Linkage Rate Between Data From Health Checks and Health Insurance Claims in the Japan National Database

Background: Japan's National Database (NDB) includes data on health checks and health insurance claims, is linkable using hash functions, and is available for research use. However, the linkage rate between health check and health insurance claims data has not been investigated. Methods: Linkage rate was evaluated by comparing observed medical and pharmaceutical charges among health check recipients in fiscal year (FY) 2009 (N = 21 588 883) with expected charges from the same population when record linkage was complete. Using the NDB, observed charges were estimated from the first published result of linking health check recipients in FY2009 and their health insurance claims in FY2010. Expected charges were estimated by combining 3 publicly available datasets, including data from the Medical Care Benefit Survey and an ad-hoc report by the Japan Health Insurance Association. Results: Only 14.9% of expected charges were linked by the NDB. The linkage rate was higher for women than for men (18.2% vs 12.4%) and for elderly adults as compared with younger adults (>25% vs <10%). Conclusions: The linkage rate in the NDB was so low that any research linking health check and health insurance claims will not be reliable. Causes for the low linkage rate include differences between health check and health insurance claims data in name format (eg, insertion of a space between family and given names) and date of birth (Japanese vs Gregorian calendar). Investigation of the causes for the low linkage rate and measures for improvement are urgently needed.