From clinical narratives to ICD codes: automatic text categorization for medico-economic encoding

Abstract

Summary We report on the design and preliminary evaluation of a system to help assignment ICD (International Classification of Disease) codes to clinical narratives. The objective of the project is to deliver an operational system to assist professional encoders of the University Hospital of Geneva in their coding task. We combine a set of machine learning and data-poor methods to generate a single automatic text categorizer, which returns a ranked list of ICD codes for any given output. In its current settings, the system is accepting three different textual fields as input: anamnesis, diagnosis and prescription fields. The combined ranking system currently obtains a precision of 74% at high ranks and a recall of about 63% for the top twenty returned codes. Although fairly promising, the impact of the tool on the institution coding efforts as well as on the billing figures (DRG: Diagnosis Related Groups) must now be evaluated.