Semantic Mapping of Clinical Model Data to Biomedical Terminologies to Facilitate Data Interoperability

The lack of use of a single terminology across all health systems has led to issues of patient data interoperability. This requires data models to be independent of the terminologies when formalising data representations. The paper discusses the principles to base a middleware system to enable semantic mapping of data in the models to formal biomedical terminologies. The principles have been tested against the MoST system, which maps Archetype data to SNOMED CT codes. At first, contextual and non-contextual methods were applied using lexical and semantic procedures to obtain matches. These automated matches were then presented as candidate mappings to clinical modelers to choose from. The aim of the research is to enable clinical modelers to quickly and efficiently codify data at the time of modeling the information through automated processes. The research intends to simplify the task of mapping thereby encouraging standardising data at source by alleviating tedious manual lookups performed traditionally by clinicians. Introduction An unusual feature of health informatics is that biomedical terminology models and clinical data models are developed by separate groups that work independently of each other. Therefore, there are inherent differences in the principles on which each of them is based as well disparities in the semantics of representation. However, there is increasing government recognition for the need to integrate the functioning of all complementing models in health information systems to achieve data interoperability and shared care. To achieve data interoperability it is important that data conforms to some standard, which is adopted by all conforming systems. In health care, formal biomedical terminologies provide such standards to record patient data. Patient data is increasingly being recorded using formal clinical information systems. Data models, such as Archetype Models, form part of such systems by providing structured and constrained representations of clinical recording scenarios. To facilitate data interoperability it is necessary to integrate the data in the models to standard terminologies like Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT). The paper discusses a methodology used to perform the first level of data standardisation i.e., data mapping. The Model Standardisation Using Terminology (MoST) system was developed to test the methodology using openEHR Archetype Models and SNOMED CT. Context and non-context methods using lexical and semantic procedures were employed to find matches. The most appropriate matches resulting from application of filtering rules were presented as candidates to the clinical modeler for mapping. A clinical modeler is a person with medical knowledge who is engaged in the task of modeling clinical data for use in information systems. The enhancement in the precision of the results by applying semantic and filtering techniques will be demonstrated. The time, speed, and quality of the results will then be evaluated against traditional manual term matching and lexical processes.