All submissions of the EM system will be redirected to Online Manuscript Submission System. Authors are requested to submit articles directly to Online Manuscript Submission System of respective journal.

An Optimum Data Warehouse for Epidemiological Analysis using the National Database of Health Insurance Claims of Japan

Author(s): Tomohide Iwao, Genta Kato*, Shigeru Ohtsuru, Eiji Kondoh, Takeo Nakayama and Tomohiro Kuroda

Background: While administrative databases for health care are increasingly used as research tools, such databases generally contain only health insurance claims data, the contents of which are insufficient for conducting epidemiological research. Creating a dataset appropriate for specific analysis requires technical expertise and familiarity with data analysis. The aim of our research is to develop a data warehouse (DW) accessible to researchers of epidemiology without this expertise.

Methods: We began by adding commonly used attributes in the epidemiological field to the National Database of Health Insurance Claims of Japan (NDB), to construct a Research Question Oriented DB. Secondly, we developed a versatile analysis unit schema by which the Research Question Oriented DW was reconstructed as perpatient units, covering demographics including sex, age group etc. We then proposed a pattern relational calculus by which research-specific attributes can be added without expert knowledge of SQL. Finally, we applied the DW in two epidemiological studies.

Results: In both studies, the coverage of attributes constructed only by the versatile analysis unit schema was limited. The versatile analysis unit schema covered 12% (3/25) of the attributes used for the one study as well as 15% (3/20) in the other study. On the other hand, the pattern relational calculus we proposed covered all remaining attributes which researchers used for their study.

Conclusion: As the versatile analysis unit schema and the pattern relational calculus were able to cover all attributes used in the two epidemiological studies, this shows that even within a limited scope, our method allows researchers who have little knowledge of SQL to tackle respective epidemiological study.

Abbreviations and Terminologies: NDB-SD: NDB Sampling Data set; DW: Data Warehouse; Shema: design of attributes in relations in the relational model theory; Relation: table with no duplicate tuple; Attribute: column name or variable name in relations; Primary key: one or more attributes that uniquely identify each tuple in a relation; Tuple: combination of attributes in a relation, almost the same meaning as row; Tuple relational calculus: logical expression used in the relational model theory; SQL: database language based on the relational model theory