Interpretation of laboratory examination results and their simple representation

Abstract

Knowledge about the causal relationship between diseases and their laboratory findings is a key component for clinical decision support systems. For efficient acquisition of such knowledge, this paper attempted to represent the interpretation of laboratory results in a guidebook for laboratory examinations. A preliminary survey revealed the structure of the knowledge compiled in the guidebook, and found essential patterns in the cause-effect relationship. We then attempted to code the knowledge, utilizing a simple cause-effect relationship between exam results and their possible causes, expressed in a disease master table. For coding of the knowledge, a two-step approach was used: first the causing disease was looked up automatically in the disease master table, and then, manually. In the study, 84.5% of the knowledge in the guidebook was identified as a candidate for the coding in the simple cause-effect relationship, and 69.1% of the knowledge was successfully coded. Failure analysis suggested that further expressive power in the representation is gained only at the cost of considerable human intervention in the knowledge acquisition, and the cost for the utilization of the resulting data. Accordingly, for a certain type of application, which might prefer simplicity over accuracy or completeness of the information, the minimalist representation could be a reasonable choice.