Analysis of Medication Error Reports
In medicine, as in many areas of research and society, technological innovation and the shift from paper based information to electronic records has created a climate of ever increasing availability of raw data. There has been a corresponding lag in our abilities to analyze this mass of data, and traditional forms and expressions of statistical analysis do not allow researchers and practitioners to interact with data in the most productive way. This is true in the emerging area of patient safety improvement. Traditionally, a majority of the analysis of error and incident reports are approached as data comparisons, and starts with a specific question which needs to be answered. Newer data analysis tools have been developed which allow the researcher to not only ask specific questions but also to “mine” data: approach an area of interest without preconceived questions, and explore the information dynamically, allowing questions to be formulated based on patterns brought up by the data itself. Additionally, the “types” of information objects that can be the objects of data analysis have been extended to include text [8][9]. Since 1991, United States Pharmacopeia (USP) has been collecting data on medication errors through voluntary reporting programs. USP’s MEDMARXsm reporting program is the largest national medication error database and currently contains well over 600,000 records. USP conducts an annual quantitative analysis of data derived from “pick-lists” (i.e., items selected from a list of items) without an in-depth analysis of free-text fields. In this paper, the application of text analysis and data analysis tools used by Battelle to analyze the medication error reports already analyzed in the traditional way by USP is described. New insights and findings were revealed including the value of language normalization and the distribution of error incidents by day of the week. The motivation for this effort is to gain additional insight into the nature of medication errors to support improvements in medication safety.