Record linkage of hospital discharge register with population register: Experiences at Statistics Netherlands

Abstract

This paper describes the first-time linking by Statistics Netherlands of the hospital discharge register (HDR), held by Prismant, with the population register (PR). This linkage is the first step in building an integrated population-based health statistics dataset at individual level, using existing register data as much as possible. The linkage of the HDR with the PR has proved to be successful: 87.6% of the HDR records could be uniquely linked to a person record in the PR, 8.7% were multiply linked, and 3.6% of the records could not be linked. The multiple linkages are a natural result of the limited resolution power of the linkage key (date of birth, sex, postal code). The 3.6% non-linkages are also plausible, given a rate of approximately 2% administrative errors in both registers. The quality of the linkage was further studied by estimating the level of mismatches (false positive linkages): this rate was estimated to be in the order of magnitude of approximately 1%, and in any case less than 3%. It was concluded that the linkage of the Dutch HDR with the PR is of good quality and that it forms an adequate basis for statistical analyses.