Ensuring semantic interoperability in the course of clinical document exchange using metadata registry related technologies

Abstract

Introduction: Data standardization is crucial to facilitate understanding and sharing data across diverse translational studies. Common data elements (CDEs) based on the ISO/IEC 11179 Metadata Registry (MDR) standard provide well-defined and structured data that are feasible to incorporate in clinical documentation in such a way that supports semantic interoperability. However, structural limitations of MDR have been an obstacle for either composing CDEs in clinical forms or interpreting them from clinical forms. Though we developed simple extended relationships, we found it only covered simple relationships. The additional semantic relationships are needed. Meanwhile, a clinical document is an essential tool to collect clinical information related to individual health. For comprehensive semantic representation and clear definition of clinical data, ISO/IEC 11179 standard based metadata, including CDEs has been used to compose clinical documents. When the decision is made to share clinical data through clinical documents, the data should be checked first for completeness and to ensure that no errors were introduced during the sharing process. The process of data validation significantly adds to the complexity of the data sharing process, but is critical to maintain the integrity of clinical data and to ensure high-quality data. Finding proper data elements from numerous data elements is essential to use them in metadata implemented clinical documents for effective semantic data exchange. It is required to develop an ontology to classify and search data elements. Methods: We reviewed the CDEs currently being use in a clinical setting to understand the inter-related data elements. We then developed use case scenarios to describe common representational challenges in the inter-related CDEs, and extended the existing composite CDEs to address the identified challenges in data presentation, and data transformation. For developing validation process, we first defined what complex clinical document is as applying the developed several semantic relationships of data elements. As considering these semantic relationships of data elements in clinical documents, we developed the process of syntactic and semantic validation of clinical documents, and specified the list of validation attributes. Meanwhile, for developing Clinical Metadata Ontology (CMO), we adopted the General Formal Ontology method with a manual iterative process comprising five steps

(1) defining the scope of each ontology, (2) identifying concepts, (3) assigning hierarchical relationships among concepts, (4) development of properties (e.g., synonyms, preferred term, and definitions) for each concept, and (5) evaluating developed ontologies. Results: We developed three types of extension to composite data element such as Repeated composite data element to resolve observational clinical data presentation challenges, and Dictionary and Template composite data elements to support knowledge data presentation and data model transformation respectively. In doing so, we defined four new constraints of CDEs in composite data element such as Dependent, Operated, Ordered, and Required. We also defined new types of the CDE such as Hybrid relationship. Base on the extension of semantic relationships of CDEs, we also developed the process of syntactic and semantic validation of clinical documents, and specified the list of validation attributes. We demonstrated and evaluated the feasibility of composite relationships as presenting a practical use case. Tree structure based CMO was developed with 200 concepts under the four first-level terms including Description, Event, Finding and Procedure. CMO has 1060 synonyms for 151 (76%) CMO concepts, and 400 definitions for 137 (69%) CMO concepts. The Web-based CMO Browser and the CMO matched BMeSH DE Browser provide convenient access to CMO and help to understand how CMO concepts are matched to DEs in the practical clinical documents (http://www.snubi.org/software/cmo). CMO is the ontology as (1) a classification scheme for data elements for clinical documents, (2) an integration tool for data elements from a diversity of clinical documents, (3) a proper clinical data-organization scheme for data elements for developing clinical information systems including PHRs, and (4) a component ontology expendably connected to other healthcare data domains such as personal lifelog data, which is supported by MELLO, and personal genomic data, which is supported by Health Avatar Project. Conclusions: This paper investigated the feasibility of representing the complex clinical data in clinical forms with the extended MDR based extended semantic composite relationships and constraints. Our results indicates that our approach is able to comprehensively represent the CDEs in two perspectives

1) the form-level data is represented by data item-level data without loss of the contextual semantic relationships between data elements and forms, 2) data integration and transformation across different standardized dictionaries or data models. The preliminary results of the present research can be used as a reference for future development of extended semantic composite relationships-applied system. It also emphasizes the extended MDR based value validation can help data error handling and provide clear error limits on data sharing. CMO is the ontology for classification of data elements. We can expect to search appropriate data elements and use them effectively in clinical documents.