Representing and sharing knowledge using SNOMED Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008) R. Cornet, K.A. Spackman (Eds) Using SNOMED CT as a Mediation Terminology: Mapping Issues, Lessons Learned, and Next Steps Toward Achieving Semantic Interoperability Sarah Maulden, MD, MS1, Patty Greim, RN, MS1, Omar Bouhaddou, PhD1,2, Pradnya Warnekar, RPh, MS1,3, Laura Megas4, Fola Parrish, PharmD5, Michael J. Lincoln, MD1,6 1 Department of Veterans Affairs, Veterans Health Administration, Chief Health Informatics Office, Salt Lake City, UT 2 Electronic Data Systems, Plano, TX 3 dNovus RDI, San Antonio, TX 4 Northrop Grumman Corporation, Chantilly, VA 5 Department of Defense, Tricare Management Activity, Falls Church, VA 6 Department of Biomedical Informatics, University of Utah, Salt Lake City, UT sarah.maulden@va.gov, patricia.greim@va.gov ABSTRACT through secure and standard-based data exchange1. When clinicians exchange data, interoperable The Clinical Data Repository / Health Data meaning is possible because clinicians share Repository (CHDR) project is a combined effort of structures of clinical practice and familiar clinical the Department of Veterans Affairs (VA) and the language2. Similarly, meaningful electronic data Department of Defense (DoD) to exchange clinical exchange requires a shared structure for transmission information between our Electronic Health Records and a common electronic vocabulary3, which yields (EHR). CHDR exchanges standardized, computable Computable Semantic Interoperability (CSI)4. CSI data, as opposed to textual data that is only human makes order checks and electronic alerts possible readable. CHDR utilizes mediation terminologies for across institutions, and is an essential component of a health data exchange. For allergy reactions data, longitudinal EHR that protects patient safety. CHDR uses SNOMED CT in conformance with Health Information Technology Standardization The CHDR project is a Congressionally-mandated, Panel (HITSP) recommendations. This paper reports combined effort which aims to exchange how we implemented this solution. standardized, computable data, as opposed to textual data that is only human readable. Computable data Business rules for mapping allergy reactions were exchange enables “semantic interoperability” and established jointly. Each agency independently permits utilization of electronic decision support mapped its legacy data to the same version of tools on the sum of local and remote data at either SNOMED CT. CHDR has since been implemented in agency6. CHDR currently exchanges pharmacy and seven locations where VA and DoD have joint allergy data elements and the agencies are working to patient care environments. Statistics on actual patient share laboratory data elements by the end of fiscal data from February-June 2007 showed a 74-99% year 2008. mediation success rate for allergy reactions data. CHDR has informed the Health IT Standards Panel Examination of mediation failures exposed issues (HITSP) that designates interoperability standards for related to mapping and SNOMED CT concept EHRs. VA and DoD use different internal data modeling. In addition, we emphasize the significance standards for allergies, and under CHDR utilize a of adherence to a detailed terminology mediation common, HITSP-specified mediation terminology. strategy, desirability of a standard SNOMED CT- CHDR exchanges pharmacy, drug allergens, and based subset for allergy reactions, and the creation of allergy reactions, and will soon exchange laboratory this subset for publication and distribution. (chemistry/hematology) data. CHDR exchange of comprehensive pharmacy information7 and drug INTRODUCTION allergy reactant information8 have been well described. The President has ordered Federal agencies to promote improved healthcare quality and efficiency 85 Representing and sharing knowledge using SNOMED Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008) R. Cornet, K.A. Spackman (Eds) The CHDR strategy for exchange of allergy reactions Business rules for mapping allergy reaction legacy (signs and symptoms) data uses SNOMED CT, in terms to SNOMED CT concepts were developed conformance with Consolidated Health Informatics jointly12. For example, SNOMED CT hierarchies (CHI) and now HITSP recommendations. We now were prioritized in order of preference for mapping as report how VA and DoD have used SNOMED CT follows: 1) Findings, 2) Disorders, 3) Morphologic successfully as a mediation terminology, and describe abnormality, 4) Observable entity, 5) Context the results. Dependent Category. Mappings from specific to more general terms (and vice versa) were avoided, METHODS because of the bidirectional nature of the data exchange. For instance, mapping “nasal burning” to Initial work for allergy reactions under the mediation “burning sensation of mucous membrane (finding)” approach included the commitment at each agency to creates either a loss of the clinical detail “nasal” normalize legacy terms, using a list of centrally when translated (for an outbound message), or forces maintained concept terms9. Allergy reactions were the translation of a general term “mucous membrane” comprised primarily of signs and symptoms, but to a specific one--“nasal”--(for an inbound message). could also include disorders or clinical conditions Local terms not found in SNOMED CT were attributable to exposure to a drug reactant. Each collected for potential submission to the SNOMED agency mapped its legacy allergy reactions data to development organization. Other mapping rules SNOMED CT10. The four-part terminology governed misspellings, qualifiers, synonyms, mediation strategy was outlined as follows11: ambiguous terms, and outdated terms. 1. Select a mediation terminology compliant Table 2 shows a sample of VA allergy reaction terms with CHI/HITSP standards (if possible). with their VA unique identifiers (VUIDs) and 2. Map each agency’s terms to concepts within SNOMED CT mappings. the mediation standard. 3. Exchange the mediation codes. Once mapping rules were established, terminologists 4. Coordinate content maintenance plans. at each agency manually mapped allergy reaction terms to SNOMED CT. VA used Apelon’s TermWorks tool and SNOMED’s CliniClue® Table 1 shows the CHI standard terminologies and browser, and DoD used the Terminology Service releases designated for the four domains at the start Bureau (TSB) and the CliniClue® browser. of the CHDR project. Table 1. CHDR Domains and Designated Standards. Domain Mediation Terminology (CHI Standard) Pharmacy RxNorm Jun 2005 Drug Allergens UMLS Jan 2005AA Allergy Reactions SNOMED CT Jan 2005 Lab (Chemistry & Hematology) LOINC 2.14 Jan 2005 Table 2. VA Unique Identifiers, Allergy Reaction Text, and Corresponding SNOMED CT Mappings. VUID VUIDText SNOMED CT ID SNOMED CT Text Blister of skin AND/OR mucosa 4637123 BLISTER 339008 (finding) Postural drop in blood pressure 4543527 ORTHOSTASIS 271648003 (finding) 4696326 ASEPTIC NECROSIS OF BONE 398199007 Aseptic necrosis of bone (disorder) 4538635 RASH 271807003 Eruption of skin (disorder) 4538640 SEIZURES 91175000 Seizure (finding) 4539274 NOSEBLEED 249366005 Bleeding from nose (finding) 2 86 Representing and sharing knowledge using SNOMED Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008) R. Cornet, K.A. Spackman (Eds) For mapping validation of allergies terms (both to translate from the mediation terminology to its reactions and reactants), two reviewers conducted native vocabulary without loss of meaning15.” three separate reviews (10 hours each for a total of 60 experts’ hours) and identified various discrepancies Mediation success rates are calculated by multiplying in about 5% of the total number of terms. All the translation success rates of each agency. When discrepancies were corrected13. An independent coded mediation fails, the CHDR project exchanges review of concepts common to both agencies was allergy reaction data as text without a mediation performed to ensure accurate translation and code. calculate expected mediation success rates14. See Table 3. RESULTS Terminology “translation” and “mediation” are Terminology translation and mediation statistics were described as follows by Bouhaddou et al.: compiled for allergy reactions data during a 5-month period in 2007. The numbers of translation and “The mediation success rate defines the percentage of mediation attempts fluctuated from month to month, data in one system that is understood and computable but generally showed an increasing trend as the by the other system. For each direction of the data project was implemented at additional sites over the exchange, inbound or outbound, there is a different 5-month timeframe. Table 4 shows translation and mediation success rate. For mediation to succeed, two mediation success rates for allergy reactions sent translations have to be successful. First, the source from VA to DoD. Table 5 shows statistics for allergy agency has to translate from its vocabulary to the reactions sent from DoD to VA. Overall, mediation mediation terminology. Then, the target agency has success rates varied from 74% to 99%. Table 3. Common and Unique Allergy Reaction Concepts Determined by Each Agency Mapping to SNOMED CT. Agency Total Common Terms Mapped Terms Unique to Each Agency Unmapped Terms VA 346 299 25 (7%) 22 (6%) DoD 456 299 47 (13%) 110 (24%) Table 4. VA-to-DoD Mediation Statistics for Allergy Reactions, Feb-June 2007.* VA-to-DoD February March April May June Total VA-to-SNOMED CT translation attempts 168 193 338 959 502 Translation failures (VA-to-SNOMED CT) 4 0 1 13 1 Total VA allergy reactions sent to DoD 164 193 337 946 501 Translation Success Rate: VA-to-SNOMED CT 98% 100% 100% 99% 100% Total allergy reactions received by DoD 164 193 337 946 501 Translation failures (SNOMED CT-to-DoD) 17 17 34 121 5 † Total VA allergy reactions sent to DoD CDR 147 176 303 825 496 Translation Success Rate: SNOMED CT-to- DoD 90% 91% 90% 87% 99% MEDIATION SUCCESS RATE: VA-to-DoD 88% 91% 90% 86% 99% *Yellow areas designate translation services performed by VA. White areas designate translation services performed † by DoD. CDR=Clinical Data Repository. 3 87 Representing and sharing knowledge using SNOMED Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008) R. Cornet, K.A. Spackman (Eds) Table 5. DoD-to-VA Mediation Statistics for Allergy Reactions, Feb-June 2007.* DoD-to-VA February March April May June Total DoD-to-SNOMED CT translation attempts 1,509 1,788 2,025 3,521 4,030 Translation failures (DoD-to-SNOMED CT) 306 467 432 432 107 Total allergy reactions sent to VA 1,203 1,321 1,593 3,089 3,923 Translation Success Rate: DoD-to-SNOMED CT 80% 74% 79% 88% 97% Total allergy reactions received by VA 1,203 1,321 1,593 3,089 3,923 Translation failures (SNOMED CT-to-VA) 1 0 8 11 69 † Total DoD allergy reactions sent to VA HDR 1,202 1,321 1,585 3,078 3,854 Translation success rate: SNOMED CT-to- VA 100% 100% 99% 100% 98% MEDIATION SUCCESS RATE: DoD-to-VA 80% 74% 78% 87% 96% *Yellow areas designate translation services performed by VA. White areas designate translation services performed † by DoD. HDR=Health Data Repository. Analysis of the causes of the mediation failures concept “systemic disease” was used at one revealed the following issues, listed in order of agency, but the other agency felt this term frequency of occurrence: added no valuable information about an allergic reaction and did not include it in its 1. SNOMED CT concept modeling issues were list of selectable reactions for use by exposed. For example, a search for providers. “nosebleed” in SNOMED CT’s CliniClue® 5. Divergent approaches to SNOMED mapping browser returns more than one option within existed between VA and DoD, despite the “finding” hierarchy: “bleeding from shared business rules. For instance, nose” vs. “nosebleed/epistaxis symptom.” “hypertension” was mapped to “finding of Another example of a modeling issue: the increased blood pressure (finding)” at one “Situation with Explicit Context” hierarchy agency, and to “Hypertensive disorder, was not addressed in the original VA/DoD systemic arterial (disorder)” at the other. mapping rules, as this hierarchy evolved within SNOMED CT after the initiation of DISCUSSION the mapping. 2. New legitimate allergy reaction terms were We begin with a list of lessons learned. added independently within each agency, which led to mediation failures in the time 1. Mapping rules must always be tailored to the interval between synchronization and specific purpose of the mapping. These rules may be updating of each agency’s files. influenced by non-terminological issues, such as the 3. Maintenance and versioning issues emerged potential for the entire message to fail if one when SNOMED CT released new versions component fails. We must recognize that mappings with new concept statuses (e.g., are often purpose- or use case-driven, as well as built “erroneous”, “limited”, “duplicate”, by semantic nuances of context. “ambiguous”) during the project. If agency updates were not synchronized, mediation 2. Even with established rules in place, there is a failures would result. clear need for continued communication between 4. Allergy reaction concepts and terms were agencies. We were unable to discern any major sometimes deemed appropriate by one consistent reason for the mapping rule violations. agency but not the other. For example, the One possibility is that VA and DoD initially used 4 88 Representing and sharing knowledge using SNOMED Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008) R. Cornet, K.A. Spackman (Eds) different mapping tools. Another is that the process within the Problem List subset, use of the Problem of finding the correct map for a term is variable and List subset to record allergy reactions (signs and influenced by syntax and linguistic features of the symptoms) may prove problematic, as is the case search engine. In several cases, the Clue browser whenever data is used for a purpose other than that yields an apparently correct result (for example, a originally intended. Consider the terms “circumoral search for “orthostatic hypotension” returns paresthesia (finding)” and “edema of pharynx “orthostatic hypotension (disorder)”) but the term is (disorder).” These terms are appropriately found located in the disorder hierarchy, rather than the within the VA/DoD Allergy Reactions subset, but not findings hierarchy (to be used in preference if within the VA/KP Problem List subset. The sheer possible). It may not be immediately apparent that an size and complexity of the Problem List subset, alternative mapping exists (“postural drop in blood compared to that of the Allergy Reactions subset, pressure (finding)”) in another hierarchy. The clinical may unnecessarily complicate data entry for knowledge, background, and familiarity with providers and result in unwanted entry of SNOMED hierarchies and features of CliniClue® inappropriate terms as Allergy Reactions. The also are likely to influence search results. Ideally, a smaller subset could enable more precise data common team, process, and toolset would be used to constraints and greater computing speed, without produce the mapping. Perhaps the mapping could sacrificing data integrity. Communication with become a service of the Standards Development HITSP is ongoing regarding this issue. We propose Organization, as is the case with RxNorm. that a new study be undertaken to evaluate the VA/KP Problem List and compare it to the VA/DoD Allergy Reaction subset, documenting content gaps, 3. SNOMED CT modeling issues were probably the areas of overlap, and suitability for use as a most difficult to address, as these require a mediation terminology. sophisticated knowledge of concept modeling and of the evolution of SNOMED hierarchies over time. In conclusion, we point out that the expense of mapping VA’s and DoD’s legacy terms (and 4. Maintenance plans for using mediation maintenance of same) was relatively substantial— terminologies need to include specific plans for even for the limited list of Allergy Reactions. As synchronizing updates to the standard reference CHDR expands to include more VA and DoD sites, terminology, in this case SNOMED CT, and also for the terminology maintenance requirements will synchronizing updates to each agency’s mapping file. continue. A significant outcome of this project is the generation Adopting the HITSP standard internally as a of a new, unique SNOMED CT subset specific for representation for allergies and reactions would be a Allergy Reactions (signs and symptoms) which could more efficient method of working toward true potentially be submitted for inclusion in SNOMED semantic interoperability. Using a phased approach, CT as an official subset. It could also be published legacy terms can be mapped to the standard, and shared among federal agencies and non-federal presented for adoption by the Standards Development partners. Organization (SDO), and eventually migrated to the standard representation itself with deprecation of In December 2007, HITSP designated the VA/Kaiser invalid legacy terms. Permanente (KP) Problem List subset (16,430 entries) as the recommended standard for allergy The use of mediation terminologies for computable reactions, a departure from previous CHI data exchange is a dynamic and evolving process. It recommendations to use the VA/DoD Allergy is prone to pitfalls, but is an effective, practical Reactions subset (864 entries)16. While many of the method for advancing the goal of semantic VA/DoD Allergy Reactions terms are contained interoperability. 5 89 Representing and sharing knowledge using SNOMED Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008) R. Cornet, K.A. Spackman (Eds) REFERENCES Affairs and the Department of Defense (DoD). AMIA Annu Symp Proc. 2007;2007:781-785. 1. Bush GW. Executive order: Promoting quality and 9. Mandel J. CHDR Fact Sheet: Terminology efficient health care in federal government Mediation vs. Common Terminology. Dept of administered or sponsored health care programs Veterans Affairs internal communication. 2007 Jan [Internet]. Washington, D.C.: Office of the Press 12. Secretary; 2006 Aug 22 [cited 2008 Jan 28]. 10. Ibid. Available from 11. Bouhaddou, Omar. CHDR and Terminology http://www.whitehouse.gov/news/releases/2006/08/2 Mediation Services. Dept of Veterans Affairs internal 0060822-2.html. communication. 2006 Sep 7. 2. Mead CN. Data interchange standards in 12. Bouhaddou O, Warnekar P, Parrish F, et al. healthcare IT-Computable semantic interoperability Exchange of Computable Patient Data Between the now possible but still difficult, do we really need a Department of Veterans Affairs (VA) and the better mousetrap? Journal of Healthcare Information Department of Defense (DoD): Terminology Management 2006 Winter; 20(1): 71-78. Mediation Strategy. Journal of the American Medical 3. Ibid. Informatics Association (JAMIA) Vol. 15 No. 2, 4. Ibid. Mar/Apr 2008. 5. CHDR Training Home Page [Internet]. 13. Ibid. Brecksville, OH: VA Learning University Office of 14. Ibid. Information National Training and Education Office; 15. Ibid. last updated 2007 Jan 4 [cited 2008 Jan 28]. 16. ANSI Public Document Library [Internet]. Available from: Washington, D.C.: American National Standards http://vaww.vistau.med.va.gov/VistaU/chdr/default.ht Institute. HITSP Summary Documents Using HL7 m. Continuity of Care Document (CCD) Component, 6. Insley M. Summary of VA/DoD v2.1; 2007 Dec 13 [cited 10 Mar 2008]; p. 49. sharing/interoperability initiatives. Dept of Veterans Available from: Affairs internal communication. 2007 May 17. http://publicaa.ansi.org/sites/apdl/Documents/Standar 7. Parrish F., Do N, Bouhaddou O, Warnekar P. ds%20Activities/Healthcare%20Informatics%20Tech Implementation of RxNorm as a terminology nology%20Standards%20Panel/Interoperability%20S mediation standard for exchanging pharmacy pecification/IS- medication between federal agencies. AMIA Annu Released%20for%20Imp.%20and%20Recognized/IS Symp Proc. 2006;2006:1057. 03%20- 8. Warnekar P, Bouhaddou O, Parrish F, et.al.Use of %20Consumer%20Empowerment/HITSP_V2.1_200 RxNorm to exchange codified drug allergy 7_C32%20- information between the Department of Veterans %20Summary%20Documents%20Using%20CCD.pd f. 6 90