Qualitative analysis of temporal information in Iliad: implications for linking Iliad to an electronic medical record as a knowledge server.

Abstract In linking Iliad to an electronic medical record (EMR) as a knowledge server, our objective is to convert the date and time stamped data from the EMR to the temporal representations existing in the Iliad knowledge base. First, an empirical analysis of Iliad's temporal information content is described: categorizing, estimating the amount (22%), and offering a temporal template to canonically represent it. Second, we analyze the effect of history and physical examination temporal information on the diagnostic performance of the system using data from 241 real cases. Third, we discuss knowledge engineering mapping techniques to convert time information found in EMRs. We conclude that the Iliad system includes a large amount of explicit and implicit time information and this information improves the system's diagnostic accuracy, both in terms of diagnosis ranking and confidence.