2006 Scientific Session Abstracts
Scientific Session Presentations & Schedule
E-Poster Presentations
Submission Compliance and Guidelines
Copyright Transfer Form (PDF)
Abstract Archive
SPiDR: A Shared Pathology Data Repository using a Service Oriented Architecture for Mission Critical Lab Results Delivery
Mark J. Routbort, MD, PhD (mark.routbort@mdanderson.org) James Fleming, Trey Elliot, Sanjivkumar Dave, Martha Thomas, Catherine Price, and Michael Riben, MD , Division of Pathology and Laboratory Medicine, Section of Laboratory Informatics, University of Texas MD Anderson Cancer Center, Houston, TX
Context: The proliferation of specialized information systems along with legacy databases complicates the development, deployment, validation, and maintenance of interfaces to enterprise clinical and research applications that desire access to laboratory results in order to support their workflow processes. Challenges include system specific HL7 implementations, interface building skill-set acquisition, and system life-cycle management and replacement, all within a context of continuing to deliver mission critical results to the enterprise. A growing demand for interfaces, a need to sunset a unsupportable legacy datastore, and an aggressive lab system replacement schedule for 2 core systems mandated a new strategy to make connections transparent and independent of source laboratory systems.
Technology: The core technologies underlying SpiDR include:
-Back-end data harvesting to the SQL Server 2000 datastore via HL7 listener/parser or transactional replication, depending on the source system
-Data modeling and schema provisioning (Altova XMLSpy)
-Provision of a middle-ware data services layer for data delivery though Web Services. The SPiDR web services are written in VB.NET for .NET 2.0
Design: The SPiDR core data layer uses several discrete databases operating in a federated design, which are then unified transparently into a single repository via the SPiDR web service. We present an open core schema for delivery of laboratory data (LabData schema) which attempts to equitably balance “consumability” with completeness. Through a single LabDataRequest schema, consumers can request clinically-oriented transactional information (“all lab data on patient X”), or perform test-focused lookups (“find the set of patients with serum glucose > 250 within the last 3 months).
Results We have loaded general laboratory data from 1973-2005, microbiology data from 1989-2005, and blood banking transfusion history data from 1985-2005. Since Dec. 2005, we have been populating SPiDR in real time with all general laboratory and microbiology data. As of Mid-May, a total of 466,177 patients (unique medical record numbers), with 2,892,349 microbiology results and 116,541,239 clinical laboratory results have been successfully loaded into SPiDR and are now available through a published web service for consumption.
Conclusion: Within a large enterprise, with multiple data customers, we have simplified the data validation, presentation, maintenance process by developing an operational datastore with a unified Web services entry point. Other laboratories may find our model useful to follow when trying to provide similar services to their data customers.