The evaluation of semantic tools to support physicians in the extraction of diagnosis codes

Abstract

Over the past few years the extraction of medical information from German medical reports by means of semantic approaches and algorithms has been an increasing area of research. Currently, several tools are available that aim to support the physician in different ways. We developed a method to evaluate these tools in their ability to extract information from large amounts of data. We tested two off-the-shelf tools that worked in a background mode. We found that the field of quality management made it necessary that these large amounts of data could be background or batch processed. Additionally, we developed a metric, based on the semantic distance of the ICD codes, in order to improve the comparison of the accuracy of the codes suggested by the tools. The results of our evaluation showed that, at present, the tools are capable of supporting inexperienced physicians, however are still not sophisticated enough to work without human interaction. © Springer-Verlag Berlin Heidelberg 2007.