Validation of Japanese Claims-Based Mortality Definitions Against Death Records from Vital Statistics: The LIFE Study.

BACKGROUND: Secondary healthcare databases are increasingly used in health-related research, but ensuring data validity is essential for generating accurate real-world evidence. This study aimed to validate mortality definitions based on Japanese medical and long-term care (LTC) claims data. METHODS: Using claims data between April 2015 and March 2023 from 13 Japanese municipalities, we developed eight algorithms to identify death. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of each algorithm were calculated using vital statistics death records as the reference standard. These metrics were calculated for the overall cohort, and also stratified by age (0-39, 40-74, and ≥75 years) and place/cause of death. RESULTS: We analyzed 1,894,123 participants (mean age: 61.2 years; women: 55.9%) with 68,966,581 person-months of claims. Specificity, PPV, and NPV were consistently high (≥0.97) across age groups, but sensitivity varied by algorithm, age, and place/cause of death. Medical claims-based algorithms identified approximately 90% of deaths in medical facilities, while the addition of LTC claims captured >95% of deaths in LTC facilities. Sensitivity for home deaths was generally low at 0.52-0.56 for individuals aged 0-39 years, which may have been due to sudden deaths (suicides or accidents) before medical services could be provided. However, home deaths among those receiving ongoing home-based care (e.g., cancer) were more accurately identifiable. CONCLUSIONS: Claims-based algorithms to identify death can be effectively used in epidemiological research in Japan by strategically applying different definitions according to type of death and target age.