Design and manage models for different types of data
Information models are essential for the representation of detailed clinical and life sciences data. Models are typically created to understand and standardize data from different sources, and can also be used to facilitate data exchange and sharing. Semedy’s Information Model solution is designed to support data engineering efforts where schemas from multiple data sources have to be represented and harmonized. This solution allows an organization to:
Represent information models and their components (e.g. data elements, value sets, concepts) as interconnected assets that can be created and managed independently
Reuse and distribute models and components within and across organizations
Instantiate information models to generate synthetic data (e.g patient demographics, diagnoses, procedures, lab tests, vitals, cancer staging details, genomic reports, etc.)
Implement automated testing scripts to validate the integrity of the models after updates to individual components
Map and reconcile local models with commonly used reference models (e.g. Fast Healthcare Interoperability Resources (FHIR), Observational Medical Outcomes Partnership Common Data Model (OMOP CDM), or Minimal Common Oncology Data Elements (mCODE™))
Semedy’s Information model solution, implemented using our Clinical Knowledge Management System (CKMS), includes preconfigured and extensible meta-models, extract/transform/load (ETL) pipelines for input and/or export, configurable views, queries, and reports. Previously defined models, data schemas, and data dictionaries can be easily loaded and periodically refreshed using our ETL framework. Commonly used models, ontologies, and terminologies are maintained and distributed by Semedy, along with guidance on how to map and reconcile local models and components to reference standards. With periodic refreshes from reference sources, or schema updates from local sources, CKMS becomes the central repository for curating, mapping, and distributing information models and associated components and metadata.
Using our CKMS platform, the demonstration will showcase mCODE™, a new set of open source reference models created for the oncology community. We will explain how all 27 mCODE profiles were implemented within CKMS, including property definitions, value sets, and terminology concepts from multiple standard sources. The complete implementation includes over 56,000 interconnected assets that can be searched, queried, and validated using the platform’s built-in semantic reasoner. We will also provide examples of synthetic data instances and mappings between mCODE and FHIR components, along with queries to identify obsolete concepts that might be present within value sets.