2017 HSR&D/QUERI National Conference

4075 — A Clinical Care Ontology for an Enhanced Electronic Health Record (EHR) and Clinical Reasoning

Lead/Presenter: Leo Obrst
All Authors: Obrst LJ (MITRE) Kazura A (MITRE) Bennett S (MITRE) Taylor M (MITRE) Granger E (MITRE) Klaus S (MITRE) Tu S (Stanford University) Ward M (VA) Nebeker J (VA)

Objectives:
The Clinical Care Ontology (CCO) provides a framework of meaning and relationships for aspects of clinical reasoning and care planning. Key conceptual nodes in the ontology include conditions, interventions, observations, experiences, and utilities/preferences. When represented in and characterized by the ontology, clinical data can be queried for decision support. One can readily see which interventions treat which conditions, and that some interventions may conflict with patient-preferred experiences. The question that this study answers is: can ontologies and associated rule formalisms significantly model evidence-based patient-centered clinical concepts and determine if they can contribute to enhanced Electronic Health Record (EHR) decision support and visualization, ultimately improving care delivery and outcomes for veterans.

Methods:
We are developing a CCO emphasizing concepts critical to patient-centricity. Additionally, we are developing more expressive rules to support automated reasoning beyond the capability of standard query languages. We are evaluating the ontology using competency questions derived from requirements of four major use cases based on assessment and diagnosis of significant co-morbid conditions, as developed by clinical experts. The central focus is on a patient-centered care plan that emphasizes goals and preferences.

Results:
We are developing and evaluating the second iteration of an OWL ontology, assessing its ability to answer use-case derived competency questions formulated as queries and rules, with accompanying reasoning. In the first sprint of the project's evaluation, the ontology answered 75% of the competency questions, and failed to answer 25%, with many of the latter occurring because of missing supporting reference data (e.g., specific medications). Further, we are developing exemplary interoperability mappings to existing terminologies and ontologies to demonstrate how they can be used to adapt, expand, or link to the ontology.

Implications:
Our development effort has produced an ontology that can answer most of the competency questions and support most of the cognitive needs identified in the use cases. The addition of a rule layer has enriched the reasoning capabilities over complex instance data, thereby indicating potentially greater clinical decision support capabilities for future systems.

Impacts:
Ontologies offer great potential to support enhanced EHR decision support and visualization, automated inferencing, and discovery of relationships.