The Application of Text Mining Software to Examine Coded Information

The purpose of this paper is to examine the use of text mining to investigate coded information. Inventory codes are usually nominal rather than ordinal. However, codes are not completely devoid of meaning; generally items in the inventory that are similar are assigned base numbers with added digits to distinguish between individual items. The standard approach to using such data in statistical models that require at least ordinal data is to create dummy variables, ie the presence or absence of an item. Instead, it is possible to treat the code as a text word, and to use text analysis to examine the data. In this paper, two examples from the healthcare industry are examined. The first examines ICD-9 codes that are used in Medicare billing to list patient risk factors and complications. The second examines medications routinely prescribed to patients in open heart surgery. Introduction Coded information is problematic using traditional statistical methods. A large retail inventory can contain hundreds and thousands of different codes. For example, the Office Depot web site contains 13 major categories of office supplies. The pencils and pens category contains another 5 subcategories with 11 further refined categories just of pens. Rollerball pens for one manufacturer have