There has been a pressing need for improving patient safety. Sizable amount of Americans do not feel safe about health care, as it is supposed to be [1]. Meanwhile, preventable medical errors that harm patients cost $17.1 billion a year which over-burdened the healthcare system [2]. Although the reasons why errors happen can be complex due to the intricate specification of the system, much attention has been drawn to how patient safety event reporting system can improve the quality and safety of health service over the past decade [3]. An outstanding event reporting system should be able to collect data that link to procedures and factors threaten patient safety in a timely manner. Nevertheless, a great number of reporting systems are suffering low quality of the data, inefficiency and ineffectiveness of data entry [3-5]. One approach aiming at improving the situation is developing a comprehensive and unified ontology for patient safety events. Over the last decade, there has seen a dramatic increase of 600% in the number of citations on ontologies in PubMed/MEDLINE [6], however, the ontology engineering is relatively lagged behind in the field of patient safety. Therefore, the development of ontology towards enhancing patient safety is in an imperative need. Ontologies or taxonomies developed specifically for use in patient safety system are not new. The Australian Patient Safety Foundation (APSF) originally reported the Australian Incident Monitoring System in 1987, and later in 1993 and 2000, APSF expanded the system twice [7]. A newly developed taxonomy endeavors to categorize major types of human error contributing to medical errors [8]. Other taxonomies or standards such as JACHO patient safety event taxonomy [9], National coordinating council for medication error reporting and prevention (NCC MERP)'s taxonomy of medication errors [10], Neonatal Intensive Care system (NIC) [11], Pediatric Patient Safety taxonomy (PED) [12], Preliminary Taxonomy of medical errors in Family Practice (PTFP) [13], Taxonomy of Nursing Errors (TNE) [14], and Adverse Event Reporting Ontology (AERO) [15] shared insights in specific domains. While these ontologies served primarily as standards of domain specific taxonomies, the rapid increase in medical information calls for a unified knowledgebase with unified language system to be used as a common denominator for sharing and learning across patient safety reporting systems. In this project, we built a semantic web ontology (Medeon) using W3C open standard Web Ontology Language (OWL) (Fig. 1). The Common Formats (v1.2) developed by the Agency for Healthcare and Research Quality (AHRQ) were employed as the taxonomy where we extracted and encoded semantic knowledge into Medeon. Recognized as a unified standard of reporting patient safety events, the Common Formats are designed to specify and collect event information, which range from general concerns to frequently occurring and serious types of the events. We chose OWL and semantic web technologies because they jointly provide unique advantages for machine understandable semantics and descriptive logic reasoning which allows us to model real-world patient safety data in a computerized system. We borrowed the hierarchical structure in the Common Formats to build the OWL classes and rephrased the narrative data in the Common Formats to construct OWL incidents and objective properties. In order to share the knowledge of the Common Formats with other domain ontologies, Unified Medical Language System (UMLS) was employed to map terminologies between different domains.
Fig. 1. A general procedure of generating Medeon. Images were adapted from AHRQ Common Formats
(https://www.pso.ahrq.gov/common) and US National Library of Medicine. (http://www.nlm.nih.gov/research/umls/)
Our ontology lay out to improve patient safety reporting system in the following aspects. The ontology
primarily serves as a knowledgebase to model the taxonomies broadly used for patient safety events. With
this knowledgebase, semantic data can be retrieved and reasoned through descriptive logic rules and
applied to text mining methods. Secondly, the use of UMLS provides a framework to encode and exchange
data between our ontology and other semantic data repositories. In the end, our ontology holds promise in
facilitating decision support in clinical research [
This project is supported by a grant on patient safety from the University of Texas System. Introduction A