Summary
The same data are viewed and handled differently by different specialists. For a physician, patient records are separate entities allowing him to treat every patient individually. A database administrator sorts out data elements from the records into tables assigning them data types and data validation rules. Researchers, studying the causality of diseases, group data into demographics, occupational, genetic, etc. Proper management of these diverse coexisting conceptualization schemes requires creating a multifaceted conceptual data model (CDM).
Methodology
This presentation discusses a methodology for developing and validating CDM as a multifaceted logical organization of data assets (DA), which is structured according to the user’s concepts and the task(s) at hand. A CDM facet is a collection of concepts that makes sense and works together in serving a defined purpose, such as database management, patient treatment or business operations. Aligning data assets with processes identifies their behavioral similarities top-down. Its bottom-up counterpart discovers structural similarities (sets of attributes). Grouping DA according to their behavioral and structural similarities creates a conceptualization scheme for a defined purpose.
CDM facilitates extracting meaningful conceptual structures from the complexity of data, enabling both proper management of DA and effective communication among diverse specialists. It also assists in identifying gaps in data capturing. In our approach each CDM facet applies to the same pool of data, presents a complete model, and consists of disjoint classes. This disjoint-ness is essential for business operations, while reporting may employ more complex relationships.
Data model and reference information model (RIM) represent two CDM facets. Data model defines how the data are stored in a database. RIM defines the basic classes of objects, such as Entity, Act and Role in HL7 RIM v3 methodology, from which we slightly diverged. The further creation of CDM facets involves identifying important business objects and subject areas, followed by creation and validation of DA groups according to their behavioral and structural similarities in every subject area. Validation is performed according to predefined rules on classes and facets, and verified by subject matter experts. Consistent color schemes assist in visual validation.
Future work
The ultimate goal is to find an adequate representation of CDM in mathematical objects with fully defined sets of operations among them for every facet as well as for transformations between facets. This methodology will enable consistent validation of CDM and assist in meaningful organization and processing of complex data and information.
Acknowledgement
This work was sponsored by the Center for Information Technology and the Office of the Chief IT Architect of the National Institutes of Health. We are grateful to Dr. John Sharp, Demetrios Kotsikopoulos, Taruna Reddy and Bill Jones for their participation in this project