Setting-Specific Factors in Achieving Widespread Adoption of Clinical Decision Support

Our initial aim was to develop a 4-stage model for CDS rule refinement, and based on this model, create and evaluate an Implementer’s Workbench (IW) for CDS knowledge formalization, adaptation, and implementation. A particular emphasis was on Setting-Specific Factors (SSFs) such as local workflow, practice patterns, site characteristics, users, EHR platform, and application and rule deployment functionality. The project required definition of knowledge representation, user interface, and taxonomy of SSFs to be used, and development of initial set of knowledge content for deployment in the area of diabetes management.

A revised goal is to work with the ONC’s Standards and Interoperability Framework’s Health e-Decisions Initiative (HeDI), to help devise a nationally adopted representation of knowledge that can be incorporated broadly by vendors and producers of EHRs into their systems. Our original intent was to work with selected vendors to show feasibility of our approach, but we believe that this work will have greater impact if it is aligned with the HeDI to influence the way all producers of knowledge content and vendors and producers of EHRs expect to receive knowledge content for execution in their systems.

As we have contributed to the definition of the Health eDecision (HeD) schema, we have created a companion model, defined as a highly modular OWL ontology, to complement and describe the information which is eventually serialized and distributed using the schema. This ontology reflects the content of the HeD schema, and the standards it was based upon (the HL7 Arden Syntax and the GELLO expression language, CREF, OMG PRR and CDSC L3), and is built on top of some widely adopted upper ontologies such as SKOS, DULCE and Dublin Core.

Having this rich formal HeD ontology, we are able to create a model-driven HeD editor. This has value in being able to support future transformation of all HeD artifacts into other formal languages and representations, such as those used in various EHR systems.

The source code, is available by following this link to GitHub. For issues or questions, please contact Davide Sottara.

The editor has a visual interactive mode of use, targeted at subject matter experts, and guides the user in the definition of an HeD document's header, expressions, triggers, conditional logic and actions.  The editor leverages the additional semantics contained in the ontology and assists the user by providing a library of commonly used templates or "primitives", also defined using the concepts in the ontology.

The editor is designed and implemented as a (web) service-oriented application, so that it can be easily embedded in broader architectures.

The current user interface is implemented using Javascript technologies:

  • it uses dynamically generated forms to collect information from the user, and relies on Google Blockly to author complex expressions in an intuitive, visual fashion.

The editor allows import of existing HeD documents, as well as creation of new ones. A document can be serialized and exported using the HeD schema, but other automated transformations are possible, based on the formal model underlying the schema, as noted above. Future developments will include the integration of the editor with a (semantic) knowledge repository.

Both the ontologies and the editor are open source, released under a liberal license (ASL2), and built using similarly liberal open source libraries and resources.