A knowledge-based approach to defining syndromes

Abstract

S i129 robust and extensible structure for rapidly characterizing, describing, and communicating nontraditional data. A Knowledge-Based Approach to Defining Syndromes Justin V. Graham, David L. Buckeridge, Zach Pincus, Michael K. Choy, Martin J, O’Connor, and Mark A. Musen Stanford Medical Informatics, Stanford University School of Medicine Syndromic surveillance can only produce meaningful results if there is a common understanding of what observations constitute a syndrome and consequently how a syndrome relates to diseases that may cause those observations. However, the constituent elements of syndromes, such as “flulike illness,” are poorly characterized and rarely explicitly defined by surveillance system developers. We describe here a preliminary ontology for the creation of bioterrorism syndrome knowledge bases that will facilitate sharing and comparison of knowledge independent of a particular system or research group. In addition, we have created an inference heuristic problem-solving method that can relate indirect measurements of disease to diseases of interest. Our ontology enables precise enunciation of forms of evidence required to diagnose a syndrome. The ontology contains six major categories: syndrome, syndrome modifier, system affected, sign/symptom, direct supporting evidence, and indirect supporting evidence. We have instantiated the ontology for the syndrome “bioweapon respiratory illness.” The inference heuristic can use the elements of this ontology to combine direct measurements into meaningful abstractions. Each sign and symptom has a defined, explicit relationship to supporting direct and indirect evidence, like measured temperature or a patient’s chief complaint. Similarly, the presence of a syndrome can only be inferred if illness within requisite body systems can be substantiated by the presence of symptoms. We propose that all developers of syndromic surveillance systems explicitly define their syndrome concepts using a standard ontology. Syndrome definitions can be stored as instantiated knowledge bases in a common central repository, permitting knowledge sharing and reuse. A Knowledge-Based Method for Surveillance David L. Buckeridge, Martin O’Connor, Justin Graham, Michael K. Choy, Zachary Pincus, and Mark Musen Stanford Medical Informatics, Stanford University School of Medicine Surveillance of prediagnostic “nontraditional” data sources (e.g., school absenteeism, pharmaceutical sales, emergency medical services calls) is expected to enhance the timeliness of epidemic detection. However, prediagnostic data are not as specific as diagnostic data, so multiple sources must be followed to reduce false-positive detections. Combined analysis of multiple nontraditional data sources requires knowledge about the relationships between data sources, but knowledge of these relationships is often qualitative and uncertain. Statistical methods perform well for focused analyses of quantitative data according to well-defined models. However, statistical models do not readily incorporate qualitative data and can become unwieldy as the number of parameters grows. A knowledge-based approach requires explicit representation of surveillance knowledge and tasks and enables knowledge to be applied to problem solving in a structured manner. Our research approach is to model the tasks involved in public health surveillance and the knowledge required to accomplish these tasks. Based on these models, we identify or develop problem-solving methods (PSMs) that accomplish surveillance tasks. This modular development approach enables controlled evaluation of different PSMs and knowledge representations in terms of epidemic detection and impact on decision making around interventions. Prototype methods have been implemented