Dongwen Wang, PhD

Associate Research Scientist

Department of Biomedical Informatics

Columbia University

 

Email:

dongwen.wang@dbmi.columbia.edu

Telephone:

(212) 305-9801

Fax:

(212) 305-3302

Address:

VC5, 622 West 168th Street, New York, NY 10032

Research

My interest of research is in knowledge representation, knowledge engineering, and their applications in healthcare, with two specific focuses: (1) development and implementation of computer-interpretable clinical practice guidelines, and (2) integration of biomedical knowledge and data from distributes sources. My research on guideline development and implementation includes a variety of work such as the translation of the published guidelines into computer-interpretable formats, development of guideline representation models, implementation of guideline execution engine, integration of guideline execution engine with the clinical information system at a healthcare institution, and delivery of interventions to improve healthcare quality. My research on knowledge and data integration includes ontology mapping, sharing of medical knowledge that are encoded in different formats, and developing the links between medical errors and cognitive tasks. Below are brief descriptions on some of my research projects in recent years.

This project is focusing on the study of the process of guideline/pathway development and implementation as well as the evaluation of the impact of a guideline-based clinical decision support system on clinician’s decision making. We are working on the development of post-CABG critical pathway and its encoding in the GLIF format. We benchmark the clinical milestones in the post-CABG pathway and study the deviation of specific patient cases from these milestones such that the potential factors leading to the improvement of post-CABG patient management can be analyzed.

We have developed a formal ontology of cognitive errors, with strict classifications of the possible errors, clear definitions of the associated cognitive mechanisms, and potential solutions to prevent the errors. Based on this formal ontology, we have created the semantic links between the medical errors in specific clinical domains and the cognitive errors that we have identified. These links can then be used in the medical error reporting process to find the possible causes of the medical errors and their potential solutions.  

This is part of our work in the VigiLens project to enhance the function of the clinical event monitor at the New York Presbyterian Hospital. To support the application of modular decision rules encoded in the Arden Syntax and clinical practice guidelines encoded in the GLIF format, we are integrating the Guideline Execution by Semantic Decomposition of Representation (GESDOR) model with the clinical information system at the New York Presbyterian Hospital. Using an incremental approach to the development of the guideline task knowledge base of the GESDOR model, we are trying to leverage the legacy systems to confer cost benefits and to improve the acceptance of guideline implementation systems.  

Sharing of computer-interpretable clinical practice guidelines is a critical issue for guideline development, dissemination, and implementation. Difficulty in the interchange between guideline representation models is a major hindrance to guideline sharing. We have developed a generic guideline execution model, the Guideline Execution by Semantic Decomposition of Representation (GESDOR) model, to map multiple guideline representation models to a set of generalized guideline execution tasks. These generalized guideline execution tasks are then used to drive the execution of guidelines encoded in different formats. Evaluation has shown that the GESDOR model can be used effectively to execute guidelines encoded in different formats, and thus, it realizes guideline sharing at the execution level.   

Representation and execution of clinical practice guidelines encoded in the GLIF format is one of my primary research interests. We have developed the GLIF3 Guideline Execution Engine (GLEE), a tool for implementation of clinical practice guidelines encoded in the 3rd version of the GLIF format. GLEE can be integrated with a healthcare institution’s clinical information system as middleware. With necessary extensions, GLEE can be used for clinical decision support, quality assurance, medical education, and guideline development.

Translation of clinical practice guidelines into a computer-interpretable format is associated with many cognitive issues. These issues are further related to the interactions between clinicians and an information system within which the guidelines are implemented. We have been working on the development of a computation model to support the cognitive tasks in guideline development and implementation. This is a collaboration work with other members of the Laboratory of Decision Making and Cognition. Our work focused on the analyses of the guideline development process as well as the use of a guideline implementation system either as a tool to provide clinical decision support in practice or as a simulation environment to provide medical education. We study how the GLIF3 guideline representation model and the tools that are based on this model, such as GLEE, can be used to provide computation support to the cognitive tasks in these processes.

This is one of the four demonstration projects funded by CDC to improve childhood immunization coverage rate through cooperation among academic medical centers, community healthcare providers, and other organizations. We developed a computer-based multi-institution immunization registry, EzVac. Components of the EzVac system include an immunization database, hospital registration systems, a Web-based registry server, WWW user interfaces, and a decision support system. We use the decision support system to check the allergies and contraindications of a patient, to generate reminders of vaccines due, to recall patients for vaccines overdue, and to forecast vaccines due in the future.

Selected Publications

 

Last updated on October 19, 2004