Symbolic Classification Methods for Patient Discharge Summaries Encoding into ICD

  • Laurent Kevers
  • Julia Medori
Conference paper

DOI: 10.1007/978-3-642-14770-8_23

Part of the Lecture Notes in Computer Science book series (LNCS, volume 6233)
Cite this paper as:
Kevers L., Medori J. (2010) Symbolic Classification Methods for Patient Discharge Summaries Encoding into ICD. In: Loftsson H., Rögnvaldsson E., Helgadóttir S. (eds) Advances in Natural Language Processing. NLP 2010. Lecture Notes in Computer Science, vol 6233. Springer, Berlin, Heidelberg

Abstract

This paper addresses the issue of semi-automatic patient discharge summaries encoding into medical classifications such as ICD-9-CM. The methods detailed in this paper focus on symbolic approaches which allow the processing of unannotated corpora without any machine learning. The first method is based on the morphological analysis (MA) of medical terms extracted with hand-crafted linguistic resources. The second one (ELP) relies on the automatic extraction of variants of ICD-9-CM code labels. Each method was evaluated on a set of 19,692 discharge summaries in French from a General Internal Medicine unit. Depending on the number of suggested classes, the MA method resulted in a maximal F-measure of 28.00 and a highest recall of 46.13%. The best F-measure for the second method was 29.43 while the maximal recall was 52.74%. Both methods were then combined. The best recall increased to 60.21% and the maximal F-measure reached 31.64.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag Berlin Heidelberg 2010

Authors and Affiliations

  • Laurent Kevers
    • 1
  • Julia Medori
    • 1
  1. 1.CENTAL - Université catholique de Louvain (UCL)Louvain-la-NeuveBelgium

Personalised recommendations