OSU Medical Center Information Warehouse

The OSUMC Information Warehouse (IW), which is a department within MedCenter IS, is a shared service that provides data integration and management, text and data mining, software training, and warehouse-oriented application development services to the entire Medical Center community. The IW compiles data from many disparate information systems throughout OSUMC (Figure 1), including patient management, billing and finance, procedures, medications, laboratory results, clinical reports, physician order entry, outcomes and demographics. In addition, data from several external sources, including regional and national health industry statistics and geographic information, are regularly incorporated into the IW.

The IW has been recognized as a core asset for facilitating translational research and advancing personalized healthcare, and as such its Research Informatics group has been focused on meeting the unique needs of the research community. In 2006, the OSU Institutional Review Board (IRB) approved a protocol recognizing the IW as an “honest broker” of clinical data. This Honest Broker protocol is a process that streamlines research access to clinical data while safeguarding patient privacy. It enables the IW to provide de-identified, limited and coded data for research purposes without requiring investigators to design initial formalized research protocols. This is particularly helpful in facilitating preliminary studies and helps reduce barriers to the exploration of higher risk hypotheses that might not otherwise be investigated. To date the Honest Broker protocol encompasses a limited subset of the entire IW clinical repository. We are exploring ways to expand the protocol to include processes that will enable safe use of new data types, such as clinical free-text and many types of medical imagery.

Some of the ongoing research-oriented projects within the IW are:

Research Data Portals

• TransChart is an operational data capturing, querying and analyzing system developed within the OSUMC transplant department for the purposes of patient care, quality assurance, reporting, etc. The IW has integrated the data from this system with other datamarts to provide more powerful and comprehensive data query and analysis capabilities.
• The Clinical Trials Minimum Dataset (CTMDS) Portal is a central data repository and communications channel for facilitating early determinations of billing requirements for clinical trials.
• The Proteomics Data Analysis Portal integrates disparate data sources to improve the workflow of electrophoresis laboratory procedures through the incorporation of analytical capabilities and historical data retrieval. Work is currently being done to determine if similar processes and techniques may also be applicable to other tests that are performed in the Special Functions Laboratory.
• The Comprehensive Wound Center Portal has adapted and integrated a suite of tools developed by the caBIG initiative to support tissue banking and analysis. The caTissue Core, caArray, caTissue Clinical Annotation Engine (CAE) and caWorkBench tools are available, and have been integrated with the IW clinical repository, allowing for the correlation of gene expression analysis with clinical phenotypes.
• The Statistical Tools Server (STS) will offer grid-based access to statistical packages such as SAS and R. The STS is a high performance system (Xeon class) that has been deployed specifically to support statistical analysis over large datasets within the IW.

Text Mining and Information Retrieval

• Oracle Text has been integrated with the UMLS Metathesaurus to provide physicians and researchers with the ability to conceptually search free-text medical reports.
• A heuristic Negation Detection Algorithm has been implemented to improve the quality of free-text medical report searches by highlighting and/or excluding negative expressions (e.g., ‘rule out’, ‘no history of’).
• The ICD9 Code Extraction leverages both text indexes created during the process of loading free-text clinical reports to enable keyword searches and conceptual search capabilities that utilize the UMLS Metathesaurus to retrieve medical record information for coding purposes through a Web interface. With this tool, patient populations can be searched and evaluated through queries including data elements such as date of discharge, medical record number, keywords, length of stay, etc.
  
  
• The PACS Image Integration Project, developed in collaboration between the IW and BMI, facilitates the extraction of medical images and associated clinical free-text reports and structured data elements (e.g., ICD9 diagnosis codes). The integration of web-based tools for visualizing and interacting with the images with the IW clinical repository and mechanisms for conceptually searching the text reports and diagnosis codes enables researchers to select image subsets by using a vast array of clinical properties, thus promoting detailed data mining and analysis.
  
• The Tissue Image/Pathology report retrieval/de-identification project provides a linkage between pathology images and their corresponding text reports. The IW provides support for handling a variety of imagery spanning from virtual microscopy slides to radiological studies such as CT, MRI, etc., as well as de-identification mechanisms.

Data Mining tools have been applied in the domains of clinical and basic research, including correlating genotype and phenotype data, predicting the effectiveness of medical procedures, tests and medications, and discovering relationships among clinical and pathological data. Current/planned mining activities include:

• SNP – Laboratory data correlation
• Laboratory test batteries
• Order sets
• Medications
• Outcome prediction

Grid Technology

• Effort is underway to adopt data grid technology from the NCI sponsored caBIG initiative. In collaboration with BMI, the IW is working to pilot projects that employ grid infrastructure in the form of both grid services and grid data in the context of an operational health care environment.
• To support both adaptability and standardization, methods have been developed for ontology-based typing and semi-automated mapping of IW data. The aims of this effort are to support the development of viable type standards, rigorously exercise the grid infrastructure in a complex real-world environment, and ultimately deploy production grade grid infrastructure across the enterprise.

For more information: Please contact Jyoti Kamal (jyoti.kamal@osumc.edu)

Project Researchers

Herb Smaltz (MedCenter IS CIO)
Jyoti Kamal, Ph.D. (Director)
Philip Payne, Ph.D (Translational Research Architect)

Project Publications