caBIG
Velos eResearch
Data Warehousing
PhenoGO
Executable Knowledge for Evidence-Based Medicine
Pressure Ulcer Risk Management Model
Vigilens
Relevant Course Projects
Meditech
caBIG (1/06 – 12/06)
Description
The University of Chicago Cancer Center (UCCRC) is a participant in the Integrative Cancer Research (ICR) Workspace of the National Cancer Institute (NCI) cancer Biomedical Informatics Grid (caBIG) initiative. The goal of the ICR Workspace is to provide tools and systems that enable the integration and sharing of clinical and basic research information among cancer researchers. As participants within this Workspace, members of the UCCRC are charged with sharing ideas and concerns of strategic importance to the program, and driving planning and process improvement discussions through the active involvement with relevant meetings and teleconferences.
Role
I am the UCCRC’s primary representative and liaison for the ICR Workspace. As such, I participate in all monthly Workspace and Translational Informatics Special Interest Group (SIG) teleconferences and bi-annual face-to-face meetings, as well as any general caBIG teleconferences and annual face-to-face meetings. I am responsible for reporting pertinent information back to the UCCRC, coordinating and documenting UCCRC participation, and submitting all required progress reports to caBIG project leaders.
top
Velos eResearch (1/06 – 12/06)
Description
Velos eResearch is a web-based, comprehensive clinical research management solution that has recently been implemented within the UCCRC. Over the past several months, all relevant data has been migrated from the previously used homegrown clinical trials management system, and investigators have begun learning and utilizing this tool for the conduct of their research.
Role
I am responsible for conducting a number of data quality assurance and reporting tasks, including:
top
Data Warehousing (1/06 – 12/06)
Description
Currently, most clinical and basics sciences researchers associated with the Department of Medicine at the University of Chicago Hospitals utilize their own individual, disparate database for storing and analyzing data. Work is currently underway to determine how to integrate such information to create a multi-purpose data warehouse.
Role
In collaboration with the Biological Sciences Division (BSD) Information Systems group, I have designed a generic database model to serve as the initial framework for such a data warehouse.
top
PhenoGO (2004 – 2006)
Description
The Gene Ontology (GO) project, in a collaborative effort to address the need for consistent descriptions of gene products from different databases, has developed three structured, controlled vocabularies (ontologies) describing gene products in terms of their associated biological processes, cellular components and molecular functions in a species-independent manner. The Gene Ontology Annotations (GOA) contain gene products (or genes) from collaborating databases annotated with GO terms and associated references to support the annotations. PhenoGO utilizes an existing NLP system (BioMedLEE), an existing knowledge-based phenotype organizer system (PhenOS), and MeSH indexing to automatically augment these annotations in GOA with additional cellular and anatomical context. The system also maps the context to identifiers that are associated with different biomedical ontologies, including the UMLS, Cell Ontology (CL), Adult Mouse Anatomy (MA), NCBI taxonomy, GO, and Mammalian Phenotype Ontology (MP).
Role
As a member of the PhenoGO research team, I contributed to the development of a cutting-edge text mining algorithm that ultimately led to the largest database of phenotypic annotations of the Gene Ontology resource. I designed and implemented advanced text mining and encoding algorithms utilizing natural language and semantic process techniques to systematically identify and associate phenotypic context with genes and gene products annotated within GOA. Additionally, I designed, coordinated and conducted a comprehensive evaluation of the underlying knowledge schema of the utilized natural language processor (BioMedLEE) as well as the content of the PhenoGO database.
Publications
Friedman C, Borlawsky T, Shagina L, Xing HR, Lussier YA. Bio-Ontology and Text: Bridging the Modeling Gap. Bioinformatics, in press.
Lussier, YA, Borlawsky T, Rappaport D, Liu Y, Friedman C. PhenoGO: Assigning Phenotypic Context to Gene Ontology Annotations with Natural Language Processing. Pacific Symposium on Biocomputing, 11:64-75 (2006).
Sam L, Borlawsky T, Rappaport D, Li J, Shagina L, Liu Y, Friedman C, Lussier YA. Towards a Eukaryotic Phenome: the PhenoGO Database System. Pacific Symposium on Biocomputing, 2006.
Borlawsky T, Rappaport D, Sam L, Li J, Liu Y, Friedman C, Lussier YA. Computational approaches for genome scale analyses of phenotypes—Bridging the phenotypic gap between clinical repositories and model organism databases. Genome Informatics Conference, Cold Spring Harbor Laboratory, 2005.
top
Executable Knowledge for Evidence-Based Medicine (2004 – 2006)
Description
Evidence-based medicine (EBM) is quickly becoming the standard for medical decision-making. Answers to physicians' patient-related questions must be integrated into clinical practice in the most concise and efficient way possible, which is challenging because of the emergence of increasingly complex medical cases involving multiple confounding conditions, and the growing number of medical information sources.
Role
I planned and implemented, utilizing Protégé, an OWL-based ontology of biomedical concepts for use in enabling information retrieval over science and clinical research knowledge sources. I coordinated project activities with and developed knowledge discovery tools to enhance current evidence-based medicine techniques using natural language processing (BioMedLEE) and semantic mapping techniques over the Cochrane Reviews with our industrial partner, John Wiley and Sons Inc. In addition, I designed, implemented and managed project-specific relational databases developed utilized FileMaker Pro and MySQL, and associated thick- and thin-client user interfaces.
Publication
Borlawsky T, Friedman C, Lussier YA. Generating Executable Knowledge for Evidence-Based Medicine Using Natural Language and Semantic Processing. AMIA Annual Symposium Proceedings, 2006.
top
Pressure Ulcer Risk Assessment Model (2004)
Description
As health care organizations attempt to reduce adverse patient outcomes and increase the quality of patient care through the early identification of those at risk for poor outcomes, one major challenge has been to decrease the incidence of pressure ulcers, a complication of bed rest often occurring in acute care settings. In addition to being painful and slow to heal, pressure ulcers are difficult to manage, and lead to increased health care costs and lengthened hospitalizations. Early identification of patients at risk for pressure ulcer development is critical to their prevention.
Role
I developed filter feature decision analysis model using heuristic statistical methods applied to a relational database of historical patient data to select 87 potentially significant features, including patient demographics, medications, and clinical visit details, that may enable automated, early detection of patients at risk of developing pressure ulcers. I utilized these attributes as input for the C4.5 decision tree induction algorithm, which classified patient risk of developing such complications. I assessed the validity of the resulting classification model, using a four-fold cross-validation technique.
Publication
Borlawsky T, Hripcsak G. Evaluation of an Automated Pressure Ulcer Risk Assessment Model. Home Health Care Management and Practice, submitted 2006.
top
Vigilens (2004)
Description
The Vigilens Health Monitor is currently being used to decrease the cost of medicine at the New York Presbyterian Hospital (NYPH) by notifying affiliated physicians of critical laboratory results. The system delivers from 250 to 450 critical alerts daily to the WebCIS portal. Vigilens and its Statistical Quality Control system (see Roles below) are used by the NYPH administration to maximize revenue by annually identifying about 3,500 uninsured patients eligible for Medicare and Medicaid for timely referral with a social worker. In addition, the Vigilens implementation has led to an 85% reduction of emergency room usage post intervention last year, resulting in an estimated reduction of emergency room utilization costs of over $1,000,000.
Role
I designed and implemented, utilizing series of complex Perl scripts, the Vigilens Statistical Quality Control (SQC) system to assess, in real time, the associated decision-support, alerting and error monitoring applications. The system generates daily, weekly and monthly reports summarizing the generated critical laboratory alerts, and on a daily basis, assesses whether the alerts from the previous day for a specific laboratory result were generated with a statistically significant greater frequency than expected. In addition, I worked closely with team members at the IBM Thomas J. Watson Research Center to develop and evaluate a custom interface for the delivery of such critical laboratory alerts over Blackberry handheld devices.
Publications
Lussier YA, Williams R, Li J, Jalan S, Borlawsky, T, Stern E, Kohli I. Partitioning Knowledge Bases Between Advanced Notification and Clinical Decision Support Systems. Decision Support Systems, in press.
Borlawsky T, Li J, Jalan S, Stern E, Williams R, Lussier YA. Partitioning Knowledge Bases Between Advanced Notification and Clinical Decision Support Systems. AMIA Annual Symposium Proceedings 2005.
PubMed Abstract
Poster
top
Relevant Course Projects (2003-2004)
Maintaining Consistency between a Concept-Based Vocabulary and a Changing Standard Controlled Terminology, Representation and Coding of Medical Data course
eID: A High-throughput Emerging Infectious Disease Literature Monitoring System, Decision Support in Biomedicine course
Integrated Public Health Laboratory Information System for the New York City Department of Health and Mental Hygiene, Public Health Informatics course
Paper
Presentation
5-Year Return on Investment for the Implementation of a Computer-Based Patient Record in a Small Group Practice, Economics of Informatics: Cost and Investment Issues in Healthcare Information Technology course
Intelligent Agent for Playing Pente, Artificial Intelligence course
top
Meditech (2001 – 2003)
Description
Meditech is a leading software vendor in the health care informatics industry, providing integrated software solutions to meet the information needs of health care organizations worldwide, including physician practices, clinics, hospitals, long-term care facilities, home health agencies, and behavioral health facilities.
Role
I was responsible for providing software service support, and coordinating software updates for administrative applications, including admissions (ADT), medical records and abstracting modules, for over sixty hospitals in the United States and Canada. I also developed custom routines for aggregating financial, clinical and demographic data to meet the regulatory requirements of the Iowa Hospital Association, as well as coordinated and managed a team of clients, management and technical staff in an initiative to address client system requirements for adhering with HIPAA regulations.
top