of health status and access to health care that are based on the multiply imputed data will be compared with analyses based on singly imputed data and analyses that delete respondents who did not report income.