Cross-Institutional Imaging Aggregation and Distribution
 
Authors:
Alberto F. Goldszal, PhD, MBA, University Radiology Group P.C.; Murray D. Becker, MD, PhD; Robert E. Epstein, MD; David M. Walor, MD; Mitchel L. Simon, MD
 
Background:
In our previous work[1], we described and evaluated a newly developed imaging exchange in which DICOM imaging data acquired at multiple (and unaffiliated) institutions were aggregated under a single patient-centric imaging database. Specifically, we discussed the implementation of a patient-centric medical imaging repository that collects diagnostic imaging studies from several healthcare facilities in the central New Jersey region, including six hospitals and eight imaging centers.

At the core of our medical imaging database, we have developed a workflow engine that interconnects, via HL7, with the hospitals ordering systems (in general, a RIS or HIS), in order to obtain bonafide patient demographic information. As a consequence of this standard-based interfacing with regional hospitals, we are also able to fetch, via DICOM, new and prior imaging studies from the acquisition site and “intelligently” pre-fetch related prior imaging exams from other regional, unaffiliated healthcare facilities. We are able to match patient records across distinct healthcare organizations through the implementation of a dynamic Master Patient Index (or d-MPI) – an algorithm that performs a probabilistic matching of patient data scattered across healthcare institutions in the absence of a common (and unique) healthcare identifier.

Once the data has been matched, it can be grouped into a single, patient-centric database for storage and distribution. This represents a major departure from traditional methods that rely on a single PACS and a common identifier (e.g., an institutional medical record number), in order to serve patient data. Methods relying on local medical record numbers fail to aggregate data across institutions and, as a consequence, only provide access to a limited set of imaging history. We believe the lack of complete patient history can lead to duplicative and often unnecessary exams or lower quality diagnoses[2, 3].
 
Evaluation:
In this work, we seek to enable real world application of diagnostic imaging data, such as secure storage and online access (anytime, anywhere) of longitudinal, geographically-distributed patient imaging data at the point-of-care by radiologists (i.e., access to new and relevant prior imaging studies for comparison, regardless of acquisition site), and clinicians (e.g., delivery of all patient’s diagnostic imaging history to primary care physicians). This imaging database can also be used by the patients themselves as a mechanism to permanently store, access, and distribute all their digital imaging studies.

We evaluate the scalability of our regional imaging exchange beyond its initial goals[1], as well as the behavior of our matching algorithm used to identify patient data across institutions with different medical record numbers and different PACS products. Furthermore, we evaluate the impact of imaging distribution from a single source, to referring physicians (as opposed to accessing piece-meal data scattered among multiple PACS databases). Ultimately, we hope to extend the imaging archival, aggregation, and distribution model to serve the patients directly.
 
Discussion:
Our previous research[1,4,5,6] has shown that up to 20% of imaging cases have relevant prior images (i.e., same anatomy, any modality) archived in a 3rd-party PACS [Fig. 1]. Often, most of these priors are not available for comparison at the point of interpretation because they are archived outside the radiologist’s own environment. Therefore, for those radiologists (or radiology groups) that serve multiple (and often unaffiliated) healthcare facilities, the only mechanism available to obtain a comprehensive view of the patient’s total imaging history is through the use of disparate, multiple hospital-centric RIS and PACS systems. This model can, obviously, severely limit radiologist productivity and efficiency and, many times, is simply impractical.

Figure 1
Figure 1: Cross-institutional overlap of imaging data (t = 4 months [Feb-May 2009], n = 200,000 imaging exams). Map shows degree of imaging data overlap across selected regional hospitals and imaging centers as a function of geographical distance.

We have developed and implemented a patient-centric imaging database of radiological studies acquired at multiple, unaffiliated healthcare organizations. The consolidation of patient records is possible due to the development and implementation of HL7-based[7] orders-results interfaces or an HL7-based registration/ADT interfaces between our system and the radiology information systems (or equivalent order-entry systems) of the imaging acquisition sites. Based on the patient demographic information (e.g., name, DOB, gender, SSN, address, etc.) available in the “order” or “registration” stream drawn from the RIS at each facility, we are able to automatically fetch (i.e., perform a DICOM query and retrieve or DICOM C-MOVE[8]) the corresponding new and old digital diagnostic imaging studies stored in the corresponding hospital’s PACS. Furthermore, we link the patient’s identity across all healthcare facilities we connect with, using probabilistic matching algorithms, enabling an all-encompassing access to relevant prior images stored across regional healthcare organizations.

 
Conclusion:
In today’s healthcare delivery system, hospitals act as isolated, non-integrated silos retaining patients’ clinical information within its own borders and, therefore, rendering the delivery of these results to a limited subset of clinicians within the hospital’s boundaries. As a consequence, physicians are forced to deliver health services with limited, and often incomplete, clinical information. Furthermore, lack of access to prior results located at other healthcare organizations lead to wasteful and unnecessary duplication of medical exams[9]. With the development of a patient-centric, longitudinal, and scalable imaging database, we hope to overcome today’s limitation and deliver the entire patient’s imaging history, regardless of imaging acquisition site, within a single easy-to-use and easy-to-access platform. Our intent is to promote better and faster access to quality medical data while reducing duplication of efforts and contributing to the overall decrease of healthcare costs.
 
References:
[1] Goldszal AF, Epstein RE, Becker MD. Reporting-Driven Workflow Orchestration and Regional Imaging Exchange. Annual Meeting of the Society for Imaging Informatics in Medicine. Charlotte, NC. June 2009.

[2] White K, Berbaum K, Smith WL. The role of previous radiographs and reports in the interpretation of current radiographs. Invest Radiol. 1994;29:263-5.

[3] Aideyan UO, Berbaum K, Smith WL. Influence of prior radiologic information on the interpretation of radiographic examinations. Acad Radiol. 1995;2:205-8.

[4] Lakhani P, Menschik ED, Goldszal AF, Murray JP, Weiner MG, Langlotz CP. Development and Validation of Queries Using Structured Query Language (SQL) to Determine the Utilization of Comparison Imaging in Radiology Reports Stored on PACS. J Digital Imaging. March 2006;Vol.19:1:52-68.

[5] Murray J, Menschik ED, Goldszal AF, Nagy PG, Flanders AE, Langlotz CP. Open Standards for Radiology RHIOs: IHE XDS-I and the Philadelphia Health Information Exchange. Presented at the 92nd Scientific Assembly and Annual Meeting of the Radiological Society of North America. Chicago, IL. November 2006.

[6] Menschik ED, Small JD, Murray JP, Goldszal AF, Horii SC. Crossinstitutional linkage of patient imaging data in the absence of a common identification system, Proceedings of the 21st Annual Meeting of the Society for Computer Application in Radiology. Vancouver, British Columbia, Canada. May 2004.

[7] http://www.hl7.org/

[8] ftp://medical.nema.org/medical/dicom

[9] Sung JC, Sodickson A, Ledbetter S. Outside CT Imaging Among Emergency Department Transfer Patients. Journal of the American College of Radiology. September 2009;Vol.6:9:626-632.