Research Master Subject Index (RMSI)
Description:
RMSI is a subproject of project Sentinel (http://sentinel.georgetown.edu) that uses WS/PIDS to manage identity domains for the Center for Clinical Imaging Research.
A Research Master Subject Index (RMSI) is analogous to an enterprise master patient index (EMPI) in that it links multiple identifiers across an enterprise; however an RMSI must also deal with the very complex issues of isolating subject populations in clinical trials and maintaining proper anonymization. In many institutions the research domain is maintained separately from the clinical domain; but in certain research programs it may be necessary to bridge the two domains to permit a researcher to access segments of a subject’s electronic medical record. The capability to correlate across multiple independent ID domains is an essential function of an RMSI.
The Person Identification Service (PIDS) [1] standard addresses the capability to handle identities in a consistent way across the healthcare continuum and coordinate and correlate identifiers between independent domains. The original PIDS specification is defined using the Interface Definition Language (IDL) of CORBA [2, 3]. It followed a rigorous design methodology based on real world requirements and validation. As part of its design, PIDS has native support for federation of the coordination and correlation of identifiers. Through Project Sentinel [4] Georgetown University has developed a Web Services-based PIDS (WS/PIDS) [5] that is uniquely suited to solving the RMSI problem. Equivalent of the CORBA IDL specified PIDS, WS/PIDS recasts the original interface specification into Web Services, preserves the semantics of original PIDS specification and extends the standard by adding extensive support for medical multimedia attributes [6].
A WS/PIDS-based RMSI has been created to support the unique research imaging environment provided by the Center for Clinical Imaging Research (CCIR) at Washington University in St. Louis. The CCIR merges state-of-the-art imaging technologies and a comprehensive IT infrastructure designed to manage clinical and translational research programs and trials in isolation from the normal clinical routine.
The primary function of the WS/PIDS-based RMSI is to provide an independent ID domain for each clinical trial supported by the CCIR. To properly manage protected health information (PHI) within a multi-use imaging research facility it is essential that the subject population of each trial or research project be maintained in isolation and that the subjects be anonymized once the imaging study is completed. Only the principal investigator (PI) responsible for the trial or project may break the anonymization. On the other hand, certain safety issues (e.g. a subject has inadvertently volunteered for too many PET studies within a specified time period) require data checks across all instances of the same subject. Similarly, once trial data have been distributed to the project personnel, a copy will be stored in a long term repository where for future data mining purposes all information for a given subject will be linked regardless of source. Finally, for some research projects it is necessary to extract either current or historical information from the research subject’s clinical medical record (with IRB approval and subject consent). In these cases it is necessary to correlate between the study specific research ID for that subject and the enterprise ID used by the BJC HealthCare system and Washington University School of Medicine.
As illustrated in Figure 1, the CCIR Correlation Manager links the clinical and research ID domains at Washington University and is the point for future cross-institution federation. The CCIR Correlation Manager interacts with both a Research Correlation Manager and a Clinical ID Manager. The Clinical ID Manager is a limited WS/PIDS implementation which only supports the Identify Person interface and serves as a wrapper for the legacy WU/BJC Clinical Enterprise Master Patient Index (EMPI). An existing XML based interface is used by the Clinical ID Manager to query the EMPI.
Figure 1. RMSI System Design Concept illustrating multiple research ID domains that are correlated within the research domain which in turn is correlated across the enterprise with the clinical ID domain.
The Research Correlation Manager correlates across multiple Research ID domains in which each domain is supported by its own Research ID Manager. Typically, human subjects are identified to the CCIR staff in order to assure the proper protocol is followed. For some sensitive protocols the human subject may only be identified to the study coordinator or PI who must accompany the subject. In these sensitive cases the subject will be known within CCIR by an alias. In all situations, once data have been collected in the CCIR it will be anonymized prior to distribution, analysis, or persistent storage. For these reasons the CCIR has been divided into two domains: identified and anonymized. In some instances the identified domain will be partially anonymized by use of aliases. We have chosen to represent the same subject with different identifiers in the identified and anonymized domains to prevent someone from taking an identifier from the anonymized domain and directly querying a system in the identified domain. Thus, each defined research project or clinical trial utilizes two ID domains and is supported by a pair of Research ID Managers. A web-based registration system allows the user to select the ID domain pair for the particular trial and identify a research subject by input of subject demographics, identification profile and trial specific ID.
The problem of cross-referencing multiple ID domains has been addressed by the IHE initiative. IHE has defined a Patient Identifier Cross-Reference (PIX) profile to support mapping identifiers across independent identification domains. This is supplemented by the Patient Demographics Query (PDQ) integration profile. Together, PIX and PDQ support most of the functionality of PIDS although the IHE model does not explicitly deal with federation of ID services between enterprises. Within our implementation we have implemented a subset of PIX/PDQ as a “connector” to link the CCIR Research Management System (based on a commercial RIS) to the RMSI. This allows, in a limited way, a direct comparison of the two technologies. WS/PIDS is more readily integrated into web-based applications using its SOAP-based web services. It inherently supports cross enterprise federation and offers low level methods for ID deprecation not found in PIX.
References:
- Object Management Group. (2001, Apr) Person Identification Service.
- Vasilescu E, Mun, Dorobantu M, Govoni, S, Gattewar P, Mun SK. “WS/PIDS:Standard Interoperable PIDS in Web Services Environments. Resubmitted with revisions for publication in IEEE Transactions on Information Technology in Biomedicine in May 2006.
- Object Management Group. CORBA.
- J. Siegel, “OMG overview: CORBA and the OMA in enterprise computing,” Commun. ACM, vol.41, no.110, pp.37-43, 1998.
- Applications of Advanced Network Infrastructure in Health and Disaster Management: Project Sentinel Collaboratory supported by contract N01-LM-3-3506 from the National Library of Medicine.
- E. Vasilescu, M. Dorobanţu, S. Govoni, S. Padh, “Representing Traits in WS/PIDS”, Distributed Diagnosis and Home Healthcare, 2006. D2H2. 1st Transdisciplinary Conference on Volume, Issue, April 2-4, 2006 Page(s): 152 - 155 Digital Object Identifier 10.1109/DDHH.2006.1624819