Linking Survey and Administrative Data with Incomplete Identifying Information

• Activity Number: 508 - Administrative Records and Record Linkage
• Type: Contributed
• Date/Time: Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Government Statistics Section
• Author(s): Marc Roemer* and Cordell Golden and Lisa B Mirel and Adam Fedorowicz and Jennifer Sayers
• Companies: AHRQ and National Center for Health Statistics (NCHS) and CDC/NCHS/OAE/SPB and NCHS and NCHS
• Keywords: record linkage ; NHIS ; MAX ; administrative records ; linked data

Abstract

Survey data enriched by linkage to administrative records remain of interest to federal agencies and policy makers for their potential use in program evaluation, policy-relevant research, and innovation. Such data linkages rely on the availability of sufficient personally identifiable information (PII) in both the survey data and administrative records. Prior to 2007, the National Health Interview Survey (NHIS) collected complete 9-digit Social Security Numbers (SSN) from survey participants. However, in attempt to address respondents' increasing refusal to provide SSN and consent for linkage, NHIS began in 2007 to collect only the last 4 digits of SSN. This change has posed new challenges for data linkages involving NHIS and requires developing new linkage algorithms to address this issue. The challenges are amplified with the linkage of NHIS to Medicaid Analytic Extract (MAX) data, which also have limited PII. This paper addresses the challenges associated with linking two data sources when limited PII is present. The paper will describe methodology for approaching this issue and present preliminary results.