Toward Automated Visual Management and Presentation of Content Recommendations for Patient Note Generation

INTRODUCTION A patient note is a clinical document, written by a physician, describing a patient under his or her care. Various kinds of notes document the patient’s status over time, and comprise the official medical record for legal and billing purposes. Composing these notes encompasses several disparate, but interdependent subtasks. Physicians gather and review previous and current patient data, such as lab results, physical exams, medications, and hospital events, to determine patient progress; select and input relevant information for the current note; and form appropriate assessments and plans for each patient. Currently, most of these subtasks are not supported by domain-informed assistance. While sophisticated systems exist for retrieving and visualizing clinical patient data (e.g., [1, 3, 5, 7, 9]), support for inserting this data into a patient note is limited, less flexible, and more complicated to use than editing notes manually in a generic word processing application. In practice, physicians often edit each current note by hand— manually searching through previous notes and copying and modifying material written for previous notes—a time-consuming and error-prone process. We are working with physicians in the New York Presbyterian Hospital (NYPH) Cardiothoracic Intensive Care Unit (CTICU) to design a user interface for patient note creation. We are developing new presentation techniques through which appropriate content can be recommended, and new interaction techniques for effectively browsing, selecting, and inserting these recommendations directly into the notes, as they are edited.