Evaluating Record Linkage Quality in the NCHS Linked Mortality Files

• Activity Number: 633
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Dean Judson*+ and Jennifer Parker and Eric A. Miller
• Companies: NCHS and NCHS and NCHS
• Keywords: record linkage ; mortality ; false positive ; false negative ; matching

Abstract

In 2013, the National Center for Health Statistics released linked mortality files linking about 2.5 million survey respondents to the 70 million death records in the National Death Index (NDI). While these data are of substantive interest for studying health endpoints (e.g., the effects of obesity on mortality outcomes), they also provide a rich data source for record linkage methodology research, and in particular, research incorporating measures of linkage quality into health studies. As with all record linkage projects, some pairs of survey-NDI record matches are of much higher quality than others. Intuitively, one would expect that the more individual components of the records match, the higher the probability that the two records are a true positive, and the fewer components match, the lower this probability and the higher the probability that they are a false positive. In the context of an analytic example of the effects of socioeconomic status on mortality using the National Health Interview Survey linked to the NDI, this research evaluates the quality of record linkage in the linked mortality files, and assesses whether such probabilities can be approximated.