Voice-enabled structured medical reporting

Abstract

A paradox exists in applications that generate Electronic Health Records (EHRs) - how can data be captured from health care professionals speaking in a natural manner and in a computer readable form? This paradox is historical. Physicians are accustomed to dictating to a tape recorder and the speech is transcribed to a paper-based report. This format makes electronic access difficult and is of low value to the health care institution's needs for research and analysis. There is considerable interest in applying speech recognition to address this paradox. Users, many of whom are uncomfortable with computers, are attracted to a system that promises to simulate a transcriptionist. Institutional requirements are fulfilled as the data is created in an electronic form. However, conflicting goals arise. Doctors have very little time, want to create reports quickly and wish to use natural narration. Health care institutions, on the other hand, want to capture information in a controlled and predictable structure. This paper describes the design of a system, currently under construction, which we have named OSSIM (Open Systems Structured Information Manager). OSSIM attempts to achieve a balance between difficult ease of use goals and demanding institutional goals for computer readable structured information.