Schemas for Telling Stories in Medical Records

To accurately support a patient's medical record, at least four interrelated models are required: a simple static one-level schema is inadequate. The models must support the medical record as a coherent story reconstructed from the sequence of recorded events within the medical record. We propose one representation which unifies all four models by a three space approach, each space acting as a schema for the space below. The three spaces assist atemporal summarisation of a patient's medical record and illustrate the difficulties of recording retrospective or contradictory observations. The approach uses a generative, descriptive subsumption-based classification formalism with a sophisticated system of semantic constraints controlling the generation of implied intensional concepts. We report our experiences in its use in a prototype clinical workstation. We believe that this model can be used for complex applications where contradictory and incomplete information is captured over time and a complex semantic constraint model is required.