Data mining techniques applied to medical information: Multiple solutions to support decision making.

摘要

The major purpose for this study is to submit possible solutions to spur the development of two new disciplines, Medical Informatics and Evidence-Based Medicine. Data mining technologies can help to explore hidden patterns and information insights from the databases. In the near future, an auto-stored medical information system (MIS, electronic medical records) will be providing information to a variety of users. It is expected that the concept of data mining and its various technologies will make the medical information system widely available and more functional.; A heart disease database with the complicated MIS-like data format of vast missing data, miscellaneous variables, and extensive existing knowledge was used to illustrate the research idea. This was accomplished through the strategies of multiple solutions and easy operating tools via four actual tasks of missing data treatment, appropriate data format selection, identification of important variables, and classification of high risk patients.; This study introduces new methods to handle the often seen difficulties when dealing with large amounts of medical data and also provides an efficient way to obtain trustworthy results. All the techniques used in this study are currently applicable with widely accessible programs. This characteristic supports the general users who therefore can encourage the development of the medical information system (MIS). The contributions of this study include: (1) An easily accessible resource for medical research, medical care decision making, and health policy design. (2) An easy and efficient formula to make compatible and trustworthy results affordable. (3) Develop new methods supplemented by other recent, less comprehensive, applications and academic substantiation to enhance the function of the MIS.