Applying data mining techniques to a health insurance information system

This paper addresses the effectiveness of two data mining techniques in analyzing and retrieving unknown behavior patterns from gigabytes of data collected in the health insurance industry. Specifically, an episode (claims) database for pathology services and a general practitioners database were used. Association rules were applied to the episode database; neural segmentation was applied to the overlaying of both databases. The results obtained from this study demonstrate the potential value of data mining in health insurance information systems, by detecting patterns in the ordering of pathology services and by classifying the general practitioners into groups reflecting the nature and style of their practices. The approach used led to results which could not have been obtained using conventional techniques.