Technological advances in medicine, particularly imaging, have resulted in early detection, objective documentation, and overall better insight into medical conditions. These advances, however, have also led to an increasingly complex medical record. Physicians now spend a significant portion of their time retrieving, structuring, organizing, and analyzing patient data, inaccurately and inefficiently: current information management systems in clinical medicine do not adequately support these functions, critical to the real-world practice of evidence-based medicine. Objective evidence, tailored to an individual patient, must be readily available to physicians as part of routine practice if true evidence-based medical practice is to become a reality. This proposal details the development and evaluation of several innovative technologies, providing solutions for the information management problems faced by physicians: 1) a distributed XML-based peer-to-peer medical record architecture, to enable portability and accessibility of patient information, regardless of geographical location; 2) a natural language processing (NLP) system for free-text medical reports, to automatically structure and characterize the contents of medical documents; 3) a phenomenon-centric data model, which supports the problem-solving tasks of the physician through explicit linking of objective findings (e.g., images, lab values) to medical problems; and 4) a time-based, problem-centric, context-sensitive visualization of the medical record, supporting a """"""""gestalt"""""""" view of the patient, with access to detailed patient data when needed. Together, these technologies will form a comprehensive system facilitating evidence-based medicine in a real-world environment. System evaluation will proceed in two parts. Technical evaluation focuses on each of the proposed technologies individually, gauging classical performance metrics: scalability of the distributed medical record; NLP precision/recall; expressibility/comprehensibility of the data model; and the usability of the new medical record user interface. Clinical evaluation will follow a time series study design (""""""""off-on-off""""""""), with implementation of the entire system in a real-world clinical environment, the UCLA Clark Urological Center. Clinical evaluation will measure the effectiveness of the system as a whole on intermediate outcomes (process of care) including the number of visits, number of procedures performed, and time to final diagnosis (disposition), as well as the impact on physician efficiency (time required to gather information and review charts).
Showing the most recent 10 out of 18 publications
