Address for correspondence Louise Pape-Haugaard, email: lph@hst.aau.dk Acknowledgement Authors are partly granted by National Board of Digital Health, Denmark and Trifork Software Solutions, Denmark. Conflicts between terminology and EHR information models as obstacles to semantic interoperability: a scientific review Louise Pape-Haugaard a , Anne Randorff Rasmussen a , Pia Britt Elberg a , Stig Kjær Andersen a a Medical Informatics Group, Department of Health Science and Technology, Aalborg University, Denmark Objectives Review of scientific literature on obstacles due to the combination of terminology systems and EHRs information models. Scientific research questions: - Which obstacles to semantic interoperability arise when reconciling terminology systems and EHRs models? References 1 Blobel B, Pharowm P. Analysis and Evaluations of EHR approaches. Methods of Information in Medicine. 2009: p. 162-169. 2 Sato L, Luhn K. CEN/ISO 13606 Pilot Study Final Report. NHS Connecting for Health; 2007. 3 Campbell J, Osornio A, Quiros F, Daniel L, Reynoso G. Semantic Interoperability and SNOMED CT: A Case Study in Clinical Problem Lists. In Kuhn KA, Warren JR, Leong T. MedInfo. Amsterdam: IOS press; 2007. Methods & Material The method used to answer the research questions is shown in Figure 1. Table 1 shows the search strategy for the literature. Boolean operators AND, OR and NOT were used in combination with keywords. Study selection, Data extraction and metaanalysis Step 1: Search and abstract reading followed by a categorization of papers. Categorization done by posing question 1: ’Which perspective is taken in this study?” Step 2: Abstracts from the category ’Terminology and EHR models’ selected for further reading and assessment. Step 3: Obstacles identified by extracting key issues in selected papers. In this case questions such as: ‘which main issue can be recognized across categories?’ 2 Literature review Results of literature search Terminology Pubmed Stating issues Adressing issues EHR model Terminology AND EHR model   Excluded Further review Approach 1 Approach 2 Approach ...n  Included Not stating issues Excluded  1 2 3 OR AND OR NOT Decision support Information model, HL7 CEN, openEHR, Archetyp*, Data model* Terminolog* SNOMED CT Findings In SNOMED CT ambiguity is emphasized in two main areas: Concept composition : ambiguity occurs when concepts are used in combination, resulting in multiple representations. Level of granularity : ambiguity occurs when concepts are granulated at different levels, resulting in semantic conflicts. In EHR information models the ambiguity issue is dependent on the development paradigm and thereby how the data model is exchanged. Data interpretation : ambiguity occurs when imported EHR data are misrepresented, resulting in misrepresentation Syntax issues: ambiguity occurs in the core of the EHR, resulting in incomprehensible data. Semantic issues: ambiguity occurs when a term used is not rigorously described, resulting in lost context. ISSUES OF AMBIGUITY happen when terminology systems and EHR information models are to interact. Multiple mappings: Observations described by an information model may relate to several concepts. Finding appropriate concepts to align with an information model is a very low-practical approach and time-demanding. Semantic ambiguity : Combining data models with terminology can lead to a conflict between the semantics of a given expression. 3 Discussion & Conclusion The solution of the overlap-problem has multiple facets. Well-documented methodological approaches are used in papers to determine whether issues can be solved. However, generic or automatic solutions are not given the same priority as proprietary solutions for legacy systems. Awareness of obstacles is important, but the scientific literature is not specific in its description of these. Issues related to the integration of terminology systems and EHR information models are closely linked to the EHR implementation process. Findings show that ambiguity is key issue. One attempt to solve this issue is to use terminology binding. 4 5 Terminology system Data interpretation Syntactical ambiguity Semantic ambiguity Level of granularity Concept composition EHR 1 EHR 2 Multiple mappings Semantic ambiguity Information model(s) Figure 1 illustrates the method used Table 1 illustrates search strategy Figure 2 illustrates issues of ambiguity Introduction An aim in eHealth is achieving semantic interoperability. Healthcare paradigms are therefore moving towards a time-, location- and treatment- independent care paradigm, where information and knowledge need to be shared among several actors in several locations. In parallel different means to achieve semantic interoperability are explored in areas as terminology modeling and EHR information models. [1,2,3] However, challenges with both terminology systems and EHR information models arise when clinical contents need to be expressed, because of multiple ways to express given concepts. 1