Mining and Modelling Temporal Clinical Data

The Clinical e-Science Framework (CLEF) demonstrator runs Information Extraction technology over textual, narrative patient notes to assemble repositories of clinical patient data for the purposes of biomedical research and clinical care. Since many important medical events in the course of a patient’s treatment are mentioned in multiple documents and most documents will only include partial descriptions of these events, constructing a coherent and complete summary of a patient’s history – what we call a patient chronicle requires an information integration step over the output of Information Extraction. In this paper we describe and evaluate an approach to information integration which is based on mining narrative patient notes for temporal properties of medically relevant events and combining these with temporal information about events as provided by the structured (i.e., non-narrative) part of a patient’s health record.