Assessing Disclosure Risk for the California Health Interview Survey Public Use Files: Practical Considerations and Current Methods

• Activity Number: 311
• Type: Topic Contributed
• Date/Time: Tuesday, August 9, 2005 : 2:00 PM to 3:50 PM
• Sponsor: Section on Government Statistics
• Author(s): M. Leeann Habte*+ and Jenny Chia and Hongjian Yu and Brandon Traudt
• Companies: University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles and University of California, Los Angeles

Abstract

Assessing disclosure risk for survey microdata files is a complex issue, and standards are still developing. National surveys have implemented various methods for assessing the likelihood that sample unique records are members of small or rare groups in the population, based on key characteristics. The California Health Interview Survey (CHIS), a large, biennial, statewide population survey of health and health care access, developed an approach for decreasing risk by introducing noise into the data file that involves: identifying and masking special unique records, imputing source demographic survey variables, and identifying and masking sample unique records likely to be members of small groups in the population. In addition, the survey implements top or bottom coding to mask outliers in continuous variables and recodes categorical variables into larger categories on a limited basis. This paper discusses the assumptions and rationale underlying CHIS' approach and its implementation in the California Health Interview Survey 2003 public use microdata files.