Limiting the Risk of Data Disclosure by Using Swapping Techniques in Variance Estimation

• Activity Number: 441
• Type: Contributed
• Date/Time: Thursday, August 12, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Sylvia M. Dohrmann*+ and Randy Sitter and Leyla Mohadjer and Jill M. Montaquila and Lester R. Curtin
• Companies: Westat and Simon Fraser University and Westat and Westat and Centers for Disease Control and Prevention
• Keywords: confidentiality ; replication ; Taylor Series linearization ; area sample

Abstract

The National Health and Nutrition Examination Surveys (NHANES) are one of the series of health-related programs conducted by the National Center for Health Statistics (NCHS). Beginning in 1999, NHANES became an annual survey with data release to the public every two years. NHANES 1999-2002 was fielded in approximately 57 locations comprising 54 PSUs that were primarily single counties. The small number of PSUs in the sample, together with characteristics available in the data files and outreach efforts aimed at increasing public awareness of the survey, led to concerns about the risk of PSU identification in the public release of the data. NCHS reviewed the data to identify disclosure risks, and requested research on variance estimation approaches to mask the PSUs. This paper will present the methods used to create alternative sets of PSU and stratum identifiers for variance estimation, the comparison of the variance estimates from each of the approaches considered, and a discussion of how the swapping approach is an improvement over the previously released method for estimating variances in NHANES.