Objective: Medical information databases provide useful Real World Evidence (RWE) and a comprehensive view of medical activities. However, since each database has limited coverage and cannot be self-sufficient, combining information from multiple databases is a useful research technique. In this study, we examined methods of estimating patient numbers by combining information from multiple databases in order to assess the respective databases and identify the respective characteristics, biases and idiosyncrasies. This process also allowed us to propose improvements in the grand design of medical information databases in Japan.
Design: Retrospective observational cohort study
Methods: We attempted to estimate the numbers of patients treated for certain diseases and the numbers prescribed a drug by three methods: i) We estimated patient numbers for seven diseases using an insurance claims database, adjusting the proportion of elderly patients according to a hospital medical records database; ii) Sales information for drug X was combined with the prescribed volume per person estimated from pharmacy claims databases to estimate the number of patients administered X; this number was divided by the prescription rate obtained from a medical claims database to calculate patient numbers for the associated disease; and iii) We examined two surveys of the National Institute of Infectious Diseases (NIID) for timely estimation of patient numbers for influenza, referring to estimates from an insurance claims database.
Results: In Method i)-iii), we proved that it is possible to estimate patient numbers for many diseases and administered drugs by effectively combining multiple medical information databases. Validation could be claimed when multiple methods lead to similar results.
Conclusion: These databases provided by government agencies and private corporations are separately managed, and there is no grand plan to integrate them into one platform. It is crucial that databases, rather than being designed to stand alone, are standardized according to widely used systems under a solid master data management strategy. This will make it easier to combine information from multiple databases and to maximize their values. Mutual use of these databases by academic researchers for epidemiological and clinical studies and by government policy makers and data scientists of pharmaceutical companies may improve the usefulness of these databases and expand their application in research.