Defining ‘sign’ and ‘symptom’ Alexander P. Cox * , Patrick L. Ray, Mark Jensen, Alexander D. Diehl State University of New York at Buffalo, Buffalo, NY, USA * apcox@buffalo.edu Abstract—The terms ‘sign’ and ‘symptom’ have proven difficult to define and represent in a biomedical ontology. Medical professionals use ‘sign’ and ‘symptom’ to refer to medically relevant information about patients; however, they do not agree on the definitions. In particular, while medical professionals agree that there is an important distinction between signs and symptoms, they do not agree on the precise nature of this distinction. It is unsurprising then that attempts to provide ontological representations of these entities have repeatedly fallen short. As an added complication, a variety of entities— including material entities, qualities, and processes—may reasonably be understood as signs or symptoms. Thus, the ontological nature of a sign or symptom raises many questions about the meanings and proper use of these terms. We discuss specific challenges to defining ‘sign’ and ‘symptom’, identify essential features of these entities, explore the ontological implications of existing definitions, and propose our own definitions. We evaluate several competing ontological representations and present our proposed representation within the framework of the Ontology for General Medical Science. The proposed representation of sign and symptom is ontologically sound, provides precise definitions of each term, and enables users to easily create customized groups of signs and symptoms. Our experience highlights general issues about developing definitions in ontologies. Keywords—sign; symptom; definition; clinical finding; OGMS; ontology I. INTRODUCTION Clinicians and other medical professionals regularly use the terms ‘sign’ and ‘symptom’ to refer to medically relevant information about patients. Yet, the use of these terms is often imprecise, inconsistent, or both. This is due, in part, to the tendency to use these terms loosely. For example, by broadly referring to both signs and symptoms as symptoms [1]. As a further complication, many medical texts—including those dedicated to the study of signs and symptoms—fail to provide even preliminary definitions for these terms [2, 3]. When definitions are provided, they are not always consistent with one another. See TABLE I for a list of definitions. Comparison of lists of signs and symptoms that are presented in the absence of definitions reveals numerous potentially inconsistent applications of ‘sign’ and ‘symptom’. According to [2], examples of symptoms include: fatigue, dizziness, fever, headache, insomnia, lymphadenopathy, night sweats, muscle weakness, weight gain, weight loss, pain, nausea, bloating, itching, sore throat, hearing loss, diarrhea, constipation, confusion, memory loss, tremor, anxiety, cough, and jaundice. According to [1], examples of signs include: jaundice, swollen joints, and cardiac murmurs. According to [4], examples of vital signs include: temperature, respirations, pulse, and blood pressure. Notice that jaundice appears on both a list of symptoms and on a list of signs. While some definitions of ‘sign’ and ‘symptom’ allow certain features of the patient to be both a sign and a symptom, others do not. Additionally, which features can be both a sign and a symptom can change based on which definition is used. Representing sign and symptom in an ontology is an ideal means by which to enforce their precise definitions and encourage their consistent application. At the same time, it emphasizes the importance of these terms to the medical community. Our goal is to precisely define the terms ‘sign’ and ‘symptom’ and to provide sound ontological representations of these entities. In doing so, we hope that our experience will serve as a primer on some of the challenges involved in developing rigorous definitions in ontologies. II. METHODS There are theoretical concerns regarding definition formation that must be considered prior to an attempt to define a term or set of terms. Definition formation is goal-driven and, as such, there are certain desiderata for what typically constitutes a “good” definition. These desiderata often depend on the type of definition one is seeking to provide as well as on the field one is working in [5, 6]. Nonetheless, we can identify certain desiderata that should hold irrespective of these concerns. In general, definitions ought to be: a) sufficiently inclusive so as to include or capture all of the actual instances of their definiens, b) sufficiently exclusive so as to exclude or discount all of the instances that are not their definiens, and c) informative enough to impart information to the audience [7]. We acknowledge that many groups may require additional desiderata. The considerations listed here are minimal desiderata for definitions. There is also an issue of conceptual priority underpinning our process. Since we acknowledge that there are general desiderata for definitions before we engage in analysis of the current literature, these concerns are conceptually prior to any considerations discovered in the process of evaluating existing efforts. We also acknowledge that there may emerge more desiderata for specific definitions or types of definitions as a result of the evaluation of a set of attempted definitions. These should also be considered when determining whether a definition is adequate. For example, if a definition meets the three initial desiderata listed above but is criticized for obscurity or inconsistency with dominant views expressed in the literature, then one should seek to find a consistent and